View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

rr
Monthly Labor Review
Reader
U.$. Department of Labor
Bureau of Labor Statistics
1975
Bulletin 1868




L ib r a r y o f C o n g r e s s C a ta lo g in g in P u b lic a tio n D a ta

Main e n try under t i t l e :
M onthly la b o r rev iew read er*
C om pilation o f a r t i c l e s on fin d in g s o f th e Bureau
o f Labor S t a t i s t i c s , p u b lis h e d betw een J a n . 19&9, and
Jan* 1975, in th e M onthly la b o r re v ie w .
S u p t. o f Docs, n o .: L 2.2:M 76/2
1 . Labor su p p ly —U n ited S t a t e s —A d d re sse s, e s s a y s ,
l e c t u r e s . 2. Labor econom ics—A d d re sse s, e s s a y s ,
l e c t u r e s . I . U n ited S t a t e s . Bureau o f Labor
S t a t i s t i c s . I I . U n ited S t a t e s . Bureau o f Labor
S t a t i s t i c s . M onthly la b o r re v ie w .
HD5724.M65
3 3 1 -1 , 1*0973
75-619070




Monthly Labor Review
Reader
U.S. Department of Labor
John T. Dunlop, Secretary
Bureau of Labor Statistics
Julius Shiskin, Commissioner
1975
Bulletin 1868

For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402, GPO Bookstore, or
BLS Regional Offices listed on inside back cover. Price $5.50
Make checks payable to Superintendent of Documents
Stock Number 029-001-01403-1
Catalog Number L2.3:1868







C O NTENTS
Page

Introduction.....................................................................................................................................................................

1

CHAPTER I. MEASURING EMPLOYMENT AND UNEMPLOYMENT
Unemployment statistics and what they mean

John E. Bregger.....................................................

6

Comparing employment estimates from household
and payroll surveys

Gloria P. G reen ......................................................

14

A 25-year look at employment as measured by
two surveys

Christopher G. Gellner ........................................

26

Analyzing the length of spells of unemployment

Hyman B. Kaitz ...................................................

36

Black and white unemployment: the dynamics
of the differential

Curtis L. G ilroy.....................................................

46

Quits in manufacturing

Paul A. Armknecht and John E. Early................

56

Job losers, leavers, and entrants: a cyclical analysis

Curtis L. Gilroy and Robert J. Mclntire..............

63

Comparing employment shifts in 10
industrialized countries

Constance Sorrentino ..........................................

68

The U.S. labor force: projections to 1990

Denis F. Johnston.................................................

78

Determining the labor force status of men
missed in the census

Deborah P. K lein...................................................

88

Discouraged workers and changes in unemployment

Paul 0 . Flaim ........................................................

95

Education of workers: projections to 1990

Denis F. Johnston.................................................

104

CHAPTER II. CHANGES IN THE LABOR FORCE

CHAPTER III. SPECIAL GROUPS IN THE LABOR FORCE
The economic status of families headed by women

Robert L. S te in ......................................................

Where women work— an analysis by industry
and occupation

Elizabeth Waldman and Beverly J. McEaddy . . .

116
124

The employment situation of Vietnam-era veterans

Kopp Michelotti and Kathryn R. Gover..............

135

Occupational characteristics of urban workers

Christopher G. Gellner ........................................

144

Employment and unemployment among Americans
of Spanish origin

Roberta V. McKay ...............................................

156

Multiple jobholding in 1970 and 1971

Howard V. Hayghe and Kopp Michelotti ..........

161




CO NTENTS-Continued
Page

CHAPTER IV. PRICE MEASUREMENT AND PRICE TRENDS
Toward comprehensive measurement of prices

Allan D. Searle .......................................................

170

Updating the Consumer Price Index— an overview

Julius Shiskin .........................................................

Ig 4

Measuring changes in industrial prices

Joseph A. Clorety, Jr............................................... 202

Determining the effects of quality change on the CPI

Jack E. Triplett....................................................... 209

The use of price indexes in escalator contracts

Francis S. Cunningham.......................................... 215

Postwar price cycles: a new chronology

Geoffrey H. Moore ................................................ 220

Prices in 1972: an analysis of changes during Phase 2

JoelPopkin

........................................................... 227

CHAPTER V. PRODUCTIVITY AND TECHNOLOGICAL CHANGE
Industry indexes of output per man-hour

Jerome A. Mark ....................................................

236

Productivity and costs in the private economy, 1973

J. R. Norsworthy and L. J. Fulco ........................ 241

Measuring productivity in the Federal Government

Charles Ardolini and J. Hohenstein...................... 248

Modernization and manpower in textile mills

Rose Zeisel .............................................................

Productivity in telephone communications

Horst Brand ........................................................... 264

Labor requirements for construction of singlefamily homes

Robert Ball and Larry Ludwig ............................

Productivity and unit labor costs in 12 industrial
countries

Patricia Capdevielle and Arthur Neef ................. 274

256

271

CHAPTER VI. WAGES AND EARNINGS
The relationship between changes in wage rates and
in hourly earnings

Victor J. Sheifer ....................................................... 284

Developing a general wage index

Norman J. Sam uels................................................ 292

Usual weekly earnings of American workers

Paul O. Flaim and Nicholas I. Peters ................. 298

Youth unemployment and minimum wages

Thomas W. Gavett

Trends in overtime hours and pay, 1969-74

Diane N. Wescott ..................................................

Occupational rankings for men and women by
earnings

Dixie Sommers

.................................................... 327

Measuring employee compensation in
U.S. industry

Alvin Bauman

....................................................... 345

Measuring annual earnings of household heads
in production jobs

Robert L. Stein and Paul M. Ryscavage




ii

................................................ 309

319

............. 353

CONTENTS—Continued
Page

CHAPTER VII. INCOME DISTRIBUTION AND PURCHASING POWER
Two measures of purchasing power contrasted

Paul M. Schwab

................................................... 364

Compensation per man-hour and take-home pay

Jack Alterman ...................................................... 376

Exploring the distribution of earned income

Peter Henle ............................................................. 386

Earnings and family income

Robert L. Stein and JaniceNeipert Hedges

...

398

CHAPTER VIII. UNIONS, BARGAINING, AND THE WORKPLACE
Changing policies in public employee relations

Joseph P. Goldberg

............................................. 412

Union membership among government employees

Harry P. Cohany and Lucretia M. Dweey .......... 422

Women’s participation in labor organizations

Virginia Berquist

When workers are discharged— an overview

Robert W. Fisher .................................................. 435

Productivity bargaining in Britain

H. M. Douty

New approach to occupational safety and
health statistics

Lyle R. Schauer and Thomas S. Ryder .............. 455

................................................. 428

........................................................ 449

CHAPTER IX. WORK SCHEDULING AND WORK LIFE
The future of work: three possible alternatives

Denis F. Johnston

Trends in labor and leisure

Geoffrey H. Moore and Janice NeipertHedges . 471

What’s wrong with work in America?— a review
essay

Harold Wool .......................................................... 480

A look at the 4-day workweek

Janice Neipert Hedges

A new type of working life table for men

Howard N. Fullerton, Jr......................................... 492




iii

............................................... 462

......................................... 487




INTRODUCTION

The Reader’s nine chapters include—

The Monthly Labor Review— the principal outlet
for the creative thinking, analytical skills, and sta­
tistical series of the Bureau of Labor Statistics and
its professional staff— is a valuable reference source
for studies of methods of collecting and compiling
labor statistics and for interpretations of their meaning
and significance. The Monthly Labor Review Reader
presents some of the best of its recent articles. The
selections were made by the staff of the Monthly
Labor Review to reflect (1) current economic or social
policy issues facing the Nation at the present time;
(2) progress in the development of statistical con­
cepts and methodology; and (3) the statistical practices
of other countries. Since the Bureau’s constant emphasis
is on the quality and integrity of its output, atten­
tion— in the selection process— was also paid to the
quality of the writing and analytical logic.
As the pioneer statistical agency in the Federal
Government, the Bureau— over its 90 years— has de­
veloped a comprehensive core of labor statistics covering
employment and unemployment, prices and living con­
ditions, wages and industrial relations, productivity
and technological change, occupational safety and health,
and special economic studies. This Reader includes
selections from each of these major areas. Although
the Monthly Labor Review itself is a mature 60
years old, the articles selected for this new volume
span roughly only the last 15 years: This restriction was
necessary to keep the Reader to manageable size
and limited to articles that are relevant to today’s
problems.
Nor does this Reader reflect the full scope of the
Monthly Labor Review's contents. Each month, the
Review reports on current developments in industrial
relations and labor law, reviews books, and presents
almost 40 pages of the Bureau’s latest statistics. The
MLR— and this Reader as well— should be of interest
to labor, business, and government officials, as well
as to research scholars, students and teachers of eco­
nomics, industrial relations, management and public
relations.




Chapter I. Measuring Employment and Unemploy­
ment: The Bureau’s two series on employment (from
the household and from the establishment surveys)
and the unemployment rate are major national economic
indicators crucial to assessments of the state of the
economy. The unemployment rate is used also to measure
progress towards meeting the goals of the Full Employ­
ment Act of 1946. More recently, the administration
of the Comprehensive Employment and Training Act
of 1973 requires that State and local area unemploy­
ment estimates that are consistent with the national
rate be used in the allocation of Federal revenue sharing
funds. The properties of these Bureau measures— along
with analyses of their trends, both in the U.S. and
other industrialized countries— are included in this
chapter, as are related series on the duration of
unemployment and the rate of voluntary separations
of workers in manufacturing industries— the so-called
quit rate. Another article discusses the relative unem­
ployment experiences of blacks and whites in the various
phases of the business cycle. A final article includes a
cyclical analysis of workers who come into and leave the
labor force— job losers, leavers, and new entrants.

Chapter II. Changes in the Labor Force, focuses on
some problems in labor force accounting— men missed
in the Decennial Census of Population and discouraged
workers who are not in the labor force because
they think they could not find a job. It also includes the
Bureau’s most recent projections of the labor force
and the educational attainment of workers to 1990.
Chapter III. Special Groups in the Labor Force,
deals with some problems and employment characteristics
of special groups of workers: Women, Vietnam-era
veterans, urban workers, Americans of Spanish origin,
and multiple jobholders.

1

of two widely used statistical measures of workers’ pur­
chasing power: the BLS series on real net spendable
earnings and the Department of Commerce’s series on
per capita real disposable personal income. These series
showed different trends from the mid-1960’s to the
early 1970’s and raise questions about the actual
course of purchasing power during the recent period of
serious inflationary pressures. Another article reports on
the increasing importance of the income of wives as a
proportion of total family income— up 3 percentage
points from 1958 to 1969. Another author reports on a
slight but persistent trend toward greater income in­
equality in this country.

Chapter V. Price Measurement and Trends: This
chapter describes improvements under way in the
Bureau’s consumer and industrial price measures. Im­
provements in concept and methodology and expansions
of scope are all covered. A comprehensive descrip­
tion of the massive program under way to overhaul
the Consumer Price Index is the subject of another
article. The development of a new comprehensive
price measure— its objective, its construction, its assets
and shortcomings, is the subject of still another article.
Other subjects covered include the use of price indexes
in long-term contracts to escalate wages and other
income payments, the cyclical behavior of price changes
during the postwar period, and price experience during
a recent period of price stabilization— 1972.

Chapter VIII. Unions, Bargaining, and the Workplace,
groups several articles on collective bargaining in the
public sector, for example, women in unions (although
their number is increasing, their participation in union
leadership positions is not) and productivity bargaining
in Great Britain. A final article in the chapter describes
the Bureau’s new survey— one of the largest in the
history of the country— to collect data on the incidence
of occupational injuries and illness in U.S. industry.

Chapter V. Productivity and Technological Change,
includes descriptions of several Bureau measures of
productivity; an analysis of productivity and cost
performance in the private sector in 1973, a year
in which long-run productivity growth slowed; pro­
ductivity in construction, textiles, and the telephone
industries, and in a substantial portion of the Federal
Government. The Federal Government measure, still in
its early stages, included more than 850 output indica­
tors, and covered 61 percent of the employment in
Federal civilian government. A comparison of produc­
tivity and unit labor costs in 12 industrial countries
is the subject of the final article of this chapter.

Chapter IX. Work Scheduling and Worklife: One
article in this chapter suggests that the role of work
in our society will be affected greatly by changes
in fertility rates: Several possibilities are presented,
ranging from the assertion that work will continue
to provide a central focus for personal satisfaction
and status achievement to the argument that our
traditional work ethic is undergoing rapid erosion,
to be displaced by new criteria of personal worth
and achievement
unrelated to work performance.
Another article reports that “little objective evidence
exists to support an inference of a rising wave of
discontent among workers, associated directly with
the nature of their jobs.” But the author goes on
to point out that “most of the available statistical
indicators are clearly much too aggregative to serve
as reliable indexes of worker discontent. Statistical
series such as productivity and labor turnover were
designed for quite different purposes.” One article in
the chapter examines the relations between labor and
leisure and another the 4-day workweek. The con­
cluding article describes two different methods to
measure the number of years, on the average, men
will work. It contrasts “generation” worklife tables
against “period” tables in an attempt to provide
a more realistic means of estimating the length of
worklife.

Chapter VI. Wages and Earnings, examines such topics
as the relationship between changes in wage rates and
hourly earnings. What are a worker’s “usual earnings,
annual earnings of household heads? What is the effect
of the Federal minimum wage on youth unemployment?
How do we measure employee compensation?” One
article discusses the limitations of existing wage meas­
ures and describes plans for a new index of change in
total employee compensation designed to measure the
full range of employment costs. Development of the new
index is now under way in the Bureau. It will be pub­
lished in stages: First, the measure of change in wage
rates is scheduled for publication in 1976 and the
measure of change in full employment costs in 1977.
Chapter VII. Distribution o f Income and Purchasing
Power: Widespread interest in welfare policies and in
measures of the economic hardship of unemployment
both point to the need for better information on
income distribution and worker purchasing power. This
chapter describes the differences in concept and scope




2

Authors of all the articles included in this new
Reader were on the BLS staff at some time in
their careers, though not always at the times the
articles were written. Virtually all of these articles
were originally published between January 1969 and
January 1975, a period in which many innovations
were made in editorial policies and typography of
the journal. Editors who served on the Monthly Labor
Review staff for at least part of that time included:
Herbert C. Morton, Editor-in-Chief; Henry Lowenstern,
Executive Editor; Georgena R. Potts and Robert W.
Fisher, managing editors; Olivia G. Amiss; Elizabeth E.




Barnes, Eugene H. Becker, Catherine C. Defina, Barbara V.
Freund, John Gusman, Mary D. Hogya, Mervyn S.
Knobloch, Constance S. McEwen, Craig E. Polhemus,
Carol A. Rosen, Louise M. Schlader, and Eugene Skotzko.
Their efforts and the devoted work of many other
anonymous BLS employees— data collectors, field
agents, clerks; statisticians, economists, programmers—
made this volume possible.
I hope the users of this Monthly Labor Review Reader
will profit from the collective wisdom contained in
the selections chosen from the Review.
JULIUS SHISKIN
Commissioner
June 1975

3




Chapter I. Measuring Employment and Unemployment




Unemployment
statistics
and what
they mean
“ J o b l e s s r a t e drops to 6.0 percent,” read a typical
headline in newspapers throughout the country on
October 8, 1971. The change— from 6.1 to 6.0 per­
cent of the labor force between August and Septem­
ber 1971— was reported accurately, but the implica­
tion of an improvement in the employment situation
was misleading. This is so because the unemploy­
ment rate is obtained by a sample survey. Any
change of one-tenth of one percent may be attribut­
able to sampling error and, therefore, not statistically
significant.
Unemployment, of course, is more than a statistic
that measures our economic well-being and the de­
gree to which immediately available manpower is
utilized. Unemployment statistics represent peo­
ple— people trying to support their families or
augment family income, people seeking their first
jobs, people changing jobs, people losing jobs, but
first and foremost, people. Whether viewed as a
measure of economic well-being or as people with
employment difficulties, however, the data are often
misused, misunderstood, and even criticized.

Data sources and concepts
To set the stage, it is first of all desirable to review
briefly the procedures by which these important sta­
tistics are collected and how the measure of unem­
ployment itself is defined. National statistics on un­
employment are derived from the Current Population
Survey (CPS), a monthly sample by personal inter­
view of approximately 50,000 households. The sur­
vey is conducted by the Bureau of the Census for
the Bureau of Labor Statistics. Persons are classified
through a series of questions which determine
whether they were employed and, if not, whether they

John E. Bregger is an economist in the Division of Employ­
ment and Unemployment Analysis, Bureau of Labor
Statistics.

From the Review of November 1971



The jobless rate in perspective:
some common misconceptions
about what the data represent,
along with pointers on how
to interpret them
JOHN E. BREGGER

were looking for work or were not in the labor force.
The data relate to the status of individuals in the
week including the 12th day of the month (the “ref­
erence” or “survey week”) and are collected in the
subsequent week (the interview week).1
Employed persons are those who perform a min­
imum of an hour’s work for pay or profit during the
reference period; also included are those who are
temporarily absent from a job or business for such
reasons as illness, vacations, or strikes, as well as
persons who work 15 hours or more a week without
pay in a family farm or business.
To be classified as unemployed, the individual
must not have worked at all during the reference
week. In addition, he must have taken some specific
steps to obtain a job in the previous 4 weeks, such
as applying directly to an employer, or to a public
employment service, or checking with friends or rela­
tives, and being available for work at the time of the
survey. Persons on layoff or waiting to begin a new
job (within 30 days) need not meet these jobseeking
requirements to be classified as unemployed. Those
persons who are neither employed nor unemployed
are “not in the labor force.” Information is collected
regularly on this group as well, many of whom are
housewives. It is worth noting that at no time during
the course of the interview is the term “unemployed”
used, and, as a consequence, the respondents them­
selves frequently do not know how they will be clas­
sified. Furthermore, no response is elicited as to
whether an individual has applied for or is receiving
unemployment compensation payments.
At the present time, each household in the survey
represents approximately 1,300 households through­
out the United States, and, similarly, one person in
the sample represents 1,300 in the population. There­
fore, a total of 5 million unemployed would be repre­
sented by 3,800 individuals. On the surface, this
appears to be a small sample for such an important
figure. However, the survey is the largest monthly

household survey in the world, some 50 times larger
than many of the national public-opinion polls, and
uses a scientifically selected sample that is studied
and reviewed continually. Moreover, the fact that
the sample yields reasonably consistent results
month after month lends credence to the procedure.
Nonetheless, the fact the data are taken from a sam­
ple does mean that a degree of sampling error exists.
The statistics on employed and unemployed per­
sons are tabulated to show a wide variety of charac­
teristics: sex, age, color, educational attainment,
marital status, household relationship, whether work­
ing full or part time or seeking full-time or part-time
work, duration of unemployment, reasons for being
unemployed, major activity (for young persons— in
school or other), and industry and occupation for
those employed and previous industry and occupa­
tion, if any, for the unemployed (industry and oc­
cupation of last full-time job lasting 2 weeks or
more). Cross-classifications of a number of these
characteristics are also available, such as by color,
sex, and age.

duction, are widely used in seasonally adjusted form,
the adjusted labor force data are more comparable
with them. There is a trade-off, however; seasonal
adjustment tends to “depersonalize” the unemploy­
ment data.
In seasonally adjusting the unemployment esti­
mates, as it has for many years, the Bureau of Labor
Statistics uses a traditional ratio-to-moving average
method.2 The unemployment rate is derived by
dividing the sum of four seasonally adjusted com­
ponents (unemployed persons 16-19 and 20 and
over, by sex) by the civilian labor force, which is
itself the sum of 12 seasonally adjusted components.3
Therefore, there are no direct seasonal adjustment
factors for the rate itself but only for its compon­
ents. However, implicit seasonal factors for the rate
may be derived as the ratio of the rate, not seasonally
adjusted, to the rate, seasonally adjusted.
Statistical significance

After the seasonal adjustment process has sorted
out the usual, recurring, and largely noneconomic
events from the more significant underlying, develop­
ments, there remains an unemployment change from
1 month to the next which may or may not be
“statistically significant.” Because the unemployment
estimates are derived from a probability sample, they
may, of course, differ from the figures that would
have been obtained if it were possible to take a com­
plete census using the same questionnaire and pro­
cedures. In other words, the data are subject to some
degree of sampling error, which must be taken into
account in evaluating changes in the data.

Seasonal adjustment

There are a number of seasonal fluctuations in em­
ployment and unemployment that occur during the
year. These include crop seasons, weather condi­
tions, opening and closing of schools, holiday buying
periods (Christmas and Easter, for example), and in­
dustry production schedules. To cite perhaps the
most dramatic shift, there is a tremendous influx of
young people into the labor market in June after
school is out— between May and June of 1971, for
example, the labor force showed a net increase of
nearly 1.9 million persons, with 1.1 million initially
unsuccessful in their job search.
To determine the economic meaning of a month’s
data relative to the previous month or months, it is
essential to differentiate between the change that
usually occurs in the month and the change, if any,
that exceeded the normal, or expected, change.
Therefore, all of the major labor force estimates are
“seasonally adjusted” to permit an easy— and more
revealing— comparison of data for 1 month with
any other.
Without this separation of the seasonal component
of changes, the continuing trend in the labor market
situation would be more difficult to discern. More­
over, since most other economic data, such as esti­
mates of Gross National Product and industrial pro­




The “standard error” is the measure of sampling
variability, that is, the variations that might occur
by chance because only a sample of the population
is surveyed. The chances are about 2 out of 3 that a
sample estimate would differ from the results of a
census by less than the standard error; the chances
are 9 out of 10 that it would differ by less than 1.6
times the standard error; the chances are 19 out of
20 that the difference would be less than twice the
standard error. In its analysis of labor force data,
the Bureau uses 1.6 times the standard error as a
basis for determining the sampling error of an esti­
mate or a change of an estimate from one point in
time to another.
For total unemployment, the
standard error is approximately 3 percent, and, at
1.6 times the standard error, the error on an estimate
of 5 million is on the order of plus or minus 150,000.
7

Thus, the term “significant,” when applied to the
unemployment numbers, is used whenever the change
in the number from one period to another exceeds
1.6 times the sampling error of the change. When an
apparent change is within this confidence interval,
the likelihood that a change actually occurred is
diminished. For example, a change in the present
level of total unemployment should exceed 150,000
from 1 month to the next to be deemed significant
statistically. Similarly, the national unemployment
rate would have to change by 0.2 percentage point
or more on a monthly basis to be significant.4
Changes that are smaller than these can reasonably
be attributed to sampling variations.
As a general rule, smaller numerical estimates
have higher relative errors. In the case of unemploy­
ment rates, the absolute error is greater when the
labor force base is comparatively smaller and also

when the rates are higher. Therefore, the jobless rate
for female Negro teenagers, of whom there were
350,000 in the labor force in 1970, would have an
exceptionally high sampling error for a month-tomonth change, whereas the error for married men
(of whom there were 38.9 million) would be quite
small. Sampling error, of course, is successively less
when the data to be compared are averaged over
successively longer time spans, such as quarterly, an­
nually, and so on. Table 1 illustrates the standard
error of change for a number of unemployment rates,
representing the major labor force groups.
Although “significant” is a statistical and there­
fore technical term in the interpretation of labor
force developments, it is often used in a more gen­
eral sense, and there may be instances in which con­
fusion arises over the use of the word. If the overall
jobless rate declines by 0.2 percentage point from
1 month to the next, for example, this is a “sig­
nificant” change statistically, in the sense that there
is a very small probability that it would have resulted
solely by chance (such as who happened to be
selected in the sample). However, the decline would
not be significant in the sense of being a very large
movement. Similarly, a change could be interpreted
as being “significant” from a policy point of view if
it reflected continued improvement or represented
a change in direction. It is clear, therefore, that care
must be taken in recognizing the multiple meanings
of the word when considering the numbers on em­
ployment and unemployment.

Table 1. Estimated error at 1.6 standard error for change
in selected seasonally-adjusted unemployment rates
Unemployment
rate in
April
1971

Monthly

Total (all civilian workers)________
Men, 20 years and o v er............. .
Wonien, 2 0 years and over..............
Both sexes,16-19 years________
Married men........ I .........................

6.1
4.4
6 0
17.2
3.1

.21
.24
.35
1.15
.21

.15
.17
.25
.83
.15

.09
.11
.16
.53
.09

Full-time workers........... ...............
Part-time workers_____ ________

5.5
9.4

.22
.70

.16
.50

.10
.32

Negro and other races.........................
Wen, 20 years and over..................

10.0
6.8
9.3
32.1

.80
.94
1.24
4.77

.56
.66
.87
3.34

.34
.40
.53
2.05

White....................................................
Men, 20 years and o v e r.......... .
Women, 20 years and over........ .
Both sexes, 16-19 years..................

5.6
4.1
5.5
15.2

.21
.24
.37
1.17

.15
.17
.26
.83

.08
.11
.17
.53

Managers, officials, and proprietors.

3.8
3.3
1.6
5.2
4.5

.24
.43
.33
.48
.73

.17
.31
.23
.34
.52

.11
.19
.15
.22
.33

Blue-collar workers.............................
Craftsmen and foremen........... .......
Operatives...... .......................... .....

7.4
4.5
8.6
10.2

.39
.52
.62
1.18

.28
.37
.44
.84

.18
.23
.28
.53

Service workers______________ _
Farm workers....................... ...............

6.3
1.8

.60
.51

.43
.39

.27
.23

Finance and service industries.......

6.3
9.6
7.0
7.5
6.3
4.0
6.5
5.3

.27
1.21
.46
.63
.68
.73
.52
.46

.19
.86
.33
.45
.48
.52
.37
.33

.12
.54
.21
.28
.31
.33
.23
.21

Government workers...........................
Agricultural wage and salary workers.

2.8
6.1

.38
1.71

.27
1.21

.17
.77

Category

Percentage error on change
at 1.6 standard error 1
Quarterly

Annual

RACE

What the overall rate doesn’t tell

Although the total unemployment rate gets the
headlines, it is not always the most meaningful meas­
ure of the situation and sometimes conceals as much
as it reveals. As a global estimate, it measures job­
lessness among all groups of workers— men and
women, white and black, young and old, urban and
rural, experienced and inexperienced. For example,
it incorporates those who experience very high rates
of unemployment, such as black teenagers, and those
with very low rates, notably married men.

OCCUPATION
White-collar workers_____________

INDUSTRY
Nonfarm wage and salary w orkers...
Construction.................................
Manufacturing_______ ________ _
Durable goods.................... .........
NonduraBle goods____________
Transportation~and public utilities.

The Bureau has continually expounded on the ag­
gregate measure of unemployment in its various
studies of the economic situation and in its monthly
“Employment Situation” press release. Key groups
are identified— by age and sex, color, persons seek­
ing full-time or part-time jobs, occupation and in­
dustry of last job, duration of unemployment (the

> For consecutive periods only; error for nonconsecutive periods slightly greater.




8

number of weeks persons are seeking work), reasons
for unemployment, and so on.
To show the importance of a thorough analysis of
disaggregated monthly unemployment figures, two re­
cent examples are presented.
1. As reported on October 2, 1970, the overall
jobless rate jumped from 5.1 percent in August to
5.5 percent in September 1970. Anyone who caught
only the newspaper headline unfortunately got a
misleading impression. However, the Bureau’s analy­
sis emphasized that the increase was wholly among
16- to 24-year-olds and adult women, many of whom
were new entrants or re-entrants to the labor force;
the jobless rate for males 25 and over— often thought
of as primary workers and also as breadwinners—
actually remained unchanged.

suggest areas for investigation. Trends among a num­
ber of worker groups should be closely watched—
for example, married men, blue-collar workers, man­
ufacturing workers, full-time workers, and those who
lost their last job.
Secondly, for those categories that are analyzed,
it is necessary to determine what changes are statis­
tically significant. As was discussed earlier, develop­
ments for particular labor force groups are generally
important only if deemed significant statistically.
Changes can also be analyzed when their movements
become statistically significant over several months
— thus, representing a short-term trend— even when
they may not be significant in consecutive months.
Third, the analyst studies those groups with higher
than average unemployment rates. This is typically
true of black workers, teenagers, women, and con­
struction workers, among others. For example, the
ratio of Negro unemployment to the total level is
nearly twice the proportion of Negro workers in the
labor force, and partly for this reason their job sit­
uation is closely watched.
Finally, it is important to view changes in unem­
ployment not just for the single month being exam­
ined, but also from a longer term perspective. Charts
can greatly assist in this determination. During the
1970 economic downturn, for example, the increase
in joblessness among adult men greatly exceeded ad­
vances among other age-sex groups, even though
their rate was comparatively low and many of the
individual monthly changes were small. Moreover,
current developments are frequently compared with
highs or lows of previous periods.

2. A few months earlier, the unemployment rate
was reported to have risen from 4.8 percent in April
to 5.0 percent in May. Following technical standards
of statistical significance, the rise of 0.2 percentage
point was barely more than a borderline change. As
the continuation of a sharp upsurge since the first of
the year, however, the increase could be viewed with
greater meaning. Moreover, the components of the
change were striking: the jobless rate for adult men
(20 and over) rose by 0.3 percentage point and that
for adult women shot up by 0.7 percentage point;
these very significant increases were countered by a
drop in the teenage rate from 15.7 to 14.3 percent,
which, of course, held the overall rate to its small
increase.
Many different interpretations could be placed on
the data that emerge from these two examples. A
purely surface analysis might suggest that the Augustto-September increase was far worse than that from
April to May, because the overall rate increased
twice as much. However, a more penetrating evalua­
tion leads to the conclusion that the May increase
could well be the more meaningful one.

Cross-currents in the labor force

It is important to note that monthly statistics of
the labor force and unemployment conceal a vast
number of movements between labor force cate­
gories. Typically, about half of the unemployed in
1 month will have found jobs or left the labor
force in the next, and about an equal number will be
newly unemployed. This is evidenced by the fact
that more than 14.5 million persons experienced
some unemployment in 1970, contrasted with an
average monthly level of 4.1 million. Similarly, there
are many people moving into and out of the labor
force each month. These “gross flows” are illustrated
in table 2, which presents March-April 1971 changes
in the employment status of the population 16 years
and over.

How is it determined which unemployment
changes are of importance? Several factors must be
considered. First of all, it is necessary to identify the
labor force groups that have the greatest economic or
social significance, both in general and in a particular
period. If employment in a certain industry or oc­
cupation is changing, the jobless rate for this indus­
try or occupation should be examined. Developments
for Negroes are often compared with those of whites.
Developments in other economic time series— strike
reports, payroll employment, retail sales, industrial
production, insured unemployment, to name a few—




9

Labor force behavior during downturns

These data are not identical with published figures
for April 1971, because they are based on the portion
of the April sample that was also in the March sam­
ple (three-fourths of the Current Population Survey
sample) and because the full estimation procedures
obviously could not be applied.5 These technical
constraints notwithstanding, the gross movements
into and out of unemployment are extensive; only
about half of the April unemployed had been job­
seekers in the previous month, whereas nearly equal
proportions were employed or out of the labor force.
Clearly, there is constant turnover in the ranks of
the unemployed.
Changes in the labor force— both in terms of level
and age-sex composition— can affect the unemploy­
ment rate even if the level of unemployment is un­
changed. For example, if the employment level in
April 1971 had been 300,000 higher, with no change
in the level of unemployment, the unemployment
rate would have rounded to 6.0 percent instead of
6.1 percent. The additional workers would have been
enough to move a rate that was already on the “low
side” of 6.1 percent down one-tenth of a percentage
point (rounded).
Over a longer period, of course, the impact can
be even more substantial if the composition of the
labor force should shift, as has occurred over the
last 15 years. In 1956, for example, adult men (20
years and older) accounted for 64 percent of the
labor force; by 1970, the ratio had slipped to 57
precent, as greater numbers of both adult women
and teenagers— groups with higher unemployment
rates than men— entered the labor market. If the
age-sex labor force distribution (10-year age groups)
had not changed between 1956 and 1970, but allow­
ing for unemployment rates for each age-sex group
to change as they did, the 1970 overall jobless rate
would have been 4.4 percent rather than the pub­
lished 4.9 percent figure.8

When business conditions begin to worsen, as dur­
ing 1970, the level of unemployment increases, of
course, while employment either grows more slowly
or declines. The labor force, the combination of
these elements, usually continues to rise, but its rate
of increase may tend to decline, as people abandon
or defer the search for work. This diminution tend­
ency is recognized in labor force theory as the “dis­
couraged worker effect.”
Another concept of labor force behavior during
cyclical downturns is the “additional worker effect.”
It holds that secondary workers are induced to enter
the labor market as breadwinners lose their jobs or
take a pay cut (or perhaps fear these circumstances
may occur). Therefore, should a worker be laid off,
his wife and perhaps a teenage son might enter the
labor force. It is extremely difficult to substantiate
the “additional worker effect” from available labor
force data. There is more concrete evidence of the
“discouraged worker effect,” and it appears to be
the larger factor.7
The above discussion suggests a very important
aspect of the working population, the fact that deci­
sions to participate or not participate in the labor
force— and thus to seek work or not seek work— are
often made for many personal reasons or in response
to factors and events about which we have no in­
formation. In other words, people cannot be assumed
to always act in a prescribed manner when it comes
to their participation in the labor force. This should
be borne in mind when one examines short-run labor
force growth. Under short-run conditions, in fact,
fits and starts in labor-force growth are more typical
than a smooth trend from month to month at an
annual rate of 1.5 to 2.0 million. This is also relevant
to an understanding of unemployment developments,
because people frequently cannot find a job as soon
as they enter the labor force. And if short-run labor
force behavior is not easily predictable, it is clear
that short-run unemployment movements are also
variable.

Table 2. Gross flows in the employment status of per­
sons 16 years and over between March and April 1971
[In thousands]

Employment status
category

Status in March
Status
in
Unem­
N otin
labor
April Employed ployed

Problems with seasonal adjustment

Because the seasonal-adjustment process is based
upon experience of past years, to the extent that
seasonal patterns or short-run labor force behavior
change, the current data may be difficult to evaluate.

force
Employed.............................................
Unemployed.........................................
Not in labor force................................




78,409
4,494
56,302

74,240
1,018
2,566

1,508
2,409
1,182

2,660
1,067
52,553

10

Table 3.

Seasonally adjusted unemployment rates as originally published and as revised in subsequent years, 1 9 6 7 -7 0
1967

Month

January............................
February......................
March..............................
April.................................
May..................................
June_____________ _
July..................................
August...........................
September................... .
October______________
November......... .............
December.........................

Origi­
nally
pub­
lished
3.7
3.7
3.6
3.7
3.8
4.0
3.9
3.8
4.1
4.3
3.9
3.7

1968

1969
1970
1971
1968
revision revision revision revision
3.7
3.7
3.7
3.7
3.9
3.9
3.9
3.8
4.1
4.3
3.8
3.7

3.7
3.7
3.7
3.8
3.9
3.9
3.9
3.8
4.0
4.2
3.8
3.7

3.8
3.8
3.8
3.8
3.9
3.9
3.8
3.8
3.9
4.1
3.9
3.7

3.8
3.8
3.7
3.8
3.8
3.9
3.8
3.5
3.9
4.1
3.9
3.8

Origi­
nally
pub­

lished
3.5
3.7
3.6
3.5
3.5
3.8
3.7
3.5
3.6
3.6
3.3
3.3

1971
1969
1970
revision revision revision
3.6
3.7
3.7
3.5
3.6
3.7
3.7
3.5
3.6
3.6
3.4
3.3

3.6
3.8
3.7
3.5
3.6
3.7
3.7
3.5
3.5
3.5
3.4
3.3

3.7
3.8
3.7
3.5
3.5
3.7
3.6
3.5
3.5
3.4
3.5
3.4

Origi­
nally
pub­
lished
3.3
3.3
3.4
3.5
3.5
3.4
3.6
3.5
4.0
3.9
3.4
3.4

1970

1970
1971
revision revision
3.4
3.3
3.4
3.5
3.5
3.4
3.5
3.5
3.8
3.8
3.5
3.5

3.4
3.3
3.4
3.5
3.4
3.4
3.5
3.5
3.8
3.7
3.5
3.6

Origi­
nally
pub­
lished

1971
revision

3.9
4.2
4.4
4.8
5.0
4.7
5.0
5.1
5.5
5.6
5.8
6.0

3.9
4.2
4.4
4.7
4.9
4.8
5.0
5.1
5.4
5.5
5.9
6.2

answer is yes, the interviewer asks, “What has . . .
been doing in the last 4 weeks to find work?” A
specific activity must be cited or the person will not
be counted as unemployed. Finally, the question,
“Is there any reason why . . . could not take a job
last week?”, is asked to ascertain if the jobseeker
was available for work rather than seeking a job for
some future period.8

This may come about due to transitional periods of
economic activity. Other complications can arise
from a shift in the timing of the survey week (whether
the 12th day is early or late in the week); in months
in which large labor force changes are taking place,
such as in June and September; possibly from un­
usual events such as strikes; the timing of holidays;
severe weather conditions; or changes in the survey
questionnaire (these, however, are rarely made). As
a consequence, the seasonally adjusted values may
exhibit erratic behavior or lack of smoothness over
certain months. As more experience is gained, after
a year or two, the new seasonal pattern will usually
emerge more clearly, and a more acceptable set of
seasonally adjusted values will become available.
This “wait-until-next-year” approach to determine
the more accurate seasonally adjusted monthly
changes is not a satisfactory answer to the policy­
maker or the newsman, whose concern is necessarily
with the present. However, it should be recognized
that revisions of the overall jobless rate rarely exceed
0.1 percentage point in the subsequent year. More­
over, since seasonally adjusted values are subject
to change, seemingly erratic movements in the cur­
rent year should be viewed as approximations of
what occurred rather than as exact measures.
In the past several years, there has been some
evidence that the overall seasonally adjusted rate
has been behaving more erratically than in prior
years. This may have come about as a result of con­
ceptual revisions and the changes in the question­
naire beginning in January 1967. Prior to 1967, an
individual not working was asked, during the course
of the survey interview, “Was . . . looking for work?”
In 1967, the question was changed to: “Has . . . been
looking for work during the past 4 weeks?” If the




1969

Although the precise effect of these changes is
difficult to quantify, there is evidence of their impact,
particularly the addition of the availability question.
The shift from an unspecified jobseeking period to
a 4-week period might well have produced a lower or
higher total, as previous respondents could have in­
terpreted “looking for work” to imply either “last
week” or some vague earlier period. Similarly, the
introduction of the specific jobseeking method re­
quirement (not asked, of course, of persons on lay­
off or those waiting to begin a new job within 30
days) may have lowered the jobless count as well,
screening out those for whom jobseeking is more a
state of mind rather than overt action.
And the effect of the “availability test” is clear,
particularly in March, April, and May, when young­
sters still in school are seeking summer jobs. All
three of the questionnaire revisions appear to have
seasonality implications, with teenagers in the spring
being the most obvious example. One result has been
a number of unaccountable movements in the overall
rate, particularly in the August-through-November
periods. Another result has been the fact that May
is now the seasonal low of the year in terms of un­
employment, instead of October. The original job­
less rate and its revisions based upon seasonal ad­
justment for the years 1967-70 are shown in table 3.
Since the 1967 alterations, the seasonal adjustment
11

process has had a chance to “settle down” somewhat,
and it appears fewer inexplicable month-to-month
jumps will be observed in 1971 and subsequent years
due to this factor. However, because changes in the
business cycle can affect seasonal patterns, it is also
likely the seasonal adjustment process has been af­
fected by the rise in the level of unemployment since
1969.
This would be most evident in months in which
very large seasonal changes occur, such as between
May and June, August and September, and Decem­
ber and January. The seasonal-adjustment process,
which at present is essentially multiplicative in na­
ture, may tend to overcompensate for these wide
variations. This accounts for a part of the unusual
drop in the jobless rate, from 6.2 percent in May to
5.6 percent in June, which was reported at the time
as somewhat exaggerating the “real” change that
took place between the 2 months.
It should be emphasized that seasonally-adjusted
values are, at best, approximations of the underlying
trend. One should therefore not expect them to be
uniformly smooth on a monthly basis or that they be
a precise reflection of cyclical movements.

for public policy development if one could make an
estimate of the number of unemployed who need
jobs. But the subjectivity of such a measure would
be exceedingly great, since in some cases it cannot
be easily determined whether an unemployed indi­
vidual really needs a job. For example, in many
families, the husband and wife are both active labor
force participants because they believe they need
the income; it cannot be arbitrarily concluded that
women with employed husbands do not need jobs.
Similarly, it seems unrealistic to count certain “mar­
ginal” workers as employed when they have jobs,
but to exclude them from the unemployed count
when they are looking for work.
On the other hand, it is also often said that the
unemployment concept excludes some persons who
should be counted as unemployed. Those who sup­
port this view point out that a number of persons
become discouraged over job prospects and cease to
look for work, even though they still desire a job.
These discouraged workers are sometimes referred
to as “hidden unemployed.” 10 Ever since the reg­
ular collection of unemployment statistics by the
Federal Government began, however, the criterion
has been that to be counted as unemployed a person
should be an active jobseeker, which lessens sub­
stantially the possibility of unemployment being a
state of mind. The President’s Committee to Ap­
praise Employment and Unemployment Statistics
recommended in its 1962 report that these workers
be identified, but it was equally firm in its belief that
they not be included in the jobless count.11 At its
behest, this category, as well as all other “not in the
labor force” groups, has been identified in the Cur­
rent Population Survey since 1967, and data are
published on a quarterly basis in Employment and
Earnings by a wide variety of characteristics. In
1970, discouraged workers as measured in the sur­
vey averaged nearly 650,000 persons,12 mostly teen­
agers and adult women.

The unemployment concepts reexamined

The question has been raised many times over the
years as to whether the unemployment concepts and
definitions should be revised. “We are counting too
many as unemployed.” “We are not counting
enough.” “Many of the unemployed don’t really want
to work and shouldn’t be counted.” “Many unem­
ployed don’t need a job.” “Many no longer looking
have just given up but are still really unemployed.”
So go some of the complaints about the current con­
cepts.
This subject has been examined almost continu­
ously since the inception of the unemployment
measurement survey in 1940. The concept that has
been accepted and used has changed very little over
this tri-decade,9 or else comparability would be a
very real problem.
Those who feel that the number counted as un­
employed is too high point to such groups as young
people, particularly those who are in school, and
married women as outstanding candidates for elim­
ination from the conceptual base. The usual justifica­
tions given are that these types of workers do not
really need a job and/or are only temporarily in the
labor force. It would be interesting and significant




Because of the vast array of statistics regularly
published on the employed, unemployed, and those
not in the labor force, it is possible for one to calcu­
late an unemployment rate based on various defini­
tions of labor force and unemployment. For example,
those who believe that labor force eligibility should
begin with age 18 and end, say, with age 69 can
exclude 16- and 17-year-olds and the over-70 group,
and then calculate a separate unemployment rate.
Similarly, estimates of discouraged workers could
be added to the unemployment and labor force totals
12

States is the maintenance of consistency over time.
This is another way of saying that the greater con­
cern is necessarily focused upon the relative, rather
than the absolute, position. In other words, it is im­
perative to know how well off the economy is each
month compared with the preceding month or some
earlier period. It is relevant to continue to examine
who should be counted as unemployed, but such
examination should not interfere with public trust
in the figures, historical continuity of the data, or
objectivity of measurement. If these three standards
are followed, unemployment statistics will continue
to provide one of the best measures of the economic
status of the Nation.
□

to arrive at a rate that includes this group, or persons
whose major activity is going to school might be
excluded.
The major criterion that has been used over time
in estimating unemployment is objectivity of meas­
urement. Need for work, intensity of desire, and fam­
ily income are all potentially subjective factors and
as such are excluded from unemployment concepts,
which are under constant review to make them as
objective as possible.
In this context, it is appropriate to conclude with
the idea that perhaps the most important aspect in
the measurement of unemployment in the United

-FOOTNOTESxFor a more detailed discussion of the Current Popula­
tion Survey and the concepts utilized, see Concepts and
M ethods Used in Manpower Statistics From the Current
Population Survey (BLS Report 313, 1967).

contained in an article by George L. Perry, “Changing
Labor Markets and Inflation,” Brookings Papers on Eco­
nomic A ctivity 3 (Washington, Brookings Institution, 1970),
pp. 411-441.

*The original data of a series “are regarded as the
product of a trend-cycle component times a seasonal com­
ponent times an irregular component. The trend-cycle repre­
sents the ‘real’ movement of the series, including cyclical
movements. The seasonal component is the annual repeti­
tive pattern which makes certain months consistently higher
or lower than others. The irregular component is a residual,
including sampling errors and short-term fluctuations which
do not follow any consistent pattern. After a satisfactory
decomposition is achieved, the seasonally adjusted series
is computed by dividing each original value by the corre­
sponding seasonal factor.” The foregoing is from ‘T he
Method o f Seasonal Adjustment for Labor Force Series,”
Employment and Earnings, February 1971, pp. 22-23.
A more technical description o f the seasonal-adjustment
method may be found in “Appendix A. The BLS Seasonal
Factor Method,” BLS Handbook of M ethods for Surveys
and Studies (BLS Bulletin 1458, 1966), pp. 222-228.

7 For a discussion of the discouraged worker and addi­
tional worker hypotheses and some indication of their rela­
tive impacts, see William G. Bowen and T. Aldrich Finegan,
The Economics of Labor Force Participation (Princeton,
N.J., Princeton University Press, 1969).
•F or further amplification of the 1967 changes, see
Robert L. Stein, “New Definitions for Employment and
Unemployment,” Employment and Earnings and Monthly
Report on the Labor Force, February 1967.
•F or a discussion of the historical background of the
conceptual framework, see President’s Committee to
Appraise Employment and Unemployment Statistics, Meas­
uring Employment and Unemployment (Washington, 1962),
Ch. I.
“ Prominent among those who support the view that
discouraged workers should be included in the jobless
counts is Professor Alfred J. Telia o f Georgetown Univer­
sity. Professor Telia has performed a considerable amount
of research with labor force models which enables him
to estimate the number o f persons not actually counted as
unemployed but who would be in the labor force under
what he defines as “full employment” conditions. See “The
Relation of Labor Force to Employment,” Industrial and
Labor Relations Review, April 1964, pp. 454—469, and
“Labor Force Sensitivity to Employment by Age, Sex,”
Industrial Relations, February 1965, pp. 69-83.

* The civilian labor force, seasonally adjusted, is the
aggregation of the four major age-sex components (male
and female, 16-19 years, and 20 years and over) for each
of three categories: agricultural employment, nonagricultural employment, and unemployment.
* In evaluating monthly estimates, the determination o f
whether a change is statistically significant is based upon
the sampling error of the estimate. It is recognized that
the seasonal-adjustment process itself is imperfect, espe­
cially on a current basis. However, the magnitude of any
error attributable to seasonal adjustment is not quantifiable
at the time the estimates first become available.

“ President’s Committee to Appraise Employment and
Unemployment Statistics, op. cit., pp. 52-56.
“ In 1967, the number of discouraged workers averaged
732,000; this dropped consecutively in 1968 to 667,000 and
to 574,000 in 1969 before rising to 638,000 in 1970 and
740,000 in the second quarter of 1971 (seasonally adjusted).
For further discussion o f these data, see “Discouraged
Workers and Recent Changes in Labor Force Growth”
(BLS Report 396, 1971).

BSee Harvey J. Hilaski, ‘T h e Status of Research on
Gross Changes in the Labor Force,” Employment and
Earnings and Monthly R eport on the Labor Force, October
1968.
•Additional amplification of structural shifts in the labor
force and their effect upon the unemployment rate are




13

Comparing
The two series generally move
in sim ilar directions,
but show different levels
due to absences, second jobs,
the census undercount, and other factors

employment estimates
from household and

GLORIA P. GREEN

payroll surveys
pling techniques and collection and, estim ation
methods, m ost of which cannot be readily m eas­
ured in terms of impact on differences in the
levels of the two series.
It should be noted at the outset that the total
nonagricultural em ployment series from the house­
hold survey is much more comprehensive than
the establishm ent series. The household series
includes— in addition to wage and salary workers—
the self-employed, unpaid workers who worked
15 hours or more during the survey week in
family-operated enterprises, and private house­
hold workers, none of whom by definition would
appear on establishment payrolls. These three
groups are readily identified in the household
survey on the basis of major industry and classof-worker designations; in 1968, they totaled
7.5 million workers. W hen these groups are
subtracted from the household estim ate of total
nonagricultural em ployment, a third series, reflect­
ing em ploym ent of wage and salary workers of
generally comparable coverage to the establish­
ment series, is obtained. In order to develop a
reconciliation, it is necessary to examine the
comparability of this derived household series
with the payroll em ploym ent series. The three
series— total nonagricultural em ploym ent (house­
hold survey), nonagricultural wage and salary
em ployment excluding private household workers
(household survey),2 and nonagricultural wage
and salary em ployment (establishment survey)—
are compared in table 1 on an annual average
basis from 1948 -68.
This study first examines the conceptual and
other differences between the two series. Second,
it attem pts to reconcile the annual levels of the
two series over the 1962-68 period 3 insofar as
known discrepancies can be quantified. (It should
be kept in mind, however, that no attem pt at
reconciliation provides a complete answer account­
ing for all of the factors that influence the levels

S t a t i s t i c s o n n o n a g r i c u l t u r a l em ployment are
key indicators of the economic health of the
N ation. Because these data serve a variety of
purposes, no one source of data can adequately
provide all of the information that is needed for
a com plete and balanced picture of the em ploy­
ment situation.
Each month, the Bureau of Labor Statistics
analyzes and publishes two independently derived
estim ates of total em ployment in nonagricultural
industries— the household series and the establish­
ment series.1 Each of these bodies of data makes
its own unique contiibution to the nonagricultural
em ploym ent picture— the household series as a
measure of the work status of individuals and the
payroll series as a count of jobs.
These series attem pt to measure different aspects
of the nonagricultural em ployment situation—
people versus jobs— but for the m ost part they
tend to show the same underlying economic influ­
ences. A t times, however, significant differences,
which m ay be confusing to the user, are observed
in the level of the estim ates, in m onth-to-m onth
changes, in the timing and extent of business
cycles, and in certain longer run trends.
In part, these differences are inherent in the
concepts and scope of the two series. Nonagricul­
tural em ploym ent measured through a household
survey cannot and should not be expected to
yield magnitudes identical with those of em ploy­
m ent measured through an employer-payroll
reporting system . Conceptual differences between
the series can usually be reconciled or explained
in large part, and a number of the differences in
coverage can be adjusted. However, there are
also discrepancies caused by differences in sam-

Gloria P. Green is an econom ist in the D ivision of E m ­
ploym ent and U nem ploym ent Analysis, Bureau of Labor
Statistics.

From the R ev iew of December 1969



14

of the two series.) Finally, em ployment levels
are examined on both an annual and a monthly
basis to gain greater insight into the divergences
in levels.

household survey as being “with a job but not at
work” in 1968. Of this total, 1.6 million or 43
percent were not paid for the time off, as shown in
table 2.

M easurable factors affectin g co m p a ra b ility

M u l t i p l e j o b h o l d i n g . Another major source of
discrepancy relates to the treatment of persons
employed in more than one job. The household
survey counts each person by work status only
once since each person is classified as either
employed, unemployed, or not in the labor force.
Employed persons holding more than one. job
during the survey week are counted and classified
in the job at which they worked the greatest
number of hours. In the establishment survey, the
total number of employed persons is overstated to
the extent that those who worked in more than one
establishment during the reference period are
counted each time their names appear on a payroll.
Workers m ay be counted more than once, for
example, because they hold down two jobs or more
concurrently or because they leave one job and
obtain another within a single reference period and
thus appear on the payroll records of both em ­
ployers. Such a situation can also arise when a
worker is continued on a payroll after leaving his
job because he is being compensated for earned
vacation time.
While it is virtually impossible to identify
persons who work at two jobs or more from payroll
records, it is usually possible to obtain this
information from the worker or a member of his
household. To gain insight into multiple jobholding, special surveys of this phenomenon have
been conducted as a part of the household survey
periodically since 1943. In M ay 1966, the m ost
recent month for which survey data are available,
approximately 2 million workers held a secondary
nonagricultural wage and salary job that would not
have been reported in the usual m onthly household
survey.4 As indicated in table 3, these additional
jobs were concentrated in trade (28 percent),
services (26 percent), and government (18 percent).
Accordingly, the household estim ate of 60 million
persons who were employed in a nonagricultural
wage or salary primary job (excluding private
household workers) during the survey week in
M ay 1966 would have to be raised by 2 million to
approximate a payroll count.5 The count for that
period would be even higher, of course, if account
were taken of the number of persons holding 3

Of the many factors that influence the levels of
the series, only a few can be quantified to any
degree. The numbers involved and their relative
effect upon the levels of the series are discussed
in the sections that follow.
U n p a id
a b se n c e s.
One major source of dis­
crepancy between the two series stem s from
different treatment of workers absent for a full
week from their jobs. The household survey includes
among the employed all persons who had jobs
during the survey week but were temporarily
absent because of illness, bad weather, vacation,
labor-management disputes or various personal
reasons, whether they were paid by their employers
for the time off or whether they were seeking other
jobs. B y contrast, the establishment series includes
only those persons on paid leave for any part of
the pay period specified in the survey. Therefore,
persons who are absent without pay for the entire
period are not included in the payroll figures. On
the average, about 3.7 million employed nonfarm
wage and salary workers were picked up in the

Table 1.

Nonagricultural employment, 1948-68
[In thousands]
Household series'

Year
Total

Total wage
and salary
employ­
ment2

Payroll
series:
wage and
salary
employment

Difference3

1948___________________
1949___________________
1950___________________
1951___________________
1952___________________
1953___________________
1954_________________

51,405
50,684
52,450
53,951
54,488
55,651
54,733

43,135
42,308
43,982
45,627
46, 465
47, 449
46,490

44,891
43, 778
45,222
47,849
48,825
50,232
49,022

1,756
1,470
1,240
2,222
2,360
2,782
2,532

1955___________________
1956___________________
1957___________________
1958___________________
1959_________________ .
1960___________________
1961_________________

56,464
58, 394
58, 789
58,122
59,745
60,958
61,333

47,838
49,518
49,745
48,876
50,330
51,487
51,690

50,675
52, 408
52,894
51,363
53,313
54,234
54, 042

2,837
2,890
3,149
2,487
2,983
2,747
2,352

1962___________________
1963___________________
1964_________________
1965___________________
1966______ ____ _ . .
1967__________________
1968___________________

62, 657
63,863
65, 596
67,594
69,859
70, 527
72,103

53,136
54, 498
56,115
58,217
60, 686
62,882
64,601

55, 596
56,702
58,331
60,815
63,955
65,857
67,860

2,460
2,204
2,216
2,598
3,269
2,975
3,259

1 Persons 14 years and over for 1948-66; 16 years and over for 1967-68.
2 Excludes private household workers.
3 Payroll series employment less household wage and salary employment.




15

Table 2. Employed nonagricultural wage and salary
workers on unpaid absences, by industry, 1962-68

attendance, and general social custom prevent
most children under 16 from working. In 1967
and 1968 household survey estim ates of non­
agricultural wage and salary workers 14-15 years
ol age averaged nearly 440,000 and 480,000,
respectively. (Prior to January 1967, official c p s
statistics on nonagricultural wage and salary
em ployment had included 14- and 15-year-olds.)
The number of young persons under 14 years of
age who are employed in nonagricultural wage and
salary jobs is not known.

(In thousands]
Industry

1962

1963

1964

1965

1966

1967

Total1______________ 1,122 1,241 1,249 1,249 1,317 1,454
Private1 __________________
Construction_____________
Manufacturing___________
Transportation and public
utilities_______________
Wholesale and retail tra d e .._
Finance, insurance, and real
estate___________ ____ _
Services2_______________
Government___ _____ ________

1968
1,629

913
95
339

993 1,006 1,024 1,065 1,168
95
114
83
92
97
365
394
377
381
460

1,314
140
488

68
201

80
239

78
227

76
237

81
242

70
256

95
285

35
163
210

36
172
247

46
188
243

37
190
225

41
194
252

41
211
284

46
240
315

1Also includes mining, not shown separately, and excludes private household workers.
2 Excludes private household workers.

E f f e c t o f t h e c e n s u s u n d e r c o u n t . Investiga­
tion of the accuracy and completeness of the
1960 Census of Population has indicated that an
estim ated total of 5.7 million persons of all ages
were missed in the enumeration. Since the decen­
nial population censuses provide the basis for
projection of current estim ates of the population,
which, in turn, serve as monthly controls for the
household survey sample, any undercount of the
population in the census can have a profound
effect upon the level of labor force and em ploy­
ment estim ates derived from that survey. A d­
vancing the ages of persons undercounted in 1960
by 7 years reveals a probable total em ploym ent
undercount of about 2.8 million persons 16
years of age and over in 1967.® Distributing the
2.8 million on a ratio basis, about 2.4 million are
estim ated here to be nonagricultural wage and
salary workers. The nonfarm em ploym ent esti­
mates based on payroll surveys are not similarly
affected, since these surveys cover all persons
on payrolls and do not depend upon probability
population controls.

additional paid jobs or more. These surveys have
been too infrequent to determine any definite
seasonal, cyclical, or secular trends, although in
recent surveys the rate of multiple jobholding has
remained substantially the same.
Even though several improvements in the
enumeration ol multiple jobholders have been
made in recent years, it is probable that no
survey has given what might approximate a
complete count. However, the extent of this
hypothetical “undercount” is uncertain. M any
persons m ay be reluctant to report secondary
jobs for various reasons, such as a distrust of the
confidentiality assurances of the survey, and
knowledge of such jobs might be deliberately
withheld from the survey interviewer. Another
possible reason for undercount is that the re­
spondent (often the housewife) m ay not be aware
of the second job or m ay not realize that the
person in question is technically on two payrolls
or more. Examples of this latter case are teachers
paid on a 12-month basis and employed in other
jobs during the summer; lawyers acting as directors
of corporations, sometimes several corporations;
school board and other government officials who
are paid for limited services rendered; persons
performing as consultants on an irregular basis,
etc. It might be possible to develop techniques to
improve the count in the latter cases, but it is
unlikely that deliberate failures to report multiple
jobholding could be uncovered.

s e r v i c e s . One minor discrepancy
between the two series which has been quantified

A g r ic u l t u r a l

Table 3. Distribution of secondary nonagricultural wage
and salary jobs, by industry, May 1966
[In thousands)
Industry division

ge

l im it a t io n s

.




___________ _________ _____

1,996

100.0

Private wage and salary.......... ............................. .............

1,640

82.2

Mining........
........................................................
Construction...................... ........................................
Manufacturing........................ ......................................
Transportation and public utilities.............................
Wholesale and retail trade__________ ___________
Finance, insurance, and real estate.............................
Services2 ..
....
......................................

9
129
183
123
551
128
517

0.5
6.5
9.2
6.2
27.6
6.4
25.9

Government................................... .......................................

356

17.8

Total____ _

While the household series
provides data on the full- or part-time em ploy­
ment status of the entire U.S. civilian noninstitutional population 16 years of age and over,
the establishment series has no age limitations,
although child labor laws, compulsory school
A

Percent

Total1

1 Data include only first additional Job.
2 Includes forestries and fisheries; excludes private household workers.

16

the adjusted payroll series. This involves a sub­
traction of the number of persons on unpaid
absence from their jobs during the survey week,
and the addition of the estim ated number of
secondary jobs, em ployment of 14- and 15-yearolds, and the estimated 1968 population undercount. The net result after allowance for these
measurable differences is to reduce the original
difference between the two series from 3.3 million
to an estim ated 340,000 in 1968.
In evaluating this difference, it should be kept
in mind that estimates of the known sources of
discrepancy are subject to considerable uncer­
tainty. It should not be assumed from this isolated
example that the net effect of allowing for all
measurable variables would always bring the two
series this close together. The net adjustment
might be greater or smaller than this figure in
another period. However, this attem pt at rec­
onciliation does illustrate the extent of the
problem, and the nature and approximate magni­
tudes of m any of the discrepancies.

on the basis of data obtained from the payroll
series stems from the different classification of
workers employed in agricultural services. The
payroll series includes them under services, while
the household series classifies them in agriculture.
They constitute nearly 90 percent of the payroll
subclassification, agricultural services, forestries,
and fisheries, and ranged from 134,000 in 1962 to
160,000 in 1968. This classification difference is
of significance primarily in comparing the services
industry component of the two series; its effect
on overall estimates of nonagricultural wage and
salary employment is of lesser importance.
N et effe c t of m easurable differences

Taking into account all of the measurable differ­
ences between the two series, it is possible to de­
velop reasonably comparable household and pay­
roll estim ates of nonagricultural wage and salary
employment. (See table 4.) However, most of the
adjustments must be made in the household data,
since they contain a wide variety of data on
other characteristics of workers, and it would
not be appropriate statistically to adjust one
series with data from another.
The adjustment technique requires the sub­
traction of the number employed in agricultural
services from the payroll series, since this group
is included in agriculture in the household series.
Except for this adjustment, all others are made to
the household data to achieve comparability with

O ther factors affecting com parability

While the foregoing factors can be quantified
to some degree, there are other factors which
influence the levels of the two series for which no
quantitative estimates can be made.
D i f f e r e n c e s i n s u r v e y c o v e r a g e . In several
respects, the establishment series is more inclusive
in terms of population coverage than is the house­
hold series. The establishment series includes
military personnel who hold civilian jobs in non­
government establishm ents during their off-duty
hours. It m ay also include some inmates of institu­
tions who are working in or outside the institution
if they are on payrolls. All military personnel and
institutional inm ates are explicitly excluded from
the household data.
Residents of Canada or Mexico who com mute to
nonagricultural jobs in the United States and thus
are included in the payroll count would also be
outside the scope of the household survey, which
covers people residing in the 50 States and the
District of Columbia. Commuting to the Uuited
States may be partly offset to the extent that some
U.S. residents are employed by establishments in
Canada or Mexico. The combined effect of these
differences, however, probably does not account
for much of the gap between the two series.

Table 4. Measurable differences between payroll and
household estimates of nonagricultural wage and salary
employment, 1968
Item

Additions
Unadjusted
totals
Adjustments
minus
reductions

67,860,000
Payroll series...........................
Less: employment in
agricultural services. . .
Adjusted payroll employ­
ment................................

160,000

Household series1....................... 64,601,000
Less: unpaid absences___
Total reductions.......

1,629,000

Plus: multiple jobholders
(est.)2.........
1968 undercount (est.).
Total additions____
Net adjustments.. _
Adjusted household
employment...................
Difference after adjustments
(payroll less household
series). . . .

Adjusted
totals and
difference

67,700,000

1,629,000
2.184.000
484,000
2.400.000

5.068.000
3.439.000
68,040,000

-340,000

1 Excludes private household workers <1,854,000).
2 Estimate includes only first additional nonagricultural wage and salary job.




17

h e s u r v e y p e r i o d . The time period covered by
the household survey is always one calendar
week; since July 1955, it has been designated as
the week containing the 12th of the month. In
the payroll series, the time reference is to the
payroll period including the 12th of the month,
which is intended to be a single week. D espite
comparability in design and intent, there are a
number of differences in reality. Some establish­
ments have 2-week or monthly payrolls (at least
one-fifth of the total) and thus are likely to reflect
more duplication due to multiple jobholding or
turnover than would be reported for a single week.
Moreover, during the longer time period a person
could be counted as employed in the payroll
series who, during the household survey week, was
either unemployed or not in the labor force.
For the Federal Government, em ployment
figures represent the number of persons occupying
civilian positions on the last day of the calendar
month plus any interm ittent workers who worked
at anytime during the month. This not only
tends to magnify the general problem resulting
from turnover but also contributes to a special
problem of comparability with the household
series, which is peculiar to December alone. T hat
month the payroll series invariably rises sub­
stantially, reflecting the hiring of temporary
postal workers for the Christmas rush period. The
household survey, on the other hand, usually does
not show a rise of similar magnitude, mainly
because the survey week occurs relatively early in
the m onth and many of the workers counted as
employed in the payroll series are not working at
the time the household statistics are collected.7
D ata permitting a correction for the deviation
from the single-week reference period in the
establishment data are not available.

of the total actually em ployed in summer schools
or on paid vacation, estim ates of the number of
regular school teachers in M ay— the last full m onth
of the school year prior to the vacation period—
are substituted in tabulations covering the vaca­
tion months.
In the household series, school teachers who
have contracts (either written or verbal) to return
to teaching in the fall would be reported during
the summer months as “with a job but not at
work” (on vacation) unless they hold other jobs
and were thus classified according to the occupa­
tion and industry of that job during the survey
week. The possibilities for multiple jobholding
among teachers during the summer months, which
would have the effect of inflating the payroll count
compared with the household count, are obvious.
However, the magnitude of this conceptual differ­
ence is uncertain.

T

a nd
e s t im a t in g
pr o c ed u r es.
The
household survey covers a scientifically selected
probability sample of 50,000, designed to represent
the entire civilian noninstitutional population. As
an early step in estim ating procedures, the sample
data (for persons 16 years of age and over) are
weighted by independently developed estim ates of
the population by age, sex, and color. These esti­
mates. are prepared by carrying forward the m ost
recent census data (1960) to take account of sub­
sequent aging of the population, deaths, immigra­
tion, and emigration.
Because the household survey is based upon a
sample, the results may differ from the figures that
would be obtained if a com plete census using the
same schedules and procedures were possible. In
this series, the relative sampling error for the esti­
mate of nonagricultural wage and salary workers
is about 200,000 at present em ploym ent levels.
This means that the chances are about 2 out of 3
that an estim ate from the sample would differ
from a complete census by less than this amount.
This estim ate of sampling error would be 400,000
if a confidence level of 19 out of 20 times is
wanted.
As in any survey, the results are also subject to
errors of response and reporting. Furthermore, as
noted earlier, that part of the population missed in
the census (the undercount) is also presumably
missed in the sample survey and subsequent
“blow-up” of the sample as well.
On the other hand, the m onthly payroll em­

S a m p l in g

The manner in
which payrolls are handled in the education system
causes a special problem of comparability between
the two series, which is peculiar to the summer
months alone. Some teachers and other educational
staff are paid on a 12-month basis; other regular
faculty members are paid only during the academic
school year (9 or 10 months) and would not ordi­
narily appear on payrolls during the summer
months. As a consequence, special treatm ent is
accorded school teachers in the establishm ent
series. Instead of using payroll reports for the sum­
mer months, which would include only that part
P roblem

o f

sch o o l




t e a c h er s

.

18

ployment estim ates are derived from reports of a
relatively large survey sample (160,000 establish­
ments having over 30 million employees), which
assures a high degree of accuracy.8 However, since
the estimating procedures employ the previous
m onth’s estim ate as the base in computing the
current m onth’s level (link relative technique),
sampling and response errors m ay accumulate over
several months. To remove any accumulated error,
the em ployment estim ates are adjusted annually
to new benchmarks (comprehensive counts of em­
ploym ent). The revision published each July also
adjusts the estimates for changes in the industrial
classification of individual establishments (result­
ing from changes in their product, which are not
reflected in the levels of estim ates until the data
are adjusted to new benchmarks).9 Another cause
of differences, generally minor, arises from im ­
provements in the quality of the benchmark data.
For the eight m ost recent benchmark revisions, the
estim ates of total nonagricultural employment
have varied from benchmarks by less than 1 per­
cent, averaging 0.3 percent.
W ithin a few industries, mostly in the service
sector, current m onthly estimates are not ob­
tained by direct reports from a sample of estab­
lishments (e.g., churches and other nonprofit
organizations), and monthly changes in these in­
dustries are based on movements shown in the
benchmark data for earlier years. Necessary ad­
justments are made at the time of adjustment to
new benchmarks. This procedure can result in
substantial error for a few individual industries,
especially those in which small establishments pre­
dominate, but its effect on the much larger non­
farm total is negligible.
The accuracy of the level of the establishm ent
series depends a great deal on the accuracy of the
benchmarks. These are primarily derived from
tax returns for unemploym ent insurance, supple­
mented by social security tax returns for small
employers and by a variety of other sources for
certain sectors. I t is possible, for example, that
errors of omission or duplication can occur in the
reporting of social insurance tax returns for
the correct period or in fitting together the bench­
mark data gathered from a number of sources.
In addition, the benchmark occurring in March
of each year can mean that the other months,
particularly the summer and fall periods, are
less precisely estimated. M any State unemploy­
ment insurance laws do not cover firms unless




they operate 20 weeks or more during the calendar
year. As a result, a number of seasonal under­
takings, such as summer resorts, hotels, and
amusement enterprises, m ay be missed by the
benchmark source and thus by the payroll sta­
tistics. The net effects of these problems are
unknown, and these uncertainties make compari­
sons of the levels of the two series even more
inconclusive.
The possibility of error in the population
censuses or the unemployment insurance bench­
marks cannot be disregarded. There is no “true”
total against which the accuracy of either can
be measured. Although the benchmark data and
the population totals are among the best statistical
measures available, as we have seen neither is
perfect.
Comparison of annual em ploym ent levels

Although measured very differently, estimates
of nonagricultural wage and salary employment
in the household and establishment series, have
exhibited similar growth patterns over the 196268 period. Em ploym ent in the household series
increased by 11.5 million (from 53.1 to 64.6
m illion ); 10 in the establishm ent series, it gained
12.3 million (from 55.6 to 67.9 million). B o th
series reflected a net expansion of about 22 percent
during the 7-year period. However, when related
to an earlier period em ployment levels reveal th at
the two series have diverged somewhat, the
difference averaging 2.7 million between 1962
and 1968 and 2.4 million over the 1948-61 period.
(Household data for 1967 and 1968 are not strictly
comparable with that for earlier years. Footnote
10 describes the differences.)
In an effort to examine some of the divergences
in levels between 1962-68, the household and
payroll estim ates by major industry groups11
have been adjusted to take into account all
measurable differences between the two series
that can be quantified on a year-to-year basis.
(See table 5.) This involves subtraction of the
number of persons on unpaid absences from thenjobs during the survey week, the addition of the
estimated number of nonagricultural wage and
salary secondary jobs to the household series, and
elimination of the number employed in agricultural
services from the payroll series. In table 5, all
quantifiable adjustments covered in table 4 have
been made on the industry data from both series,
with the exception of the undercount for every

19

Table 5.

Comparison off payroll and household estimates of nonagricultural wage and salary employment, 1962-68
[In thousands]
Payroll series
Industry and year

Total wage and salary employment:
1962 ..........................................
1963 ..........................................
1964 ..........................................
1965 ..........................................
1966 ..........................................
1967 ..........................................
1968 ..........................................
Private wage and salary employment:
1962 ------------------------------1963 ..........................................
1964 ..........................................
1965 ..........................................
1966 .......... .............................
1967 ..........................................
1968 ..........................................
Mining:
1962 ..................................
1963 ..................................
1964 ..................................
1965 ...................................
1966 ..................................
1967 ..................................
1968 ..................................
Construction:
1962 ..................................
1963 ..................................
1964 ..................................
1965 ..................................
1966 ..................................
1967 ..................................
1968 ..................................
Manufacturing:
1962 ..................................
1963 ..................................
1964 ..................................
1965 ..................................
1966 ..................................
1967 ..................................
1968 .................................
Transportation:
1962 ..................................
1963 ..................................
1964 ..................................
1965 ..................................
1966 ..................................
1967 ..................................
1968 ...................................
Trade:
1962 .................................. .
1963 ..................................
1964. ..................................
1965 .................................. .
1966 .................................. .
1967 ..................................
1968 ..................................
Finance:
1962........................................
1963.......................................
1964 ...................................
1965. ..................................
1966...................................
1967.......................................
1968........................................
Services:»
1962 ..................................
1963 ..................................
1964 .................................. .
1965 ..................................
1966 ..................................
1967 ..................................
1968 ...................................
Government:
1962........................................
1963 .............................. .......
1964 ..................................
1965 ..................................
1966 ...................................
1967 ..................................
1968 ..................................

Household series

Total i

Agricultural
services*

Adjusted
total
(excluding
agricultural)

Total3

Unpaid
absen­
ces

Additional
paid jobs*

55,596
56,702
58,332
60,815
63,955
65,857
67,860

134
139
143
148
152
155
160

55,462
56,563
58,189
60,667
63,803
65, 702
67,700

53,136
54,498
56,115
58,217
60,686
62,882
64,601

1,122
1,241
1,249
1,249
1,317
1,454
1,629

1,867
2,132
2,005
2,018
1,996
2,108
2,184

53,881
55,389
56,871
58,986
61,365
63, 536
65,156

2, 460
2,204
2,217
2,598
3,269
2,975
3,259

1,581
1,174
1,318
1,681
2,438
2,166
2,544

46,706
47,477
48,735
50,741
53,163
54, 4b9
56,015

134
139
143
148
152
155
160

46,572
47,338
48,592
50, 593
53,011
54,304
55,855

44.433
45,405
46,752
48,594
50, 340
51,737
53,012

913
993
1,006
1,024
1,065
1,168
1,314

1,549
1,808
1,582
1,626
1,640
1,718
1,778

45,069
46,220
47, 328
49,196
50,915
52,287
53, 476

2,273
2,072
1,983
2,147
2,823
2,722
3,003

1,503
1,118
1,264
1,397
2, 096
2,017
2,379

650
635
634
632
627
613
610

650
635
634
632
627
613
610

541
525
512
505
524
536
508

12
8
8
11
15
18
20

12
7
8
7
9
9
9

541
524
512
501
518
527
497

109
110
122
127
103
77
102

109
111
122
131
109
86
113

2,902
2,963
3,050
3,186
3,275
3,208
3,267

2,902
2,963
3,050
3,186
3,275
3,208
3,267

2,990
2,980
3,103
3,253
3,283
3,238
3,337

95
95
83
92
97
114
140

143
181
133
118
129
126
130

3,038
3,066
3,153
3,279
3,315
3,250
3,327

-8 8
-1 7
-5 3
-6 7
-8
-3 0
-7 0

-136
-103
-103
-9 3
-4 0
-4 2
-6 0

16,853
16,995
17,274
18,062
19,214
19,447
19,768

16,853
16,995
17,274
18,062
19,214
19, 447
19,768

16,983
17,582
17,986
18,726
19,793
20,182
20,362

339
365
377
381
394
460
488

188
258
192
225
183
182
184

16,832
17,475
17,801
18, 570
19, 582
19,904
20, 058

-1 3 0
-587
-712
-6 6 4
-579
-735
-594

21
-4 8 0
-527
-508
-3 6 8
-4 5 7
-2 9 0

3,906
3,903
3,951
4,036
4,151
4,261
4,313

3,906
3,903
3,951
4,036
4,151
4,261
4,313

3,866
3,888
3,942
3,965
4,048
4,150
4,284

68
80
78
76
81
70
95

101
92
103
120
123
129
133

3,899
3,900
3,967
4,009
4,090
4,209
4,322

40
15
9
71
103
111
29

7
3
-1 6
27
61
52
-9

11,566
11,778
12,160
12,716
13,245
13,606
14,081

11,566
11,778
12,160
12,716
13,245
13,606
14,081

10, 239
10, 484
10,800
11,280
11,504
11,872
12,210

201
239
227
237
242
256
285

495
598
566
530
551
588
605

10, 533
10, 843
11,139
11,573
11,813
12, 204
12, 530

1,327
1,294
1,360
1,436
1,741
1,734
1,871

1,033
935
1,021
1,143
1,432
1,402
1.551

2,800
2,877
2,957
3,023
3,100
3,225
3,383

2,800
2,877
2, 957
3,023
3,100
3,225
3,383

2,649
2,728
2. 849
2,953
2,954
3,156
3,271

35
36
46
37
41
41
46

125
115
107
119
128
142
147

2,739
2, 807
2,910
3,035
3,041
3,257
3.372

151
149
108
70
146
69
112

61
70
47
-1 2
59
-3 2
11

7,894
8,186
8,566
8, 939
9, 399
9,944
10, 432

7,164
7,219
7.561
7,912
8,233
8,603
9,040

163
172
188
190
194
211
240

485
557
473
507
517
542
570

7,486
7,604
7,846
8,229
8, 556
8,934
9,370

864
1,106
1,148
1,175
1.318
1,496
1,552

408
582
720
710
843
1,010
1,062

8,890
9,225
9,596
10, 074
10, 792
11,398
11,846

8,703
9,093
9,363
9,623
10,346
11,146
11,590

210
247
243
225
252
284
315

318
324
423
392
356
390
406

8,811
9.170
9, 543
9,790
10,450
11,252
11,681

187
132
233
451
446
252
256

79
55
53
284
342
146
165

8,028
8,325
8,709
9,087
9,551
10.099
10, 592
8,890
9,225
9,596
10, 074
10, 792
11,398
11,846

134
139
143
148
152
155
160

Adjusted
to payroll
basis9

Payroll minus
household
(unadjusted)

Adjusted payroll minus
adjusted house­
hold series

May. These surveys were notconducted in 1967 and 1968; data represent rough approxi­
mations calculated on the basis of May 1966 survey results.
* Equals households series (excluding private household workers), less unpaid
absences plus additional paid jobs.
9 Excludes private household workers; includes forestries and fisheries.

• Based on March 1968 benchmark data.
* Derived as 90 percent of agricultural services, forestries and fisheries, SIC 07-09.
J Excludes private household workers.
< Includes only secondary nonagricultural wage and salary paid iobs. Data for 1962 6 reflect actual results of multiple jobholding surveys conducted during the month of




Differences

20

expansion in trade and service activities in recent
years, which has provided increased opportunities
for part-time work to persons already employed.
Another development in the job market that would
logically increase the supply of part-time workers
is the continuing downtrend in the full-time
workweek in various sectors.
Payroll em ploym ent in the service industries
was 1.5 million higher than the household count
in both 1967 and 1968. In 1968, there were 10.6
million workers in the service industry in the pay­
roll series compared with 9.0 million workers
reported in the household series. Adjusting the
household series for unpaid absences (240,000)
and estimated secondary jobs (570,000), and the
payroll series for workers in agricultural services
(160.000) , results in a residual difference of about
I . 1 million. Aside from the probability of under­
statem ent of dual jobholding in the household
survey, some of the residual difference can possibly
be traced to benchmark problems. Since the
industry tends to have smaller units and a fast rate
of turnover among firms, deviations from bench­
marks m ay be sizable. In addition, the timing of
benchmarks (March) could mean that seasonal
firms, which are typical in services, are in­
adequately accounted for both in the benchmark
and throughout the year as a moving constant.
The direction of this suspected deviation is not
known, however.
Employees on government payrolls totaled
over 11.8 million in 1968, compared with about
I I . 6 million in the household series, a difference
of 260,000.12 The usual adjustment technique to
the household data does not significantly narrow
the difference, however, as the number of
secondary jobs added (about 410,000) was
nearly offset by the number of unpaid absences
(320.000) , resulting in a remaining difference of
160,000. One source of discrepancy can be traced
to the payroll series for the Federal Government,
which as explained earlier, counts all civilian
employees on the rolls on the last day of the
calendar month and any interm ittent workers
who worked at any time during the month; this
contrasts with the household count covering a
single reference week. It is not possible to indi­
vidually discuss Federal, State, and local govern­
ment employment differences, however, because
the two subsectors are not separately identified
in the household series tabulations.
Household survey estimates of wage and salary

year and 14- and 15-year-olds for 1967 and 1968;
data by industry, which would permit these two
adjustments, are not available. For this reason,
the adjusted household totals in tables 4 and 5
will not coincide.
Adjusting the household series to a payroll basis
does not fully reconcile the overall levels of the
series but does tend to reduce existing differences.
For example, in 1968, total nonagricultural wage
and salary employment, as measured by the
household survey, was 64.6 million, compared
with 67.9 million in the payroll survey, a net
difference of 3.3 million. The number of unpaid
absences to be subtracted from the household
figures totaled 1.6 million, while the estimated
secondary-job count to be added was 2.2 million.
Subtracting the 160,000 wage and salary workers
in agricultural services from the payroll series
results in an adjusted payroll count of 67.7 million.
Thus a residual difference between the two series
of 2.5 million workers remains.
In light of the totally different sampling,
collection, and estim ating m ethodology used in
the two series, it is perhaps more noteworthy that
there is a high degree of consistency between some
of the industry estimates. This is true not only for
large sectors like manufacturing and government
but also for some of the much smaller groups,
particularly transportation and public utilities,
and finance, insurance, and real estate. On the
other hand, certain industries show distinct
problems. Those industries with relatively ex­
tensive dual jobholding are, by and large, the
industries in which the largest differences persist.
M ost significant in contributing to the overall
disparities between the two series are the m o v e ­
ments in trade and services.
I n d i v i d u a l i n d u s t r i e s . Differences in estimates
of em ploym ent in trade during the 1962-68 period
accounted for over 50 percent of the net difference
in total nonagricultural wage and salary em ploy­
ment. In 1968, the discrepancy amounted to 1.9
million workers, as there were an estimated 14.1
million in the payroll series compared with 12.2
million persons in the household series. After
adjustment for unpaid absences (280,000) and
dual jobholding (600,000), the difference totaled
nearly 1.6 million.
It is quite probable that the number of dual
jobholders enumerated in the household survey is
somewhat understated, due to the continued




21

Table 6. Nonagricultural wage and salary employment,
monthly, 1968

employees in manufacturing industries exceeded
the number of workers in manufacturing in the
payroll series during the entire 7-year period.
The largest difference, nearly 740,000, occurred in
1967, when household estim ates of nonfarm wage
and salary employment in manufacturing totaled
20.2 million workers compared with 19.4 million
workers in the payroll series. Of all the industry
divisions, manufacturing has the largest count of
persons on unpaid absences, numbering 460,000
in 1967, and more than offsetting the small num­
ber of persons holding secondary jobs (180,000).
As a result, the gap between the two series was
reduced to nearly 460,000. Considering the size
of the industry and the magnitude of the adjust­
ment for pay status, residual differences in 1967,
as well as in the other years under discussion, are
probably not significant.
Household estim ates of wage and salary em­
ployees in the construction industry also exceeded
the number of workers on construction payrolls
each year during the 7-year span. In 1968, how­
ever, there were 3.3 million in the household
series, slightly more than the number in the
payroll series. While the differences are not
significant either before or after adjustment, they
can partly be attributed to the fact that the pay­
roll series covers workers in contract construction,
while the number reported in the household series
may not accurately differentiate between contract
construction, force-account, and
speculative
construction.13 Aside from the problem of defining
construction wage and salary workers, other
problems probably stem from the number of
small, short-lived firms in the industry.

[In thousands]

Month

January______________
February_____________
March_______________
April________________
May _______________
June_________________
July_________________
August_________ _____
Septem ber.. __________
October______________
November____________
December____________

65, 765
66,115
66,475
67,170
67, 465
68, 470
68, 036
68,205
68,610
68,959
69, 247
69, 805

Payroll
Household
minus
series2 household
series

62, 740
63,313
63, 446
63,752
64, 263
65,104
65, 734
65, 876
64, 599
65,125
65, 358
65, 902

3,025
2, 802
3,029
3,418
3,202
3, 366
2,302
2,329
4,011
3,834
3, 889
3,903

Month-t o-month
cha nge
Payroll
series
3- l,9 2 0
350
360
695
295
1,005
-434
169
405
349
288
558

Household
series
3-l,4 7 7
573
133
306
511
841
630
142
-1,277
526
233
544

1 Based on March 1968 benchmark data.
2 Excludes wage and salary workers in private households.
3 Change from December 1967 to January 1968.

in the trade and service sectors. Of the 4.0 million
net differential recorded in September, for ex­
ample, 2.0 and 1.7 million were noted in trade
and services, respectively. On the other hand,
estim ates for most of the other industries were
surprisingly consistent.
W hile month-to-m onth changes in each series
can be explained in large part, divergences between
the two in any given month are more difficult
to reconcile than divergences in the annual
averages. T he only difference between the two
series that can be quantified and therefore
adjusted on a m onthly basis (on both a total and
industry-by-industry basis) is the count of unpaid
absences. Such an adjustment should theoreti­
cally elim inate one major source of disparity but,
unfortunately, it also widens the gap in em ploy­
ment levels. Moreover, while year-to-year changes
in the two series have at least been in the same
direction, though not of the same magnitude,
significant differences in m onth-to-m onth changes
occur both in direction of m ovem ent and in size,
which cancel out to a certain extent in comparisons
of annual averages.
As an illustration of differences in the direction
of change, payroll em ploym ent decreased by
about 430,000 workers (from 68.5 to 68.0 million)
in July 1968, as declines in trade, manufacturing,
and governm ent countered gains in the other
industry sectors. In contrast, em ploym ent in the
household series increased by 630,000 workers
(from 65.1 to 65.7 million), as pickups were reg­
istered in all industry sectors except mining and
manufacturing, both of which remained unchanged
from June. Similarly, from August to September,

M onthly com parisons

In addition to a reconciliation of the annual
levels of the two series, further insights into the
problem of comparability may be gained by look­
ing at month-to-month movements in em ploy­
ment levels. An examination of published m onthly
household and payroll estimates of nonagricultural wage and salary employment levels for
1968 reveals divergences between the two series
ranging from 2.3 million in both July and August
to 4.0 million in September. (See table 6.) This
compares to the 3.3 million differential for 1968
on an annual average basis.
As in the annual comparisons, the largest
disparities in overall monthly levels took place




Payroll
series1

22

ularly in July, when vacations are common and
many vacationing employees are not eligible for
pay. Though diminishing in recent years, the num­
ber of persons on unpaid vacation is still consider­
able. The household count of employed nonagricultural wage and salary workers on vacation in
July 1968 totaled 7 million, and more than 1.6
million of these persons did not receive pay for the
time off. These absences were concentrated
primarily in manufacturing, trade, services, and
government.
Similarly, variations in the two series are also
reflected by the relatively high incidence of ab­
sences due to bad weather and illness, particularly
in the winter months and during periods of major
industrial disputes.14 The 1967 auto strike, in­
volving over 150,000 workers, accounted for a
significant swing in the two series between
September and November 1967. When unpaid
absences are subtracted from the household series,
the month-to-month movem ents and consequently
the seasonal pattern more nearly resemble those
of the payroll series, particularly with respect to
July and September. (See table 7.)
Because of the design and timing of the multiple
jobholding survey— conducted in M ay of most
years since 1962 as part of the regular household
survey— the effects of multiple jobholding on
em ploym ent levels from m onth to m onth cannot
be quantified. It is likely, however, that changes
in em ployment levels during certain seasons of the
year can, in part, be attributed to wide seasonal
swings in the extent of dual jobholding. For
example, taking additional jobs at Christmas time
by persons who are already employed would be
reflected in the establishment series but not in the
household series, since workers holding more than
one job are classified according to the major
activity only.15 M ultiple jobholding could largely
account for the sharp rise in em ploym ent in the
establishment series in December and the sub­
stantial decline in January, while changes in the
household series are much smaller, although
generally in the same direction. During the past
7 years, the Novem ber-to-Decem ber increase in
nonagricultural wage and salary em ployment has
averaged 490,000 in the establishment series
compared with about 325,000 in the household
series. B y the same token, the decline in January
has averaged about 1.7 million in the establish­
ment series compared with about 1.2 million in
the household series.

employment in the household series declined by
1.3 million workers (from 65.9 to 64.6 million),
with unemployment cutbacks in all industry sec­
tors except government. On the other hand, the
payroll series showed a gain of 400,000 (from 68.2
to 68.6 m illion); em ploym ent pickups in manufac­
turing, transportation and public utilities, trade,
and government offset declines in mining, construc­
tion, finance, insurance and real estate, and
services.
Much of the disparity in monthly em ployment
levels is inherent in the different seasonal patterns
of the two series, although both series are subject
to the same general seasonal fluctuations. The
amplitude of their seasonal factor— original level
divided by the seasonally adjusted level— ranges
from a little over 98 percent to over 101 percent.
At current em ploym ent levels, this 3-percentage
point differential implies a seasonal expansion and
contraction of approximately 2 million employees
during the year for each series.
The two series move fairly consistently during
the first half of the year, as each shows steady
em ploym ent increases through June. (During the
January-June 1968 period, m onthly payroll em­
ploym ent increases averaged 540,000 compared to
470,000 in the household series.) The only real
difference in the first half of the year is that the
payroll series has its seasonal low in February
while the household series’ low occurs in January.
After June, however, certain significant dif­
ferences emerge. The establishment series dips
sharply between June and July, then resumes its
upward m ovem ent, reaching a seasonal peak in
December. The household series continues upward
from June to a seasonal peak in August, drops
significantly in September, then rises slightly dur­
ing the remainder of the year, although December
edges off marginally. In part, these differences
can be attributed to different treatment in each
series of unpaid absences and dual jobholding.
However, other factors discussed earlier undoubt­
edly exert some influence on employment levels.
Seasonality caused by vacation-taking appears
an important reason for monthly variations in
unpaid absences, particularly in the summer
months. For example, in July and August 1968,
unpaid absences totaled 2.7 and 2.9 million,
respectively, compared with 1.2 million in both
M ay and September. The exclusion of workers on
unpaid absences from the establishment series
leads to a seasonal dip in the summertime, partic­




23

Table 7. Current seasonal adjustment factors for non­
agricultural wage and salary employment

Similarly, there m ay be seasonal fluctuations
in the number of workers who normally hold both
agricultural and nonagricultural jobs and spend
more time at one type of work or the other in any
one month, depending on the demands of the
harvesting season. Such workers are counted in
their nonagricultural jobs each month in the
establishm ent series, but by their major activity
during the survey week in the household series.
As noted earlier, different treatm ent of school
teachers and other educational staff creates di­
vergences in em ploym ent levels, particularly in
the summer months. In similar fashion, the en­
trance and exit of young persons into and out of
the labor market during the summer have a
significant effect on employment levels. The
decline in em ployment in the household series in
September reflects primarily the large number of
youngsters leaving the labor force to return to
school. Part of the explanation as to why the pay­
roll series actually increases in September is
that it m ay not be as responsive to changes in
teenage em ploym ent as the household survey;
as noted earlier, a number of summer resort,
hotel, and amusement undertakings where em­
ploym ent of youth is prevalent m ay be missed
by the benchmark source and thus in the current
em ploym ent estimates.

Month

January.. _______________
February..................................
March..’.......................... .......
April____________________
May............. ...........................
June.
.............................
July..........................................
August___ ______ ________
Septem ber................. ..........
October__ ______________
December_______________

Payroll
series
(implicit)

98.6
98.4
98.8
99.6
99.9
101.0
100.1
100.2
100.6
100.8
100.8
101.4

Household
series
(implicit)

98.3
98.9
98.9
99.3
99.4
100.7
101.7
101.7
99.5
100.1
100.7
100.6

Household
(excluding
unpaid
absences)
98.4
99.0
99 4
99.7
100.1
100.4
100.0
99.8
100.2
100.7
101.1
101.3

the postwar business cycles for comparative pur­
poses. The data, however, should be used with an
awareness of their coverage, concepts, m etho­
dology, and limitations.
The household survey places its primary em­
phasis on the work status of individuals and
relates this status to other characteristics, such
as age, sex, color, educational attainm ent, and
marital status. However, it is not well suited to
providing detailed information on the industrial
distribution of em ploym ent and because of this,
em ploym ent levels by industry are not published.
T he payroll survey provides practically no infor­
mation on personal characteristics of workers
(except sex), but is an excellent source for de­
tailed industrial and geographic em ploym ent data.
It also provides hours aVid earnings data directly
related to the em ploym ent figures. Moreover, the
payroll series usually measures m onth-to-m onth
change more precisely than the household series
and therefore is more reliable for current analysis
of em ploym ent changes. Therefore, the payroll
and household surveys may be regarded as sup­
plementary and complementary. B oth serve a
useful purpose, and neither should be discarded
in favor of the other.
□

Selecting th e best series

In addition to explanations of the reasons for
differences between the two series, m any users
m ay wish to know which is the “best series”
under given circumstances. A meaningful answer to
this question can be given in terms of the purpose
for which the data are required. For cyclical
analysis, neither series should be overlooked. Each
has a long history, with data available for all of

F O O T N O TE S
1 Household data are collected in a national sam ple sur­
vey of approxim ately 50,000 households (called the Current
Population Survey) conducted m onthly by the Bureau of
the Census for the Bureau of Labor Statistics. A detailed
description of this survey appears in Concepts and Methods
Used in M anpower Statistics From the Current Population
Survey ( b l s R eport 313, 1967).
Establishm ent data are based on payroll reports from a
sam ple of 160,000 establishm ents em ploying over 30 million
nonagricultural wage and salary workers, collected by
State agencies in a cooperative program w ith b l s . For a
more detailed discussion of this survey, see the technical




24

note in Employment and Earnings, published m onthly by
b l s , and bl s Handbook of Methods for Surveys and Studies
( b ls Bulletin 1458, 1966), chapter 2.
2 All subsequent references to nonagricultural wage
and salary em ploym ent in the household series in text,
and tables, exclude private household workers.
3 For a discussion and reconciliation of differences
between the tw o types of em ploym ent series in 1961 and
earlier years, see Chapter V' and Appendix I of Measuring
Employment and Unemployment, President’s C om m ittee to
Appraise Em ploym ent and U nem ploym ent Statistics,
1962.

4 See H arvey R. Hamel, “ M oonlighting— An Economic
Phenom enon,” Alonlhly Labor Review, October 1967, pp.
17-22, reprinted as Special Labor Force Report No. 90.
5 M ultiple jobholding surveys were not conducted in
1967 and 1968. For purposes of reconciliation, rough ap­
proximations of the number of workers employed in second­
ary nonagricultural wage and salary jobs in these years
have been calculated on the basis of the M ay 1966 survey
results. In 1967 and 1968, it was estim ated that these
workers totaled 2.1 and 2.2 million, respectively.

lem was elim inated by asking an additional question in
the household survey of all persons reported as selfem ployed in a nonfarm business as to whether the business
was incorporated. The effect of this additional question
was to reduce the average level of nonfarm self-em ploym ent
by about 750,000 (in 1966) and to raise nonagricultural
wage and salary employment by a corresponding amount.
In a third change, which wTas relatively insignificant, per­
sons absent from their jobs during the survey week and
seeking other jobs were shifted from the unemployed to
the em ployed status (with a job but not at work). The
effect of this change increased total em ploym ent (in 1966)
by about 80,000, m ost of whom were nonagricultural wage
and salary workers. The net results of these changes were
to reduce the differential between the tw o series by ap­
proxim ately 400,000 in 1967 and subsequent years. The
fact that the 1968 differential equaled th at of 1966
(table 1) suggests th at a further (inexplicable) widening
occurred in the latter year.
11 The household survey has never been used to measure
industry em ploym ent, and industry absolutes are not
published. Such data are subject to sam pling errors to a
much greater extent than are the totals. In addition,
industry data are also largely affected by response and
classification errors to the extent th at respondents m ay
not accurately report the industry of em ploym ent. This
could be expected to occur m ost frequently between
construction and manufacturing (force-account construc­
tion, belonging in the latter industry) and wholesale trade
and manufacturing (where respondents report the latter
instead of the form er). These data are provided here,
however, because th ey contribute to an understanding of
com parative m ovem ents in the totals.
12 U ntil July 1969, when the March 1968 benchmarks
were introduced, the deviation between the tw o series
had been nearly 3 times this amount. Because the
payroll em ploym ent estim ate for the Federal Governm ent
is a complete count, the series is not subject to benchmark
revision; the benchmark adjustm ent for State and local
government em ploym ent (based on the Quinquennial
Census of Government) is performed at 5-year intervals.
As a consequence, any deviations in current em ploym ent
levels from actual benchmark levels tend to accumulate.
13 If construction activities are classified in other in­
dustries in the payroll series, this could also account for
minor reconciliatory differences. However, the differences
are probably concentrated in manufacturing and trade
(force-account construction) and finance, insurance, and
real estate (speculative construction). Force-account
construction refers to construction work performed by an
establishm ent primarily engaged in some business other
than construction, for its own account and use, and by
its own employees.
14 Workers on strike for the entire reference period arc
counted as em ployed (with a job but not at work) in the
household series but are not so counted in the payroll
series.
15 The c p s survey week in December is usually 1 week
earlier than the payroll survey reference period, which
would also have an im pact upon the probable extent of
m ultiple jobholding as well as upon divergences between
the two surveys for the month.

6 This estim ate was made using a “comparability assump­
tion” by Denis F. Johnston and James R. W etzel in their
article, “ Effect of the Census Undercount on Labor Force
E stim ates,” Monthly Labor Review, March 1969, pp. 3-13.
It is assumed here that the estim ated undercount for 1968
was the same as that for 1967. This undercount is an
extrem ely rough approximation and should not be accorded
a high degree of accuracy.
7 Because of special processing and enumeration prob­
lems of the Current Population Survey in December,
primarily due to the Christmas season, it is usually con­
ducted 1 week earlier, i.e., the week containing the 5th
day instead of the week containing the 12th. This earlier
survey week may also result in the missing of a consider­
able number of temporary sales workers in retail trade,
accounting for still another source of deviation in the two
series.
8 A discussion of the sam pling and estim ating procedures
together w ith estim ates of sam pling variability for both
series are published m onthly in Employment and Earnings.
9 These adjustm ents generally mean th at the em ploy­
ment series have been revised back to the previous com­
plete count and forward to the current m onth’s estimate.
For a detailed discussion of the adjustm ent of payroll em­
ploym ent levels to new benchmark levels, see “ BLS Estab­
lishment E m ploym ent Estim ates R evised to March 1968
Benchmark L evels,” July 1969 Employment and Earnings.
19 For 1967 and 1968, annual average household totals
are not precisely comparable to those for earlier years.
Im provem ents in the methods of measuring labor force
data initiated in January 1967 have clarified and sharpened
the household statistics. (A detailed discussion of these
conceptual changes can be found in “ New Definitions for
Em ploym ent and U nem ploym ent,” reprinted from the
February 1967 Employment and Earnings and Monthly Re­
port on the Labor Force.) Three particular changes affected
the household count of nonagricultural wage and salary
employment and consequently the differences between the
two series. First, the exclusion of 14- and 15-year-olds, as
discussed earlier, reduced total nonagricultural em ploy­
ment and at the same tim e widened the disparity between
the two series in subsequent periods. Secondly, a shift to
wage and salary em ploym ent of persons erroneously clas­
sified as self-em ployed had the effect of reducing the gap
between the series. In essence, estim ates of the selfemployed, particularly in the trade, miscellaneous service,
and construction industries, had been too high prior to
1967, because some persons were enumerated as selfemployed, who actually operated their own incorporated
enterprises and were therefore listed on the payrolls as
salaried officers of the corporation. This classification prob­




25

A 25-year look
at employment
as measured
by two surveys

Industry employment estimates based
on household interviews
and on payroll records
move similarly but differ
in levels and other specifics
CHRISTOPHER G. GELLNER

ment have tended to show much more smoothness
and stability historically than the household esti­
mates largely because of these factors but also be­
cause of different procedures used in compiling the
two series.
This does not mean, however, that the household
estimates are not a useful measure of employment in
major industries. The household series, in fact, has
some unique virtues. It is the only source providing
detailed insights into employment by industry in
terms of age, color, occupation, and similar charac­
teristics. It also extends coverage to some industries,
such as agriculture and private household services,
not covered by the payroll estimates. Furthermore,
it provides unemployment estimates by industry and,
thus, is the only source from which comparisons of
both employment and unemployment can be made
on an industry basis. Because both series of esti­
mates, are important, this article contrasts their be­
havior during the 1948-72 period for each of the
major industry groups, describing and explaining the
differences exhibited.

H ow h a v e t r e n d s in the data developed in the
payroll employment series and the household em­
ployment series of the Bureau of Labor Statistics
compared over a considerable period of time? The
computer has made it possible to analyze data on
nonagricultural wage and salary employment in
eight major industry groups and for the nonfarm
economy as a whole over a 25-year period, 1948 to
1972. This study of a large number of observations
shows that movements in the two series are strik­
ingly similar despite their separate sources.
The series compared

Information on the number of jobs and rate of
growth of employment by industry is developed by
the Bureau of Labor Statistics from data collected
by the Bureau of the Census in monthly interviews
at 45,000 households (the Current Population Sur­
vey— CPS) and from payroll data collected by BLS
monthly from 155,000 business establishments (the
Current Employment Statistics program— CES).
Data from the payroll series have generally been
considered more reliable than those from the house­
hold series. There are two reasons for this. First,
BLS believes that the industry of a worker can be
determined more precisely from authentic payroll
records of establishments whose industry classifica­
tions are periodically reviewed than from the an­
swers of household respondents. Second, the payroll
figures are derived from a much larger sample than
the household data— about 40 percent of payroll
employment while the household data are derived
from interviews conducted at less than a tenth of a
percent of the households in the universe. Move­
ments in the monthly payroll estimates of employ-

General trends and differences

The underlying movements of the industry data
from each series have been comparable over time,
but the series often differ in the levels of the esti­
mates, in month-to-month and annual changes, in
the timing and extent of business cycles, and in
other shortrun trends. These differences arise be­
cause the two series actually measure different phe­
nomena— the Current Population Survey counting
the number of persons employed and the Current
Employment Statistics program counting the number
of jobs occupied.1 This results in different treatment
in each survey of persons holding more than one job
and of those on unpaid absence. For example, in
the CPS, persons working during the survey week in
more than one job are counted once and are classi-

Christopher G. Gellner is an economist in the Division of
Employment and Unemployment Analysis, Bureau of
Labor Statistics.

26
From the Review of July 1973



fied according to the job in which they worked the
greatest number of hours. By contrast, in the estab­
lishment survey, workers on the payroll of more
than one business establishment would be counted
every time their names appeared on payrolls. Also,
people who have jobs but are temporarily absent
from them during the survey week because of ill­
ness, bad weather, vacation, a labor management
dispute, or for various other reasons are counted
nonetheless as employed— with a job but not at
work— in the household survey even if they did not
receive pay for this period of absence. In the payroll
survey, workers on unpaid absence are not counted.
The two surveys also differ in coverage, sources
of information, methods of collection, and estima­
tion procedures. These and other differences have
been comprehensively discussed and, wherever pos­
sible, quantified in two earlier analyses of the two
series. For a thorough discussion of the overall dif­
ferences which are only cursorily touched on in this
article, these two sources should be consulted.2 This
article also differs from the earlier analyses in that it
focuses on the historical trends of the two series,
but does not attempt a further reconciliation of their
differences. (For such reconciliations, see the two
earlier studies.)
The establishment survey covers only wage and
salary workers on the payrolls of nonagricultural es­
tablishments. The household survey includes all per­
sons who worked at least 1 hour for pay or profit
during the survey week— whether as wage and sal­
ary workers in establishments or private households
or as self-employed workers— as well as those who
worked 15 hours or more without pay in familyoperated enterprises. In this article, the household
survey estimates of employment by industry have
been adjusted to exclude private household workers,
the self-employed, and unpaid family workers. This
eliminates major coverage problems and facilitates a
more direct comparison with the payroll estimates.
However, differences caused by the two surveys’
contrasting treatment of multiple jobholding, work­
ers on unpaid absences, and other factors such as
the population undercount in the household survey
and the inclusion of persons under 16 years of age
in the establishment survey still affect the levels.3

clearly show (1 ) the two series have remarkably
similar trends over time, and (2 ) the payroll series is
significantly smoother, in terms of month-to-month
fluctuations. This fact is attributable primarily to the
much larger size of the establishment sample. It is
also due in large part to use of a link-relative tech­
nique to estimate monthly payroll employment. This
means that the previous month’s estimate is used as
a base in computing the current month’s estimate.4
In addition, the payroll series are adjusted once a
year to a benchmark— a complete and independent
count of employment for each industry.5 As a result
of this process the series is considered relatively free
of error over the long span. It is still possible, how­
ever, that the benchmark counts for certain indus­
tries, such as construction and services which
abound with short-lived and seasonal firms, may be
imperfect. On the other hand, the error on consecu­
tive months change for total nonagricultural employ­
ment estimated from the household survey is on the
order of plus or minus 225,000 workers based on
the current size of the sample (45,000 households).
Since the sample was considerably smaller during
the 1950’s and early 1960’s, the errors on monthto-month changes were larger.
Chart 1 shows that during the entire 1948-72 pe­
riod total nonagricultural wage and salary employ­
ment was consistently higher in the payroll than in
the household series. One possible cause for this gap
may be the so-called population undercount. The
household survey findings are applied each month to
a population count projected forward from the de­
cennial Census. There is evidence that a sizable
number of persons have been missed in each of the
recent censuses. Consequently, household survey es­
timates also tend to understate the true level of em­
ployment.
Both series have shown virtually the same growth
pattern over the period. Despite the short-term de­
clines in employment during some recessions both
series showed nearly the same average annual rate
of growth (compounded) during the period— 2.1
percent for the household estimates and 2.0 percent
for the payroll. Thus, there was only a slight nar­
rowing of the absolute gap between the series over
the years (2.3 million in 1948 compared to 2.0 mil­
lion in 1972). It is noteworthy, however, that this
narrowing was not of a long-term nature, occurring
largely during the recovery stages of the 1970-71
recession. In addition to this cyclical development,
which also occurred to some extent during previous

Computer-drawn charts are presented in this arti­
cle, tracing the monthly movements of the house­
hold and payroll series over the 1948-72 period for
total nonagricultural wage and salary employment
and in eight major industry divisions. The charts




27

Chart 1. Comparison of household and establishment survey employment— total nonagricultural industries, wholesale
and retail trade, and manufacturing— seasonally adjusted, 1 9 4 8 -7 2

Employment in thousands




Employment in thousands

16,000

15.000

14.000

13.000

12.000

11,000

10,000

9,000

28

recessions, the adjustment of the household series to
new population controls introduced in January 1972,
based on the 1970 census, had the effect of raising
the household figures for nonfarm wage and salary
employment by 250,000.® During 1961-69, years of
uninterrupted economic expansion, there had been a
widening of the actual gap to about 2.7 million. In
percentage terms, however, the gap between the two
series has narrowed somewhat over the past 25
years. In 1948 the payroll estimate exceeded the
household estimate by 5.4 percent; in 1972, it ex­
ceeded it by only 2.9 percent.
Differences in levels and movements in the two
series are clarified through an examination of indi­
vidual industry data. For example, the net difference
between the aggregate levels from the two series has
stemmed from differences in the estimation of trade
and services employment.

workers are picked up in the Current Population
Survey sample each month and estimates derived
from the sample are subject to considerable sam­
pling variability. Because of the small mining sam­
ple, a month-to-month change must exceed 50,000
before one can be 90-percent confident that a
change in employment levels occurred. By contrast,
the payroll series is subject to very little sampling
error. The sharp drops in the payroll figures in October-November 1949 and January-March 1950, as
well as the declines in 1966, 1968, and 1971, are
attributable to strikes. Employees on strike are not
included in the establishment series if they are off
the payroll during the entire reference week. In the
household survey, on the other hand, workers on
strike are counted as employed— with a job but not
at work. Consequently, the household figures did
not drop during heavy strike periods.

Patterns by industry

Construction. Wage and salary employment in con­
struction showed nearly identical trends and rates of
growth in the payroll and household series between
1948 and 1972. Generally, the household estimates
were slightly above the payroll estimates, with the
exception of relatively short spans in 1951-52 and
1971-72. Why the two series diverged in the early
1950’s and 1970’s, yet were relatively close other­
wise, is difficult to ascertain. Residential construction
reached record levels in the early 1950’s and 1970’s
and may have caused the divergence. Because of dif­
ferences in coverage, the household survey may have
been more likely to reflect employment increases dur­
ing periods of rapidly expanding construction activ­
ity, particularly in the residential area. (See chart 2.)
In the household survey all persons whose major
job is reported to be construction work are counted
in the construction industry. By contrast, the payroll
survey includes only workers employed by bona fide
contractors as defined in the Standard Industrial
Classification (SIC) system. It omits so-called
“force-account” construction workers— those em­
ployed by an establishment, the main product or
service of which is something other than contract
construction. Therefore, workers performing con­
struction in industries such as real estate develop­
ment, steel, automobiles, and trade are not included
in the contract construction classification but rather
in the industry called for by their establishment’s
main product. The extent of this classification differ­
ence is probably minimal during “normal” times but

Mining. The only industry group to show a decline
in employment during the 1948-72 period was min­
ing. Chart 2 shows that the decline in the payroll se­
ries was gradual and long term, occurring mostly
between 1948 and 1962. Laborsaving technology in
metal and coal mining accounted for the industry’s
employment shrinkage.
The household estimates of mining employment
have behaved more irregularly, but most of the de­
cline also occurred between 1948 and 1962. A
phenomenally large part of the decline occurred in
early 1954 and was probably related to a change in
the Current Population Survey sample introduced at
that time.7
Since 1962, the payroll series has remained rather
stable. On the other hand, the household figures
held steady until late 1970 and then rose enough to
close the gap which had existed during most of the
25-year period. Due to the recent relative narrowing
of the gap between the series, the average rate of
decline of mining employment over the period was
somewhat greater with respect to the payroll series
than the household series (table 1).
Month-to-month and short-term disparities be­
tween the two series with respect to mining employ­
ment are partly explained by (1 ) the relatively high
sampling variability of the household data, and (2 )
the two surveys’ different treatment of workers on
strike. Since the mining industry is comparatively
small, only a small number— about 500— of mining




29

Chart 2. Comparison of household and establishment survey employment— construction, finance, insurance, and real
estate, and mining— seasonally adjusted, 1 9 4 8 -7 2

Employment in thousands

Employment in thousands




30

Table 1. Household and payroll estimates of nonagricultural wage and salary employment by industry groups, selected
years 1948-72

Item

Total
nonagri­
cultural wage
and salary

Mining

Construction

Manu­
facturing

Transpor­
tation and
public
utilities

Wholesale
and retail
trade

Finance,
insurance
and real
estate

Services

Government

HOUSEHOLD SURVEY
1948___________________
1950___________________
.
1960_______________
1970___________________
1971....................... ..............
1972___________________

42,603
43,493
51,235
67,691
68,209
70,729

871
829
552
499
551
583

2,421
2,501
2,962
3,525
3,668
3,890

15,477
14,903
16,552
20,224
19,119
19,437

4,141
3,943
4,033
4,476
4,431
4,553

8,474
9,029
10,092
12,945
13,789
14,392

1,658
1,754
2,567
3,578
3,696
3,934

4,300
4,745
6,542
10,000
10,188
10,611

5,261
5,789
7,935
12,424
12,764
13,329

Change in level, 1948 to
1972_________ ________

+28,126

-2 8 8

+ 1,469

+3,960

+ 412

+ 5,918

+2,276

+6,311

+ 8,068

Average annual rate of
growth (compounded)___

+ 2.1

-1 .7

+ 2 .0

+ 1.0

+ 0.4

+ 2 .2

+ 3 .7

+ 3.8

+ 4 .0

1948___________________
1950___________________
1960___________________
1970___________________
1971___________________
1972___________________

44,891
45,222
54,234
70,593
70,645
72,764

994
901
712
623
602
607

2,169
2,333
2,885
3,381
3,411
3,521

15,582
15,241
16,796
19,349
18,529
18,933

4,189
4,034
4,004
4,493
4,442
4,495

9,272
9,386
11,391
14,914
15,142
15,683

1,829
1,919
2,669
3,688
3,796
3,927

5,206
5,382
7,423
11,612
11,669
12,309

5,650
6,026
8,353
12,535
12,856
13,290

Change in level, 1948 to
1972______ ___________

+27,873

-3 8 7

+ 1,352

+3,351

+ 306

+ 6,411

+2,098

7,103

+7,640

Average annual rate of
growth (compounded)___

+ 2 .0

-2 .0

+ 2 .0

+ .8

+ .3

+ 2.2

+ 3 .2

+ 3.7

+ 3 .6

PAYROLL SURVEY

may increase during booms in residential construc­
tion when developers may do more of their own
construction work.8
Also during boom periods in homebuilding,
many small firms or associations of individuals enter
the industry “to get a piece of the action.” Many
may not register under the unemployment compen­
sation laws (even if required to do so by State law)
because of the added expense or because in the past
they were too small to qualify. The payroll survey
only covers firms or establishments registered under
State unemployment insurance programs. Even if
registered under the unemployment insurance laws,
however, firms entering the industry would not be
picked up in the payroll sample until after the an­
nual complete census of establishments— the bench­
mark. Thus, employment in firms which did not re­
main in operation until the next benchmark would
never be included in the payroll count.

Manufacturing. Chart 1 shows that during most of
the 1948-60 period the payroll estimates of wage
and salary employees in manufacturing slightly ex­
ceeded the number of workers in manufacturing es­

Thus it is possible that the divergence between
the two series in 1971-72 may have been related to
the surge in residential homebuilding, which pushed
housing starts to record levels. The divergence in
1951-52 cannot be connected as easily to housing
starts— since residential housing starts peaked in
1950, then fell somewhat in the ensuing 2 years

timated in the household series. In 1961-62, the se­
ries moved together, and since 1962, there has been
a reversal of the previous pattern, with the
household employment levels exceeding the payroll.
Nevertheless, both series showed fairly similar aver­
age annual rates of growth for the 1948-72 period
— 0.8 percent for the payroll estimates and 1.0 pcr-




when the gap between the series developed.
But the recent divergence between the two series
may not be tied solely to increased construction ac­
tivity in the residential sector. When the series
began to drift apart in 1969, residential construction
was not particularly strong. Furthermore, when
housing starts showed strong spurts at other times
between 1948 and 1972, such as in 1954-55, 1959,
and 1963-64, the household series did not show an
inordinate degree of increase compared with the pay­
roll data. Compared with the booms in the early
1950’s and 1970’s, however, the increased residen­
tial construction activity during these other periods
was relatively mild.

31

Chart 3. Comparison of household and establishment survey employment— services, government, and transportation
and public utilities— seasonally adjusted, 1 9 4 8 -7 2

Employment in thousands

Employment in thousands

13,000

12,000
Household survey
Payroll survey

11,000

4,250

4,000

3,750

1948

1950




1952

1954

1956

1958

1960

32

1962

1964

1966

1968

1970

1972

cent for the household (table 1). The two surveys
also showed about the same rates of employment
growth for both the durable and nondurable goods
sectors of manufacturing. For both sectors, payroll
estimates of employment were generally a little
above the household estimates during the 1950’s but
moved below them in the early 1960’s. This reversal
in position occurred first in nondurable goods and
was followed shortly thereafter by durable goods.

in the industry exceeds the number of secondary
jobholders.
Transportation and public utilities. Employment
growth in transportation, communications, and pub­
lic utilities as measured by both surveys fluctuated
widely during the 1948-72 period. Apart from cycli­
cal effects, the payroll series exhibited considerably
more stability than the household series, which
shows a significant degree of month-to-month varia­
tion. Trends in both series during the period were
similar, although the payroll series generally was
above the household series, particularly during the
1950’s. As in manufacturing, industry classification
revisions introduced in 1960 may have contributed
significantly to the levels of the household data for
transportation more closely approximating those of
the payroll series (chart 3 ).
Chart 3 also indicates that both employment se­
ries moved downward during the 1950’s. Long-term
declines in employment in railroad transportation
and local and interurban passenger service were the
major causes of this. After the 1960-61 recession,
employment expanded relatively rapidly until the
1970-71 economic slowdown. Gains in trucking and
air transportation, communications, and public utili­
ties figured importantly in the growth of the 1960’s.
However, because of declines in the 1950’s, the 25year growth in the industry was slight— 410,000 in
the household survey and 300,000 in the payroll.

In terms of cyclical developments, chart 1 also
shows that both series of data on manufacturing em­
ployment have behaved similarly. Between 1962 and
1969, a period of sustained economic growth, both
expanded at the average annual rate of about 3 per­
cent, and both series plummeted at a particularly
rapid pace during the 1970-71 recession.
Current differences between the two series in lev­
els of manufacturing employment can be ascribed in
large part to the contrasting ways of treating work­
ers on unpaid absence and with more than one job.
Manufacturing has a relatively large number of un­
paid absences— more than any other industry group
— yet a comparatively small number of secondary
jobholders. Since persons on unpaid absence are
counted as employed only in the household survey,
theoretically the household estimates should exceed
the payroll figures. However, as chart 1 shows, this
has been the case only since the two series switched
relative positions in the early 1960's.
In 1960, the industrial (and occupational) classi­
fication was revised in the household series. The
purpose of this revision was to improve the quality
of the data in terms of industry detail by making
household industry groups conform to the 1957 re­
vision of the Standard Industrial Classification
(SIC) system. Following this revision, an additional
question was added to the CPS questionnaire in
1960: What is the name of the employer of each
worker in the household? The purpose of this ques­
tion was to improve the accuracy of industry report­
ing. In troublesome classification cases, this question
permitted matching the individual’s firm with an ap­
propriate list of industry classification codes. It is
reasonable to assume that the increase in the house­
hold measurement of manufacturing employment re­
sulting from the addition of this question and the
use of the new classification system improved the re­
liability of the household estimates substantially,
since, theoretically, the household series should have
been yielding estimates above those of the establish­
ment series because the number of unpaid absences




Trade. Historically, the payroll estimates of employ­
ment in wholesale and retail trade have been sub­
stantially above the household estimates for the in­
dustry. In fact, the industry has generally accounted
for about half the net difference between the two
surveys in total nonagricultural wage and salary em­
ployment. The difference in trade employment did,
however, vary during the 1948-72 period, ranging
from 400,000 in 1950 to 2.2 million in 1969. By
1971, the gap had narrowed to 1.3 million. Despite
occasional disparities in growth in the short term,
the annual rates of expansion of both sets of esti­
mates over the 1948-72 period were identical— 2.2
percent (table 1).
A significant amount of moonlighting occurs in
trade and is the major reason the payroll estimates
have consistently exceeded those of the household
survey. Probably, the increase in moonlighting
causes the two series to diverge during business ex­
pansions and, conversely, the contraction of dual
33

Finance, insurance, and real estate. Employment in
this group of industries, as measured by both the
payroll and household series, expanded continuously
during 1948-72. Employment shown in the payroll
series rose consistently while the household series
fluctuated markedly, undoubtedly due to sampling
variability. (See chart 2.) There was, in addition,
some disparity in the rates of growth of the series.
The average annual rate of increase of the house­
hold survey was 3.7 percent, exceeding the payroll
survey’s 3.2 percent, the largest single industry dif­
ference during the period between the two series in
terms of rates of growth. As a result, the gap be­
tween the two series, which had been substantial
during the 1950’s, was completely closed by 1972.

jobholding causes them to converge during cyclical
downturns. Another, relatively minor, reason for the
gap between the two series is that the payroll sur­
vey, in contrast to the household survey, includes
some military personnel and inmates of institutions
employed in trade. Moreover, many wage and salary
workers in the industry were misclassified as selfemployed in the household survey before 1967.9
This happened because some owners of small retail
businesses, or salesmen with a loose relationship to
their organization, may regard themselves or may be
considered by the respondents in their households as
self-employed, although they are listed as salaried
officers or salesmen on the payroll of the business.
Services. As with trade, payroll estimates of wage
and salary employment in services historically have
been substantially higher than those of the house­
hold estimates. As chart 3 shows, the gap between
the two series gradually widened after the late
1950’s. Over the entire 25 years, it ranged between
0.5 and 1.5 million workers. When combined with
that in trade, this gap has accounted for the entire
net difference in household and payroll levels of
nonagricultural employment since World War II.
The average annual rate of increase of the two
series over 1948-72 was nearly identical— 3.7 per­
cent for the payroll series and 3.8 percent for the
household, a faster rate than any other industry
group except government during the 25-year period.
Differences in household and payroll estimates of
employment in the service industries stem in large
part from many of the factors that account for the
gap in trade: The number of dual jobholders in the
industry has consistently exceeded the number of
workers on unpaid absence. The gradual widening
of the difference between the two series is probably
due to increased dual jobholding. Another minor
factor is that services, like trade, is also an industry
in which members of the Armed Forces are likely to
hold jobs during off-duty hours.

Government. Among the major industry groups,
government (Federal, State, and local) posted the
largest increase in employment between 1948 and
1972, both in terms of the absolute change and in
average annual percent changes (table 1). The
household series, however, increased at a somewhat
faster rate than the payroll series, such that the
small gap between the two series that prevailed
throughout much of the period was completely
closed by 1972.
Persistent discrepancies between the levels of the
two series can be traced in part to the different
times when monthly employment in the Federal
Government is sampled in each survey, in addition
to a different treatment of multiple jobholders and
unpaid absences. The monthly household estimates
reflect government employment in a single week in
the month (the reference week). By contrast, the
payroll series for the Federal sector counts all civil­
ian employees on the rolls on the last day of the
month plus all intermittent employees who worked
during the month. Thus, some persons counted as
employed in government in the payroll survey were
not so classified in the household survey. There is at
present no method devised which can quantify the
effect of this difference in reference periods. How­
ever, since turnover among Federal employees is rel­
atively low, it should be fairly small. A major
exception would be in December when many
temporary workers are hired by the U.S. Postal
Service.
Another difference between the two series stems
from the treatment of teachers during the summer.
In the payroll survey, teachers are counted as em­
ployed regardless of whether they are paid only dur­
ing the school year or on a 12-month basis. There­

Another source of discrepancy is that service em­
ployment in the payroll series includes workers en­
gaged in agricultural services— about 250,000 in
1972— who are classified in agriculture in the
household survey. Working in the other direction
the payroll benchmark, because of its nature and
timing (March), may not pick up employment in
some of the short-lived or seasonal firms (mostly
businesses in resort areas open only during the sum­
mer months).




34

fore, teachers taking jobs during their summer
vacation would be counted twice. In the household
series, teachers would be counted only once, either
as an employed teacher (with a job but not at
work) or as employed in the job obtained over the
summer. The overall effect of these procedures
would be to lower the household level of govern­
ment employment relative to that of the payroll
level.

or more payrolls during the reporting period. On an
industry-by-industry basis, there may be other fac­
tors intrinsic to each industry which have been the
major causes of the discrepancies between levels.
Generally, however, the two series showed the same
long-term trends and rates of increase between
1948-72, both for total nonagricultural employment
and for employment in major individual industries.
Moreover, the cyclical movements of the two series
during the 25-year period also have been very much
alike, especially with respect to turning points.
Although some inconsistencies between the series
continue to prevail, each possesses unique qualities.
Since the payroll series is derived from reports of
industry establishments, it furnishes extremely relia­
ble employment estimates by industry. The house­
hold survey, on the other hand, provides demo-graphic and labor force detail not available from the
employer reports.
□

B e c a u s e o f d i f f e r e n c e s in methodology and con­
cepts, the payroll and household employment series
by industry cannot be expected to yield the same
magnitudes, even when the differences in coverage
have been eliminated. In terms of total nonagricultural wage and salary employment, the household
survey levels have historically been lower than the
payroll levels, primarily because the payroll series
counts workers more than once if they are on two

FOOTNOTES
1 To be more precise, the payroll series counts the total
number of persons appearing on the payrolls of business
establishments at any time during the survey week. Thus,
any job held by more than one person during the week
would be counted more than once.

(i For an explanation of the changes and an indication of
the differences in the household employment estimates re­
sulting from adjusting the CPS to population controls
based on information from the 1970 Census, see “Revisions
in the Current Population Survey,” Employm ent and Earn­
ings, February 1972.

- President’s Committee to Appraise Employment and
Unemployment Statistics, Measuring Employment and Un­
employment (Washington, 1962), chapter IV and appendix
I; and Gloria P. Green, “Comparing employment estimates
from household and payroll surveys,” Monthly Labor Re­
view, December 1969, pp. 9-20.

7 In February 1954, the CPS sample was expanded from
68 to 230 sample areas, although the overall sample size of
25,000 households was retained. Contemporaneously, a sub­
stantially improved estimation procedure (composite esti­
mate) was introduced which took advantage of the large
overlap in the sample from month to month. These two
changes improved the reliability of most of the major sta­
tistics by an amount equivalent to that of doubling the
sample size.

1 For a comprehensive discussion of the differences be­
tween employment data from the household and
establishment surveys, including the effect of the population
undercount in the household survey and the inclusion of
14- and 15-year-olds in the payroll series, see the sources
cited in footnote 2.

s According to a survey by the National Association of
Home Builders of 450 major U. S. homebuilders, who in
1971 planned to build one-fourth of all housing units, 80
percent were involved in activities other than homebuilding.
The largest group (61 percent) were in land development
See “Major Homebuilders in 1971,” Economic N ews N otes
for the Building Industry (National Association of Home­
builders, 1971).

1 For a description of the “link relative” technique used
in estimating monthly payroll levels of employment, see the
Technical Note in any recent issue of Employment and
Earnings.
•r’ A BLS benchmark is a comprehensive count of the num­
ber of workers on the payrolls of business establishments. It
is derived from a complete census of all establishments
covered by State unemployment compensation laws with
supplementary information from a number of Federal and
private agencies. Annually the payroll figures are updated to
reflect information from the March benchmarks of the
previous year. For a description of the payroll benchmarks,
see “BLS Establishment Estimates Revised to March 1971
Benchmarks Level,” Employment and Earnings, October
1972.




!*In 1967, a clarifying question was added to the CPS
questionnaire which asked all persons reported as self-em­
ployed whether or not the business was incorporated. Oper­
ators of small incorporated enterprises were claissified as
wage and salary workers instead of self-employed. See
Robert L. Stein, “New Definitions for Employment and
Unemployment,” Employment and Earnings and Monthly
Report on the Labor Force, February 1967.

35

Analyzing
the length of
spells of
unemployment

New findings show
average jobless spells
are shorter than previously
believed, but there are
more of them
HYMAN B. KAITZ

H ow long does a person remain unemployed on
average? A simple question, yet one that cannot

visit over 52,000 households throughout the
United States and ask about the labor force
status during the preceding week of all household
members 16 years of age and older. (The pre­
ceding week, which includes the 12 th of the
month, is called the reference week.) Individuals
looking for work during the reference week arc
asked how long they have been unemployed. On
the basis of these responses, a distribution of
unemployed persons by duration of unem ploy­
m ent to the end of the reference week is published
regularly, together with the average duration of
unemployment. The published material represents
a cross section of the unemployed, for the m ost
part before their unem ploym ent spell ends.
Let us begin by examining regularly published
distributions of unem ploym ent by duration for
the average reference week in each of the years
from 1948 through 1969.
In table 1, we find, for example, that during an
average reference week in 1969 about 133,000
people were unemployed more than 27 weeks.
This number represents the long-term unemployed
in 1 week, but it cannot easily be converted into
an estim ate of the number of long-term unem­
ployed persons in a month or year for several
reasons. M any of these people remain among the
long-term unemployed from week to week and are
counted repeatedly, while others find work or
leave the labor force, and still others enter the
long-term unemployed from among those who
previously had lesser amounts of unemployment.
A t the other end of the scale, we know that among
the 1,629,000 with less than 5 weeks of unemploy­
m ent in the average reference week in 1969, many
will ultim ately become long-term unemployed.
The long-term and short-term unemployed can
be separated if we consider only the average num-

be easily answered despite the wealth of data
available on the unemployed.
For many years the Bureau of Labor Statistics
has been reporting regularly an estimate of the
average duration of unemployment for those who
are unemployed in a particular month. During
1969 this figure was about 8 weeks.
Now it is possible to supplement this measure of
the duration of unemployment. This article de­
scribes a method for estimating the number and
the average length of all of the spells of unem­
ploym ent completed during the year. During 1969,
for example, this estim ating procedure indicates
that, on the average, a person who became un­
employed remained unemployed for about 5 weeks.
The differences between these two averages will
be explained below, but it is important to note
here that the two averages differ primarily because
they measure two essentially distinct groups.
It should also be pointed out that each measure­
ment is an estim ate since at the present time we do
not have so-called longitudinal surveys that follow
unemployed individuals week-by-week during
their spells of unemployment. Instead, we must
infer the length of the spells from a series of snap­
shots (surveys) at monthly intervals.
Because these estim ates are derived by using
analytical techniques relatively unfamiliar to labor
force analysts, a detailed development of the
method used is presented in the appendix.
Earlier analysis

Each m onth during the week including the 19th
(the survey week), Census Bureau interviewers
H ym an B. K aitz is chief of the D ivision of S tatistical
Standards, Bureau of Labor Statistics.

From the Review of November 1970



36

Table 2. Distribution of completed unemployment spells
by duration

ber who left unemployment in the reference week.
The average number who end unemployment spells
of varying lengths in an average week multiplied
by the number of weeks in a year yields the
estimated annual number of spells by duration.
Table 2 contains these data.
In table 2 we find, for example, 570,000 spells of
27 weeks or longer in 1969. Because of their length,
these spells probably correspond closely to the
number of people who were jobless this long in 1969,
since it is unlikely that they had more than
one spell in that year. Among the 24 million spells
of less than 5 weeks, many were completed by
individuals who had more than one spell. Conse­
quently, the shorter spells cannot readily be con­
verted into a corresponding number of people.
Similarly, the estimated total of 32 million spells
of unemployment ending in 1969 correspond to a
smaller number of people with some unemploy­
ment in that year since some people experienced
more than one spell.

|ln thousands]
Number of weeks
Year

1948________
1949________
1950________
1951________
1952________
1953________
1954.... ..
1955________
1956________
1957________
1958________
1959________
1960________
1961________
1962________
1963________
1964________
1965________
1966............
1967________
1968________
1969________

It should be emphasized that the data in tables
1 and 2 constitute two distinct ways of looking
at the unemployed by duration, and both are
essential to an understanding of patterns of un­
employment. However, while we have been ac­
customed to looking at the type of data presented
in table 1, the estimates in table 2 are new.
Table 1. Distribution of the unemployed by duration of
unemployment (up to the reference week)
[Annual averages in thousands]
Number of weeks
Total
Less
than 5
1948______
1949_
......
1950______
1951.. . ..
1952......... . . .
1953................
1954
1955____
1956..
___
1957_______
1958____
1959________
1960________
1961.. _____
1962______ .
1963________
1964____
1965_____
1966
1967______
1968____
1969............

2, 278
3,634
3,287
2,054
1,883
1,834
3, 533
2, 852
2,750
2,859
4,601
3,738
3,852
4,713
3,913
4,070
3,787
3,366
2,878
2,975
2,816
2, 831

1,300
1,756
1,450
1,177
1,135
1,142
1,605
1,335
1,412
1,408
1,753
1,585
1,719
1,806
1,659
1,751
1,697
1,628
1,573
1,634
1,594
1,629




5-10

505
863
754
421
390
368
811
598
594
651
958
778
823
964
812
877
798
707
573
675
613
627

11-14

164
331
301
153
126
114
305
217
211
240
438
335
353
411
323
354
319
276
206
218
197
200

15 26

193
428
425
166
148
132
495
366
301
321
785
469
503
728
534
535
491
404
287
271
256
242

Average
duration
27 and (in weeks
over
116
256
357
137
84
78
317
336
232
239
667
571
454
804
585
553
482
351
239
177
156
133

25, 580
30,450
25, 510
26,010
24, 540
25, 580
28, 60C
27, 300
28, 680
26,190
32, 350
31,220
33, 590
34,230
32,900
33,210
33, 860
33,670
35,900
31,110
32, 340
32,10C

Less
than 5

5-10

11-14

15-26

18,880
20,310
16,870
20,700
19,150
20,470
18, 720
19,720
21,010
18,150
21,890
22,190
24,190
23, 060
22,980
22, 590
24,130
25, CC0
28,410
22, 090
23,980
23,67C

4,190
5,150
4,150
3, 000
3, 510
3, 370
5, 300
4,360
4,480
4,420
4.160
4,200
4,280
5,220
5,230
5, 500
5,090
4 ,61C
4,420
5,680
5, 390
5, 360

900
1,630
1,270
780
680
630
1,320
860
960
1,130
1,470
1,190
1,380
1,330
1,150
1,320
1,200
1,120
900
1,160
1,020
1,110

1,140
2,320
2, 050
1,070
850
790
2, 020
h 360
1,420
1,650
2,650
2,180
2, 320
2,450
1,930
2,220
2,010
1,810
1,370
1,500
1,320
1,390

27 and
over
470
1,040
1,170
460
350
320
1,240
i; 030
810
840
2,180
1,460
1,420
2,170
1,610
1,580
1,430
1,130
800
680
630
570

Average
duration
(in weeks)

4.6
6.2
6.7
4.1
4.0
3.7
6.4
5.4
5.0
5.7
7.4
6.2
6.0
7.2
6.2
6.4
5.8
5.2
4.2
5. C
4.5
4.6

In table 3, the distribution of unemployment
by duration found in the average cross section is
compared with the distribution of average spells
of unemployment in 1969.
The contrast between the two duration dis­
tributions shown in table 3 is a marked one.
While only 2.8 million people were unemployed in
an average week of 1969, 32 million spells of un­
employment occurred during the year. The dis­
tribution of completed spells is more heavily
skewed toward the shorter durations than is the
case in the cross-section distribution. In the latter,
a little more than half of the unemployed in an
average Aveek are shown with fewer than 5 weeks
of unemployment, while almost three-fourths of
completed spells fall in this interval. As a result,
the average duration of unemployment spells of
4.6 weeks is only six-tenths of the average duration
in the cross section. The greater skewness and
loAAer average duration of completed spells is due
to the probability of leaving unemployment being
inversely related to the length of unemployment.
(See the appendix for detailed deA^elopment of
this point.)
The structure of unemployment Avhich emerges
from the data on spells is quite different from
previous ideas based on the characteristics of
cross-section data. Unem ploym ent duration is
much shorter on the a\'erage than had previously
been thought the case. B y far, the bulk of those
who become unemployed experience spells of only
a few weeks. These spells undoubtedly reflect,

Two views of the unem ployed

Year

Total

8.
10.
12.
9.
8.
8.
11.
13.
11.
10.
13.
14.
12.
15.
14.
14.
13.
11.
10.
8.
8.
8.

37

Table 4. Weekly continuation rates for 1969

among other things, the influx of young people in
the summer, interm ittent looking for work by
marginal workers, and seasonal activities. These
spells are not necessarily terminated by finding a
job. During 1969, for example, about half of those
leaving unemployment found work and the re­
mainder left the labor force. (These proportions
are inferred, under equilibrium conditions, from
the known fact that people who became unem­
ployed were drawn, in approximately equal pro­
portions, from the employed and from those out
of the labor force.) M any people look for short
periods because they must go back to other
duties such as attending school or keeping house.
Longer spells are more likely to be terminated
when jobseekers become discouraged and no
longer look for work. In addition, multiple short
unemployment spells are experienced by some
people, which add up to a substantial amount of
unemployment for them during the year, with
concomitant low annual earnings.
D ata on completed spells offers a variety of
avenues for analysis. The rest of this article
considers two: (1) implications in the distribution
of unemployment for one’s chances of leaving the
unemployed; and (2) the behavior of spells
of unemployment under changing economic
conditions.
To begin, we examine the distribution of unem­
ploym ent by duration under relatively stable
conditions with respect to one’s chances of leaving
the unemployed. We call this probability the “con­
tinuation rate” and it is fully explained in the
appendix. It is the probability that a person un­
employed n weeks remains unemployed for an
additional week. Rates for 1969 are in table 4.
In this table we find, for example, that for
people who have had 5 weeks of unemployment
to the reference week, almost four-fifths (78 per­
cent) will go on to experience a sixth week, and

Average cross
section
2,831,000
100.0

32,100, 000
100.0

Under 5 weeks___________ ____
5-10 weeks__________________
11-14 weeks_________________
15-26 weeks_________________
27 weeks and over_____________

57.6
22.1
7.1
8.5
4.7

73.7
16.7
3.5
4.3
1.8

Average duration__________ _______

8. 0 weeks

4.6 weeks




n = l _______________________________________________
2______________ ________________ ____ ____ .
3....... ............................. ............................................
4_______ ____ ____________________________
5_______________ _______________ _________ ____
6________________ ___________ _____
7................................................................................ .............
8____________________________________ __________
9_______________ _____________

.76
.67
.72
.75
.78
.80
.82
.84
.86

Average of 10 to 14_____ ____ _________
Averege of 15 and over._________ ____ _______ ____ ____

.89
.92

Spells and the business cycle

The basic data for the next analysis is given in
table 5. In this table we find, for example, that 3.51
percent of the civilian labor force was unemployed
in an average week of 1969, and 21.8 percent were
in their first week of unemployment.
The data are given in percentages to remove the
effects of population growth and cyclical labor
force changes from the corresponding numbers
which might otherwise have been used. The per­
cent of unemployed with 1 week of unem ploy­
ment in the average cross section for each year is
the same as the percent of unemployed in the aver-

Completed spells

Total: number__________________
percent___________________

Continuation rates

so on. We note that continuation rates generally
increase with the length of unemploym ent already
experienced. Two separate reasons for this phe­
nomenon may be offered, although this m ay not
exhaust the possibilities. Since these continuation
rates are calculated for all of the unemployed
combined, they are undoubtedly influenced by
the heterogeneity present among the unemployed
in their separate chances of leaving the unem­
ployed. For example, if we had two groups among
the unemployed, each with a constant but different
continuation rate, the cross-section distribution
for both combined would show the continuation
rate increasing with the length of unemployment.
A second possible hypothesis is that of feedback
effect in which the longer a person is unemployed,
the less chance he has of reemployment or of
otherwise leaving the ranks of the unemployed
the relative importance of these two factors' m ay
be investigated by considering the duration dis­
tributions for unemployed with very specific
characteristics, such as age, sex, color, occupation
family role, and so on. If each of these groups is
sufficiently homogeneous, increasing continuation
rates will reflect only the feedback effect. More
investigation of homogeneity is needed.

Table 3. Comparison of distributions of unemployment
by duration, 1969
Weeks of unemployment

Weeks of unemployment experienced up to reference week

38

Table 5. Data for business cycle analysis, 1 9 48 -69

age week who are beginning new periods of un­
employment. This percent (s) was regressed on the
unemployment rate (u) for the 22 years with the
following results:
s= 3 3 .6 1 7 —3.140u
( 12. 6)

? = .8 8 2 ,

Year

1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969

D — W = 2.33

The inverse relationship is very significant, as
indicated by the high correlation coefficient and
the t value of 12.6 for the coefficient of u. There is
no particular serial correlation as indicated by the
value of the Durbin-W atson (D-W) coefficient.
One alternate form of this equation was com­
puted with two additional independent variables:
the unem ploym ent rate lagged 1 year, and a
dummy variable equal to unity for 1967, 1968,
and 1969 to pick up possible effects of a change in
the labor force questionnaire in 1967. These added
variables proved not to be significant. Other
regression forms were not considered in the
present analysis. It should be noted that changes
in the age-sex-color composition (heterogeneity) of
the unemployed over the postwar period may
well affect this aggregate relationship. Also, since
the observations are annual averages, phases of
the business cycle are not well articulated in it.
Nevertheless, the aggregate relationship does
give some approximate results of interest to us.
First of all, increases in the unemploym ent rate
are accompanied by declines in the percent of
unemployed starting new spells, as the historical
data indicate. Specifically, a one-percentage-point
increase in the unemploym ent rate is accompanied
by a three-pcrcentage-point decline in the percent
of the unemployed starting new spells. As detailed
development in the appendix shows, the average
duration of spells is equal to the reciprocal of the
proportion with new spells (100/s). The number of
spells initiated in a year is given by the product
of the unemployment rate (u) and the percent of
unemployed with 1 week of joblessness in the
reference week. As noted, this number excludes
the effects of population growth and cyclical labor
force changes.
T otal unem ploym ent (u) is equal to the product
of the number of new spells (us) and the average
duration per spell (100/s). Consequently, small
percentage changes in the unemploym ent rate
from one period to another are approximately
equal to the sum of the percent changes in the
number of new spells and in the length of the




Percent of unemployed
in first week of
unemployment in
average reference week

21.6

16.1
14.9
24.3
25.1
26.8
15.6
18.4

20.0

17.6
13.5
16.0
16.8
14.0
16.2
15.7
17.2
19.2
24.0

20.1
22.1
21.8

average spell. Table 6 presents such percent
changes for increments of 0.5 in the unemployment
rate. All estim ates come from the equation
discussed earlier.
It must be emphasized that these results derive
from the form of the relationship specified. Until
a better form of the relationship is determined,
these results should be considered tentative and
should be evaluated against common sense con­
siderations. We note that at low unemployment
rates (between 3.0 and 3.5), the percent of newly
unemployed changes more rapidly than the
average spell length. This is not inconsistent
with the assumption that in a tight labor market,
the ranks of the unemployed contain a higher
than average proportion of people only marginally
connected with the labor force who swell the
unemployed for relatively short periods.
As unem ploym ent rises (up to about 5.5 per­
cent), the rate of increase in the length of average
spells rises, but the rate of increase in new un­
em ployment spells declines. Above an unemploy­
ment rate of 5.5 percent, the length of the average
spells increases faster than the unemploym ent
rate, with accompanying declines in the number
with new spells. This may reflect the conversion
of a potential series of short spells for some workers
into one long spell, with some workers being dis­
couraged enough to leave the labor force, keeping
them from entering the ranks of the unemployed
with new spells. It undoubtedly reflects more
substantially the behavior of workers with strong

39

Table 6. Percent of unemployed with new spells and average
spell length for selected values of unemployment rates

attachm ent to the labor force. The particular
sequence in which unemployment rates change
through the business cycle m ay also have effects
on the patterns of new spells and spell duration
which should be explicitly considered. In general,
the changing mix of workers with various degrees
of marginality at different levels of the unem ploy­
ment rate, and its contribution to these results,
can only be investigated through a disaggregated
analysis.
B oth of the brief analyses presented in the latter
part of this article are intended to serve mainly
as examples of what can be done with data based
on or related to spells of unemployment. T hey are
not intended to present substantial analyses in
their own right.
There are m any avenues of research. We plan to
explore current estim ates of completed spells

Unemployment rate
Level

Average duration

Percent
change

3.0_________
3.5.
4.0.

Weeks

6.9
4.42

14.3

7.5
4.75

4.5.

8.0
5.13

11.1

.792

6.4
.842
4.1
2.2
0.4
.899

10.6
6. 77

8.3

—1.4
.887

11.9
7. 57

9.1

.896

6.12

6 .0 .

.726

9.6

9.1

Percent
change

.877

5.58

5.5.

Proportion

8.8

10.0

6.5.

Percent
change

4.13
16.7

12.5
5.0.

Unemployed with new spells

-3 .2
.859

month by month. Satisfactory estim ates should
be possible from the available data but their
development awaits further empirical work.

APPENDIX
We first consider unemployment in a “steady
state” or equilibrium condition, in which the level
of unemployment remains the same from week to
week. The number of people leaving the ranks of
the unemployed each week is balanced by the
addition of an equal number of newly unemployed,
and the distribution of the unemployed by dura­
tion (in weeks) remains constant. This assumption
of a steady state in unemployment would approxi­
m ately represent the years 1968 and 1969, for
example, apart from seasonal change.

these people remain unemployed? The answer is
that in a steady state condition, their average
unemployment subsequent to the reference week
will be the same as their average period of unem­
ploym ent thus far. T hat is, their average com­
pleted spell of unemploym ent will be twice their
average period of unemploym ent thus far. The
reasoning for this m ay be given in nontechnical
terms. All those in the first category may be classi­
fied in subgroups by the length of their completed
spells. Consider, for example, the subgroup which
will ultim ately complete spells exactly 14 weeks in
length. Under the steady-state hypothesis, w ithin
the reference week some of these people will have
just begun their spells, while others are just con­
cluding theirs, with the remaining people in be­
tween. For this group we should therefore expect
to find that on the average they have already had
7 weeks of unemploym ent up to the reference
week.
In the same w ay we would expect to find that
on the average, each of the other subgroups is
halfway through its spells of unemploym ent. For
all of these subgroups combined we would there­
fore expect to find that, on the average, they are
halfway through their spells. To put it another
way, the average length of completed spell for
all people in the reference week will be twice the
average duration up through the reference week.

Three groups of unem ployed

Three categories of the unemployed are sep­
arately and explicitly considered in this analysis:
(1) all those with some unemployment in the
reference week; (2) those who have just begun
spells of unemployment in the reference week; and
(3) those who have just completed spells in the
reference week. These categories are not mutually
exclusive, since the first category includes those
in the other two categories and those with 1-week
spells will be in both the second and third cate­
gories.
The first category is the only one we observe
directly and for which we have the distribution
by weeks of unemployment. The question is
sometimes asked: H ow long on the average will




40

If the average duration up through the reference
week is 8 weeks, then we can say that these
people will ultim ately complete their spells with
an average of 16 weeks. (These computations
are subject to a small adjustment as will be noted
in the examples below.) This result is summarized
in the following statem ent:
Under equilibrium conditions, w ithin a given
week, the unemployed are halfway through their spells
of unem ployment on the average; thus, their average
duration of completed spells will be twice the average
duration of the unemployment already experienced to
the given week.
This method of reasoning does not, however,
tell us, for example, how m any people will have
14-week spells. The number in the reference week
with 7 weeks include those with 7 weeks or more
in their spells, but some people who will have 14week spells will also come from those listed with
fewer than 7 weeks of unemployment in the ref­
erence week. As will become evident, there is little
need to further pursue the question of the duration
of unemploym ent spells of people in the first category. We note that people in that category have
begun their spells at different times and will con­
clude them at different times, but they all are
unemployed in the reference week. The findings
for this category cannot be generalized to cover a
longer reference period, such as a month.
In order to extend our analysis to cover unem­
ployment in the other two categories, it will be
helpful to work with some simple duration
patterns. There is only one unemploym ent pat­
tern for which the average completed spell for all
people in the first category is the same as that
for the people in the other two categories: when all
the unemployed have exactly the same length of
spell, for example 8 weeks. Consider such an equi­
librium pattern with two people becoming unemTable A - l.

ployed, and two people leaving unemployment
each week. The unemploym ent distribution in the
reference week is shown in table A—1.
If we make our calculations to a single point of
time— for example, to the middle of the reference
week (the left hand column of table A - l ) — we
find that the average elapsed duration of unem­
ploym ent up to that time for all the unemployed
is 4 weeks. B ut we have specified that every un­
employed person will go on to experience 8 weeks
of unemployment. The relationship between these
two averages is therefore in accord with the general
rule previously stated.
M ore re a lis tic pattern

We next look at a more typical pattern of un­
employment by duration in a given reference week.
In a steady state, in any given reference week,
more people will be shown as having 1 week of
elapsed unemployment than as having 2 weeks;
more will have 2 weeks than 3, and so on. In
general terms, we need say only that the number
with n weeks of unemployment will not exceed the
number with in-1) weeks, whatever the value
of n. This statem ent allows us to include the
pattern presented in table A -l.
This generalization of the longitudinal pattern
of unemployment is consistent with reality as the
following argument show's. Suppose there are 100
people in the reference wreek with 4 weeks of un­
employment. In the following w'eek, at most, 100
people will now' be showrn with 5 w eeks of unem­
ployment. It is more likely that some will have
withdrawn from the ranks of the unemployed for
some reason, so that the number with 5 w'eeks of
unemployment will be less than 100. That the
number with 5 wreeks of unemployment in the
week following the reference week equals the
number with 5 w'eeks of unemployment in the
reference week is an essential characteristic of the
steady-state distribution, which has the same
duration pattern from week to week. Table A -2
shows howr this pattern keeps regenerating itself
in the case of this second simple duration pattern.
Each w eek five new people become unemployed.
Four of these will continue to remain unemployed
for a second week. Three of the four will remain
unemployed for a third week, and so on. Finally,
one person will remain unemployed through a
fifth week and then leave. N o one is unemployed

First simple duration pattern

Duration of unemployment (in weeks) to—
Middle of reference
week

End of reference
week

.5
1.5
2.5
3.5
4.5
5.5
6.5
7.5

1
2
3
4
5
6
7
8

Total........................................................................




Number of unemployed in
reference week

2
2
2
2
2
2
2
2
16

41

for more than 5 weeks in a spell. In this pattern
(which is longitudinal in nature contrasted with
the cross section or latitudinal pattern in the
reference week), we follow a particular group or
cohort from 1 week to the next as indicated by the
diagonal arrows. Any cohort followed in this way
for 5 weeks will have the same duration distribu­
tion as in the cross section in a single week. As
already noted, in a steady state, the cross-section
and longitudinal distributions are identical.
To compute the desired average durations we
can go through a simple arithmetic exercise. If we
take as our reference point the middle of the week
(see left hand column of table A -2), then the
average elapsed duration in the cross-section dis­
tribution is 1% weeks for the 15 unemployed
people. If we take the first cohort of five people
with 1 elapsed week to the end of the reference
week, then the average subsequent unemploy­
ment for this group from the middle of the refer­
ence week to the end of their respective spells is
2% weeks. Of these five people, one drops out after
1 week, so he has an additional duration of x/2
week. Another drops out at the end of the follow­
ing week with a total additional duration of 1%
weeks, and so on. In the same way we follow the
second cohort of four people and find that their
average additional duration of unemployment is 2
weeks. Similar calculations for the remaining three
original cohorts yield average additional durations
of iy2, 1, and y2 weeks, respectively. The average
additional duration for all five cohorts combined is
1% weeks— the same as the cross-section duration,
confirming the rule stated earlier. The final aver­
age length of completed spell for all of those with
unemployment in a selected reference week (the
first category) is twice this number or 3 % weeks.
Table A -3 shows how we obtain the duration dis­
tribution of completed spells for those who are
leaving unemployment in the reference week or for
those who enter unemployment.
The preceding discussion shows that we can
use the cross-section pattern in a single week to
deduce the experience of a group of unemployed
persons from the time they become unemployed
to the end of their spells. B y taking first differences
in the cross-section distribution (second column
from left) we get the number who leave unemploy­
ment by the length of their completed spells
whether begun or ended in the reference week.
I n a steady state condition, the pattern of com­
pleted spells fo r those who enter unemployment at the




42

Table A-2.

Second simple duration pattern

Duration of unemploy­
ment (in weeks) to—
Middle of
reference
week

End of
reference
week

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

Total________ .

Number of unemployed in reference week

Week 1

Week 2

Week 3

Week 4

Week 5

5
4
3 \ ^

1
15

^ 4

\

3 \

^ 3

2 \

^

^ 1

^ 1
15

^ 1
15

2

^ 1
15

15

same time is the same as fo r those who leave unem ­
ployment at the same time.
Because our specified cross-section distribution
includes individuals with different lengths of com­
pleted spells, we find in table A-3- that the cross
section of unemployment by duration differs from
the pattern of spells that will occur for all persons
with some unemployment, and from the pattern
for people entering or leaving unemployment.
Continuation rates

Before we examine actual cross-section duration
distributions of the unemployed, it will be helpful
to develop some additional properties of steadystate distributions. We start with a concept
already familiar in actuarial and demographic
analysis (the “survival rate” ) which is designated
here as the “continuation rate.” For our purposes,
the continuation rate (rn) is the proportion of
people unemployed n weeks who continue to
remain unemployed for the n-{-lst week. Clearly
rn must fall somewhere between zero and unity
and may change as n changes, or remain constant.
In a steady-state distribution, the value of rn
for a specified value of n does not change from
1 reference week to another. Consider the con­
tinuation rates associated with the two simple
distributions we have alread}'- discussed. (See table
A -4).
Apart from the fact that rn falls between zero
and unity, there is no other requirement it must
satisfy, other than that a value of zero ends the
pattern. We reject the case in which rn is constant
and equal to unity, since this gives us a distribu­
tion with infinite average duration, and no
movement out of the unemployed. However, we
admit consideration of duration patterns with

Table A-3. Second simple duration pattern by cross
section and completed spells of unemployment

It is apparent that, apart from a constant, (1—r),
the two distributions are the same, and they will
consequently have the same distributions and the
same averages. B ut only in the case of a constant
continuation rate will the two patterns coincide.
A critical difference between distributions by
duration of unemployment for all unemployed
persons and for those beginning or ending their
unemployment spells must be noted. As already
indicated, the results for persons in the first cate­
gory cannot be aggregated or averaged over vari­
ous time periods easily, but those for the other two
categories can. For example, spells begun (or con­
cluded) in 1 week do not overlap information for
spells begun (or ended) in any other week. An
answer to the rather imprecise question of what
is the character of unemployment spells in a year
can be provided by taking all those with some
unemployment in the first week of the year and all
those beginning unemployment in the other 51
weeks of the year, in order to develop unduplicated
averages or aggregates for the year. The character­
istics developed in this way would approximate the
spells for those becoming unemployed in the aver­
age reference week of the year. If the question were
stated in terms of all the spells begun or ended
during the year, it could be answered directly in
terms of the data for those beginning or ending
unemployment in the reference week.

Distribution by completed spells
Duration of
unemployment
(weeks)

l . _ _____ _________
2________________
3________________
4____
5________________
Total________
Average durations__

Cross section
(number of un­
employed in
reference week)

All persons
with some
unemploy­
ment

Persons
beginning
spells

5
4
3
2
1

1
2
3
4
5

1
1
1
1
1

1
1
1
1
1

15 (people)

15 (people)

5 (spells)

5 (spells)

156 (to middle of
reference week)

Persons
ending
spells

3
3
( o end of spell

infinite and reasonable averages which m ay be
represented by mathematical functions which
theoretically allow durations of infinite length.
This situation is the same as in ordinary statistical
practice when we assume that variables may be
represented by a normal distribution, which has
infinite tails to the left and to the right.
In order to understand the characteristics of
actual data, we shall need some additional in­
sights into completed spell distributions.
In a steady-state condition, with a constant con­
tinuation rate (less than u n ity), the cross-section
distribution pattern to end oj the reference week is
the same as the pattern oj completed spells fo r those
either beginning or ending their spells at the same
time.
The reasoning behind this statem ent is as fol­
lows. L et ip represent the number of people in
the cross-section distribution with 1 week of
completed unemployment; u2, those with 2 weeks;
and un, those with n weeks. Thus, u2 also represents
that part of the Ui who go on to have a second week
of unemployment, and so on. If r= co n sta n t con­
tinuation rate, then by definition, u2H-Ui =
r = . . . = u n+1-Hiin whatever the value of n.
Consequently, u2= r u 1, . . ., un+1= r u n whatever
the value of n. The number of unemployed in the
reference week who complete a spell of 1 week is
Ui —u2==Ui (1—r). Similarly, the number who
complete spells of n weeks is un—un+1 = un (1—r).
The two distributions m ay be set down together
as follows:
Wteks of
unemployment

Cross-section
distribution

Two additional general statem ents which are
needed for our use can be made for any duration
distribution:
The number oj new spells oj unemployment in any
period is the sum oj the number oj people in each week
in the period identified as having completed 1 week
oj unemployment.
I n a steady-state distribution, the average duration
oj completed spells is equal to the ratio oj the total
number oj unemployed in the reference week to the
number with 1 week oj unemployment.
This last statem ent is confirmed by a simple
algebraic formulation. Let un be the number of
people in the cross-section distribution with n
completed weeks of unemploym ent through the
end of the reference week. Let 11! = the number
with 1 week and u = to ta l number of unemployed’
in the reference week. Then u = T ^ u n. The number

Completedspell
distribution1

1

Ui

u i (1 —r )

2

uz

uz (1—r)

n

u„

un (1-r)




O ther general characteristics

71=1

43

of people who complete spells n weeks in length is

m ay well look different for a time from that for
people concluding spells. Finally, the average
duration of completed spell for those in the first
category may be temporarily less than or more
than twice the average in a given cross-section
distribution. However, in the process of aggregat­
ing or averaging to annual levels, these short­
term anomalies are smoothed away, so that in
this article it was legitimate to consider the aver­
age cross-section distribution for the year as
equivalent to a longitudinal distribution, and to
derive the corresponding distributions of com­
pleted spells by duration.
The methodology for obtaining completed spell
distributions m ay also be used for estim ating these
distributions month by month from the household
survey data to show how they change in response
to economic and other influences.
The balance of this appendix is confined to
description of the procedure for estim ating the
distribution by duration of completed spell for
1969. Cross-section distributions by duration are
ordinarily published within the class intervals
shown in table 1. Unpublished data are available
for a few categories of unemployed for single
weeks of duration. However, this detailed fre­
quency pattern reveals that: (1) some biases are
present in the recall of the first several weeks of
unemployment, and possibly elsewhere; and (2)
frequencies tend to bunch up at durations which
are multiples of a month; that is, quarters and
half years. It was therefore deemed advisable to
attem pt to fit smooth mathematical functions to
the data within the published class intervals to
derive the estim ates of the completed spell dis­
tributions. A major indication of the erratic nature
of some of the frequencies lay in the variation of
continuation rates above unity, and in wide fluc­
tuations in adjacent values of the continuation
rates.
No single mathematical curve appeared to fit all
of the intervals sim ultaneously. On the other hand,
reasonable results were obtained by fitting curves
to two or three adjacent intervals at a time. A
logarithmic normal curve was fitted to the data for
the two bottom intervals and used to estim ate the
number with unemployment of 1, 2, 3, or 4 weeks.
The top two intervals were fitted with an expo­
nential function that was used to estim ate the
number of people with 27 weeks of unemployment.
Treating the bottom three intervals together, we

un—un+i for all values of n.
Consequently, the average spell duration is the
sum of the number of spells weighted by the length
of spell divided by the total number of spells.
Average duration = X

j

71 = 1

Ul

Expand the numerator in term by term d etail:
S

n(un—un+1) = u , —u2+ 2 ( u 2—u3)
+ 3 ( u3—u4)-f- . . .
= u ! + u 2-t-u3+

. . . = u,

Thus the average spell duration=u-f-U!.
T est this method on the two sample duration
distributions we have examined in tables A - l and
A -2. For the first one, u = 1 6 , and ^ = 2 , so the
average duration of completed spells is 16/2 = 8
weeks. For the second one, u = 1 5 , and u, = 5, so
the average duration of completed spells is
15/5 = 3 weeks. In the special case of a distribution
with a constant continuation rate (r), the average
duration of completed spell is 1/(1 —
r) which is also
the average duration through the end of the refer­
ence week of the cross-section distribution.
Applications to data

Before we can apply the findings developed so
far to actual household survey data, we must
emphasize one particular point: some of these
results apply most accurately to steady state
distributions. In the real world, as unemploy­
ment grows or declines or changes in distribution
by duration, these properties may hold less pre­
cisely. For example, 4 weeks after there are heavy
layoffs, we m ay find that the number of people
with 4 weeks of unemployment is greater than
the number with 3 weeks in a given cross-section
distribution. In a steady state distribution this
result is impossible. On the other hand, in the
corresponding longitudinal distribution by dura­
tion after the layoffs, the number with 4 weeks
of unemployment will be no greater than the
number in the preceding week with 3 weeks of
unemployment. The cross-section and longitu­
dinal distributions may thus be temporarily differ­
ent from each other. Similarly, the completed
spell distribution for those starting new spells




44

Table A-4. Continuation rates applicable to patterns in
tables A -l and A-2

fitted a mixed exponential (a weighted sum of
two exponentials) to estimate the number of per­
sons unemployed 5 weeks, 11 weeks, and 15 weeks,
respectively.
At this point two additional general statem ents
must be made to help us in our computations and
in our understanding of the results.
I n a steady state distribution, the number oj spells
involving from m to n weeks of unemployment is
equal to the number oj unemployed persons in the
cross-section distribution with m weeks oj un­
employment m in u s the number with n - f 1 weeks.
The reasoning here is similar to that employed
before in deriving the simple estimate for the
average duration of completed spells. As already
indicated, the number of people in the crosssection who complete spells of exactly n weeks
in length is u n — u n+i, where u„ is the number
of people with n elapsed weeks of unemployment
through the reference week. The number of
persons with completed spells of from m weeks
through n weeks is therefore (u„ — um+i) +
( U m+ i —

U m +2)

4*

• • •

4"

(*l n

U n + l)— U m

Continuation rates
Duration of unemployment to end of
reference week (in weeks)

1__________________ ___________
2_____ ____ _____________ ______
3________ _____ ________________
4______________________________
5____________________ __________
6______________________________
7..................... ........................................
8______________________________

Table A-2

1.00
1.00
1.00
1.00
1.00
1.00
1.00
0

0.80
0. 75
0. 67
0. 50
0

employment. Possible explanations for this pattern
have already been given. In accord with the
general rule just stated, we expect in this situa­
tion that the completed spell distribution would
have a smaller average duration than the crosssection distribution.
Since we want to derive the distribution of
spells within the same intervals as for the crosssection distribution, we shall need estimates of
the numbers of people in the cross section with
the following specified weeks of unemployment,
in accord with the estim ating procedure de­
scribed earlier: iq, u5, un, Ui5, and u27.
The derived estim ates are generally satisfac­
tory, although they are undoubtedly susceptible
to improvement. Other mathematical functions
might also be considered in the curve-fitting
process. The estimation of the number of people
with 1 week of unemployment was particularly
critical and difficult. The logarithmic normal fit
gave results which were closest to the actual data
in the bottom interval for individual weeks and
gave acceptable estim ates for the continuation
rates. Nevertheless, the estimated total number
of spells in each of the years may be subject to
an estimating error of 5 percent or more. This
margin of error also will be present in the duration
of average spell because of their direct connection.
Estim ates for the rest of the duration distribu­
tions should be reasonably good otherwise.
An excellent single reference is D . J. Bartholo­
mew, Stochastic Processes in the Social Sciences
(New York, John W iley and Sons, 1967), which
also has an extensive bibliography. Two other
references are R. F. Fowder, Duration oj Unemployment on the Register oj Wholly Unemployed
(London, H. M. Stationery Office, 1968), and
B. Craig, Development oj M anpower Statistics
(Paris, Organization of Economic Cooperation
and Developm ent, 1969).
Q

u n + l-

I n a steady state distribution, i j the duration
pattern has an increasing (decreasing) set oj con­
tinuation rates associated with increasing weeks oj
duration, then the average duration oj completed
spell will be less than (more than) the average dura­
tion in the cross-section duration distribution.
In other words, if the likelihood of a person
continuing to be unemployed increases the longer
he is unemployed, then the average duration of
completed spells will be less than the average
duration found during the reference week. The
opposite is true when he is less likely to continue
to be unemployed the longer he has been un­
employed.
Increasing continuation rates will yield rel­
atively higher frequencies of spells at the lower
end of the distribution than for a distribution
with constant continuation rates, while a distri­
bution with constant continuation rates will have
relatively higher frequencies at the lower end than
a distribution with declining continuation rates.
In each of these two comparisons, the distribution
with the greater weight at the lower end will
clearly have a lower average duration than the
other. Since the constant continuation rate distri­
bution has the same average for completed spells
as for the cross-section, the results follow.
Table 4 of the article showrs that continuation
rates generally increase with the length of un-




Table A-l

45

Black and white
unemployment: the
dynamics of
the differential
h e w id e l y a c c e p t e d v ie w
in the economic litera­
ture is that the unemployment situation of black
workers improves relative to that of whites when
the demand for labor is strong and deteriorates visa-vis whites when the demand for labor slackens.
Yet, observed changes in the ratio of black-to-white
unemployment rates— roughly 2-to-l throughout
most of the post-World War II period— suggest that
the unemployment situation of blacks improves com­
pared with that of whites when the demand for labor
slackens and deteriorates when the demand for labor
is strong. Why is it that changes in the ratio of blackto-white unemployment rates appear to run counter
to the generally accepted view? Can the apparent
contradiction be reconciled?
In attempting to analyze the disproportionate share
of unemployment experienced by black workers 1 and
to compare changes in unemployment among blacks
and whites, most analysts use the ratio of black-towhite unemployment rates which will be called the
relative unemployment differential,2 following the
terminology used in the literature. Others have fo­
cused on the difference between the rates which is
called the absolute unemployment differential. This
article introduces another measure— ratio of the per­
centage-point changes in the unemployment rates of
blacks and whites— which is termed the incremental
ratio. Because no single statistical measure can be
expected to tell the whole story, the incremental ratio
can be used in concert with existing measures in de­
scribing the relative incidence of black and white
unemployment. Its main contribution is to enable the
analyst to describe more accurately the changing

T

Curtis L. Gilroy is a labor economist in the Division of
Employment and Unemployment Analysis, Bureau of Labor
Statistics. Roberta V. McKay, an economist in the same
division, assisted in the preparation of this article.

the Review of February 1974
DigitizedFrom
for FRASER


46

New analysis adds support to beliefs
that blacks fare better
in prosperity than in other
phases of business cycle
CURTIS L. GILROY

unemployment burden of the two groups during the
business cycle.3
No attempt will be made in this paper to describe
the employment situation of black workers, which is
well documented;4 nor will an attempt be made to
investigate the causes of excessive black unemploy­
ment.'' Rather, this paper, while emphasizing the
usefulness of the two popular measures of black and
white unemployment differences and exploring some
inconsistencies between and limitations in them, seeks
to demonstrate that the incremental ratio is prefer­
able in comparisons of changes in the incidence of
black and white joblessness over the business cycle.
In particular, this study will use the incremental ratio
to support the hypothesis that blacks are affected
more (proportionate to their numbers) than whites
by changes in the demand for labor.
The measures

In comparing unemployment of black and white
workers, it is concluded that blacks experience a
disproportionate share of unemployment if the rela­
tive unemployment differential (B /W ) is greater
than 1.00 and the absolute differential between the
rates (B -W ) is positive.
Neither of these popular measures is completely
satisfactory by itself, however, in measuring changes
in the incidence of unemployment. A comparison of
the relative differential at two points in time can
easily create a false impression if it measures changes
in unemployment rates with widely divergent bases.
On the other hand, a comparison of the absolute dif­
ferentials in rates does not describe the relative
change in unemployment attributable to each group.
The problems in both measures may be demonstrated
as follows: Consider, for example, a situation in
which blacks and whites have initial unemployment
rates of 8.0 and 4.0 percent, respectively. If both

It does not show the change in relative unemployment
rates but rather depicts the absolute change in unem­
ployment rates and expresses it in relative terms; that
is, (Bt2 — B tl) / (W .2 — Wtj) m A B /A W . The
prime advantage of the incremental ratio is that it
takes into account the widely different bases from
which were measured the changes in the rates in
percent terms. For example, it shows that during the
1970-71 recession, proportionate to their share of
the labor force, 14 persons were added to the already
high unemployment rolls of blacks for every 10 that
were added to the unemployment of whites—
A B / AW = (6.3 - 9.2) / (3.3 - 5.4) = 1.4.
(See table 1.)
Let us translate this incremental ratio into numbers
of unemployed people. Assuming that the sizes of the
black and white labor forces were 1,000 and 10,000
respectively (blacks make up, in fact, about 10 per­
cent of the labor force), the number of unemployed
blacks would increase from 63 to 92, whereas the
number of jobless whites would rise from 330 to 540.
Thus, on a per-thousand basis, the unemployment
rolls of blacks increased by 29 workers over the last
recession; that for whites by only 21 workers.
By contrast, a comparison of the relative unem­
ployment differential over that period shows an im­
provement in the black unemployment situation as
the black-white ratio declined from 1.91 at the peak
of the business cycle to 1.70 at its trough. This
occurred because the percentage increase in the un­
employment rate for whites exceeded that for blacks.
The decline in the ratio is misleading, however, be­
cause it does not account for the greater increase,
proportionately, in unemployment among blacks than
whites. Blacks are then worse off relative to whites
even though the relative differential decreased.
If, during the prosperity phase of the cycle, black
unemployment fell from 9.2 to 6.3 percent and that
for whites dropped from 5.4 to 3.3 percent, the rela­
tive differential would rise. Looking solely at changes
in the relative differential, one could argue that the
economy should be in a continuing state of recession
to allow the relative differential to fall to the ideal of
1.00. On the other hand, the incremental ratio of 1.4
would indicate that blacks experienced a greater
decrease in unemployment (proportionate to the size
of their labor force) than their white counterparts
during the cyclical upswing.
The apparent inconsistency in the measures dis­
appears, however, when the incremental ratio takes
on a value greater than the initial value of the rela-

rates decrease by the same absolute amount, say 2
percentage points, the relative unemployment differ­
ential would rise from 2.0 to 3.0, while the absolute
differential in the rates would remain at 4.0 percent­
age points. If both the white and black rates were
halved (whites from 4.0 to 2.0 and blacks from 8.0
to 4.0) the relative differential would remain un­
changed at 2.0 and the absolute differential would
decline from 4.0 to 2.0 percentage points. Finally, if
the white rate were reduced from 4 to 1 percent and
the black rate from 8 to 4 percent, the relative dif­
ferential would increase from 2.0 to 4.0 while the
absolute differential would fall from 4.0 to 3.0 per­
centage points. The policy implications emanating
from this last example are somewhat unclear. Short
of out-and-out equality of black and white unemploy­
ment rates, there is a question as to whether blacks
are better off, in that they experienced a greater
decline in their rate vis-a-vis whites, or worse off, in
that their relative position deteriorated.
The question therefore remains: Which measure is
the more appropriate? Many analysts, in relying on
the relative differential, have pointed to a relative
improvement in the black employment situation
whenever the differential decreased. Other writers
argue that the situation may not have improved at
all and insist that what is important is not so much
that blacks have higher unemployment rates but
whether an increase affects them more adversely than
whites, say from the peak of a business cycle to its
trough.6
A comparison of the relative differential in rates
in two distinct periods— denoted by (B /W )ti and
(B /W )t2— is useful but its usefulness is limited to a
comparison of those periods when economic condi­
tions are similar. It is informative, for example, as a
measure of long-term changes between two periods
with similar overall unemployment rates. The rela­
tive differential in rates is most appropriate in meas­
uring relative unemployment burdens of blacks and
whites at a point in time. The absolute differential
between rates— (Btj — Wtl) or (Bt2 — Wt2) — im­
plies the correct unemployment relationship between
time periods, but it does not go far enough. To gauge
more precise relative changes in unemployment an­
other measure is desirable— the incremental ratio.
The incremental ratio

The incremental ratio incorporates the favorable
aspects of the relative and the absolute differentials.




47

five differential. For example, during the expansion­
ary phase of the 1960’s, black unemployment fell
from 12.4 to 6.3 percent while white joblessness de­
clined from 6.1 to 3.3 percent (table 1). The incre­
mental ratio was 2.2, and the relative differential
declined from 2.03 to 1.91. In this case the value
of the incremental ratio (2.2) exceeded that of the
initial relative differential (2 .0 3 ). To clarify its
meaning, then, if the incremental ratio exceeds 1.0,
the black unemployment rate is changing more than
the white rate; to produce a decline in the relative
differential over time, its value must exceed that of
the initial value of the black unemployment rate
divided by the white rate (B-f-W typically must be
greater than 2 .0 ).
Table 1.

Three limitations of the incremental ratio, how­
ever, are important to consider. First, the ratio is not
relevant to measuring unemployment differences at
one point in time. The relative and absolute differ­
entials in rates are most useful here. Second, the
incremental ratio is inapplicable over very short pe­
riods of time when any change in a group’s unem­
ployment rate (for example, A B /A W = (8-7) /
(4-4) = 1 /0 = undefined) is unlikely. Third, using
the incremental ratio to measure changes in unem­
ployment over long periods of time also may be
misleading since it would tend to hide the short-run
ups and downs in the economy for which it is a most
useful measure.
Despite the black-white unemployment rate ratio

Peak-to-trough and trough-to-peak changes in unemployment rates by color, by age and sex, 1954-70
Peak

Trough

Age and sex
July

April

1957

1958

Total, 16 years and over:
White.................................................................
Black.................................................................
Incremental ratio......................................

3.7
7.9

6.4
13.4

Both sexes, 16-19 years:
White................... .....................................
Black.........................................................
Incremental ratio..............................

10.5
19.8

Men, 20 years and over:
White________ - .....................................
Black....................... ................................
Incremental ratio..............................
Women, 20 years and over:
White.........................................................
Black.........................................................
Incremental ratio..............................

Trough

May

February

1960

1961

2.7
5.5
2.0

4.7
10.0

6.1
12.4

14.8
25.4

4.3
5.6
1.3

12.9
24.8

3.1
7.0

5.9
13.8

2.8
6.8
2.4

3.7
6.8

5.9
10.5

2.2
3.7
1.7

Trough

Peak

August

July

1954

1957

Total, 16 years and over:
White................................................................
Black................................................................
Incremental ratio......................................

5.6
10.3

3.7
7.9

Both sexes, 16-19 years:
White.........................................................
Black.........................................................
Incremental ratio..............................

13.2
16.9

Men, 20 years and over:
White.........................................................
Black.........................................................
Incremental ratio..............................
Women, 20 years and over:
White.........................................................
Black.......... ..............................................
Incremental ratio..............................




Peak
Over-theperiod
change

Over-theperiod
change

Peak
Over-theperiod
change

Trough
Over-theperlod
change

November

November

1969

1970

1.4
2.4
1.7

3.3
6.3

5.4
9.2

2.1
2.9
1.4

15.6
30.7

2.7
5.9
2.2

10.6
23.6

15.4
32.3

4.8
8.7
1.8

3.9
9.1

5.3
11.6

1.4
2.5
1.8

2.1
3.7

4.0
6.5

1.9
2.8
1.5

4.3
8.3

5.7
10.2

1.4
1.9
1.4

3.4
5.5

5.1
7.8

1.7
2.3
1.4

Trough

Peak

Trough

Peak

February

February

1961

1969

Over-theperiod
change

Over-theperiod
change

April

May

1958

1960

- 1 .9
-2 .4
1.3

6.4
13.4

4.7
10.0

- 1 .7
-3 .4
2.0

6.1
12.4

3.3
6.3

- 2 .8
- 6 .1
2.2

10.5
19.8

- 2 .7
- 2 .9
1.1

14.8
25.4

12.9
24.8

- 1 .9
-.6
.3

15.6
30.7

10.6
23.6

-5 .0
- 7 .1
1.4

5.0
10.2

3.1
7.0

- 1 ,9
- 3 .2
1.7

5.9
13.8

3.9
9.1

-2 .0
- 4 .7
2.4

5.3
11.6

2.1
3.7

- 3 .2
- 7 .9
2.5

5.2
9.2

3.7
6.8

- 1 .5
- 2 .4
1.6

5.9
10.5

4.3
8.3

- 1 .6
- 2 .2
1.4

5.7
10.2

3.4
5.5

- 2 .3
- 4 .7
2.0

48

falling during the trough-to-peak period in the 1960’s,
the ratio characteristically rises in times of prosperity
and falls in recessionary periods. This is evident from
chart 1, which traces the ratio of the major age-sex
groups over several recent business cycles. By con­
sidering only the ratio, we would be led to believe
that blacks are generally better off relative to whites
in cyclical downturns than they are when economic
activity is strong. However, this is not the case.
Blacks become worse off than their white counter­
parts in recessions, as the incremental ratio shows.
This is supported by James Tobin, who has presented
the conventional view, in his statement:
People who stand at the end of the hiring line and
the top of the lay-off list have the most to gain
from a tight labor market. It is not surprising that
the position of Negroes relative to that of whites
improves in a tight labor market and declines in a
slack market.7
Chart 1.

Peak to trough; trough to peak

Characteristic of recessions is a slackening in the
demand for labor; a feature of every recovery is an
increase in the demand for labor services. Black
workers have experienced a greater absolute increase
in their unemployment rate than white workers in
almost all cyclical downturns (peak-to-trough) and
a greater absolute decrease in their unemployment
rate than whites in the recovery periods (trough-topeak). (See chart 2.) This differing cyclical unem­
ployment experience of blacks relative to whites is
revealed by the incremental unemployment rate
ratio, which is consistently greater than 1.0.
The peak-to-trough and trough-to-peak changes in
the seasonally adjusted unemployment rates for white
and black workers are shown in tables 1 and 2 for
selected age, occupational, and industry groups. The

Black-white unemployment rate ratio at peaks and troughs of business cycles, 1954-73

NOTE: Unemployment rates for the peaks and troughs were obtained
by averaging the three seasonally adjusted monthly observations centered




at each turning point of the cycle,

49

unemployment rates for the peaks and troughs were
obtained by averaging the three monthly observations
made at each turning point of the cycle.8 Constrained
by the availability of unemployment data by color for
the various age groups, our observations of three
peak-to-trough and trough-to-peak movements occur
during 1954-73; for occupations and industries, ob­
servations covering two peak-to-trough and one
trough-to-peak movements during 1959-73 are pre­
sented.9
The phenomenon of blacks being more adversely
affected than whites by business downturns has been
mitigated over the last several business cycles. For
every successive peak-to-trough period, black work­
ers have shared less of the increase in unemployment.
In the 1957-58 downturn, for example, proportionate
to the size of their labor force, 20 black workers were
added to the unemployment totals for every 10

white workers; however, during the 1969-70 reces­
sion, only 14 blacks became jobless for every 10
white workers. (See table 1.) Moreover, proportion­
ately more blacks than whites have left the ranks of
the unemployed with each successive recovery pe­
riod. During the 1954—57 recovery, 13 black workers
for every 10 white workers left the unemployed
ranks, while in the 1961-69 period of prosperity,
there was a decrease of 22 unemployed black work­
ers for every 10 white workers.
Among the age-sex groups, adult black men have
borne the brunt of the increase in unemployment in
the cyclical downturn but have also experienced pro­
portionately more of the decline in unemployment
when economic activity picked up. While adult black
men and women shared about equally in the increase
in unemployment in the most recent downturns— the
incremental ratios are similar in size— black women

Chart 2. White and black unemployment rates, all workers, 16 years and over, 1954-73
[Seasonally adjusted quarterly averages]

1954

1956




1958

1960

1962

1964

50

1966

1968

1970

1972 1973

do not benefit as much as black men in the upswing.
In most of the major occupational groups, the
ratio was smaller in the most recent recession than
in the 1960-61 downturn. (See table 2.) This indi­
cates that black workers suffered a smaller increase
in joblessness than white workers during the 1969-70
slowdown. This occurred particularly among clerical
and sales workers, operatives, and service workers—

those occupations in which two-thirds of the black
labor force are employed.
An example of the particular usefulness of the in­
cremental ratio arises from examining the experience
of the professional and managerial occupational
group. The negative sign of the incremental ratio for
this group indicates that the rates for blacks and
whites moved in opposite directions. In this case, not

Table 2. Peak-to-trough and trough-to-peak changes in unemployment rates by color, by selected occupation and indus­
try groups, 1959-70
Peak

Trough

Occupation and industry
May
1960

February
1961

Professional, technical, and managerial:
White_______________________________
Black________________________________
Incremental ratio......................................

1.5
2.8

2.0
3.6

Clercial and sales workers:
White................................................... ............
Black........ ........................................................
Incremental ratio..................... ....... .........

3.6
7.5

4.6
9.8

Craft and kindred workers:
White.......... - ....................................................
Black.................................................................
Incremental ratio......................................

4.3
10.4

9.1
15.3

Operatives:
White.................................................................
Black.................................................................
Incremental ratio......................................

7.7
10.5

11.5
17.4

Nonfarm workers:
White.................................................................
Black...................................... .........................
Incremental ratio......................................

11.2
13.1

Service workers:
W hite..............................................................
Black.................................................................
Incremental ratio......................................

Peak
Over-theperiod
change

Trough
Over-theperiod
change

Trough

Peak

February
1961

November
1969

Over-theperiod
change

Novembw.»
1969

November
1970

0.5
.8
1.6

1.1
2.1

1.8
1.4

0.7
-.7
- 1 .0

2.0
3.6

1.1
2.1

- 0 .9
- 1 .5
1.7

1.0

2.7
5.6

4.3
8.4

1.6
2.8
1.8

4.6
9.8

2.7
5.6

- 1 .9
- 4 .2
2.2

2.0
3.2

4.1
5.5

2.1
2.3
1.1

9.1
15.3

2.0
3.2

- 7 .1
-1 2 .2
1.7

3.8
6.9
1.8

4.4
5.2

7.7
9.3

3.3
4.1
1.2

11.5
17.4

4.4
5.2

- 7 .1
12.2
1.7

19.8
20.6

8.6
7.5
.9

7.0
6.9

10.2
11.7

3.2
4.8
1.5

19.8
20.6

7.0
6.9

-1 2 .8
-1 3 .7
1.1

5.5
9.4

6.8
11.8

1.3
2.4
1.8

3.4
6.3

5.5
7.9

2.1
1.6
.8

6.8
11.8

3.4
6.3

- 3 .4
- 5 .5
1.6

Construction:
White.................................................................
Black.................................................................
Incremental ratio......................................

10.9
17.2

24.0
30.1

13.1
12.9

5.4
7.1

9.2
14.1

3.8
7.0
1.8

24.0
30.1

5.4
7.1

-1 8 .6
-2 3 .0
.7

Manufacturing:
White.................................................................
Black................................................................
Incremental ratio......................................

5.4
11.3

8.7
19.2

3.3
7.9
2.4

3.3
5.1

6.3
10.2

3.0
5.1
1.7

8.7
19.2

3.3
5.1

—5.4
-1 4 .1
2.6

Trade:
White.................................................................
Black.................................................................
Incremental ratio......................................

5.3
12.1

7.3
14.3

2.0
2.2
1.1

3.3
7.0

5.3
9.1

2.0
2.1
1.1

7.3
14.3

3.3
7.0

-4 .0
- 7 .3
1.8

Services:
White.................................................................
Black.................................................................
Incremental ratio......................................

3.8
9.1

5.7
12.8

1.9
3.7
1.9

2.9
5.9

5.0
7.6

2.1
1.7
.8

5.7
12.8

2.9
5.9

- 2 .8
- 6 .9
2.5

Government:
White................................................................
Black.................................................................
Incremental ratio......................................

1.9
3.6

2.2
5.6

.3
2.0
6.7

1.5
3.9

1.9
5.2

.4
1.3
3.3

2.2
5.6

1.5
3.9

-.7
- 1 .7

Occupation

2.3
2.3
4.8
4.9

1.0

Industry




1.0

51

2.4

only did blacks experience a lower incidence of un­
employment relative to whites around the bottom of
the cycle, but their incidence of joblessness actually
decreased in the recession while that for whites rose.
This was due in large part to the increase in overall
unemployment during the last recession being rela­
tively more concentrated in the professional and tech­
nical occupations; for example, in aerospace, elec­
tronics, and other defense-related industries in which
only a small proportion of black workers were em­
ployed. Throughout the period of recovery, the
incremental ratios show that blacks, proportionate to
the size of their labor force, experienced a greater
decrease in unemployment than white workers. In
each of the occupational groups, the incremental
ratio was in excess of 1.0.
Patterns similar to the occupational ones exist for
the major industry groups from peak to trough and
the trough to peak. Although the incremental ratio
was greater than 1.0, substantial decreases in the
ratios occurred within manufacturing, service, and
government over the last two recessions. In the re­
covery period, black unemployment decreased at a
faster rate than white with the exception of the con­
struction industry, where black workers became un­
employed in greater relative numbers than white
workers and have been less likely than whites to be
rehired.
Black-white differences over time

The degree to which changes in overall economic
conditions affect the unemployment rates of various
labor market subgroups has been the concern of
several analysts.10 Few, however, have attempted to
systematically measure the impact on the various
demographic subgroups by color.11 This section will
show the impact of the business cycle on the unem­
ployment rate of black and white workers and will
measure the extent to which the jobless rates of
blacks have been affected proportionately more than
those for whites over time. Thirty regressions were
run by the selected age-sex, occupational, and indus­
try subgroups by color. The dependent variable (Y )
was the subgroup unemployment rate by color, and
the independent variables— (X i and X 2) — the un­
employment rate of males 35-44 years old, and time,
respectively.12 By using the jobless rate of men 35-44
years old as a proxy for the changing level of eco­
nomic activity, the regression equations will permit
the estimation of what the unemployment rates of




52

various subgroups would be as economic conditions
change. The labor market becomes “loose” or “tight”
roughly coincident with changes in aggregate demand
defining the various phases of the business cycle.
Quarterly data from the Current Population Survey
were used covering the years 1954—73 for the age-sex
groups and 1959-73 for occupations and industries.
The results of the regressions appear in table 3.
All the equations show a high degree of statis­
tical significance and demonstrate a close and posi­
tive relationship between the surrogate measure for
economic change (henceforth referred to as the
“prime age unemployment rate” ) and the incidence
of unemployment among the various labor market
subgroups. The extent to which a given change in
the prime rate would affect the subgroup rate is
substantially greater for blacks than for whites in all
equations. For example, a change of one percentage
point in the prime rate (X x) results in a change in
the same direction of 1.05 percentage points for
adult white males but an increase of 2.26 percent­
age points in that for adult black males. Among
both blacks and whites, the coefficients are larger—
indicating greater movement in rates— for bluecollar than for white-collar workers, since the for­
mer group, which has proportionately more semi­
skilled and unskilled labor, is considerably more
affected by changing business conditions.13 More­
over, the coefficients of the goods-producing relative
to the service-producing industries are larger, due
to the faster growth of the latter and the fact that
they are cyclically less sensitive.
A time variable was included in the regression
equations to show the extent to which a trend could
be discerned in the various unemployment rate
series. Although care must be taken in interpreting
the effect of the time variable because it tends to
include all factors varying with time, the results of
the regressions do show a worsening in the overall
employment situation for both blacks and whites.
The sign of the coefficient (b2) of the time variable
(X 2) is positive, which indicates that over the period
1954-73, the subgroup unemployment rates have
trended upward. A significant exception was that for
adult black men, whose rate has trended downward
secularly. The upward trend was greatest in teenage
unemployment, both among blacks and whites; the
upward trend among teenage blacks was significantly
greater than for all other groups and confirms the
fact that much of the black unemployment problem
is a youth problem as well; that is, a greater num-

Table 3. Regression results showing relationships between unemployment rates of various labor market subgroups,
prime male unemployment rate, and time
Dependent
variable

Color

«)

(bi)

(bj

(r,)

(s)

D-W

Age-sex-color

Dependent
variable

Color

(a)

(bi)

(bj)

(r»)

Operatives...........

White

-.0 0 7
(.01)
.423
(.40)
2.255
(1.23)
4.287
(2.54)

1.881
(12.15)
2.937
(11.46)
2.466
(5.60)
3.014
(7.42)

.033
(3.67)
.025
(1.66)
.017
(.69)
-0 .2 5
(1.06)

.77

.85

1.91

.77

1.41

2.24

1.803
(5.04)
5.215
(7.80)

.941
(10.94)
1.317
(8.19)

6.445
(4.52)
.619
(.60)
3.038
(1.01)
10.191
(1.81)

1954(1)—1973(1)
Black
Both sexes, 16
years and over.. White
Black
Men, 20 years and
over................... White
Black
Women, 20 years
and over........ .

White
Black

Both sexes, 16-19
years.................. White
Black

.680
(4.96)
2.957
(8.68)

1.024
(32.09)
1.852
(23.38)

.019
(12.84)
.019
(5.31)

.043
(.34)
1.657
(4.48)

1.051
(35.84)
2.256
(26.25)

.007
(5.48)
-.0 1 8
(4.71)

1.339
(8.58)
3.643
(8.37)

.814
(22.42)
1.240
(12.26)

.016
(9.58)
.018
(3.83)

4.941
(8.52)
7.603
(5.40)

1.817
(13.48)
2.875
(8.79)

.062
(10.20)
.233
(15.71)

.93

.24

1.11

Nonfarm laborers White

.89

.60

.85

Black

.95

.22

1.25

Service workers,
excluding private
household_____ White

.94

.65

1.15

.87

.27

1.00

.68

.77

.74

.72

1.02

.75

.77

2.48

1.20

Black

Manufacturing... White
Black'
Construction____

White
Black

Wholesale and
retail trade___

Black
Clerical and sales
workers............

White
Black

Craftsmen and
kindred workers. White
Black

-.3 7 1
(2.52)
.864
(1.63)

.456
(12.85)
.603
(4.72)

.021
(10.43)
.008
(1.11)

.560
(2.55)
1.614
(1.60)

.774
(13.77)
1.470
(6.16)

.026
(8.04)
.048
(3.43)

-.2 3 5
(.28)
1.249
(1.00)

1.383
(6.85)
2.285
(7.63)

.010
(.86)
-.0 2 7
(1.59)

.67

2.23

2.22

.46

2.42

2.05

.019
(3.91)
-.0 1 0
(1.12)

.72

.47

1.90

.71

.88

1.47

2.482
(7.23)
3.081
(12.40)
2.637
(3.65)
4.955
(3.66)

-.1 1 8
(5.95)
.006
(.43)
-.0 1 0
(.24)
-.2 1 9
(2.81)

.82

1.89

.99

1.086
(2.40)
3.977
(3.45)

1.116
(10.24)
2.210
(7.96)

.026
(4.22)
.029
(1.84)

1.003
(3.21)
3.105
(3.56)
.379
(1.74)
1.010
(1.73)

.894
(11.89)
1.726
(8.22)
.388
(7.40)
.726
(5.15)

.025
(5.88)
.021
(1.74)
.015
(5.02)
.014
(5.11)

Industry-color

1959(1)—1973(1)

White

D-W

1959(0-1973(1)

Occupation-color
Professional,
technical, and
managerial____

(s)

.76

.20

2.15

.35

.70

1.60

.78

.31

2.18

.42

1.34

1.97

White
Black

Services, excluding private
household____

White
Black

Government......... White
.56

1.11

2.07

.70

1.65

2.04

Black

.82

1.37

2.13

.31

3.98

2.10

.52

7.44

1.44

.68

.60

2.19

.60

1.53

1.57

.73

.41

1.47

.62

1.16

1.90

.50

.29

2.19

.37

.77

i.60

NOTE: t values are in parentheses.

many women and young workers who characteris­
tically have higher unemployment rates.
The coefficient of determination (r2) shows the
extent to which the regression equations explain the
variations in the unemployment rate of the sub­
groups. In general, a greater amount of the variation
in the white than of the black subgroup unemploy­
ment rates could be explained by the regression
equation.14 The greater unexplained portion of the
variation in black unemployment (in addition to
greater sampling variability) is undoubtedly due to
such structural factors as educational deficiencies,
growth rates of various labor force subgroups, re­
gional distribution, employment discrimination, and
the like. The standard error of the estim ate(s),
which is a measure of how much, on the average,
the actual unemployment rate deviates from that
calculated from the regression equation is larger
for the black regressions than those for the white.

ber of teenagers entering the labor market relative
to job opportunities available to them.
During 1959-73, the trend in unemployment
among the various occupational groups has been
generally upward. Those downward trends that oc­
curred for blacks in the blue-collar and service
occupations, were statistically insignificant. The sig­
nificant upward 'rends were predominant in the
white-collar professions, which may partly be ex­
plained by the difficulty in securing employment for
a more highly educated and trained labor force at a
time when there were severe cutbacks— as in 197071— in the aerospace, electronics, and other defenserelated industries.
Unemployment appears to have become more
concentrated among the service-producing indus­
tries for both blacks and whites and less prevalent
in the goods-producing sector. This may be due in
part to the service-producing sector’s having attracted



53

The proposition that blacks are affected relatively
more than whites by changes in the demand for
labor is further substantiated by a comparison of
the coefficients (b x) of the prime male unemploy­
ment rate (X j). Expressing these coefficients as
the ratio Bbi/W bi (where B and W denote the
black and white coefficients), the relative change
in black and white unemployment rates can be
seen. This is clearly the incremental ratio. Since
Bbi and Wbj are consistently positive, and Bbi >
Wbj, the ratio is positive and greater than 1.00.
(See table 4.) This means that over the long period,
whenever business conditions have deteriorated,
black unemployment has risen proportionately more
than white; however, when conditions improved,

blacks have left the ranks of the unemployed in
greater relative proportions.
W e m a y c o n c l u d e from our analysis that in re­

covery and downturn, blacks are affected rela­
tively more than whites by changes in the demand
for labor. The differing cyclical unemployment ex­
perience of blacks relative to whites is more ac­
curately portrayed by a measure introduced in this
article— the incremental ratio. Utilization of this
ratio of the difference between the unemployment
rates of blacks and whites (A B /A W ) reconciles
the inconsistency in patterns which emerges when
comparing black and white unemployment rates at
two points in time using the traditional measures—
the absolute unemployment differential (B-W ) and
the more popular relative unemployment differential
(B /W ).
A primary advantage of the incremental ratio is
that it takes into account the widely different bases
from which the changes in unemployment are meas­
ured. Consequently, use of the incremental ratio is
desirable because it permits a fuller understanding
of the dynamics of black and white unemployment
over the business cycle. For example, a narrowing
of the traditional ratio, which generally occurs dur­
ing recessions, has created a misleading impression
that blacks were less affected than whites by in­
creases in unemployment. The incremental ratio
shows, in fact, that black unemployment has risen
proportionately more than that for whites during
such periods; this finding was buttressed in this
article by regression and variance analysis. Likewise,
a widening of the differential is characteristic of
recovery periods. This is also misleading because it
implies that blacks become worse off than whites
when the economy expands. The incremental ratio,
however, shows that black unemployment has de­
creased proportionately more than white unemploy­
ment during such periods.
□

Table 4. A comparison of black and white coefficients
of the independent variable Xi (male prime unemployment
rate)

Subgroup

Black
coefficient
Black
White
-!-White
coefficient coefficient coefficient
(Wb,)
(Bb,)
(Bbl/Wbl)

Age-sex
Both sexes, 16 years and over_____________
Men, 20 years and over.................. ..................
Women, 20 years and over________________
Both sexes, 16-19 years______________ _ _

1.852
2.256
1.240
2.875

1.024
1.051
.814
1.817

1.809
2.147
1.523
1.582

.603
1.470
2.285
2.937
3.014
1.317

.456
.744
1.383
1.881
2.466
.941

1.322
1.899
1.652
1.561

4.955
3.081

2.637
2.482
1.116
.894
.388

2.093
1.241
1.980
1.931
1.871

Occupation
Professional, technical, and managerial..........
Clerical and sales workers________________
Craftsmen and kindred workers___________
O peratives...____ ___ ______ _____
Nonfarm laborers...______ ________ _____
Service workers, excluding private household.

1.222
1.400

Industry
Construction ______________ __________
Manufacturing_____________ __________
Wholesale and retail trade________________
Services, excluding private household______
Government_____ _______ ______________

2.210
1.726
.726

SOURCE: Table 3.

-FOOTNOTES1 Statistics for members of the black and other U.S. mi­
nority races— called “Negro and other races”— are used to
indicate the situation for black workers. Blacks constitute
89 percent of the larger group.

Investment in Human Capital and the Nonwhite-White U n­
employment Differential, unpublished Ph. D. dissertation,
State University of New York (Binghamton), 1973.
3 This procedure was suggested by Paul O. Flaim who
utilized it in an unpublished Bureau of Labor Statistics
analysis, “The Negro-White Unemployment Relationship,”
March 1970.
4 See Black Americans, a chartbook, Bulletin 1699 (Bu­
reau of Labor Statistics, 1971); Gloria P. Green, E m ploy­
ment in Perspective: The Negro Employment Situation, Re­

2 In a pioneering study, Harry Gilman examines the cycli­
cal variability of the relative incidence of black and white
unemployment in “The W hite/Non-W hite Unemployment
Differential,” in Mark Perlman, ed., Human Resources in
the Urban Economy (Washington, D.C., Resources for the
Future, Inc., 1963), pp. 75-113. See also Curtis L. Gilroy,




54

Econometric Study,” Review of Economics and Statistics,
May 1965, pp. 137-149.
11 Barbara R. Bergman and David E. Kaun looked at
detailed age groups by color in their Structural Unemploy­
ment in the United States (U.S. Department of Commerce,
Economic Development Administration, 1967) pp. 77-81.

port 391 (Bureau of Labor Statistics, 1971); and The Social
and Economic Status of the Black Population in the United
States, 1972, Current Population Reports, Series P -23, No.
46 (Bureau of the Census, 1973) and similar Census reports
in previous years.
c See, for example, Gary S. Becker, The Economics of
Discrimination (Chicago, 1967); Harry Gilman, “Economic
Discrimination and Unemployment,” American Economic
Review, December 1965, pp. 1077-1096; Ralph E. Smith
and Charles C. Holt, “A Job Search-Turnover Analysis of
the Black-White Unemployment Ratio,” in Industrial Rela­
tions Research Association, Proceedings of the Twenty-Third
Annual Winter Meeting, December 1970, pp. 76-86; Lester
Thurow, Poverty and Discrimination (Washington, The
Brookings Institution, 1969).

12 The equation took the form U = a -f- biXi + b2X 2
where U represents the subgroup unemployment rate; a, a
constant term; b; and b2, coefficients of the independent
variables Af, and X t\ and X t and X 2, the unemployment rate
of “prime” age males 35-44 years old and time respectively.
The use of the “prime” age male unemployment rate as a
cyclical indicator rather than the more popular aggregate
unemployment rate was to avoid having much the same
variable on both sides of the equation. See Bergman and
Kaun, p. 78.
13 See Malcolm S. Cohen and William H. Gruber, “Varia­
bility by Skill in Cyclical Unemployment,” Monthly Labor
Review, August 1970, pp. 8-11.
14 The cyclical variability of the various black and white
unemployment rate series can be seen by comparing the
standard deviations of the respective series. If blacks are
more affected than whites by changes in the demand for
labor, it will be reflected in a greater variability in their
unemployment rate series through a larger standard devia­
tion. The standard deviation, while reflecting the particular
cycle amplitudes, will also include additional variability.
Some of this variability is due to the fact that the black and
white unemployment rates have different size bases. When
the variances of two such series are to be compared, a meas­
ure of relative variance may be more useful because it ex­
presses the magnitude of the variation relative to the size of
the quantity that is being measured. If the absolute varia­
bility of the unemployment rate is assumed to depend in part
upon the average level of unemployment, then the standard
deviation as a percentage of the mean— coefficient of varia­
tion— is an appropriate measure. In all cases, the standard
deviations as well as the coefficients of variation were greater
for the black than for the white series.
Some of this variability will be due to sampling error.
For the white series, the Current Population Survey is large
enough to make the random sampling error variance small.
On the other hand, the unemployment rate series for black
workers rest on smaller samples for which the sampling
error variance is substantially larger. Harry Gilman found
that, even after adjusting for differing sample size among
blacks and whites, black workers still had a greater absolute
variability in their unemployment rates than whites in 64
percent of the intermediate occupational groups studied.
Those occupations within which the variability was similar
for both blacks and whites were the higher skilled profes­
sional jobs where blacks are relatively few in number. See
Gilman, “The W hite/Non-W hite Unemployment Differen­
tial,” pp. 90-92.

6 Harry Gilman, for example, has dismissed the relative
unemployment differential (B /W ) as a desirable index, in­
stead using the difference between white and black unem­
ployment rates (B -W ) at two points in time as the appro­
priate measure. See Harry J. Gilman, “The W hite/Nonwhite Unemployment Differential,” p. 92.
7 James Tobin, “Improving the Economic Status of the
Negro,” Daedulus, Fall 1965, p. 406.
8 The cycle turning point dates used are those defined by
the National Bureau of Economic Research. They are, peakto-trough: July 1957-April 1958; May 1960-February 1961;
and November 1969-November 1970.
Three-month averages were computed to smooth out in­
herent sampling variability (particularly among blacks be­
cause of the relatively small sample size from the Current
Population Survey) and to mitigate somewhat the discrep­
ancy which may occur between the NBER cycle turning
points and the turning points in unemployment.
9 The year 1959 was chosen because that is the year the
data first became available for occupations and industries by
color. The choice of 1959 as a starting date for this analysis
may be questioned, however, because it is viewed by some
as not a representative year in that it was sandwiched in
between two recessions. In addition, a major steel strike in
that year may have adversely affected the employment situa­
tion for blue-collar workers.
10 See, for example, Paul M. Ryscavage, “Impact of Higher
Unemployment of Major Labor Force Groups,” Monthly
Labor Review, March 1970, pp. 21-25; Robert A. McMil­
lan, “What Happens When the Unemployment Rate
Changes?,” Economic Review, Federal Reserve Bank of
Cleveland^ June-July 1972, pp. 3-16; Vladimir Stoickov,
“Increasing Structural Unemployment Re-examined,” Indus­
trial and Labor Relations Review, April 1966, pp. 368-376;
Comment by Arthur Butler, “Identifying Structural Unem­
ployment,” and Reply by Stoickov in Industrial and Labor
Relations Review, April 1967, pp. 441-446; and Lester
Thurow, “The Changing Structure of Unemployment: An




55

Quits
in manufacturing:
a study of
their causes

The rate of voluntary separations
is a good economic indicator;
the reasons for quitting are
changeable and derive from workers’
attitudes toward the economy
PAUL A. ARMKNECHT AND
JOHN F. EARLY

ments, since quit decisions are made by individual
workers.
Since the beginning of the series, the quit rate has
exhibited a median lead of 15 months at the business
cycle peak and a median lead of 1 month at the
trough. This long lead and the desirable statistical
properties of the series make it a good forecaster of
possible downturns in the economy.

L ab o r m o b il it y is the sine qua non for the efficient

allocation of labor factors in the production process.
The only reliable labor mobility data available on a
continuing and current basis are those reported by
the Bureau of Labor Statistics in its monthly series
on labor turnover in manufacturing, particularly the
rate of voluntary separations. This article undertakes
to lay a foundation for the use of the series in cur­
rent economic analysis and to discover the reasons
for variation in the quit rate over time and among
industries.

Time series regression

The highly cyclical nature of the quit rate has
already been noted, but the literature on the subject
has developed a controversy over the question of
whether the rate also has a trend.2 It has been
argued (1 ) that there is no trend in the rate, (2 )
that there is a decline in the rate because of non­
transferability of pensions and other fringe benefits
— the so-called industrial feudalism hypothesis, and
(3 ) that there has been a decline in the quit rate
because of endemic factors, such as the aging of the
work force. Our study supports the view that there
has been no trend.

The quit rate as a cyclical indicator

In the post-World War II era, the quit rate in
manufacturing has been a smooth, well-behaved se­
ries that has rather consistently led the business cycle
at its peak and coincided with it at the trough. (See
chart 1.) A test of its adequacy as an economic
indicator by means of the methods adopted by Geof­
frey H. Moore and Julius Shiskin1 placed it on a
par with the most commonly accepted indicators. Of a
possible summary score of 100, the quit rate scored
71, compared with 69 and 65, respectively, for the
layoffs and total accession rates. Tables 1 and 2
show the smoothness and small revisions in the quit
rate which are two of the important factors contrib­
uting to its quality as an indicator. These desirable
traits may arise, in part, from the fact that while the
BLS labor turnover survey is based on a sample of
approximately 38,000 establishments, the true size of
the sample underlying the quit rate estimate is the
10.4 million workers employed in these establish-

To determine whether there has been any measur­
able trend in the quit rate in the past two decades, a
number of time-series regression models were tested,
using both quarterly and annual data. Only the final
equations for the quarterly model will be presented
and discussed here. A more detailed description of
other hypotheses tested and of statistical difficulties
that had to be overcome will be found in a forthcom­
ing BLS staff paper. The following is the two-stage
least-squares estimate of the model which explains
the data best over time. All insignificant terms, in­
cluding the constant, have been removed.

Paul A. Armknecht and John F. Early are economists in the
Division of Industry Employment Statistics, Bureau of
Labor Statistics. An earlier version of this article was pre­
sented at the meeting of the American Statistical Association
in Montreal, Canada, on August 16, 1972. A more detailed
study will appear in a forthcoming BLS staff paper.

From the Review of November 1972



(1)

qt = .238 A(ht) + .405 D(ht) + .310 hUl
(.074)
(.064)
(.043)
R 2 = .763
c

56

Durbin-Watson = 2.18

Chart 1.

Manufacturing quit rate, seasonally adjusted, 1 9 4 7 -7 1

Rate

NOTE: Peaks (P) and troughs (T) refer to business cycle turning points determined by the National Bureau of Economic Research.

The standard errors are contained in parentheses,
and the following definitions apply:
qt

cally much more rigorous than that employed by
some who used a rather impressionistic mode of
analysis. Second, the new hire rate seems to be a
more appropriate measure of the cyclical swings in
job availability and security than were the variables
used by Pencavel to remove cyclical effects. The ab­
sence of a trend in the quit rate does not mean that
there have been no long term shifts in the patterns of
mobility. We will, in fact, show later that there have
been some rather dramatic shifts. But the absence of
a trend does suggest that, on the average, the manu­
facturing worker is no more or less mobile in seeking
new employment than he was in the years immedi­
ately following World War II.
It will be noted that for the current quarter the
effects of the new hire rate have been divided into
two parts— the increases, or “absorption,” and the
decreases, or “disabsorption.”3 There appears to be
a distinct asymmetry of behavior here. A decline in
hiring during the current quarter will depress the
propensity to quit by 70 percent more than a similar
expansion in hiring would have increased it. In short,
the manufacturing worker is very cautious and can
have his confidence shaken much more readily than
restored. Such behavior helps explain the difference

= the change in the quit rate in quarter t.

A (h t) = the positive change in the new hire rate
in quarter t, zero if the change was
negative.
D(ht) = the negative change in the new hire
rate in quarter t, zero if the change was
positive.
ht_, = the change in new hires in the quarter
previous to t.
The new hire rate explains the quit rate so well
probably because it is a measure of the jobs available
and of job security, and it seems quite likely that the
more jobs there are and the more secure a worker
feels the more inclined he will be to seek a better
paying job. As already indicated, the constant term
in this equation was not significant, which means that
there was no constant change in the quit rate for the
past two decades— that is, there was no trend. Our
model differs from the models used by those who
have found negative trends in the quit rate in at least
two important ways. First, our model was statisti­




57

Table 2.

in the leading behavior of the quit rate at business
cycle peaks and troughs.
One other hypothesis that we wanted to test was
whether there was an additional, forward-looking attitudinal factor in the determination of the quit rate.
It was our hypothesis that the workers’ decisions to
quit were based not only on recent hiring practices,
but also on their views of the future, which might
depart from past experience. We further hypothe­
sized that this future expectation about the condition
of the labor market would also be closely tied with
the workers’ consumption plans. As a result we ex­
pected two things: that the quit rate and the savings
rate should be positively correlated; and that, even
after the removal of current and past hiring effects,
there should remain an unexplained portion of varia­
tion in the quit rate that would correlate positively
with the growth of aggregate economic activity in the
following quarter.
Our first test found a significant positive correla­
tion between the savings ratio and the quit rate. The
second resulted in the following equation, where
G (Y t+i) is the rate of growth of the real Gross
National Product in the next quarter:
(2)

qt =

.1 6 0 A ( h t) +
(.077)

R*C — .781

Change

Average monthly change.............
Average monthly revision............
Percent revision to change_____

=

Labor turnover economic indicators, 195S-71
Quits

Layoffs

Acces­
sions

New
hires

Average percent change:
Original series____ ______ _______
Seasonal factors............... .................
Seasonally adjusted series...... ...........
Irregulars______________________
Trend-cycle____________________

18.61
18.02
3.87
3.41
1.86

15.20
12.87
8.09
6.94
2.62

16.59
16.61
4.18
3.72
1.16

19.23
19.24
4.27
3.48
2.14

Irregular/trend-cycle ratio____________

1.83

2.65

3.21

1.63

Number of months of cyclical dominance
(MCD)....... .......................................... .

2

3

4

2

NOTE:.These statistics for the layoff and accession rates differ slightly from those
published by the Bureau of the Census in Business Conditions Digest since seasonal
adjustment methods used by the bureaus differ.




0.5
0
9.0

0.3
.1
27.2

As noted above, changes in the quit rate over time
seem to be largely caused by changes in economic
factors as well as expectations about future changes.
But there still remain questions about the causes of
the variations in quit rate behavior among industries.
One should certainly expect low-paying indus­
tries to experience higher quit rates since their em­
ployees are most likely to find higher paying jobs and
have less to lose by quitting. Industries that are hir­
ing large numbers of new employees may experience
higher quit rates since workers will be less concerned
about job security. Highly seasonal industries may
offer lower job security, attract the casual worker,
and, as a result, show a higher proportion of quits.
In addition to the characteristics of the industry,
characteristics of the workers may also contribute to
quit behavior. Women, for instance, may either ex­
hibit a casual attachment to the labor force and thus
have low opportunity costs associated with high quit
propensities, or they may believe that they will face
discrimination in hiring and thus be reluctant to quit.
Production workers, who are generally affected more
than other workers by seasonal and cyclical changes
in the economy, may exhibit greater propensities to
quit since the nature of their work is marked by such
problems as work hazards, lack of opportunity for
promotion, poor supervision, and low wages, all of
which weigh more heavily in their evaluation of their
jobs. On the other hand, it may be true that the
lower education of the production worker may
impede his mobility by reducing his knowledge of the
market.

2.11

Measure

0.7
.1
9.0

Layoffs

Cross-section regression

Equation 2 preserves the essential characteristics of
equation 1. The G (Y t+I) term has a significant posi­
tive coefficient, indicating the presence of a forwardlooking attitude on the part of workers in their quit
decisions. The only difference between equations 1
and 2 is the spread between the absorption and
Table 1.

0.8
.1
11.9

Quits

disabsorption coefficients for new hires. This increase
further emphasizes the cautious nature of the Ameri­
can manufacturing worker. When future expectations
are indirectly entered into the equation, it becomes
even more difficult to restore lost confidence unless
expectations for future growth reinforce current
improvements.

-019 G ( Y t+I)
(.007)

Durbin-Watson

Total
New hires
accessions

1 Less than .05.

.424 D (h t) + .292 h t_t
(.062)
(.042)
+

Labor turnover rate revisions, 1966-69

58

Procedure. To test these hypotheses we ran ordinary
least-squares regressions for each year from 1959
through 1971, using annual averages for the 94 in­
dustry groups in manufacturing for which the Bureau
of Labor Statistics publishes labor turnover data.
The model we used regressed the quit rate for each
industry ( Q ) on the average hourly earnings of
production workers (E ), the ratio of production
workers to all employees (P ), the amplitude of
the seasonal factors for employment (S ),4 the net
new hire rate— calculated as the difference between
the quit and new hire rates— ( Hn), and the ratio of
women to all employees (W ) for that industry.
An equation was estimated for each year, using
index forms of the data with the manufacturing aver­
age for that year as the base to remove secular trends
from some of the data. The final regression coeffi­
cients were transformed to beta coefficients. This
transformation was made for the purpose of allowing
for differences in variation among the variables. (See
table 3.) Those coefficients which are not signifi­
cantly different from zero are in parentheses. With
these transformations of the data it was possible to
establish the importance and direction of each varia­
ble in determining the interindustry variation in quit
behavior. It is interesting to note that for 1960 the
results we obtained were very similar to those ob­
tained by Pencavel using a somewhat different
model.5

factor determining interindustry variations in volun­
tary separations is the relative level of earnings. Next
in order of importance are relative net hires, followed
closely by the relative proportion of female employ­
ment. Finally, in the latter years of the decade, the
relative proportion of production workers proved
to be significant, while variations in seasonality were
of minimal significance in all but a few key years.
Earnings versus security

Pecuniary motivations cause relatively high lev­
els of voluntary separations in low paying industries.
Skill requirements in such jobs are generally low.
Such positions are readily available to new or inex­
perienced workers, only to be vacated as soon as the
workers develop some skill and become aware of
other job opportunities. In high paying industries,
voluntary turnover is lower because of the low proba­
bility of obtaining a better paying job.
The earnings variable may also reflect other re­
lated market phenomena. For example, industries
with relatively low wage levels may be highly com­
petitive, labor-intensive industries where cost con­
scious entrepreneurs have minimal regard for human
capital. In such situations poor working conditions
reflected in the low levels of earnings may also ex­
plain quit behavior. On the other hand, industries
with higher wage levels may be highly unionized, in
which case unionization may be a contributor to the
higher earnings level as well as better working condi­
tions. In addition, the greater importance of human
capital in these latter industries may give manage­
ment a stake in reducing turnover.
Earnings differentials may reflect, in part, skill and
age differentials among industries. However, when
variables for occupational and age differences among
industries were introduced by Pencavel, the results
were highly insignificant. Such industry occupational
and age composition items are only available from
the decennial census and could, of course, have
changed substantially during the 1960’s.
An examination of the coefficients for earnings in
table 3 reveals that this pecuniary factor has become
an increasingly important one in determining interin­
dustry variations in quits. With the exception of peri­
ods of economic recession in the manufacturing sec­
tor (1960-61, 1967, and 1970-71), there has been
a steady progression in the importance of this varia­
ble over the decade of the 1960’s. The slight decline
in relative importance for this factor in times of

Findings. The results substantiate our qualitative as­
sessments of the relationship between quits and the
explanatory causes, even down to the indeterminacy
of the role of women and production workers in
overall quit behavior. By far the most important

Table 3. Beta coefficients for variables in cross sectional
analysis, 1959-71
Year
1959..................................
1960..................................
i961................................
1962..................................
1963..................................
1964......... .................... .
1965..................................
1966............................... .
1967..................................
1968...............................
1969..................................
1970..................................
1971..................................

E

P

-0.579
- .559
- 452
- .602
- .695
- .788
- .856
- .874
- .844
- .911
- .943
- .914
- .904

(0.048)
(.030)
(.026)
(- .011)
(- .016)
(- .005)
(.100)
.209
.192
.180
.149
.164
.164

S
0.143
(.026)
(.040)
(.000)
(.054)
(.041)
(.074)
.116
(.046)
.127
.121
(.165)
(.110)

Hn

W

0.371
.394
.422
.408
.364
.386
.301
.298
.300
.184
.253
.224
(.126)

(0.114)
.249
.272
.172
(.056)
(- .075)
- .219
- .287
- .197
- .234
- .229
- .188
- .289

NOTE: The variables in this table are: E= average hourly earnings of production
workers; P = ratio of production workers to all employees; $ = seasonal amplitude;
H=net new hire rate; W= ratio of women to all employees.
Numbers in parentheses indicate insignificant coefficients.




59

tion of the relationship changed. In 1965, the pro­
portion of female workers again became a significant
factor, but the relationship with quits was inverse. As
the relative proportion of women workers among
industries increased, the quit rate decreased.
There were several important undercurrents in the
labor market during the last half of the 1960’s which
could account for this reversal. The manufacturing
labor market became very tight. This in part was due
to the Vietnam war buildup, which increased the
demand for war related goods, generated more in­
come which increased demand for consumer goods,
and produced a manpower shortage arising from the
increased manpower needs in the military services.
As a result, there was a large influx of women into
the labor force. In addition, demographic factors
may have had their effect as there was evidence of a
slight “marriage squeeze” in 1963 and a more drastic
one beginning in 1966.6 This would also account for
the rapid increase in labor force participation among
women as well as declines in labor force withdrawal
for reasons related to marriage.
Social, cultural, and technological changes are also
quite relevant to this shift in quit behavior among
women. Such factors as the social approval and safer
methods of contraception, increasing educational at­
tainment, antidiscriminatory legislation, introduction
of labor saving equipment for household and office
use, and many others have led to the acceptance of
the modem woman as a productive worker and eco­
nomic competitor.
Despite the decline in the attitude that a “woman’s
place is in the home,” sex discrimination in hiring still
may serve as a deterrent to voluntary job mobility
for women. Since social and technological changes
have lessened the necessity for the casual attachment
of women in the labor force, the previously men­
tioned discrimination factor would seem to be a
more plausible explanation of the women’s influence
on voluntary separations in recent years.

business cycle downswings reflects the shift in impor­
tance from wage betterment to job security motiva­
tions.
The job security factor itself tends to show a grad­
ual decline over the decade as net new hires become
a less important variable, although this trend is also
interrupted during periods of cyclical downturns. The
shifts in degree of importance between the pecuniary
and job security factors over the years tested, point
out the counterbalancing relationship of these two
factors in workers’ motivations to leave their jobs
voluntarily.
In analyzing these two trends, one must remember
that the years studied are the only postwar period
characterized by prolonged economic growth. There­
fore, the increasing importance of pecuniary factors
and the decreasing importance of job security may
have been influenced to a degree by this extended
period of growth. To some extent, the expansion of
industrial centers from urban to suburban areas re­
sulting in extended labor market areas has probably
increased the worker’s knowledge of opportunities
within the market. Increasing educational attainment
and mass communication also may have increased
the information reaching the worker. Such informa­
tion makes the jobholder’s behavior more consistent
with the neoclassical concept of “economic man”
trying to increase his earnings and consumption
power under the constraint of his pains for laboring.

Women workers
In the manufacturing sector women tend to have
higher quit rates than men, partially owing to the
fact that industries with a high proportion of women
employees are also among the lower paying ones.
Hence, part of the reason for differences in quit
propensities between the sexes is the concentration of
women in lower paying jobs. Our model, however,
takes account of earnings differentials, so that it can
measure more accurately the true effect of women’s
employment as a factor in determining variations in
quits among industries. Considering the beta coeffi­
cients shown in table 3, one can see that the role
women play in determining quit propensities under­
went a drastic reversal during the last decade. From
1960 to 1962 the proportion of women employed in
an industry was a significant factor directly affecting
the frequency of quits. As the relative proportion of
women increased so did the quit rate. In the next 2
years their effect was not significant, but the direc­




Production workers and seasonality
The relative concentration of production workers
does not emerge as a significant factor until 1966. It
was in this period that demographic factors became
important in labor supply. Many young workers bom
during the postwar period entered the labor market.
With a tight market and low skill requirements,
members of this group were available for many semi­
skilled production line positions which may not have

60

been entirely to their liking. The sudden change in
age composition and labor supply may account for
part of this shift. Still another factor, somewhat re­
lated, is that in the tight market job information was
diffused more widely to a workforce of increasing
education and sophistication. This situation resulted
in better knowledge of alternative opportunities and
made it possible for the worker to behave more like
the classical economic man. Combined with disillu­
sionment of the young, job satisfaction among pro­
duction employees may also have declined. As the
labor market slackened in 1969 and 1970 the pro­
duction worker effect became less important, as is
borne out by the coefficients in table 3. Even in a
very slack labor market the greater propensity of the
production workers to quit remained, indicating the
presence of a shift in the basic pattern of manufactur­
ing quit behavior.
Finally, we come to the question of seasonality.
As our beta coefficients indicate, it is the least impor­
tant of our variables and proves to be significant only
in the years when the manufacturing business cycle is
at a peak (1959, 1966, 1969). This fact suggests
that seasonality becomes an important factor only
when jobs and alternative opportunities are plentiful.
The combined effects of these trends and shifts in
the individual variables are manifested in differences
among the various equations. We tested all pairs of
regression equations based on Chow’s test for differ­
ences between pairs of equations.7 We noted that
there are no significant differences among equations
which are separated by 1 or 2 years. There are no
significant differences in the quit experience among
the 15 pairs of equations preceding 1965, and there
are only two significant differences among the pairs
which lie entirely in the latter half of the period. Of
the remaining 36 pairs of equations that span both
subperiods, however, there are only three which do
not exhibit a significant difference in quit experience,
and these are separated in time by 1 or 2 years. We
can safely conclude, therefore, that the changes in
the effects of the individual variables resulted in a
sudden, dramatic shift in the overall basis for the
interindustry quit rate variation in the middle of the
last decade.

for all manufacturing which make it a good economic
indicator, the variations in the quit rate through time
and the sources of these variations, and the differ­
ences in quit rates among industries and the changing
bases for these differences. Through these analyses
we have obtained several results which should be
helpful in the examination of the quit rate itself, the
functioning of the labor market, and the economic
situation as a whole.
• The total manufacturing quit rate is a statisti­
cally reliable and well behaved series. Preliminary
estimates are revised only rarely and only in the most
unusual cases does this revision exceed 0.1 of a per­
centage point. The seasonally adjusted series is quite
smooth and serves as a reasonably reliable economic
indicator.
• Workers are very conscious of job security and
can have their confidence easily shaken, while resto­
ration of that confidence is quite difficult. As a result,
the changes in the quit rate may precede aggregate
economic activity by as much as five quarters during
periods of prosperity, but remain quite close to
movements in the total economy during periods of
slowed economic activity. Worker assessment of job
security seems to be built largely on the behavior of
the labor market during the past two quarters or so,
with extra weight being given to recent adverse de­
velopments. The variations in hiring among indus­
tries explains some of the variation in quits, although
this effect has been declining in recent years, with the
exception of recession years. This result suggests that
a worker draws his clues to the labor market situa­
tion not only from the closest period in time but also
from the situation that exists in the plant and indus­
try in which he is employed. The decline in the
importance of job security in the interindustry varia­
tions suggests that, with time, the worker’s horizons
are broadening and he keys his behavior to wider
economic occurrences, although there is some rever­
sion to the most immediate clues during times of
uncertainty and insecurity.
• The quit rate may be the best summary measure
of manufacturing workers’ attitudes, which in turn
make an important contribution to aggregate demand
and the course of the total economy. It is possible
that the observed correlation of quits and future ag­
gregate economic activity arises from the fact that an
uncertain worker is a cautious consumer. Such a
dynamic of aggregate demand suggests that the pub­
lic policy of creating jobs in time of slack economic
activity will do more than just increase aggregate

Summary and conclusions
We have viewed voluntary separations in Ameri­
can manufacturing industries from three different
perspectives: the properties of the average quit rate




61

demand through the usual accelerator-multiplier
principles: it will also serve to restore the confi­
dence of the worker as consumer and thereby in­
crease aggregate demand in a shorter period of time.
• Through time, the average worker has based his
decision to quit on different factors, and the impor­
tance he has attributed to each of them has been
changing. But he seems to have retained essentially
the same risk-taking posture which is modified only

by changes in the availability of jobs. The absence of
a secular decline in the quit rate, the increasing im­
portance of earnings levels in quit decisions, and the
sudden emergence of the production worker’s greater
propensity to quit, all suggest that there are no struc­
tural shifts taking place in the economy which would
impede the mobility of labor. The data on women,
however, suggest that there still remain some struc­
tural deficiencies in the labor supply process. □

1 Geoffrey H. Moore and Julius Shiskin, Indicators of
Business Expansions and Contractions (N ew York, Colum­
bia University Press, 1967).

5 In the Pencavel model (equation IA ) the beta coeffi­
cient for the earnings variable was —0.428, for the female
ratio 0.227, and for the hiring variable (accessions lagged)
0.321. The R* value for this equation is 0.778. His equa­
tion also contained a significant unionization variable and an
insignificant one for earnings variability. (See Pencavel, op.
cit., p. 21.)

2 For example: Ewan Clague, “Long-Term Trends in Quit
Rates,” Employment and Earnings, December 1956, pp.
iii-ix; Arthur Ross, “Do We Have a New Industrial Feudal­
ism?,” The American Economic Review, December 1958,
pp. 903-920; John E. Parker and John F. Burton, Jr.,
“Voluntary Labor Mobility in the U . S. Manufacturing
Sector,” Proceedings of the Twentieth Annual Winter M eet­
ing of the Industrial Research Association, pp. 61-70; John
H. Pencavel, An Analysis of the Quit Rate in American
Manufacturing Industry (Princeton, Industrial Relations
Section, Princeton University, 1970).
3 This type of formulation has been suggested, in a some­
what different context, by Lester C. Thurow, “The Changing
Nature of Unemployment,” Review of Economics and Sta­
tistics, May 1965, pp. 137-149.
4 A detailed discussion of this method is presented in The
BLS Seasonal Factor M ethod, which is available upon re­
quest at the Bureau o f Labor Statistics.




62

6 The marriage squeeze occurs when there is an abund­
ance of women o f marriageable age over men of marriagea­
ble age. See Current Population Reports, Series P -25, No.
388, U.S. Bureau of the Census, for a more detailed explana­
tion.
7 G. C. Chow, “Tests for Equality Between Sets of Coef­
ficients in Two Linear Regressions,” Econometrica, July
1960, pp. 591-605. The test outlined by Chow uses the
F-ratio. The numerator is the difference between the sum of
squared residuals from the regression o f the pooled data less
the sum of the squared residuals for the individual
regressions. The denominator is the latter sum. Both numer­
ator and denominator are adjusted for degrees o f freedom.

New models trace shifts among
job losers, leavers, and entrants
during economic
downturns and recoveries
CURTIS L. GILROY AND ROBERT J. MclNTIRE

W h a t h a p p e n s to unemployment when there is a
significant drop in economic activity? To what ex­
tent does it increase, how fast, and what happens
to four groups that make up the unemployed— job
losers, job leavers, reentrants, and new entrants into
the labor force— the groups denoted by the Bureau
of Labor Statistics’ “reasons for unemployment.”
This study shows that job losers account for the
greatest increase in unemployment and that their
response is more immediate than for other groups.
A recent Monthly Labor Review article 1 provided
a description of the unemployed by indicating which
ones have lost their last job (job losers), voluntarily
quit their last job (job leavers), reentered the labor
force after a period of absence (reentrants), or
entered the job market for the first time, never before
having held a full-time job (new entrants).

As that article emphasized, the job-loser group is
of particular interest to analysts. First, this category
comprises the largest single grouping of the un­
employed by reason (about 40 percent) and is the
most cyclically sensitive. Second, the plight of job
losers is viewed by some as being more acute than
that of the other groups, since over one-half of the
job losers are heads of households who generally
bear substantial family responsibility. Third, job
losers are the only group of unemployed whose
joblessness stems, not from voluntary action, but
from forces outside their control (that is, decisions
by employers).

To remedy some of these deficiencies, the first
approach in this followup study was to design a
new model which takes the form
Y t = oc + j9Qt -f \T t + c,

(1)

where Y represents the number of unemployed by
reason; oc is the constant; /3 and A are the coefficients
of Q and T (a cyclical indicator and a time variable,
respectively); and e is the error term.
The index of industrial production (1 9 6 7 = 1 0 0 )
compiled by Federal Reserve Board is used here
as a coincident cyclical indicator. This eliminates
the problem which may arise when much the same
variable appears on both sides of the equation.

Curtis L. Gilroy is a labor economist in the Office of Cur­
rent Employment Analysis, and Robert J. Mclntire is a
mathematical statistician in the Office o f Systems and
Standards, Bureau of Labor Statistics.

63



the “unemployment rate” for each of the groups,
as well as their proportion of total unemployment,
exhibits different cyclical patterns was shown in the
original article by a simple regression equation
(Y t = oc - f /?Qt + tt). In that equation, Y repre­
sented either the rate or proportion of the un­
employed for each group, oc was the constant term,
/? was the coefficient of Q, a cyclical indicator (the
overall jobless rate), and e was the error term.
This basic model, while useful, did not adequately
specify the relationship between movements of the
various groups of the unemployed and changes in
the level of economic activity. First, the use of the
aggregate unemployment rate as a proxy for the
business cycle was questionable, since the dependent
variables (particularly job losers and reentrants)
comprise much of the independent variable (the
overall jobless rate). Second, as the low DurbinWatson statistics suggested, serious correlation prob­
lems existed. Third, no time variable was incor­
porated into the model. Finally, an additional
problem arose because the rates of unemployment
by group were used instead of the actual number of
unemployed by group as a dependent variable.2
A new model

Because the main purpose of the previous article
was an analysis of the changing characteristics of
the unemployed by reason over the 1967-72 period,
only cursory examination was given to the cyclical
behavior of the groups. The average extent to which

From the Review of November 1974

Job losers, leavers,
and entrants:
a cyclical
analysis

A time variable was included in the regression
equation because a linear trend is assumed to be
prominent in the various series, independent of
cyclical movements— an upward trend related to
growth in the size of both the labor force and out­
put over time. Thus, a correction for trend is neces­
sary to obtain a more accurate measure of cyclical
effects.
The results from this initial step appear in table 1,
part (a ), and show that as industrial production
rises (falls), signaling a pickup (downturn) in eco­
nomic activity, the number of unemployed in all of
the categories falls (rises). A comparison of the
standardized (beta) coefficients indicates that job
losers are the most cyclically sensitive because they
experience the greatest relative change.
Even in this form, however, the equations gen­
erated low Durbin-Watson statistics, the patterns of
residuals indicating that serial correlation was still
present. This typically occurs when observations are
made over time and the effect of a disturbance
occurring at time period t carries over into the next
period, t -f- 1. It is crucial, because it violates a
basic assumption of ordinary least-squares estima­
tion, that the disturbances be independent of one
another; that is, that the error terms be generated
randomly. The regression equations above assumed
there was no serial correlation, that rho (£), the
coefficient of serial correlation, equals zero (where
serial correlation is assumed to be first order of the

form et = pet-i + vt). If, in fact, the disturbances
are not independent, the least-squares estimators
would not be the best linear unbiased estimators.
They would lose efficiency, because the dependence
among the disturbances reduces the effective num­
ber of independent pieces of information in the
sample. Thus, conventional formulas for carrying
out tests of significance or constructing confidence
limits with respect to regression coefficients would
lead to incorrect inferences. If such positive serial
correlation exists, standard errors would be under­
estimated (t-statistics overestimated) and confidence
intervals narrowed. In short, the test would be likely
to show more statistical significance than it really
should.
In an attempt to overcome this problem, the re­
gressions were rerun using the method of first dif­
ferences, with the equation taking the form
(Yt — Y,.j) = ex' + /3 (Q, — Q,_i) +

f'

(2)

where a ' = 0 and e' = et — et-i- This method recog­
nizes that there is serial correlation but further as­
sumes that the true value of p is unity. The results
indicated negative serial correlation (high DurbinWatson) and continued to cast doubt on the signifi­
cance of the regression coefficients for all equations.
The serial correlation indicates that the func­
tional form of the model may have been misspecified
or that relevant independent variables may have
been missing. The authors investigated various func­
tional forms for the independent variables without

Table 1. Regression results showing relationships between unemployed persons by reason and the index of industrial
production and time, 1967-73
[Numbers in thousands]

(a) Dependent variable (Yt)

Constant

Job losers________ ______ ______

9,275 39
(32 18)
672 96
(6.11)
3,509.75
(17.55)
1,165.76
(9.86)

Job leavers. ...................................
Reentrants....................... .......................
New entrants................................................................

(b) Dependent variable (Yt —

Job losers..................................

Constant

p -

715

Job leavers.............

p - 566

Reentrants........ .........

p - 361

New e n tra n ts....................

p - 592

2,301 74
(14 18)
269 84
(3 17)
2,196 57
(13 07)
447.94
(5.16)

NOTE: t — statistics are in parentheses.




—83
(28
-3
(2
—27
(13
—8
(7

98
74)
04
73)
16
40)
48
07)

R*

S

36 68
(38 56)
4 64
(12.77)
15 52
(23 52)
6.82
(17.49)

.949

112.77

55

—1 34

.825

43 08

84

- .2 4

.894

78.24

1 16

- 90

861

46 25

.79

- 54

R*

s

DurbinWatson

.769

74 78

2 21

<«t ~ P «t-i) (Tt - p T t.J

—72 20
(12 46)
—2 59
(1.30)
—26 74
(9 99)
-7 .9 1
(3.64)

DurbinWatson

Tt

34 10
(16 52)
4 67
(6 89)
15.73
(17.67)
6.74
(9.07)

(«t - P « t- J

—1.13

.572

35 59

2 15

- .1 7

826

67.23

2 14

-.8 5

.611

36.70

2.07

- 45

Qt* and (Qt — pQt_j)* = standardired (beta) coefficients.

64

• t*

achieving any significant reduction in serial correla­
tion. This paper reports on the results of adding
lagged monthly differences as independent variables.
But, for a given model, with a given set of variables,
a method exists for obtaining improved estimates of
the coefficients if the value of p, the coefficient of
serial correlation, can be estimated.
An estimate of p, which we label p, was obtained
from the regression procedure on the original equa­
tion Yt = oc - f - ftQt -f--ATt -f- €t. Constructing new
variables, then, the following equation was run for
each of the groups:

The model with lagged monthly differences

Although the credibility of the equations has been
enhanced, one might rightly inquire as to whether
the response of the various groups to changes in
economic activity is immediate, or whether it is
drawn out over a number of time periods.
To test for lagged responses, the following model
was run for each of the groups:
Y t = oc -f- <pQt-t + fix (Qt — Qt-i) -}- £2 (Qt-i — Qt-2 )
(Q t -2 — Qt-3 )
/34 (Qt-n — Qt-i) + A Tt + f t
(4)

+£3

where Y represents, as before the number of un­
employed by reason; (Qt — Qt-i) . . • (Qt-3 —
Qt t) are month-to-month differences lagged from
the most recent monthly change (Q t — Qt-i) back
through the fourth previous month (Qt-3 — Qt-4).
The coefficients of these lagged differences represent
the relatively “short-run” cumulative effects of a
“once-and-for-all” change in economic activity; that
is, ft, measures the effect after 1 month, ft2 measures
the accumulated effect after 2 months, and so forth.
Four successive periods were included in the model
after investigation with up to six periods. These
investigations indicated that the short-run effects of
a change in Q on each of the reasons groups run
out by the fifth month; that is, monthly differences
beyond the fourth previous month did not have
significant coefficients and did not change the results
for the first four periods. The variable Qt-4 serves
to represent the cyclical influence not accounted for
by the short-run differences, and therefore is tanta­
mount to the relatively “long-run” effect.
Because there appeared to be serious auto-correla­
tion problems in the regression equations run in this
form, the model was reestimated with first differences
adjusted using an estimate of the coefficient of first

(Y t — pYt) = cc* + $ (Q t — pQt-j)
+
\ (T t — p IY i) +
(3)

where oc* = oc (1 - p) and f* = e, - p«t.x.
These results are shown in table 1, part (b ), and
are not materially different from the original esti­
mates. The coefficients ft and % estimated with the
rho transformed variables, are improved estimates
of ft and A in the original equation. The coefficients
are more efficient and more reliable for tests of
significance.
Run in this form, the regression equations indi­
cated that, on the average, 72,000 workers (or twothirds of the overall increase in unemployment)
would lose their jobs with a 1.0 absolute decrease
in the industrial production index. The smallest
increase in unemployment would occur among job
leavers, who would account for only 2 percent of
increased joblessness. This finding is consistent with
workers’ reluctance to voluntarily leave their jobs
when the economy is weakening, and is in line with
the behavior of the quit rate in manufacturing which
falls in times of job scarcity. In fact, with respecifi­
cation of the model, the coefficient of Q was no
longer statistically significant for job leavers.

Table 2. Regression results showing relationships between unemployed persons by reason and ‘short-run’ and ‘long-run’
changes in the index of industrial production and time, 1967-73 1
[Numbers in thousands)
Dependent variable
< * t-frt-i>
Job losers................

p

=.

760

Job leavers_______

p

= .572

Reentrants..............

p

=.131

New entrants_____

p

=

292

Constant

(«t-4 - f a t J

X,

x 5.

X,

X,

2,236 46
(15.22)
268 85
(2 92)
3,180 03
(18 71)
996 80
(10 85)

—85 25
(13 66)
—2 76
(1 26)
—29 13
(14.66)
—11 02
(8.36)

—25 98
(2.84)
2 37
(0.46)
1 28
(0 14)
7 21
(1.47)

—61 17
(5 93)
—4 32
(0 79)
-1 4 92
(1.63)
6 46
(1 30)

—57 47
(5.44)
4 33
(0 79)
-2 7 69
(3.04)
- 2 68
(0.54)

—75 51
(7.42)
- 9 82
(1.84)
-2 1 52
(2 38)
0 15
(0 03)

NOTE: t — statistics are in parentheses.




37.69
(17.30)
4 88
(6.78)
16 44
(26 21)
7 26
(17 26)

x .= I(Q t- i - Q t - J - p ( Q t - . - Q t J ]
Xj = [(Q.-2 - Q , J - p(Q,-a - Qt-*)]
X4 = P f . - Q t J - p(Qt -4 - Q t j ]

1 The "short-run” lagged independent variables, are:

X, = [(Qt - Q«.J -

(Tt - f r t - i )

- QtJ1

65

R'

s

DurbinWatson

.794

62 58

2.14.

567

35 62

2 09

915

62 02

1 98

848

34.01

1 92

order serial correlation as shown by
(Yt — pYt-i) = oc* 4-

0 (Qt-4

response is distributed over the four periods. The
instantaneous response by employers to a down­
turn in business activity (here defined as an abso­
lute decrease of 1.0 in the index of industrial pro­
duction) is to lay off 25,000 workers. In the next
month, an additional 36,000 workers lose their jobs
for a total of 61,000 after 2 months. In the “longrun”— after 4 months— 85,000 workers are laid off.
The number of unemployed job leavers, on the
other hand, is not cyclically sensitive, and the group’s
short-term response to changing economic condi­
tions is erratic. In the long run, the increase in the
number of job leavers due to a drop in industrial
production is the smallest of all the groups. This is
consistent with the findings in the previous section.
Reentrant unemployment, however, clearly re­
sponds to cyclical swings in the economy. As in the
case of job losers, the effect is distributed over
several months; unlike job losers, however, there is
little instantaneous response. There is virtually no
effect in the current period to a 1.0 decrease in the
production index. In the second month, however,
15,000 of the additional unemployed workers are
reentrants; in the “long run,” there are nearly 30,000
unemployed reentrants. This development probably
reflects the fact that a greater proportion of the
labor force reentrants must pass through the un­
employment stream when the demand for labor
slackens.
The number of new entrants appear to be some­
what sensitive to cyclical swings, although their
response is less pronounced and rather different.
The positive, although insignificant, instantaneous
response reflects perhaps the fact that new entrants
are motivated in the short run by phenomena other
than changing levels of economic activity. The
impact of the business cycle appears to take hold
after several months, however, as indicated by the
significant and negative long-run response. Presum­
ably, uncertain economic conditions do not encour­
age additional new entrants to the labor force. But
it does seem reasonable that, if economic conditions
are deteriorating, new entrants unemployed in pre­
vious months are more likely to remain unemployed
and any additional new entrants are more likely to
be unemployed.
The results of this study are not strictly compar­
able to those of the study of this type previously
reported in the Monthly Labor Review because of
conceptual and methodological differences. How­
ever, both studies show job losers, the largest group,

— p Q ts ) + fa. [(Qt — Qt-0

— £(Q»-i — Q*-»)l + • • • -+- ^ (Tt — pTt-i) + e*

(5)

where cc* = « (1 — p ), «* = «t — pct-i, and all
coefficients are estimates of the coefficients in the
original lagged equation; that is, the coefficient cal­
culated for <f> in equation (5 ) is an estimate of 4> in
equation (4 ) , and so forth for the other coefficients.
The results appear in table 2 and conform in general
to what would be intuitively expected.
The number of unemployed job losers shows a
substantial degree of cyclical sensitivity and the

Chart 1. Actual and projected numbers of unemployed by
reason, October 1973-July 1974




66

to be most affected by changes in economic activity.
Application to recent data

Although the purpose of this article is not so
much to develop a predictive model as to obtain
better estimates of selected coefficients, the results
of the equation, when applied to dates for the most
recent economic slowdown, seem to track fairly
well with the actual changes in the number o f un­
employed by reason. This can be seen from chart 1,
which traces the predicted changes and the actual
changes over the October 1973 to July 1974 period.
October 1973, in this case, coincides roughly
with the beginning of the energy crisis which caused
a very abrupt slowdown in economic activity in
some sectors. Because of this abrupt slowdown,
which caused a sudden decline in the industrial pro­
duction index, the actual number of unemployed job
losers increased rapidly from October to February,
but then remained relatively stationary through July.
As shown in the chart, the projected number of
job losers— derived by using the coefficients of the
rho adjusted model with lagged monthly differences
— rose much more gradually. Although its steepest

rise occurred in the first months of 1974, when the
effects of the energy crisis were becoming more
widespread, it did not catch up with the level of
the actual series until April. There was then a rela­
tively large gap between the actual and the pro­
jected series from January to March.
However, given the unusual nature of the recent
slowdown, and the fact that the regression line
measures average movements, this gap is not too
surprising. Compared with previous slowdowns, the
one which occurred last winter had a serious impact
on only a limited number of industries— principally
automobile factories and gasoline stations. It was
the sharp cutbacks in these industries which caused
the surge in the number of unemployed job losers.
Other industries were hardly affected during this
period, and this may account for the fact that the
number of unemployed reentrants, although pro­
jected to rise gradually, did not show an actual
increase until spring. In sum, given the abrupt
nature and industry concentration of the recent
slowdown, the divergences between the projected
and actual series appear to be within reasonable
bounds.
□

-FOOTNOTES-

1 Curtis L. Gilroy, “Job losers, leavers, and entrants:
traits and trends,” M onthly Labor Review, August 1973, pp.
3-15.
2 The main drawback of the use o f component rates is
that each rate is really not a rate in and of itself. For a
true unemployment rate, the numerator and denominator
must consist of groups with like characteristics. For exam­




67

ple, the “true” unemployment rate for job losers would be
the number of job losers divided by the job-loser labor force,
not the entire civilian labor force. But the job-loser labor
force does not exist; it has no meaning. Thus, use of the
component rates, though an interesting and sometimes useful
breakdown of the aggregate unemployment rate, is little
more than a tautology.

Comparing
employment shifts
in 10 industrialized
countries

Canada, the United Kingdom,
Belgium, the Netherlands,
and Sweden, like the
United States earlier, have become
primarily service economies
CONSTANCE SORRENTINO

Generally, with a nation’s economic development
and its progress in industrialization, the distribution
of the employed population shifts from agricultural
to industrial activities, particularly manufacturing,
and further from these sectors to service activities.

services (transportation, communication, public
utilities, trade, finance, public administration, private
household services and miscellaneous services1).
Employment in government enterprises is classified
according to the sector appropriate to the output
of the enterprise. In addition to information for the
three broad breakdowns, separate figures are pro­
vided for manufacturing.
Foreign country data were adjusted to United
States concepts wherever significant conceptual dif­
ferences existed. The adjustments made as well as
the sources of the data are discussed in an appendix.

The United States emerged as the world’s first
service economy— over 50 percent of employment
in service industries— shortly after World War II.
With some lag, the other industrial nations of the
world appear to be following that pattern. By 1970,
6 of the 10 industrial countries had over half their
civilian employment in the service sector.
Sectoral shifts in employment largely reflect differ­
ing rates of change in demand and productivity.
In turn, these changes affect overall rates of change
in productivity and economic growth. Manpower
shifts from the low productivity farm sector to higher
productivity sectors result in increases in productivity
growth. On the other hand, shifts from the industrial
sector to the services sector generally have a mod­
erating effect on overall productivity growth.
This article presents data on comparative civilian
employment by sector in 10 developed countries at
5-year intervals from 1950 to 1970. The data for
1950, and perhaps for 1955, were affected by the
recovery from wartime conditions in many of the
European countries and Japan. These recovery ele­
ments may have distorted the usual relationship in
some countries. Certainly, the substantial gains in
industrial employment experienced by Italy, Japan,
and Germany from 1950 to 1970 should be viewed
against this background.
Data are provided for three broad sectors: (1 )
agriculture, forestry, hunting and fishing (called
“agriculture” in the text); (2 ) industry (comprising
mining, manufacturing, and construction); and (3 )

Shifts in employment

A vast reallocation of sectoral manpower took
place in Canada, France, Italy, Germany, Japan,
and Sweden during 1950-70. More moderate shifts
occurred in Belgium, the Netherlands, the United
Kingdom, and the United States. In general, employ­
ment disparities among the 10 countries narrowed
significantly.
Table 1 provides employment data by economic
sector, and table 2 shows percent distributions of
employment by sector. Data for 1950 were not
available for France’s industry and services; there­
fore, the French data cover only the period from
1955, except for agriculture.
Agriculture. Employment in agriculture declined in
all countries, usually quite rapidly. In conjunction
with the growth in total employment in most coun­
tries, this resulted in a significant fall in agriculture’s
share of employment.
Large differences among countries in the propor­
tion of employment in agriculture narrowed between
1950 and 1970. In 1950, agriculture dominated in
Italy and Japan, accounting for 40 percent of all
workers. One-fifth to one-fourth of total employ­
ment in Canada, Germanv. and Sweden was in agri-

Constance Sorrentino is an economist in the Division of
Foreign Labor Statistics and Trade, Bureau of Labor Statis­
tics.

From the Review of October 1971



68

culture. The United Kingdom had, by far, the lowest
proportion of workers in agricultural activities, at 5
percent. The United States and Belgium were next,
at 12-13 percent.
By 1970, agriculture accounted for more than
10 percent of employment in only France, Italy,
and Japan. The United Kingdom continued to have
the smallest proportion, 3 percent, and the United
States and Belgium followed closely with about
5 percent.
In most countries, the rate of decline in agricul­
tural employment accelerated in the sixties (table 3).
Table 1.

Agricultural employment in Italy fell by 3 percent
a year between 1951 and 1960 and at almost
6 percent a year from 1960 to 1970. The only
exception was Canada, where the decline slowed
from 3.7 percent a year in the 1950’s to 2.6 percent
in the 1960’s.
Movement out of agriculture generally makes
additional manpower available for industry and
services. However, rural to urban migration in Italy
and Japan actually tended to curb the total labor
supply during 1950-70. Many women and children
who formerly worked as unpaid farm laborers with-

Civilian employment by economic sector, 1950-70

[In thousands]

Year

United
States

Belgium

Canada

France

Germany

Italy

Japan

Nether­
lands

Sweden

United
Kingdom

34,940
39,250
43,370
46,200
50,150

3,575
3,815
4,019
4,349
4,477

3,424
(*)
3,558
3,704
3,852

22,608
23,527
24,256
25,327
24,710

15,070
14,070
12,800
10,500
8,500

572
525
465
388
340

795
(»)
570
432
325

1,228
1,122
1,028
847
711

8,200
9,910
12,380
15,010
17,850

1,465
1,592
1,678
1,845
1,818

1,323
(*)
1,431
1,565
1,480

10,507
11,235
11,462
11,739
11,081

6,180
7,530
9,430
11,450
13,730

1,103
1,177
1,241
1,331
1,309

1,069
(*)
1,128
1,215
1,089

8,194
8,852
9,122
9,242
9,026

11,670
15,270
18,190
20,690
23,800

1,538
1,698
1,876
2,116
2,319

1,306
(*)
1,557
1,707
2,047

10,873
11,170
11,766
12,741
12,918

Civilian employment
1950 *................................................. ..................................
1955........................................................ .............................
1960........................... ..........................................................
1965...................................... ................................................
1970 *....................................... ...........................................

58,920
62,171
65,778
71,088
78,627

3,402
3,414
3,438
3,608
3,670

4,976
5,364
5,965
6,862
7,879

18,752
18,727
18,712
19,560
19,967

22,869
(*)
25,954
26,699
26,327

19,098
19,701
19,877
18,915
18,698

Agriculture1*4*
1950 ' .....................................................................................
1955....................... ................................... ................. .........
I960......................................................................................
1965......................................................................................
1970'........ ...........................................................................

7,268
6,551
5,572
4,477
3,566

430
361
300
230
191

1,139
954
795
694
613

5,631
5,041
4,189
3,480
3,009

5,183
(*)
3,623
2,966
2,533

8,510
7,624
6,470
4,884
3,639

Industry9
1950*.......................................................................... .........
1955............................................................................ .........
1960......................... ................................ ...........................
1965....................................................... ...............................
1970 ' ...................................................................................

19,850
21,825
21,995
24,311
26,066

1,584
1,612
1,584
1,670
1,621

1,722
1,850
1,906
2,233
2,377

(*)
6,849
7,136
7,819
7,918

9,854
(J)
12,449
13,183
12,899

5,702
6,540
7,267
7,594
8,048

Manufacturing
1950 *................. ..................... ............... ............... ...........
1955.......................................................................................
1960.......................................... ..........................................
1965............... .....................................................................
1970*........................................................................... .

15,448
17,097
17,149
19,190
20,737

1,165
1,191
1,201
1,278
1,249

1,316
1,373
1,471
1,636
1,790

(*)
5,043
5,240
5,570
5,662

7,415
(’)
9,718
10,288
10,306

4,448
4,773
5,344
5,518
5,954

Services'
1950 *.....................................................................................
1955............................ .......................................................
I960.......................................................................................
1965.....................................................................................
1970'..................................................................................

31,800
33,796
38,212
42,301
48,994

1,388
1,441
1,554
1,708
1,858

2,116
2,560
3,264
3,934
4,888

1 1951 data lor Italy.
1 Not available.
1 1969 data for Belgium, France, Germany, and the Netherlands. Data for other
countries, except the United States, are preliminary estimates for 1970. For the United
States, 1970 data are final.
4 Includes forestry, hunting, and fishing.
9 Manufacturing, mining, and construction.
'Transportation, communication, public utilities, trade, finance, public administra­




(*)
6,837
7,387
8,261
9,040

7,832
(*)
9,882
10,550
10,895

4,885
5,538
6,141
6,437
7,012

tion, private household services, and miscellaneous services.
NOTE: Wherever significant conceptual differences occur, data have been adjusted
to U.S. concepts. Modifications have also been made so that the data for each country
reflect a compatible time series.
SOURCE: Organization for Economic Cooperation and Development, Labor Force
Statistics (various issues); International Labor Office, Yearbook of Labor Statistics
(various issues); and national statistical publications. Some data based partly on
estimates.

69

Table 2.

Percent distribution of civilian employment by economic sector, 1 9 5 0 -7 0
United
States

Belgium

Canada

France

Germany

Italy

Japan

Nether*
lands

Sweden

43.1
35.8
29 5
22.7
16.9

16 0
13.8
11 6
8 9
7.6

23.2

23.5
25.2
28.5
32.5
35.6

41.0
41.7
41.8
42.4
40.6

38.6
(•)
40.2
42.3
38.4

17.7
19.2
21.7
24.8
27.4

30.9
30.9
30.9
30.6
29.2

31.2
32.8
28.3

33.4
38.9
41.9
44.8
47.5

43.0
44.5
46.7
48.7
51.8

38.1
(J)
43.8
46.1
53.1

United
Kingdom

Agriculture2
19501.....................................................................................
1955.......................................................................................
1960.......................................................................................
1965.......................................................................................
1970 4.....................................................................................

12.3
10.5
8.5
6.3
4.5

22.9
17.8
13.3
10.1
7.8

12.6
10.6
8.7
6.4
5.2

30.0
26.9
22.4
17.8
15.1

22.7
(’)
14.0
11.1
9.6

44.6
38.7
32.6
25.8
19.5

A

11.7
8.4

Industry*
1950 1.....................................................................................
1955.......................................................................................
1960.......................................................................................
1965.......................................................................................
1970 4.....................................................................................

33.7
35.1
33.4
34.2
33.2

46.6
47.2
46.1
46.3
44.2

34.6
34.5
32.0
32.5
30.2

&

38.1
40.0
39.7

43.1
(*)
48.0
49.4
49.0

29.9
33.2
36.6
40.1
43.0

Manufacturing
19501.....................................................................................
1955.......................................................................................
I960.......................................................................................
1965.......................................................................................
1970 4.....................................................................................

26.2
27.5
26.1
27.0
26.4

26.4
25.6
24.7
23.8
22.7

34.2
34.9
34.9
35.4
34.0

(')
26.9
28.0
28.5
28.4

32.4
3%
38.5
39.1

23.3
24.2
26.9
29.2
31.8

Services
1950».....................................................................................
1955.......................................................................................
1960......................................................................................
1965.............................................. - .......................................
1970 4.....................................................................................

54.0
54.4
58.1
59.5
62.3

40.8
42.2
45.2
47.3
50.6

42.5
47.7
54.7
57.3
62.0

34.2
(*)
38.1
39.5
41.4

25.6
28.1
30.9
34.0
37.5

4 1969 data for Belgium, France, Germany, and the Netherlands.
• Manufacturing, mining, and construction.
SOURCE: Calculated from data in table 1.

1 1951 data for Italy.
* Includes forestry, hunting, and fishing.
' Not available.

drew from the labor force entirely when their
families left agriculture. Thus, the female participa­
tion rate declined in both countries.2 In most other
countries, this effect was outweighed by the increas­
ing number of married women entering the labor
force when their children reached school age.
Industry and manufacturing. Industrial employment
rose in all countries during 1950-70. However,
in six countries, the increase did not keep pace
with overall employment expansion; consequently,
the proportion in industry actually declined. Canada’s
industrial sector experienced the greatest loss in
share, falling over 4 percentage points. The industrial
sectors in Belgium and the United Kingdom lost
about 2 percentage points, while losses of less than
1 percentage point occurred in the United States,
the Netherlands, and Sweden. (See table 4.)
The industrial sectors in Italy and Japan experi­
enced substantial gains of 12-13 percentage points
in share of total employment. France and Germany




(’)
36.5
39.5
42.2
45.3

70

were the only other countries having industry gain
in employment share from 1950 to 1970. In the
period from 1960 to 1970, France, Germany, Italy,
and Japan were also the only countries with industry
employment rising as a proportion of total employ­
ment. However, the 1960-70 gain in Germany was
only 1 percentage point, after a rise of almost
5 percentage points in the earlier decade.
Japan’s industrial sector grew most, with employ­
ment more than doubling between 1950 and 1970,
greatly outpacing the rise in total civilian employ­
ment. Italy’s industrial employment rose 41 percent
but increased in relative size even more than
Japan’s because total civilian employment declined.
In the United States, industrial activities now
account for one-third of total employment, about
the same as in 1950. By 1970, all other countries
except Canada had more than a third of their man­
power in industry. With almost half of total employ­
ment in industry, Germany had the highest relative
proportion. The United Kingdom, the Netherlands,

the Organization for Economic Cooperation and
Development (OECD) and the United Nations,
trade in manufactured goods during 1955-68 in­
creased faster than manufacturing output in all major
industrial countries except the United Kingdom and
the United States.3

Italy, and Belgium had over 40 percent of their
employed manpower working in industry.
Employment in manufacturing grew slightly faster
than employment in overall industry in the United
States, Belgium, Germany, Japan, and the United
Kingdom. Manufacturing growth matched industrial
growth in Canada. In the remaining countries, rapid
increases in construction employment pushed indus­
trial growth ahead faster than manufacturing growth.
Foreign trade is an important factor in analysis
of comparative trends in manufacturing employment.
In many other countries, exports of manufactured
goods are larger relative to total GNP than in the
United States and have grown faster than U.S.
exports in the postwar years.
Postwar expansion of Japanese, Italian, and
German exports of goods has been a particularly
important factor in their rapid rise in manufacturing
employment. Japanese manufacturing employment
expanded from 18 to 27 percent of total employ­
ment between 1950 and 1970; Italy’s from about
23 to 32 percent of the total; and Germany’s from
32 to 39 percent. In contrast, U.S. manufacturing
employment remained at about 26 to 27 percent
of the total. U.S. exports of goods had a slower rate
of increase than those of any other country studied
except the United Kingdom. According to data of

Table 3.

Services. Prior to the shift from a predominantly
industrial to a predominantly service economy
shortly after World War II, over half of the em­
ployed population in the United States was produc­
ing goods— agricultural commodities and industrial
products. Until 1958, the United States was the
only industrialized country with over half its employ­
ment in services. Then Canada crossed the 50
percent level. The United Kingdom joined them
around 1965. Since 1965, Belgium, the Netherlands,
and Sweden have also become predominantly service
economies. Only France, Italy, Japan, and Germany
continue to have more workers employed in the
production of goods than of services. Japan and
France appear likely to become service economies
during the 1970’s.4 But Italy and Germany will
probably not shift until later, because service
employment constitutes only 38 percent of total
employment in Italy and 41 percent in Germany.
In 1970, the United States and Canada had about

Average annual rates of change in employment by sector, 1950-70, 1950-60, and 1960-70
Sector

United
States

Belgium

Canada

France

Germany

Italy

Japan

Nether­
lands

Sweden

United
Kingdom

1.8
- 2 .9
4.0
4.1
3.6

1.2
- 2 .8
1.1
.9
2.2

0.6
- 4 .6
.6
.1
2.3

0.4
- 2 .8
.3
.5
.9

2.2
- 1 .6
4.2
4.3
4.5

1.2
- 2 .1
1.4
1.2
2.0

0.4
- 3 .4
.8
.5
1.8

0.7
- 1 .8
.9
1.1
.8

1.5
- 4 .2
3.7
3.8
2.7

1.2
- 3 .5
.9
.6
2.3

0.8
- 5 .3
.3
-.4
2.8

0.2
- 3 .8
- .3
-.1
.9

1950-70 1
Civilian employment.............................................................
Agriculture.................. .................................................
Industry.................... ........................................ .........
Manufacturing........................................................
Services..........................................................................

1.5
- 3 .6
1.4
1.5
2.2

0.4
- 4 .4
.1
.4
1.6

2.3
- 3 .2
1.6
1.6
4.3

0.3
- 3 .4
(')
(*)
(l)

0.7
- 3 .8
1.4
1.7
1.8

- 0 .1
- 4 .6
1.8
1.6
1.9

1950-601
Civilian employment.............................................................
Agriculture....... .....................................................—
Industry.........................................................................
Manufacturing........................................................
Services..........................................................................

1.1
- 2 .7
1.0
1.1
1.9

0.1
- 3 .7
0
.3
1.1

1.8
- 3 .7
1.0
1.1
4.4

(4)
- 3 .3
(*)
(*)
(*)

1.3
- 3 .7
2.4
2.8
2.4

0.5
- 3 .1
2.7
2.1
2.6

1960-70*
Civilian employment.............................................................
Agriculture.....................................................................
Industry........................................................................
Manufacturing........................................................
Services................................ .-......................................

1.8
-4 .6
1.7
1.9
2.5

0.7
- 5 .1
.3
.4
2.0

2.8
- 2 .6
2.2
2.0
4.1

1 1950-69 for Belgium, France, Germany, and the Netherlands; 1951-70 for Italy.
* Not available.
* 1951-60 for Italy.




0.7
- 3 .8
1.2
.9
2.3

0.2
- 4 .1
.4
.7
1.1

- 0 .6
- 5 .9
1.0
1.1
1.3

4 Less than .05 percent per year.
' 1960-69 for Belgium, France, Germany, and the Netherlands.
SOURCE: Calculated from data in table 1.

71

Table 4 . Change in share of total employment by sector,
1 9 5 0 -7 0 1

rate of increase in productivity. These two factors
combined to cause the sharp reduction in agricul­
tural employment discussed earlier.
Industry grew fastest in output in all countries
except the United States, where services led in
growth. Of the countries covered in table 5, Germany
and Italy had the sharpest increases in industrial
output. Output in Japan’s industrial sector probably
grew faster than in either of these countries, however,
as overall gross domestic product in Japan increased
at a rate of almost 10 percent a year during 1952-69.
Productivity growth in industry was rapid in most
countries, although not as rapid as in agriculture.
Growth in services output was generally rapid,
and outpaced growth in total output in the United
States, Belgium, and Sweden. However, productivity
increases in services were relatively low, resulting
in the sharp rise in services employment in most
countries.5
In the United States, the level of productivity
in agriculture is well below that in industry and
services. The industry sector has the highest level
of productivity of the three sectors, and services
ranks second. It should be recognized, however,
that within each sector productivity can vary con­
siderably. In services, for example, productivity
ranges from a level lower than agriculture in
miscellaneous services to a level 4 times that of
agriculture in finance, insurance, and real estate.
Furthermore, in the industry sector, mining pro­
ductivity is over double that of manufacturing, while
output per man-hour in construction is well below
manufacturing. In general, the relationships dis­
cussed above also hold true for most of the foreign
countries studied.6
Movement of employment out of the low pro­
ductivity agricultural sector, which occurred in all
countries studied, to sectors with higher output per
worker tends to raise the level of productivity in
the whole economy, and hence, contributes to
economic growth. The following tabulation is based
on OECD calculations of the contribution to eco­
nomic growth of shifts in employment between
sectors during 1955-68.7 The proportion of the
total increase in output attributed to such shifts
in eight countries was:

[Percentage points]
Industry
Country

United States.......................................
Belgium................................................
Canada.................................................
France..................................................
Germany...............................................
Italy......................................................
Japan....................................................
Netherlands.........................................
Sweden................................................
United Kingdom...................................

Agrlculture

- 7 .8
-7 .4
-1 5 .1
-1 1 .8
—13.1
—25.1
—26.2
- 8 .4
-1 4 .8
- 2 .5

Services
Total

Menu*
facturlng

-0 .5
- 2 .4
-4 .4
3.1
5.9
13.1
12.1
-.4
-.2
- 1 .7

0.2
-.2
- 3 .7
1.5
6.7
8.5
9.7
- 1 .7
-2 .9
.3

8.3
9.8
19.5
8.8
7.2
11.9
14.1
8.8
15.0
4.2

1 1950-69 for Belgium, Germany, Netherlands; 1955-69 for France; 1951-70 for
Italy.
SOURCE: Calculated from data in tab le 1.

62 percent of their employment in services. The
next highest proportion was in Sweden with 53
percent. Canada experienced the most dramatic
increase in services, with employment growing at
4.3 percent a year and going from about 43 to 62
percent of total employment between 1950 and 1970.
Thus Canada accomplished in 12 years what it took
the United States 25 years to do— move from half
to over three-fifths of total employment in services.
During the 1950-70 period, the services sector
expanded more rapidly than the industrial sector
in all countries except Japan. In Italy, the annual
rate of growth in services was only slightly higher
than that in industry. In the other countries, however,
the rate of growth in services was much faster than
in industry.
Output and productivity trends

Changes in employment structure are the net
result of varying rates of change in demand and in
productivity. Other things being equal, increased
output requires more employees. However, pro­
ductivity increases reduce the number of workers
required for a given output.
Nine countries’ rates of growth in real output
(gross domestic product) by sector during 1950-69
are presented in table 5. Data by sector on the
growth of output per employed person (includes
wage and salary employees, unpaid family workers,
and the self-employed) are also provided. Constant
price data by sector are not available for Japan.
In general, the agricultural sector experienced the
slowest rate of growth in output and the fastest




United States ........................................................................
B e lg iu m ...................................................................................
Canada ...................................................................................
F r a n c e ..................................................................................
Germany ...............................................................................
Italy .......................................................................................
Netherlands ......................
United K in g d o m ...................................................................

72

11.3
12.9
15.3
18.3
14.9
36.8
10.3
5.2

Table 5.

Average annual rate of change in o u tp u t1 and in output per employed person by sector, 1 9 5 0 -6 9
Output per employed person

Output
Country 2

United S ta te s.................................................. .......... . .
Belgium 3__________ _______________ ______ ___
Canada « . . . ................... .............. ................ ..................
France 5...........................................................................
Germany................. .........................................................
............................................................................
Italy
Netherlands4...................................................................
Sweden...........................................................................
United Kingdom...............................................................

Total

Agriculture

Industry

Services

Total

Agriculture

Industry

Services

3.8
3.6
4.8
5.6
6.7
5.4
5.1
4.0
2.6

1.0
1.7
2.0
1.6
2.6
2.6
2.7
.4
2.2

3.8
3.8
5.6
6.6
8.0
7.6
6.0
5.0
2.9

4.2
3.7
4.8
5.6
5.8
5.0
4.8
4.4
2.2

2.3
3.3
2.4
5.1
5.9
5.6
3.9
3.5
2.1

4.7
6.4
5.2
5.4
6.6
7.0
5.5
5.1
5.1

2.2
4.1
3.8
5.5
6.5
5.7
4.9
4.3
2.5

2.0
1.9
.4
3.6
4.0
3.1
2.6
2.3
1.3

1 Gross domestic product at constant (1963) market prices for the United States,
France, Germany, and Sweden and at factor cost for all other countries.
2 Not available for Japan.
2 1956-68.
4 1950-68.
* 1955-69.

• 1951-69.
SOURCE: Calculations based on output data from the Organization for Economic
Cooperation and Development, National Accounts of OECD Countries, 1950-1968
(Paris, OECD, 1970) and estimates for 1969; and employment data comparable to
the statistics in table 1.

Although the OECD did not make calculations for
Japan, it is probable that the shift from agriculture
contributed substantially to Japanese economic
growth, as was the case in Italy. The effect of the
shift was, of course, strongest in countries where
agriculture acounted for a large proportion of
employment in the base year, 1955.
Not all sectoral shifts necessarily result in in­
creased growth of productivity for the economy.
The movement into services employment, which
has occurred in all countries studied, generally
tends to reduce the rate of growth in productivity.
As table 5 indicates, the service sector has a slower
rate of growth in productivity than agriculture and
industry in all countries. Movement out of agricul­
ture into many of the service industries represents
a shift to a sector with a higher level of productivity,
but lower rate of growth in productivity. Movement

from industry to services generally represents a
shift to a sector with a lower level of productivity
and a lower rate of growth in productivity.
Projected shifts in the structure of the U.S.
economy during the 1970’s are not likely, therefore,
to promote a faster rate of productivity growth.
Sectors with typically low rates of growth are ex­
pected to expand employment faster than those with
high rates.8 For example, rapid expansion of employ­
ment in miscellaneous services is expected. This
sector has both a low rate of growth in productivity
and a relatively low level. OECD projections for
foreign countries also indicate a slowdown in the
contribution to growth in productivity of sector
shifts in the 1970’s. In Italy, the contribution of
sectoral shifts to economic growth in the present
decade is expected to drop to less than half the
level estimated for 1955-68.9
□

-FOOTNOTES1 Miscellaneous services include hotel, repair, recreational,
personal, medical, legal and educational services.

statistics on employment and wages. The largest sector
affected is government services, where only Belgium and
Germany make some allowance (necessarily arbitrary) for
productivity increases. Clearly, it would have been desirable
to show government services separately from other services,
but comparable data are not available. Also, the use of
employment data rather than man-hours may somewhat over­
estimate labor input and therefore underestimate productivity
in services, as compared with other sectors, because there is
probably more part-time work in services. Canadian authori­
ties suggest this may be one reason for the apparent small
increase in output per employed person in Canadian service
industries.

2 In Italy, the female labor force participation rate dropped
from 32 percent to 25 percent between 1960 and 1970; in
Japan, it fell from 57 percent in 1955 to 47 percent by 1970.
In contrast, female participation in the U.S. labor force rose
from 36 percent in 1955 to 38 percent in 1960 and to 43
percent by 1970.
* Organization for Economic Cooperation and Develop­
ment, The Growth of Output, 1960-1980 (Paris, OECD,
1970), p. 61.
* According to the Statistics Bureau of the Prime Minister’s
Office, Japan will probably have half of its employed popuation in services by 1975.

8 For rough calculations of relative output per employed
person by sector in foreign countries, see OECD, op. cit., p.
36.

BPart of the slow productivity rise in services reflects a
measurement problem: in the absence of better methods,
output in constant prices in services is often measured by




7 OECD, op. cit., p. 39. In the OECD calculations, it was
assumed that the overall increase in output is the sum of

73

increases in four independent components: (1 ) agricultural
output; (2 ) output attributable to the growth in productivity
in industry and services; (3 ) output due to the increase in
total employment; and (4 ) output due to shifts in employ­
ment between sectors.
The increase in output due to shifts in employment be­
tween sectors was based on the following assumptions: (1 )
output in agriculture at the end o f the period would have
been the same even if labor had not left this sector; and
(2 ) productivity in industry and services would have been
the same whether or not labor had moved into these

sectors. Given these assumptions, the sectoral shift effect is
measured by the difference between actual output at the
end of the period and output as it would have been with
end-period productivity in industry and services, but the
same percent distribution o f employment as the beginning of
the period.
* Patterns o f U.S. Economic Growth (BLS Bulletin 1672,
1970).
* OECD, op. tit., p. 92.

Appendix: concepts used and limitations of data

Sources. With the exception of the United Kingdom,
the employment data used in this study refer to
total civilian employment; that is, wage and salary
workers, unpaid family workers, and the selfemployed. Data for the United Kingdom exclude
unpaid family workers; however, such workers ac­
count for a very small fraction of total employment.
Employment statistics for the United States and
Canada were derived solely from sample surveys of
households. Statistics for most recent years (gen­
erally 1960 onward) for Germany, Italy, Japan, and
Sweden are from household surveys while data for
the 1950’s were from other sources, such as popula­
tion censuses and official estimates based on censuses
and various other sources. Data for France, Belgium,
and the Netherlands are derived from official esti­
mates by their statistical offices. Employment figures
for the United Kingdom are based on compulsory
national insurance statistics and official estimates of
the self-employed. Distribution of the self-employed
by sector for the United Kingdom was estimated
by the Organization for Economic Cooperation and
Development.
For all countries, employment in government
enterprises is classified according to the sector appro­
priate to the output of the enterprise. This is the
procedure followed by the International Standard
Industrial Classification (ISIC).1 In the BLS estab­
lishment survey, government-operated establishments
are classified in a separate economic division, as
provided for in the U.S. Standard Industrial Classi­
fication (SIC). However, the U.S. employment data
in this report are on the ISIC basis as regards the
treatment of government enterprises. These data
are obtained from the U.S. labor force survey in
which the economic division claimed by the respond­
ent is recorded. Thus, a person working in a




government-operated manufacturing establishment is
classified in the manufacturing sector.
Output (gross domestic product) figures used in
this article are based on OECD definitions; there­
fore, data for the United States differ somewhat
from the statistics published by the U.S. Office of
Business Economics (OBE). The major difference
is that OECD figures include an estimate of capital
consumption by government, whereas no such esti­
mate is made by OBE in the national accounts.
Figures for output per employed person in the
United States presented in this article differ from
the indexes regularly published by the Bureau of
Labor Statistics. BLS data relate solely to the pri­
vate economy whereas the data in this article include
output and employment in the government sector.2
BLS figures on output per employed person are
based on output figures from OBE. In contrast,
output figures used in the calculations in this article
are based on the OECD definitions. There are also
some minor differences in the employment data
used. BLS data for the farm sector are somewhat
different in coverage from the agricultural statistics
used in this report. (The agricultural sector includes
forestry, hunting, and fishing as well as farming.)
Adjustments. Certain modifications in the basic
employment data were necessary for greater com­
parability among countries. Military personnel are
included in the basic employment data in some
foreign countries and allocated to the services sector.
Adjustment to omit the military has been made
in all cases. Unpaid family workers who work fewer
than 15 hours are excluded from employment data
in the United States, but usually are included in
other countries. Adjustments were made to exclude
such workers from the Japanese and Italian data.

74

comparisons and for trends over time than the
figures regularly published for each country.
The average annual rates of change in tables 3
and 5 were calculated by the method of selected
points. In this method, the growth rate is obtained
by connecting the logs of the beginning and terminal
values of the period of years considered with a
straight line (this trend line is given by the compound
interest rate formula). Therefore, the growth rates
are affected by the selection of the terminal years.
This study omits the effects of intrasectoral shifts
in civilian employment. There have been large
differences in the employment experience of indus­
tries within the major sectors studied. There is
considerable heterogeneity in the service sector. For
example, the U.S. has seen a dramatic growth in
services supplied to businesses, and in educational
and health services; at the other end of the scale,
the number of persons employed in domestic house­
hold service and in transporation has been sharply
reduced.
Comparisons in this study relate to the number
of persons employed. Distribution of hours by
sector would have shown somewhat different results
since trends in and levels of hours may differ.
A significant portion of the increase in U.S. services
employment, for example, was in part-time work.
The U.S. services sector has shown a consistent
postwar decline in average hours paid for. The
industry sector, on the other hand, showed very
little change in hours paid for from 1950 until a
rise in overtime hours occurred in 1964-65.4 Office
of Business Economics’ figures on the number of
full-time equivalent employees by sector eliminate
the effects of part-time work on the sectoral dis­
tribution of employment in the United States. The
1950 and 1970 percent distributions by sector of
full-time equivalent wage and salary workers and
the self-employed according to OBE statistics are
as follows:

Numbers of unpaid family workers in the other
countries were not large enough to make a signifi­
cant difference in employment proportions; con­
sequently, no other adjustments for unpaid family
workers were made.
No adjustment was made for the varying lower
age limits used for employment statistics in different
countries. The lower age limit for U.S. data was 16,
while the limit in the other countries was 14 or 15.
Other adjustments to achieve consistency of the
employment data within certain countries were made.
For the United States, adjustments were made in
1950 and 1955 data to reflect changes in survey
definitions occurring in 1957, when persons on
temporary layoff and persons waiting to begin a
new job were shifted from the employed to the
unemployed count. Since 14- and 15-year-olds were
excluded from the U.S. labor force by 1967 changes
in definition, data for all years prior to 1967 were
adjusted to exclude them. However, no adjustments
were made in the U.S. data for inconsistencies in
the series resulting from the introduction of data
from the decennial censuses into the estimation
procedure in 1953 and 1962, the inclusion of Alaska
and Hawaii in 1960, and the shifting from un­
employed to employed status in 1967 of persons
absent from their jobs during the survey week and
seeking other jobs.3
Adjustments were also made for such breaks in
the comparability of time series as the 1967 introduc­
tion of a redesigned labor force survey in Japan and
the introduction of three different Standard Indus­
trial Classification systems (in 1948, 1958, and
1968) in the United Kingdom. Also, where popula­
tion census data were used in lieu of labor force
survey statistics for 1950, these data were adjusted
to a compatible basis with the survey data based
on a comparison of survey and census data in a
year when both were available.
Limitations. The adjustments discussed above have
accounted for all major definitional differences in
employment statistics between countries and all
significant time series differences within countries.
However, it should be emphasized that only approxi­
mate comparability was achieved among countries.
In some instances, it was necessary to make adjust­
ments based on incomplete data or on overlapping
data in one year which may not be fully applicable
to other years. Nevertheless, the adjusted figures
provide a more accurate basis for international




Agriculture ...................................................
Industry ..........................................................
Services ..........................................................
Government enterprises .............................

1950

1970

11.8
34.9
52.0
1.3

4.5
32.2
61.7
1.7

The United States employment data used in this
report show a similar percent distribution by sector.
(See table 2.) It.should be remembered that em­
ployment in government enterprises is distributed
according to the sector most appropriate to the
output of the enterprise in tables 1 and 2.

75

By limiting the analysis to civilian employment,
two important segments of the labor force are
omitted— military personnel and the unemployed.
The size of a nation’s military forces can have major
economic implications, but the forces themselves
are not considered to be engaged in economic
activity. The unemployed are omitted because by

------------------

definition they are not productively engaged and
their association with a particular economic sector
is tenuous at best. In the United States, these ex­
cluded groups represented 7 to 10 percent of the
labor force between 1950 and 1970. Elsewhere, the
military and unemployed generally accounted for
much smaller proportions of the labor force.
□

APPENDIX

1
United Nations, International Standard Industrial Classi­ from the 1960 census reduced the population by 50,000 and
fication of A ll Economic Activities, Statistical Papers, Series
labor force and employment by 200,000. The inclusion of
M, No. 4, Revision 2 (N ew York, United Nations, 1968).

Alaska and Hawaii beginning in 1960 resulted in an increase
of about 500,000 in the population and about 300,000 in the
labor force. The 1967 shift of persons absent from their jobs
and seeking other jobs to the employed category increased
total employment by about 80,000.

*BLS figures On the private economy do include output
and employment o f government enterprises, but exclude
public administration.
’ Beginning in 1953, population levels were raised by
about 600,000 and labor force and employment raised by
about 350,000; beginning in 1962, the introduction of figures




*

Patterns of UJS. Economic Growth (BLS Bulletin 1672,

1970), p. 13.

76

Chapter II. Changes in the Labor Force




The U.S.
labor force:
projections
to 1990

Special Labor Force Report shows
work force expanding to 101.8 million by 1980;
rate of growth is then expected to decline,
with labor force reaching 107.7 million
by 1985 and 112.6 million by 1990
DENIS F. JOHNSTON

D uring the 1970’s, the total labor force of the
United States is estimated to expand by 15.9 mil­
lion, reaching 101.8 million by 1980, according to
the latest projections of the Bureau of Labor Statis­
tics. This increase implies an average annual growth
rate of 1.7 percent, about the same as the average
annual rate for the 1960’s. After 1980, the rate of
growth is expected to decline, averaging only 1.0
percent a year during the eighties. At this decelerated
rate, the labor force is estimated to reach 107.7
million by 1985 and 112.6 million by 1990.
Projected changes in the labor force are of neces­
sity closely related to changes projected in the size
and age composition of the working-age population
— those 16 and over. Projected changes in labor
force participation rates (the percent of the popula­
tion in the labor force) for specific age-sex groups
are also significant, but their impact is overshad­
owed by the effect of the projected population
changes. Between 1970 and 1990, for example, 89
percent of the projected change in the male labor
force and 68 percent of that of the female labor
force can be attributed to projected population
changes. Only among men 65 and over, and women
20 to 24 and 45 to 54, do projected changes in
labor force participation rates have a greater effect
on the labor force than changes in population.
This article presents projections of the total labor
force of the United States, by age and sex, for 1980,
1985, and 1990.1 It includes a discussion of past
trends, as background for the analysis of changes
implied by the projections, together with a brief
summary of the assumptions which underlie the

projections and the methods employed in their de­
velopment.
Changes in the 1970’s

The projected 1980 labor force differs markedly
from the actual labor force of 1960 and 1970 in its
composition by age and sex. The median age of the
labor force, which declined from 40 to 38 years
during the 1960’s, is expected to fall still more rap­
idly during the present decade, reaching 35 years by
1980. The major factor in this decline is the sharp
rise in the number of young adult workers aged 25
to 34 years— from 17.7 to 26.8 million— as the
“baby boom” generation moves inexorably through
the life cycle. This age group— one-fifth of the labor
force in 1970— is estimated to make up over onefourth of the labor force 10 years later (table 1).
These projections also indicate a continuing in­
crease in the proportion of the labor force that are
women— from 36.7 percent in 1970 to 38.5 percent
in 1980. This projected increase is much less pro­
nounced, however, than the rise since 1960, when
32.1 percent of those in the labor force were women
(table 2 ). Two major reasons may be cited in sup­
port of the more modest growth projected for
women workers during the present decade. First, the
largest changes in the female population in the
1970’s are in the age group (25 to 34 years) whose
labor force participation rate is lower than for those
age groups where the largest population increases
occurred in the 1960’s. Second, the unusually rapid
increase in women’s labor force participation rates
during the past decade is associated with the precip­
itous decline in the birth rate. These projections as­
sume that no further drastic declines in birth rates
are in prospect. Thus, the labor force participation
rate of women 25 to 34 years old, which rose from
35.8 percent in 1960 to 44.8 percent in 1970, is_
projected to rise only 5.4 percentage points during

Denis F. Johnston is Senior Demographic Statistician,
Office of Manpower Structure and Trends, Bureau of
Labor Statistics. William V. Deutermann, Jr., of the Divi­
sion of Labor Force Studies, assisted in developing the sta­
tistical materials.

78
From the Review of July 1973



Table 1. Total population, total labor force, and labor force participation rates, by age and sex, actual 1960 and 1970
and projected 1980, 1985, and 1990
[Numbers in thousands]

Total population, July 1

Labor force participation rates,
annual averages (percent of
population in labor force)

Total labor force, annual averages

Sex and age group
Actual

Projected

Actual

Projected

Actual

Projected

1960

1970

1980

1985

1990

1960

1970

1980

1985

1990

1960

1970

1980

1985

1990

121,817
21,773
67,764
32,279

142,366
32,257
71,777
38,333

167,339
37,463
84,740
45,136

175,722
34,405
94,028
47,289

183,078
31,643
103,309
48,126

72,104
12,720
46,596
12,788

85,903
19,916
51,487
14,500

101,809
23,781
61,944
16,084

107,716
22,184
69,202
16,330

112,576
20,319
76,421
15,836

59.2
58.4
68.8
39.6

60.3
61.7
71.7
37.8

60.8
63.5
73.1
35.6

61.3
64.5
73.6
34.5

61.5
64.2
74.0
32.9

59,420
5,398
2,880
2,518
5,553
11,347
11,878
10,148
7,564
4,144
3,420
7,530
2,941
4,590

68,641
7,649
3,937
3,712
8,668
12,601
11,303
11,283
8,742
4,794
3,948
8,395
3,139
5,256

80,261
8,339
4,111
4,228
10,666
18,521
12,468
10,781
9,776
5,263
4,513
9,710
3,633
6,077

84,285
7,141
3,515
3,626
10,305
20,540
15,409
10,630
9,874
5,129
4,745
10,386
3,852
6,534

87,911
7,045
3,373
3,672
9,021
21,040
18,378
11,922
9,424
4,787
4,637
11,081
4,065
7,016

48,933
3,162
1,322
1,840
4,939
10,940
11,454
9,568
6,445
3,727
2,718
2,425
1,348
1,077

54,343
4,395
1,840
2,555
7,378
11,974
10,818
10,487
7,127
4,221
2,906
2,164
1,278
886

62,590
4,668
1,887
2,781
8,852
17,523
11,851
9,908
7,730
4,558
3,172
2,058
1,289
769

66,017
3,962
1,603
2,359
8,496
19,400
14,617
9,744
7,716
4,421
3,295
2,082
1,322
760

68,907
3,901
1,530
2,371
7,404
19,853
17,398
10,909
7,307
4,112
3,195
2,135
1,365
770

82.4
58.6
45.9
73.1
88.9
96.4
96.4
94.3
85.2
89.9
79.5
32.2
45.8
23.5

79.2
57.5
46.7
68.8
85.1
95.0
95.7
92.9
81.5
88.0
73.6
25.8
40.7
16.9

78.0
56.0
45.9
65.8
83.0
94.6
95.1
91.9
79.1
86.6
70.3
21.2
35.5
12.7

78.3
55.5
45.6
65.1
82.5
94.4
94.9
91.7
78.1
86.2
69.4
20.0
34.3
11.6

78.4
55.4
45.4
64.6
82.1
94.4
94.7
91.5
77.5
85.9
68.9
19.3
33.6
11.0

62,397
5,275
2,803
2,472
5,547
11,605
12,348
10,438
8,070
4,321
3,749
9,115
3,347
5,768

73,725
7,432
3,828
3,604
8,508
12,743
11,741
12,106
9,763
5,257
4,506
11,433
3,780
7,653

87,078
8,057
3,969
4,088
10,401
18,442
12,903
11,625
11,307
5,966
5,341
14,343
4,595
9,748

91,437
6,910
3,397
3,513
10,049
20,301
15,741
11,407
11,492
5,804
5,688
15,537
4,942
10,595

95,167
6,776
3,243
3,533
8,801
20,750
18,524
12,695
10,934
5,396
5,538
16,687
5,267
11,420

23,171
2,061
801
1,260
2,558
4,159
5,325
5,150
2,964
1,803
1,161
954
579
375

31,560
3,250
1,324
1,926
4,893
5,704
5,971
6,533
4,153
2,547
1,606
1,056
644
412

39,219
3,669
1,427
2,242
6,592
9,256
6,869
6,537
5,057
3,055
2,002
1,239
758
481

41,699
3,203
1,247
1,956
6,523
10,339
8,560
6,542
5,213
3,033
2,180
1,319
814
505

43,669
3,188
1,205
1,983
5,826
10,678
10,219
7,364
5,003
2,853
2,150
1,391
864
527

37.1
39.1
28.6
51.0
46.1
35.8
43.1
49.3
36.7
41.7
31.0
10.5
17.3
6.5

42.8
43.7
34.6
53.4
57.5
44.8
50.9
54.0
42.5
48.4
35.6
9.2
17.0
5.4

45.0
45.5
36.0
54.8
63.4
50.2
53.2
56.2
44.7
51.2
37.5
8.6
16.5
4.9

45.6
46.4
36.7
55.7
64.9
50.9
54.4
'57.4
45.4
52.3
38.3
8.5
16.5
4.8

45.9
47.0
37.2
56.1
66.2
51.5
55.2
58.0
45.8
52.9
38.8
8.3
16.4
4.6

BOTH SEXES
Total, 16 years and over.
16 to 24 years........................
25 to 54 years..................
55 years and over............ .........
MEN
Total, 16 years and over.
16 to 19 years...................
16 and 17 years.............
18 and 19 years..................
20 to 24 years........... ..........
25 to 34 years______ ______
35 to 44 years___________
45 to 54 years................. .........
55 to 64 years......... ..........
55 to 59 years....................
60 to 64 years__________
65 years and over............ .........
65 to 69 years..... ..............
70 years and over_______
WOMEN
Total, 16 years and over.
16 to 19 years....... ....................
16 and 17 years.................
18 and 19 years.................
20 to 24 years............................
25 to 34 years............................
35 to 44 years........................ .
45 to 54 years______________
55 to 64 years............. ..............
55 to 59 years.....................
60 to 64 years........... .........
65 years and over__________
65 to 69 years......... ...........
70 years and over..............

SOURCE: Population and labor force data for 1960 are from Special Labor Force

Current Population Survey estimates. Projected population data are from Current
Population Reports, Series P-25, No. 493, Series l .

Report 119 and differ slightly from later estimates. Corresponding 1970 data are from

the current decade, reaching 50.2 percent in 1980.
The large increase in the number of young adult
workers, and the continued rise in the number of
women in the labor force, are the salient features of
the changes in prospect during the remainder of the
present decade. However, the changes which are
foreseen in the other age-sex groups of the work­
ing-age population are also significant (table 3).

This development has important implications for the
absorption of these new young labor force entrants
into the Nation’s economy. During the 1960’s when
the number of teenage workers was rising by about
240,000 a year, on average, teenage unemployment
fluctuated between 12.2 and 17.2 percent (on an
annual average basis). In contrast, the size of the
teenage labor force is estimated to increase by only
about 70,000 a year, on average, during the current
decade. This slower pace of increase should enhance
the effectiveness of measures designed to reduce the
unemployment rate among teenagers. (See chart 1.)

First, the teenage labor force, which increased
from 5.2 million in 1960 to over 7.6 million in
1970, is projected to increase still further, but at a
slower pace, reaching a peak in the late 1970’s.
Thereafter, this group may be expected to decline
slowly in number, reaching 8.3 million in 1980.




Second, the group aged 20 to 24 is projected to
continue to grow rapidly in size during the current

79

decade, but again at a slower pace than during the
1960’s. This group increased by an average of
480,000 a year during the 1960’s, but is expected to
increase at the more moderate pace of 320,000 a
year during the current decade, reaching 15.4 mil­
lion workers by 1980.
Third, the group aged 35 to 44, which was the
same size in 1970 as in 1960, is projected to in­
crease by 1.9 million during the current decade,
reaching 18.7 million in 1980, as the larger number
of persons born between 1935 and 1944 moves into
this age group of workers.
Fourth, the group aged 45 to 54, which increased
by 2.3 million between 1960 and 1970, is projected
to decline by nearly 600,000 during the present dec­
ade, reaching 16.4 million workers in 1980. At that
time, this group will consist mostly of the relatively
small number of persons born between 1925 and
1934— the “depression” cohort.
Fifth, the Nation’s older workers (aged 55 and
over) are projected to continue to increase in num­
ber at a somewhat more moderate pace during the
current decade. This group increased by about 1.7
Table 2.

million during the 1960’s, and is expected to in­
crease by an additional 1.6 million during the
1970’s, reaching 16.1 million workers in 1980.
Within this group, the number of workers aged 65
and over is projected to remain nearly constant, ris­
ing from 3.2 million in 1970 to 3.3 million in 1980.
This trend is in contrast to the steady increase in
the size of the population 65 and over, which is ex­
pected to grow from 19.8 million in 1970 to over
24 million by 1980. Projected reductions in the
labor force participation rates of persons in this age
group yield a nearly constant number of workers de­
spite the substantial increase in the population.
Comparison with earlier projections

In general, the present set of labor force projec­
tions differs from the previous BLS study in two
major respects. First, the participation rates for men
in all age groups are now estimated to decline over
time, reflecting the observed downward movement
over the 1955-72 period. 2 Second, the participation
rates for women are considerably higher than those

Distribution of total labor force, by age and sex, actual 1960 and 1970 and projected 1980, 1985, and 1990
Percent distribution

Number (in thousands)
Sex and age group

Actual

Projected

Actual

Projected

1960

1970

1980

1985

1990

1960

1970

1980

1985

1990

Total, 16 years and over___
16 to 24 years.................... .............
16 to 19 years..........................
20 to 24 years_____ ____ _
25 to 54 years_____ . . ............
25 to 34 years_____________
35 to 44 years_____ _____
45 to 54 years_____ ___ . . .
55 years and over________ _____
55 to 64 years...........................
65 years and over__________

72,104
12,720
5,223
7,497
46,596
15,099
16,779
14,718
12,788
9,409
3,379

85,903
19,916
7,645
12,271
51,487
17,678
16,789
17,020
14,500
11,280
3,220

101,809
23,781
8,337
15,444
61,944
26,779
18,720
16,445
16,084
12,787
3,297

107,716
22,184
7,165
15,019
69,202
29,739
23,177
16,286
16,330
12,929
3,401

112,576
20,319
7,089
13,230
76,421
30,531
27,617
18,273
15,836
12,310
3,526

100.0
17.6
7.2
10.4
64.6
20.9
23.3
20.4
17.7
13.0
4.7

100.0
23.2
8.9
14.3
59.9
20.6
19.5
19.8
16.9
13.1
3.7

100.0
23.4
8.2
15.2
60.8
26.3
18.4
16.2
15.8
12.6
3.2

100.0
20.6
6.7
13.9
64.2
27.6
21.5
15.1
15.2
12.0
3.2

100.0
18.0
6.3
11.8
67.9
27.1
24.5
16.2
14.1
10.9
3.1

Median age.....................................

39.9

38.2

35.2

35.8

37.0

Total, 16 years and over___
16 to 24 years..............................
25 to 54 years...... .................... .......
55 years and over.................... .......

48,933
8,101
31,962
8,870

54,343
11,773
33,279
9,291

62,590
13,520
39,282
9,788

66,017
12,458
43,761
9,798

68,907
11,305
48,160
9,442

67.9
11.2
44.3
12.3

63.3
13.7
38.7
10.8

61.5
13.3
38.6
9.6

61.3
11.6
40.6
9.1

61.2
10.0
42.8
8.4

Median age________ ____

39.7

38.2

35.2

35.8

36.9

Total, 16 years and over___
16 to 24 years............... ............... .
25 to 54 years............... .................
55 years and over_____________

23,171
4,619
14,634
3,918

31,560
8,143
18,208
5,209

39,219
10,261
22,662
6,296

41,699
9,726
25,441
6,532

43,669
9,014
28,261
6,394

32.1
6.4
20.3
5.4

36.7
9.5
21.2
6.1

38.5
10.1
22.3
6.2

38.7
9.0
23.6
6.1

38.8
8.0
25.1
5.7

Median age____ _________

40.3

38.2

35.1

35.9

37.1

BOTH SEXES

MEN

WOMEN




80

The uses of projections
. . . The basic distinction between a projection and
a forecast reflects the purpose it is intended to serve
rather than the method o f its preparation or the
degree o f understanding w hich it reflects. A forecast
may be defined as a projection which has been
selected as representing the “most likely” outcom e
in situations w hose determ inants are insufficiently
known or controlled to permit outright prediction.
Its distinguishing characteristic is the elem ent o f
judgm ent or decision which is necessary in making
such a selection. If projections are race horses, the
forecast is the horse you decide to bet on.
W hereas projections may serve a number o f func­
tions, the basic function on a forecast is to delineate
the m ost probable outcom e in a specified area of
concern over a specified period in the future. T he
need for a forecast does not arise until and unless
the user must com m it him self to a definite plan o f
action extending into the future. G iven such a co m ­
mitm ent, the preparation or adoption o f som e kind
o f forecast is inescapable.
Projections in general, and econom ic-dem ographic
projections in particular, may be used to m eet a
number o f purposes. First, they are m ost com m only
designed to fulfill an anticipatory function— allow ing
the user to anticipate the probable m agnitude or
im pact o f som e probable or postulated set o f condi­
tions or changes at som e future time. . . .
Second, projections— or the forecast which is se­
lected from am ong them — are an essential input for
planning and program developm ent. If our plans and
programs are rational, they must be future-oriented,
and they must therefore incorporate som e systematic
appraisal o f the environm ent in which these plans
are likely to operate in the future. . . .
Third, projections are an essential— though som e­
tim es im plicit— ingredient in program evaluation.
A ttem pts at program evaluation, especially in areas
involving social behavior, com m only encounter the
problem that program benefits cannot be estimated
with nearly the confidence or accuracy that sur­
rounds estim ates o f program costs. The social re­
searcher recognizes in this difficulty the truism that
the im pact o f any social program is entangled in a
web o f cross-im pacts reflecting the totality o f inter­
actions occurring in the society. One way to avoid
this difficulty is to project the course of d evelop­
m ents which might be anticipated in the absence of
the particular program, so that com parison o f this
projection with actual post-program outcom es may
yield an estimate, how ever crude, of program im pact
or “benefit.”
Fourth, projections may be viewed as essential
links in a chain o f conjecture; each projection in­
cludes am ong its underlying assum ptions certain




conditions which are derived from a prior projection,
and most projections are likely, in turn, to provide
inputs to other projections. . . .
Fifth, projections serve a public inform ation fu n c­
tion. Our justifiable concern with the m anipulative
and propagandistic elem ents which may be found in
projections prepared for public effect should not
obscure the fact that projections, when freed o f such
influences, have a unique educational value. . . .
Finally, projections serve an exploratory or h eu­
ristic function, insofar as they may be developed in
order to delineate the probable (or p ossible) con ­
sequences o f alternative sets o f initial conditions and
determ ining factors. W hile the ch ief value o f such
exercises may be educational, they m ay be o f c o n ­
siderable practical value to the decisionm aker as
well. T o the extent that they expand his awareness
o f the “degrees o f freedom ” which he enjoys in a
given situation, they may prompt his consideration
o f alternative solutions which he might not other­
wise have recognized.
Each o f these six functions provides a perspective
from which to suggest a course o f action in “build­
ing upon” the available econom ic and dem ographic
projections. H ow ever, it is the last o f these functions
which most clearly reflects the nature and potential
value o f projections in their purest sense, and it is
the fulfillm ent o f this function which m ost nearly
im plies a capacity to carry out the other functions
as well.

. . . T o build upon econ om ic and dem ographic
projections, it is necessary to recognize the different
purposes for w hich projections are developed and
the different strategies which are called for in pur­
suing these purposes. From the standpoint o f the
technician, the necessary strategy is straightforward:
we need to integrate our econom ic and dem ographic
m odels, incorporating additional indicators o f rele­
vant social processes, so as to d evelop more inclusive
social system s m odels. But for the decisionm aker and
social critic alike, a different strategy m ust be
em ployed— one w hich recognizes in the failures o f
past predictions not the need for im proved analytical
system s, but rather the existence o f opportunities for
the expression o f hum an values w hich alone give
m eaning to our decisions.
— D e n is F . J o h n s t o n ,
“Building on Economic and Demographic Projections,”
a paper presented at a meeting of the
Society o f Actuaries, Toronto.

81

Table 3.

Net changes in total labor force 16 years old and over, by age and sex, 1960-70, 1970-80, and 1980-90
Net change
(in thousands)

Average annual rate of change1
(in percent)

Percent change

Sex and age group
1960-70

1970-80

1980-90

1960-70

1970-80

1980-90

1960-70

1970-80

1980-90

13,799
2,422
4,774
2,579
10
2,302
1,871
-1 5 9

15,906
692
3,173
9,101
1,931
-5 7 5
1,507
77

10,767
-1 ,2 4 8
-2 ,2 1 4
3,752
8,897
1,828
-4 7 7
229

100.0
17.6
34.6
18.7
.1
16.7
13.6
-1 .2

100.0
4.4
19.9
57.2
12.1
-3 .6
9.5
.5

100.0
-1 1 .6
-2 0 .6
34.8
82.6
17.0
-4 .4
2.1

1.75
3.81
4.93
1.58
(2)
1.45
1.81
-.4 8

1.70
.87
2.30
4.15
1.09
-.3 4
1.25
.24

1.01
-1 .6 2
-1 .5 5
1.31
3.89
1.05
-.3 8
.67

5,410
3,672
1,317
421

8,247
1,747
6,003
497

6,317
-2 ,2 1 5
8,878
-3 4 6

39.2
26.6
9.5
3.0

51.8
11.0
37.7
3.1

58.7
-2 0 .6
82.5
-3 .2

1.05
3.74
.40
.46

1.41
1.38
1.66
.52

.96
- 1 .7 9
2.04
-.3 6

8,389
3,524
3,574
1,291

7,659
2,118
4,454
1,087

4,450
-1 ,2 4 7
5,599
98

60.8
25.5
25.9
9.4

48.2
13.3
28.0
6.8

41.3
-1 1 .6
52.0
.9

3.09
5.67
2.18
2.85

2.17
2.31
2.19
1.90

1.07
-1 .3 0
2.21
.15

BOTH SEXES
Total, 16 years and over..............................
16 to 19 years.................... ...........
20 to 24 years..............................
25 to 34 years_____ _____ ____
35 to 44 years__________ . _
45 to 54 years..... ...........
55 to 64 years.......................
65 years and over..............
MEN
Total, 16 years and over__
16 to 24 years______
25 to 54 years_______
55 years and over..............

... .

WOMEN
Total, 16 years and over....... ......................
16 to 24 years...................... ...............................
25 to 54 years_____ _______ ________ _______
55 years and over........................................ ...........
1Compounded continously.

2 Less than .05 percent.

formerly estimated, although the rate of increase is
generally slower than that observed during the 1960
decade (table 4 ). The net effect of these changes is
to reduce the 1980 male labor force by 1.0 million
(in comparison with the previous BLS projection)
and to raise the female labor force by 2.1 million,
for a net increase of 1.1 million (from 100.7 to
101.8 million workers).
The direction of both of these major changes is
the same as that of earlier revisions in the BLS
projections (as shown in table 4 ). However, unlike
the earlier revisions, the present projection does not
hold the participation rates for men in the central
working ages (25 to 54) at a constant level. Instead
it allows these rates to edge downward slowly, on
the assumption that the observed reductions between
1955 and 1972 are not attributable to cyclical fac­
tors, but rather reflect a long-term secular trend. As
has been noted in previous BLS projections, the
projected declines in the participation rates of
younger men (16 to 24) are assumed to reflect the
net effect of continued growth in school enrollment,
while the declines projected among men 55 and
over reflect a long-term trend toward earlier retire­
ment— an option which is increasingly supportable
by virtue of the improved terms and increased cover­
age afforded by a host of private and public pension
plans and personal savings. 3




In regard to the upward revision in the participa­
tion rates for women, three major points should be
made. First, the current projection implies a sub­
stantial reduction in the rate of increase of partici­
pation rates of women under 35. This is particularly
noticeable in comparison with the very rapid gains
observed among women in this age group during the
1964-72 period, when their participation rates in­
creased by 10 percentage points, reaching 50.4 per­
cent in 1972. The projected gain over the following
8-year period (1 9 7 2 -8 0 ) is only 2.5 percentage
points. As noted previously, the more modest
growth projected in the labor force participation
rates of these younger women reflects the judgment
that the extraordinary growth observed during the
past decade was accelerated by certain factors which
are not expected to have a significant impact in the
future. The most important of these is the rapid de­
cline in fertility that occurred during the 1960’s. Be­
tween 1960 and 1972, the general fertility rate de­
clined from 118.0 to 73.4— a drop of 38 percent.4
Since the presence of young children in the home
limits the availability of mothers for work outside
the home (ceteris paribus), this reduction in fertility
allowed a growing proportion of young women to
enter the labor force. In addition, the Vietnam
buildup of the late 1960’s afforded unusually favor­
able employment opportunities for these women.

82

Chart 1. Change in labor force (annual average) over
successive decades, 1960 to 1990, by age group

flects in part the slow increase in labor force partici­
pation among women 45 to 54 years old observed
during the past decade. In addition, it is felt that the
very large increases projected in the number of
young women workers 25 to 34 years old may have
a limiting effect on the employment opportunities of
older women.
Finally, the projection for older women (55 and
over) shows a small increase in their rate of labor
force participation during the remainder of the cur­
rent decade. This projected increase occurs only
among women 55 to 64 years old; the long-term de­
cline in participation among women 65 and over is
expected to continue. Although the projected labor
force of women 55 and over in 1980 is practically
identical with the previous BLS projection, the pro­
jected participation rates are somewhat lower, re­
flecting the stabilized rates observed in the recent
past. This apparent discrepancy is accounted for by
the larger size of the population of women 55 and
over currently projected for 1980.

Annual average change (in millions)
Age

-.2

0

.2

.4

.6

.8

1.0

Changes in the 1980’s

1 An increase of one thousand per year, on average.

Also, in the late 1960’s the number at or near the
median age at which women married for the first
time was considerably larger than the number of
men 2 to 3 years older than themselves whom they
would normally have married. This temporary im­
balance was exacerbated by the Vietnam buildup,
thus inducing considerable delay in marriage. Each
of these factors is assumed to have had a strong
positive influence on the participation rates of young
women in the recent past, and none of these factors
is expected to be operative in the future.
Second, the current projection implies a more
moderate reduction in growth of the participation
rate among women 35 to 54 years old. Between
1964 and 1972, their participation rate increased by
5 percentage points, reaching 52.7 percent in 1972.
The corresponding increase for 1972-80 is only 2.0
percentage points. This slower projected growth re­




83

The outstanding feature of the projected 1980-90
increase in the total labor force is the slower pace of
growth— from an average annual rate of 1.7 percent
in the 1970’s to 1.0 percent in the 1980’s. At this
reduced rate, the labor force is projected to increase
by 10.8 million during the 1980 decade, reaching
107.7 million by 1985 and 112.6 million by 1990.
Also significant is the expected shift in the locus of
major expansion, from the 25- to 34-year-old group
in the 1970’s to the 35- to 44-year-old group, during
the 1980’s. The latter group, whose number is pro­
jected to increase by about 190,000 a year, on aver­
age, during the current decade, is projected to grow
by nearly 900,000 a year, on average, during the
1980’s. One manifestation of this shift is the esti­
mated rise in the median age of the labor force—
from 35.2 years in 1980 to 37.0 years in 1990.
The number of young workers (16 to 24 years
old) is projected to decline by nearly 350,000 a
year, on average, during the 1980’s from 23.8 mil­
lion in 1980 to 22.2 million by 1985 and 20.3 mil­
lion by 1990— only 400,000 more than their num­
ber in 1970. (See chart 2.) However, this younger
group in 1990 is expected to differ sharply from
that of 1970, with nearly 500,000 fewer men and
900,000 more women workers— reflecting the as­
sumed continuation in both the downward trend in

Table 4.

Comparison of current labor force projection with earlier BLS projections, 1980 and 1985

[In thousands]
Total labor fo rc e 16 years old and over, by age and sex

1980
Sex and age group

Current
projection

SLFR
1191

1980
SLFR
492

(1)

(2)

(3)

101,809
23,781
61,944
16,084

100,727
23,130
61,377
16,220

62,590
13,520
39,282
9,788

39,219
10,261
22,662
6,296

1985
Differences

Current
projection

1985
SLFR
119'

Difference
( 6 ) - (7)

(6)

(7)

(8)

( D - (2)

( 1 ) - (3)

(4)

(5)

99,942
22,554
60,431
16.957

1,082
651
567
-1 3 6

1,867
1,227
1,513
-8 7 3

107,716
22,184
69,202
16,330

107,156
22,242
68,525
16,389

560
-5 8
677
-5 9

63,612
13,690
39,983
9,939

64,061
13,888
39,893
10,280

-1 ,0 2 2
-1 7 0
-701
-151

-1,471
-3 6 8
-611
-4 9 2

66,017
12,458
43,761
9,798

67,718
13,179
44,542
9,997

-1,701
-721
—781
-1 9 9

37,115
9,440
21,394
6,281

35,881
8,666
20,538
6,677

2,104
821
1,268
15

3,338
1,595
2,124
-381

41,699
9,726
25,441
6,532

39,438
9,063
23,983
6,392

2,261
663
1,458
140

BOTH SEXES
Total, 16 years and over........
16 to 24 years________
25 to 54 years_______
55 years and over..............................
MEN
Total, 16 years and over____
16 to 24 years..................
. __
25 to 54 years.....................
55 years and over........ ......................
WOMEN
Total, 16 years and over____
16 to 24 years...................................
25 to 54 years____ _____________
55 years and over_______________

1 Sophia C. Travis, "The U.S. labor force: projections to 1985,” Monthly Labor

Monthly Labor Review, February 1965, pp. 129-40, reprinted as Special Labor Force
Report 49.

Review, May 1970, pp. 3-12, reprinted as Special Labor Force Report 119.
2 Sophia C. Travis and Denis F. Johnston, “Labor Force Projections for 1970-80,”

Table 5. Effect of alternative fertility assumptions on projected total labor force of women 16 to 49 years old, by age,
1980, 1985, and 1990 ’
[In thousands]
1985

1980

19902

Sex and age group
Series
D

Series
E

Series
F

Series
D

Series
E

Series
F

Series
D

Series
E

Series
F

101,138

101,809

102,166

106,932

107,716

108,247

112,119

112,576

113,031

62,590

62,590

62,590

66,017

66,017

66,017

69,102

68,907

68,834

38,548
1,425
2,228
6,372
4,770
4,104
3,593
3,225
3,203
9,628

39,219
1,427
2,242
6,592
5,038
4,218
3,632
3,237
3,205
9,628

39,576
1,429
2,253
6,730
5,176
4,268
3,646
3,241
3,205
9,628

40,915
1,245
1,943
6,307
5,167
4,689
4,588
3,904
3,384
9,688

41,699
1,247
1,956
6,523
5,505
4,834
4,641
3,919
3,386
9,688

42,230
1,247
1,964

43,017
1,356
1,971
5,643
5,042
5,116
5,202
4,931
4,052
9,704

43,669
1,205
1,983
5,826
5,387
5,291
5,268
4,951
4,054
9,704

44,197
1,149
1,991
5,965
5,646
5,416
5,307
4,963
4,056
9,704

BOTH SEXES
Total, 16 years and over_________ _____________
MEN
Total, 16 years and over..............................................
WOMEN
Total, 16 years and over......... __............._..................
16 and 17 years.....................
......................... ..............
18 and 19 years_______ __________________ _____ ___
20 to 24 years......... ............................ ................ _.............. .
25 to 29 years________________ ____________________
30 to 34 years............. ...........................................................
35 to 39 years_______ ________________________ ____
40 to 44 years...........................................................................
45 to 49 years______ _____________ ________ _____ _
50 years and over_______________ ______ ____ _____ _

1 As currently defined by the Bureau of the Census in Current Population Reports,
Series P-25, No. 493, Series D implies an ultimate completed cohort fertility rate of
2/500, that is, 1,000 women would have, on average, 2,500 births throughout their
childbearing period. Series T. implies a corresponding rate of 2,100, and Series F
implies a rate of 1,800. The basic projections in this report are based on the Series E




6,686

5,743
4,920
4,668
3,927
3,387
9,688

population projections.
2 The differences in the projected male labor force in 1990 are due to differences
among the three series in the number of births projected for 1973 and 1974—cohorts
which would be 16 and 17 years old in 1990. The projected female labor force 16 and
17 in 1990 is similarly affected.

84

the participation rates of young men and the up­
ward trend for young women.
Workers in the 25- to 34-year-old group are esti­
mated to continue to increase in number during the
1980’s but at a much slower pace than in the
1970’s, reaching 29.7 million by 1985 and 30.5 mil­
lion by 1990. Moreover, this gain is expected to
occur primarily during the first half of the 1980
decade, with an annual average increase of 600,000
a year, in contrast to an increase of only 160,000 a
year, on average, between 1985 and 1990.
The prospects among workers 45 to 54 years old
imply a reversal of the trend foreseen for the cur­
rent decade— from an annual average decline of
nearly 60,000 in the present decade to an average
gain of 180,000 a year in the 1980’s. Meanwhile,
the smaller number of persons born in the 1925-34
period will be moving into the 55- to 64-year-old
age group, whose labor force numbers are therefore
expected to decline by nearly 50,000 a year, on
average.
Finally, the outlook for workers 65 and over dur­
ing the 1980’s is for a slow but steady increase in

number (20,000 a year), as the assumed continuing
decline in their participation rates is more than
offset by the continued rise in the underlying popu­
lation of older persons— from 24 million in 1980 to
25.9 million in 1985 and 27.8 million by 1990.
The sex distribution of the projected labor force
is not expected to change greatly in the 1980’s. The
proportion of workers who are women is expected
to rise from 38.5 percent in 1980 to 38.7 percent in
1985 and 38.8 percent in 1990. This stabilization
reflects primarily the changing age composition of
the working-age population during the decade, with
declines in the number of young women and very
small increases in the number of women 45 to 64.
years old— the two age groups whose participation
rates have been relatively high (chart 3).
Alternative projections

The alternative projections shown in table 5 de­
scribe the estimated effect of specified changes in a
single variable (fertility) upon the size and age-sex
distribution of the projected labor force.5 Table 5
shows the projected total labor force of women 16
to 49 years old, by age, for 1980, 1985, and 1990,
under three alternative assumptions concerning fer­
tility: Series “D,” “E,” and “F.” As is explained in
the following section on methodology, Series E
(defined as 2,100 births per 1,000 women) is the
series adopted for the basic set of projections in this
report; it represents a level of fertility whereby each
generation is barely replaced by the next one, so
that the population eventually stops growing (except
for immigration). Series D implies a higher fertil­
ity rate of 2,500 births per 1,000 women, while Se­
ries F implies a lower rate of 1,800 births per
1,000 women. Thus Series E is somewhat closer
to F than to D. In developing these alternative
projections, the assumed participation rates for
women with and without children under 5 years old
are the same for each series; the only difference
among the three series is the difference in the pro­
portions of the population of women with and with­
out children under 5. Series D implies a larger
proportion of women in each childbearing age group
with children under 5, while Series F implies a
lower proportion, with Series E falling in between.
The effect of these alternative fertility assump­
tions (ceteris paribus) can be illustrated by examin­
ing the 1980 projection. As noted previously, the
basic Series E projection yields a total labor force

Chart 2. Age-sex profile of total labor force, 1970 actual
and 1990 projected

|

j

1990 excess over 1970
1970 excess over 1990

AGE GROUPS
Men




Women

7.5

85

Chart 3. Labor force participation rates of women, by
age, 1960, 1980, and 1990

of 101.8 million. A shift to Series D has the effect
of reducing the female labor force (and thus the
total labor force) by about 670,000, while a shift to
Series F increases the labor force by about
360,000. Thus, the range of the projected variation
in the size of the labor force, as we move from Se­
ries D to Series F, amounts to about 1.0 mil­
lion, or 1 percent of the basic projection for 1980.
Among all women workers, however, that range
amounts to 2.6 percent of the basic projection, and
among women in the principal childbearing ages
(16 to 4 9 ), it amounts to 3.5 percent of the basic
projection.0
It should be noted, parenthetically, that the pre­
vious projections assumed continuation of the Series
C fertility levels (the level which approximates
the actual fertility rate of the mid-1960’s). Since
that time, fertility has declined to its present level,
which is close to Series E. On the basis of the
above calculations, the shift from Series C to Se­
ries E would account for an increase in the size of
the female labor force of about 700,000 in 1970.
Thus, the “error” in the fertility assumption alone
accounts for over one-third of the 1.9 million under­
estimate of the 1970 female labor force in the BLS
projections prepared in 1964.7

[Percent of total population in total labor force]

Percent of
population in
labor force

Methods and assumptions

The projections in this report reflect anticipated
changes in the demographic composition of the pop­
ulation of working age, combined with our judg­
ments as to the changes which might be expected in
the labor force participation rates of the several
age-sex groups in the population. The predominant
factor in these projections is the anticipated change
in the size and age-sex composition of the popula­
tion. The projections assume no drastic changes in
the propensity of the several population groups to
seek work. They also assume a generally favorable
demand situation, together with the absence of
major wars or other major social or economic dis­
turbances. Finally, the projections assume no major
legislative or social changes which would alter the
conditions under which individuals choose to enter
or remain out of the labor force, or which would
alter the prevailing definitions of “labor force,” “em­
ployment,” or “unemployment.” 8
The general approach is to extrapolate observed
trends in the participation rates of each age-sex
group to the terminal date of the projection (1990),




to to
19 24

to
29

to
34

to to
39 44

to to
49 54

to to to and
59 64 69 over

Age. groups

and to apply the projected rates to the projected
population to obtain the labor force. The major
steps in this procedure are as follows:
Step 1. Annual average rates of labor force participa­
tion (the percent of the total population in the total

86

labor force) were obtained for each year, 1955
through 1972, for men and women separately in the
following age groups: 16-17, 18-19, and 5-year
groups thereafter to 70 and over. By means of lin­
ear regression, the average annual change in the
participation rates of each age-sex group over the
1955-72 period was obtained. That average annual
change, times 5, was taken as representative of the
average observed 5-year change in the participation
rate of each age-sex group.

pation rates among women with children under 5
and those without children under 5 to the participa­
tion rates for all women in the specified age groups
were estimated and projected. The projected ratios
were then used to obtain projected participation
rates for women with and without children under 5,
by age, to 1990.
Step 5. The percentages of women in each age
group (16 to 49) who would have children under 5,
consistent with the fertility levels of the Bureau of
the Census’ Series C, D, E, and F projections of
population (as given in Current Population Reports,
Series P-25, No. 493) were estimated for the years
1975, 1980, 1985, and 1990. These percentages
were then applied to the projected total population of
women in these ages to obtain the number of women
with and without children under 5 for the target
years.

Step 2. Each of the observed 5-year changes in
participation rates was then gradually reduced by a
constant proportion for successive 5-year periods,
so as to reduce all changes to approximately zero in
50 years (that is, in 10 5-year periods). Such a
“tapering” of trends is designed to prevent the oc­
currence of future rates that might otherwise fall
outside plausible (or possible) limits. It also reflects
the assumption that each rate is moving toward some
asymptotic level which can only be defined arbi­
trarily. To accomplish this reduction, a constant
multiplier (M ) was applied to each observed average
5-year change to obtain the projected change over
the first projected period (1 9 7 0 -7 5 ). That change
was again multiplied by M to obtain the projected
change for the next 5-year period (1 9 7 5 -8 0 ), and
so on to 1985-90. For example, the largest observed
5-year change was —4.64 percentage points (among
males 65 to 6 9 ); the multiplier (M ) was assigned
a value such that 4.64 X M10 < 0.05. In this case,
M = .63. Similarly, the smallest observed 5-year
change (among women 65 to 69) was —.22; here,
the appropriate value for M is .84.

Step 6. The projected participation rates for women
with and without children under 5 (as obtained in
step 4) were then applied to the projected numbers
of women with and without children under 5 (by
age), yielding a projected labor force consistent with
the Series C, D, E, and F population projections.
Step 7. An analysis of recent trends in the fertility
of American women and of information relating to
the fertility expectations of young married women
led to the decision to adopt the Series E projec­
tions for the basic set of labor force projections. Ag­
gregating the projected labor force of women 16 to
49, by presence of children under 5, and dividing by
the corresponding population produced a final pro­
jected set of participation rates for all women 16 to
49 for the target years, consistent with Series E
population projections.

Step 3. The projected 5-year change for each age-sex
group, 1970-75, added algebraically to the 1969-71
average labor force participation rate for the speci­
fied group (used as a base) yields the projected par­
ticipation rate for 1975. Repeating this procedure
yields projected participation rates for 1980, 1985,
and 1990.

Step 8. On the assumption that changes in fertility
would not affect the participation rates for men or
for women 50 and over, the projected labor force
for these latter groups was obtained by multiplying
the projected population by the projected participa­
tion rates obtained in step 3.
□

Step 4. For women in the childbearing ages (16 to
4 9 ), the trends in the ratio of the observed partici­

-FOOTNOTEScolor or race, a category included in the earlier report, is
not yet available, and will be published in a forthcoming
report in 1974. The new projections are based on the Series
E projections of population, as given in the Current Pop-

1 These projections supersede those which were presented
by Sophia C. Travis in “The U.S. labor force: projections
to 1985,” M onthly Labor Review, May 1970, pp. 3-12, re­
printed as Special Labor Force Report 119. Information by




87

Determining the
labor force status
of men missed

Special Labor Force Report
describes pilot use of a
new technique for securing
labor force data
in urban poverty areas

in the census

DEBORAH P. KLEIN

R ecent attempts have been made to obtain
heretofore unavailable social and economic data
about men missed in the census—especially men
from minority groups between 20 and 50 years
of age who are estimated to have high rates of
undercount. The studies, which were conducted
in New Haven, Conn., Central Harlem in New
York, N .Y ., and Trenton, N.J., used a “casual
interview” technique. This approach consisted of
interviews in bars, poolrooms, restaurants, on
street comers, park benches, and similar
locations. This article discusses the results of the
new approach, which is one way to obtain more
extensive social and economic data for those
parts of the population that have been difficult
to folly enumerate in censuses.

Completeness of coverage varies by age and sex,
as well as by race and ethnic group. Coverage is
proportionately better for children than adults,
and better for females than for males. A relatively
large number of persons over 65 years of age were
missed. The highest rates of undercount were found
among men 20-34 years old and 50-54 years.
Women’s undercount rates are lower than men’s
at every age below 50, and it is possible that age
misstatement accounts for part of the undercount
for women at the older ages.
The census data provide benchmarks for pre­
paring monthly population estimates between cen­
suses. These estimates are used to weight the data
from the Current Population Survey (cps) which
provide monthly statistics on economic charac­
teristics of the population. Thus, any undercount
in the decennial census is transmitted to the intercensal statistics and may affect the reliability of
the published labor force data.
The Bureau of Labor Statistics tried to quantify
the possible effect of the undercount on national
unemployment rates. Two different assumptions
about the labor force status of uncounted persons
were used to analyze population data that had
been adjusted for the estimated undercount.8
Under the “comparability” assumption, missed
persons were assigned the same labor force status
as counted persons in the same age-sex-color group.
Under the “poverty-neighborhood” assumption,
missed individuals were assigned the characteris­
tics of persons living in^ urban poverty areas and
in the same age-sex-color cell.
Regardless of which assumption was used, the
resulting estimates of labor force size and employ­
ment were substantially larger when account was
taken of the missed persons. Distributions by age,
sex, and color changed only slightly, but the levels
were higher than those indicated by the published

B ackground: th e undercount

It is estimated that about 3 percent of the
population was missed in the 1960 census. All the
studies undertaken to estimate the number of
persons missed indicate that the undercount rate
(percent of persons missed) varies significantly by
race, age, and sex. The 1960 census enumerated
98 percent of white persons but only 90 percent of
persons of other races,1 according to Census
Bureau estimates.2 The total number of unenu­
merated persons has been estimated to be 5.7
million, of whom 38 percent were members of races
other than white. Thus, while the number of
uncounted white persons is greater than the num­
ber of uncounted persons of other races, the
proportion of white persons missed is considerably
smaller than the proportion of persons of other
races.
Deborah P. Klein is an economist in the Urban Employ­
ment Studies Group of the Office of Manpower and
Employment Statistics, Bureau of Labor Statistics.

From the Review of March 1970



88

obtain the names and addresses was *'‘casual inter­
views,” that is, interviews conducted in casual
settings such as bars, poolrooms, and on street
corners. A second source was lists obtained from
establishments, such as restaurants, laundries, and
hospitals, which often hire large numbers of lowpaid workers. Another aspect of the b l s research
project was to compare the effects of conducting
an undercount study in conjunction with a com­
plete census count, and conducting such a study
without a complete population count.
The pilot program was conducted in two areas—
the Negro poverty areas of New Haven, Conn.,
and the Central Harlem area of New York, N .Y .
New Haven was selected because it was the site of
a pretest of the 1970 decennial census. The b l s
research project was timed to follow shortly after
this pretest. In New York (which had had no
recent census), two sources—employers’ lists and
casual interviews—were used to obtain names and
addresses. The target populations in both areas
were Negro men between the ages of 20 and 50
years, because the estimated rates of undercount
were highest for this group.4
In New Haven, where casual interviews were
the only source of names and addresses, the Census
Bureau was able to check the names and addresses
obtained from the casual interviews against the
listing obtained in the census pretest. Followup
interviews were conducted at households where the
names and addresses obtained in the casual inter­
views could not be matched with records of that
census pretest. These interviews inquired about
the whereabouts of the individual in question. In
New York, the procedure called for a household
interview at every address obtained from either

figures. The national unemployment rate was not
appreciably different from the published one under
either assumption. I t would have required a very
high undercount rate, coupled with a grossly
higher unemployment rate among the uncounted
persons, for the national published unemployment
rate to have been significantly in error (table 1).
In some local areas, the undercount m ay con­
stitute a greater proportion of the population than
it does in the N ation as a whole. In these areas,
including the estim ated undercount might make a
significant difference in labor force data as well as
population data. I t has been suggested that
undercount rates are highest in crowded urban
poverty areas and sparsely populated rural areas.
Particular concern has been expressed about the
quality of the population and labor force estimates
for the N ation’s largest cities. A t the city level we
do not know what percentage of the population is
missed and what the characteristics of these
missed people are. The demographic analysis which
is used to obtain national estim ates has not been
done on a local level, primarily because adequate
birth, death, and migration rates are available only
on a national basis.

Bureau research on missed persons
The issue of severe unemployment in urban
poverty areas highlighted the fact that failure to
obtain information about all residents could sig­
nificantly affect the labor force data for these areas.
Consequently, the Bureau of Labor Statistics
designed a pilot research program to improve sta­
tistics for urban poverty areas. This program,
which included among its aims the gathering of
more information about persons not counted in
household surveys in these areas, was conducted in
the spring of 1967. The undercount portion of the
program, which complemented the previous work
of the Census Bureau in this area, was concerned
with identifying the labor force characteristics of
men not enumerated in household surveys, such
as the decennial census or the c p s .

Table 1. Effect on the unemployment rate of including
omitted persons under selected assumptions, by color
and sex, X967
Unemployment rate
Color and sex

The basic procedures of the b l s undercount
study were to obtain a set of names and addresses
through some source other than a household sur­
vey; to determine whether the individual would be
reported in a household survey; and then to com­
pare the characteristics of those individuals re­
ported by the household to the characteristics of
those who were not reported. One source used to




Official
estimate

Poverty neighborhood
assumption

Comparability
assumption
Omitted
persons

Adjusted
rates

Omitted
persons

White:
Both sexes..............
Male.......................
Female....................

3.4
2.7
4.6

3.4
3.1
3.7

3.4
2.8
4.6

5.1
4.9
5.7

3.4
2.8
4.6

Negro and other races:
Both sexes..............
Male.......................
Female....................

7.4
6.0
9.1

6.0
4.9
7.4

7.2
5.9
9.0

6.9
6.5
8.1

7.3
6.2
9.0

Source: Monthly Later teview, March 1969, tables on pages 10,11, and 12.

89

Adjusted
rates

the casual interview or the establishment lists. The
household interview was used to determine whether
the person would be listed in a household interview
such as the c p s , and to ascertain whether those
persons not listed as household members were part
of the undercount. This method revealed itself to
be considerably less effective than the method of
conducting such a study in conjunction with a
census count.
The New Haven study identified 39 cases of
persons missed in the census pretest. Obviously,
the number of cases was too small to permit
inferences about the characteristics of all un­
counted Negro men in urban poverty areas. The
results, however, were significant in providing some
insight, albeit inconclusive, into the social and
economic characteristics of the undercount, and
into a method which would increase identification
of the uncounted persons. (Even fewer cases were
found in New York.5)
The primary finding of the study was that the
labor force status of the undercount group was
very much like the labor force status of their
neighbors who were counted. (See table 2.) Two
significant differences between the enumerated
group and the undercount group support the
hypothesis that men are missed in census counts
because they do not have family responsibilities
and, as a result, frequently shift their places of
residence. The differences were: (1) the under­
counted group tended to have more casual
attachments to their places of residence; that is,
the proportion of those who had lived at their last
place of residence 1 year or less was nearly 4 times
higher for the undercount group than for the
enumerated group, and (2) a large proportion of
the undercount group had never been married. In
addition, the New Haven study suggested that
economic and social characteristics of under­
counted Negro men in urban poverty areas could
be identified through the technique of obtaining
names and addresses through casual interviews
following a complete census of the area.

The primary finding of the Trenton study
substantiated the tentative conclusions of the New
Haven study—the labor force status of men who
are not counted in a census is similar to that of
men who are counted. (See table 3.)
The Trenton Model Cities Agency used a
slightly revised version of the questionnaire
designed by b l s for use in the New Haven study.
The schedule covered the areas of educational and
marital status, age, place of birth, residential
history, labor force status, occupation, earnings,
and hours worked.
Unlike the New Haven study, in the Trenton
study there was no followup probe at addresses
which were unmatched in the census record. In
New Haven, there had been a complete census
count, a series of casual interviews, a matching of
names, and then a followup household interview.
The address of each person who had not been
enumerated in the census pretest was visited and
the respondent was asked about the individual in
question. If the respondent acknowledged that the
individual did live at the address, then that person
was considered to be part of the undercount. The
Trenton study omitted this followup household
interview. The Census Bureau classified all persons
whose names could not be matched with enumera­
tion lists and whose addresses were within the
enumeration district as persons missed in the
census pretest.
The Trenton survey was about twice as large
as the one in New Haven. Over 900 names and
Table 2. Comparison of selected characteristics of casual
interview respondents in New Haven
(Percent distribution]

Characteristics

Marital status:
Married..................
Separated.................
Widowed or divorced.
Never married..........
Information not
available...............
Years at residence:
1 year or less............
More than 1 year.......
Information not
available...............
Labor force status:
Employed.................
Unemployed.............
Unemployment rate..
Not in labor force___
Information not
available...............

Trenton undercount study
A recent study in Trenton, N .J., employed the
casual interview technique used in the b l s New
H aven Undercount Project already described.
The study was undertaken by the Trenton M odel
Cities Agency to gain additional information about
the situation of persons in poverty areas.




Number of responses.......

Total
Persons
Persons not matched
in census records
persons matched in
in
census records
casual (men counted
study in the census) Men found in Men not found
field followup
in field
(undercount)
followup
48
16
5
30

33
18
0
46

40
23
4
32

2

2

3

1

21
70

12
75

46
51

27
68

9

12

3

5

78
11
13
8

78
11
12
8

77
13
14
8

77
12
13
7

3

3

3

3

249

39

219

507

Source: BLS Report 354, pp. 25-26.

90

57
9
6
25

Table 4. Comparisons of selected characteristics of
casual interview respondents in Trenton
(Percent distribution]

addresses were obtained from casual interviews in
Trenton. These names were divided into three
groups. The first group consisted of 283 names that
were matched with the census lists; that is, men
who were enumerated in the census. The second
group consisted of 290 names that could not be
matched with census lists but whose addresses were
within the city limits. This group was considered to
be part of the undercount. The third group (350
names) contained persons whose enumeration
status was unclear; they may or may not have
been enumerated. Included in this group were
schedules that could not be matched because of
problems in address classifications and schedules
which arrived past the deadline for Census Bureau
checking. There were about 150 additional sched­
ules that could not be classified because their
addresses were outside Trenton city limits.
Duplicate schedules (which typically occurred
when more than one enumerator interviewed the
same individual) were also excluded from the
tabulations.
When the characteristics of the persons inter­
viewed in a casual setting and counted in the
census were compared with the characteristics
of the persons interviewed in the same area and
not counted in the census, the general finding was
that the two groups were very similar in regard to
labor force status. The unemployment rate for
the missed individuals was almost identical to
that of their counted neighbors. Furthermore, the
unemployment rate for the men whose enumeration
status was not known (the men for whom no
census match could be made) was about the same
as the others. The rate of nonparticipation in the
labor force was somewhat larger for the under-

Characteristics

Afe:
Less than 20 years..............
20-29.................................
30-39.................................
40-49.................................
50 or more.........................
Information not available....
Marital status:
Married.............................
Separated, widowed or
divorced.........................
Never married....................
Information not available....

Labor forct status

Employod...........................
Unemployed.......................
Unemployment rate___
Not in labor force...............
Information not available....

83
9
10
6
1

88
9
10
4
1

79
9
10
9
2

Number of responses..........

923

283

290

<•>

8
46
20
20
5
1

5
40
28
23
3
<•)

44

58

35

39

18
34
4

13
26
3

21
41
3

21
34
6

Years at residence:
Leu than 1........................
lor 2................................
3 or more...........................
Information not available....

13
22
62
3

9
20
69
2

14
25
58
4

15
21
59
4

Number of responses.................

923

283

290

350

Another characteristic in which the unenumerat­
ed and the uncounted were similar was educational
attainment. Among the men interviewed in the
Trenton study, about 30 percent of each group
had not attended high school and about 65 percent
had not graduated from high school.
Despite the similarity of the two groups in terms
of labor force status and educational attainment,
there were some characteristics in which they
differed. The uncounted group was somewhat
younger, less likely to be married, and more mobile.
(See table 4.) The differences in age distribution,
of course, affected the other characteristics.
Furthermore, the mobility aspects could reason­
ably be expected to affect enumeration. Young,
unmarried men are more likely to shift their living
arrangements, thus making it difficult to enu­
merate them.
The Trenton study added another dimension to
the undercount question; it was possible to tabu­
late the results by race and ethnic group. The
sample was approximately three-quarters Negro;
another fifth were persons with Spanish surnames
(primarily of Puerto Rican birth); the remainder
were other Caucasians, some Orientals, and a few
men of undetermined race. The Spanish surname
group was younger, and less likely to have ever
been married, than the Negro group. M ost of the

Men who
could not
be
classified
85
10
10
5
0)
350

i Less than 1 percent




6
28
24
36
6

count group than for the enumerated. However,
the rate was less than 10 percent for both groups.

[Percent distribution]

Men classi­ Men classi­
fied as
fied as
enumerated part of the
in the
undercount
census
pretest

6
38
24
26
5
1

Men who
could not
be
classified

>Less than 1 percent

Tabl« 3. Comparisons of labor forco status of casual
intorviaw respondents in Trenton

Total men
in study

Men
Men
classified
classified
Total
as
as
men in enumerated
part of
study
in the
the
census
undercount
pretest

91

men with Spanish surnames were born in Puerto
Rico; the Negroes were evenly divided between
those born in New Jersey and those born in a
southern State. The persons with Spanish sur­
names were newer to the area than the Negroes.
In the Trenton study, the unemployment rate
for Puerto Ricans was significantly higher than
the rate for Negroes. The differential was main­
tained for each of the classification groups,
although the extent of the differential varied. (See
table 5.)

percentages were 35 and 58, respectively. Length
of time at current residence is another variable
which may distinguish between the enumerated
and the unenumerated. In New Haven, 46 percent
of the undercount had lived at their current
residence 1 year or less, compared with 12
percent of the enumerated; in Trenton the rates
were 14 and 9 percent, respectively. While the
differences were greater in New Haven, they were
in the same direction as in the larger Trenton
study. The general finding seems to be that a
married man living with his wife at a stable
address is more likely to be reached in a household
survey than a single man who moves frequently.
The significant conclusion that can be drawn
from these studies is that the labor force status of
the uncounted is very similar to that of the counted
in the same urban poverty area. (See tables 2
and 3.) In New Haven, the unemployment rate
for the enumerated was quite close to that of the
uncounted; in the larger Trenton study, the rates
were virtually identical.
Equally important, from the standpoint of
evaluating published unemployment statistics, is
the implication that enumeration of all persons in
an urban poverty area would not significantly
change the unemployment rate for that area—and
perhaps this is true for other areas as well. If this
is true, it would provide greater credence to the
estimates of labor force size and employment pre­
pared by Johnston and Wetzel and discussed
earlier. (See table 1 and discussion on page 26.) The
findings of the studies in Trenton and New Haven
provide evidence that could support either the
comparability assumption or the poverty neighbor­
hood assumption. If similar studies were conducted
in nonpoverty areas, it might become apparent
which assumption is more valid.

Characteristics of the undercount
Both the Trenton and New Haven studies were
primarily methodological; that is, they were
designed to test the feasibility of using the casual
interview technique to collect data about persons
ordinarily missed in a census of an urban poverty
area. The data obtained from these surveys are not
sufficient to describe the characteristics of all men
living in urban poverty areas—counted or un­
counted in a census—because the data were
limited to two areas and we do not know whether
the casual interview technique reaches a represent­
ative sample of the local population. However,
some conclusions may be drawn about the relation­
ships between the characteristics of counted and
uncounted persons in urban poverty areas.
The significant social relationships deal with the
ties of counted and uncounted men to a particular
family and residence. In both New Haven and
Trenton, the major difference between the group
of men who would have been enumerated in a
census and those who would not was in the
strength of these ties. (See tables 2 and 4.) For
example, in New Haven, only 33 percent of the
undercount were married, compared with 57
percent of the enumerated. In Trenton, the
Table 5.

Labor force status of casual interview respondents, by race or ethnic group, Trenton

(Percent distribution]
Men with Spanish surnames

Negro men
Labor force status

Classified as
enumerated
in the census
pretest

Total

Classified as
part of
the undercount

Unclassified

Total

Classified as
enumerated
in the census
pretest

Classified as
part of
the undercount

Unclassified

Employed...................................................
Unemployed..................... ..........................
Unemployment rate................................
Not in labor force.........................................
Information not available..............................

85
8
9
5
1

88
6
7
5
1

81
9
10
8
3

88
9
9
3
0

75
17
18
9
1

80
19
19
1
0

73
12
14
16
0

70
18
20
10
1

Number of responses................ .................

695

203

231

261

188

70

51

67




92

Characteristics of the method
An evaluation of these studies indicates that
the technique of conducting casual interviews in
conjunction with a complete census count merits
serious consideration in any attempt to collect
data on missed persons. This data collection tech­
nique produced, for the first time, information
about the economic characteristics of men missed
in a census.
There are several advantages to the casual inter­
view technique. First, it can reach persons not
usually contacted in household surveys. Whether
the individual is missed because his entire house­
hold is not located, because he does not maintain a
a stable relation with any one household, or
because his household chooses not to acknowledge
his presence, the casual technique offers a prospect
of reaching him. Thus, this technique is suitable
for identifying the characteristics of persons sub­
ject to various types of undercount. Second, it can
be employed selectively; that is, it can be directed
to a specific group by the designation of the inter­
view locations and instructions to the interviewers.
Furthermore, the questionnaire can be designed
specifically for the selected group. For example,
the choice of language and the approach of the
enumerators can be tailored to fit the target popu­
lation. Third, the use of the casual interview
technique permits the enumerator to speak directly
to the desired respondent during the initial con­
tact. In household and other random surveys, on
the other hand, the initial contact is often made
with the wife, roominghouse owner, or other per­
son. When a follow up with the desired respondent
is not possible, the data that was obtained from
the secondary respondent is less reliable than data
from the desired respondent would have been.
Fourth, the casual nature of the questioning and
the relaxed atmosphere of the interview locations
may induce candid responses. There is some evi­
dence that this technique can obtain information
of a kind not readily available from household
surveys. For example, the New Haven study
obtained information about illegal activities that
had not been available from regular household
surveys. Fifth, the technique is a relatively in­
expensive method of obtaining a large number of
responses in a short period of time. The elimination
of callbacks to locate specific individuals resulted
in a lower cost per schedule than in household
interviews.




93

A major disadvantage of the casual interview
technique is that it does not provide a sample with
a scientifically delineated universe. This makes it
difficult to establish the representativeness of the
survey findings. This objection is partially blunted
when the survey is done in conjunction with a
complete census count. Under these circumstances,
the individuals reached were members of the
census universe, although not necessarily a random
sample of this group. Despite this objection, the
advantages of selectivity, direct access, candid
responses, and low expense appear to make this
technique a useful tool for determining the
characteristics of the undercount.
There is no set requirement for the type of
enumerator to use for casual interviews. In New
Haven, all of the interviewers were men experienced
in field work and familiar with the area of enumera­
tion. In Trenton, the interviewers were young men
and women of various ages with some survey
experience. Both male and female enumerators
were successful. Although experience with using
nonindigenous interviewers in these situations is
limited, a strong case could be made for the use of
interviewers who are indigenous to the area.
Variation in the hours of enumeration served to
prevent labor force bias. It appeared best to
interview during day and evening hours, and over
the weekend where that is possible.
The samples in both New Haven and Trenton
were not designed to be representative of the city
as a whole but rather of specific areas—minority
group poverty areas. The enumerators were
instructed to interview men from minority racial
or ethnic groups between the ages of 20 and 50.
This group was selected because of undercount
rates estimated to be very high. In New Haven,
the 500 men were primarily Negro; in Trenton the
900 men were primarily Negro and Puerto Rican.
The data indicate that in each city about 10
percent of the men had ages outside of the bound­
aries set. However, this percentage was substan­
tially lower than it would have been had there
been no attempt to restrict the sample.
In each city, the casual interviews were con­
ducted in poor areas, and the sites were such places
as bars, restaurants, poolrooms, street corners, and
park benches. In New Haven, this was done to
increase the percentage of unemployed and
marginally employed men (working in low-skilled,
low-paying jobs) because it had been suggested
that these men constituted a disproportionate

share of the undercount, whose characteristics
were the focus of these studies. In Trenton, the
sections of the city where interviews were con­
ducted yielded a similar sample of men.
Thus, any differences between the men in each
sample and the total population of their city would
reflect the method of sample selection and would
have no necessary correlation with the social and
economic distribution of the undercount or the
population of that city. However, the character­
istics of the sample group are not atypical of other
samples that have been drawn from urban poverty
areas.8
The small sample size and the restricted nature
of the selection process have precluded the drawing
of any definitive conclusions about the characteris­
tics of all persons not counted. We have no way of
knowing whether the characteristics of unenumer­
ated men reached through the casual interview
technique are typical of the entire undercount.
There are two reasons for this uncertainty. First,
the characteristics of the undercount in other

geographic areas, economic strata, or age groups
may be very different from the characteristics of
the undercount in an urban poverty area. Second,
even within an urban poverty area, the technique
of casual interviews may not reach all of the
undercount. For example, there may be some men
who never go to bars or stand on street comers.
However, the quality of the findings that have
been made thus far suggests that additional
studies should be undertaken. The question now
is whether the insights thus far obtained from
studying the undercount among minority groups in
urban poverty areas would be supported in similar
or dissimilar studies of other groups in other areas.
Wider application of the method described above
may bring us closer to obtaining a better definition
of the characteristics of the undercount, better
understanding of the reasons for the undercount,
insight into techniques that might reduce the
extent of undercount, and a better appreciation of
published data that is affected by
the
undercount.
□

■FOOTNOTES
1 R efers to N egroes, O rientals, and A m erican In d ian s.
N atio n w id e, N egroes m ak e up ab o u t 92 p ercen t of races
oth er th a n w h ite, an d a higher proportion in urban povertyareas.

and E li S. M arks and Josep h W aksberg, “ E v a lu a tio n of
C overage in th e 1960 C ensus of P op u la tio n th rou gh C aseb y-C ase C h eck in g,” in D a v id M . H eer, ed ., op. cit.
* S ee D e n is F . J o h n sto n and J a m es R . W etzel, “ E ffect
of th e C ensus U n d ercou n t on L abor F orce E stim a te s ,”
Monthly Labor Review, M arch 1969, p p . 3 -1 3 .

* F or sources of e stim a tes an d m ore d etail, see Ja co b S.
S iegel, “ C om p leten ess of C overage of th e N o n w h ite P op u ­
la tio n in th e 1960 C ensus and C urrent E stim a te s, and
S om e Im p lic a tio n s,” in D a v id M . H eer, ed ., Social Statistics
and the City (C am bridge, M ass., J o in t C enter for U rban
S tu d ies of M a ssa ch u setts In s titu te of T ech n o lo g y and
H arvard U n iv e r sity , 1968). A sum m ary of th e m eth od s
u sed to e stim a te th e e x te n t of th e un d ercou n t w ill be found
in b l s R ep ort 354, Pilot and Experimental Program of the
Urban Employment Survey. F or a m ore d etailed descrip tion ,
see Jacob S. Siegel and M elv in Zelnik, “ An E v a lu a tio n of
C overage in th e 1960 C ensus of P op u lation b y T ech n iq u es
of D em ograp h ic A n alysis and by C om p osite M e th o d s,”
1966 Proceedings of the Social Statistics Section, American
Statistical Association; L eon P ritzker and N . D . R o th w ell,
“ P rocedural D ifficu lties in T ak in g P a st C ensuses in Pre­
d o m in a n tly N egro, P u erto R ican , an d M exican A reas,”




4 For a d etailed d escrip tion of th is research, see
R ep o rt 354, cited in fo o tn o te 2.

bls

* In N e w Y ork on ly th ree cases of u n d ercou n t w ere id en ti­
fied. B ecau se of th is sm all num ber an d b ecau se of th e large
num ber of u n located addresses, m ean in gfu l com p arison s
b etw een fou n d and m issed persons cou ld n o t be m ade.
T here w as considerable d ifficu lty in lo c a tin g ap a rtm en t
d w ellers in th e m u ltiu n it te n a m e n ts w ith poor or n on ­
ex iste n t te n a n t id en tification ty p ic a l of th e p o v e r ty areas
in N ew Y ork and oth er large cities.
6 T ab les p rovid in g d eta iled d a ta on th e ch aracteristics
of th e resp on d en ts in th e T ren ton stu d y are a v a ila b le from
th e B ureau of L abor S ta tistic s.

94

Discouraged
workers and
changes in
unemployment

First time series analysis of data from the
Current Population Survey indicates
the number of discouraged workers
rises as unemployment increases
PAUL 0. FLAIM

U

n t i l a couple of decades ago, the many millions
of working-age persons outside the labor force were
of limited concern to labor economists and policy­
makers, either as a potential source of manpower or
as a possible threat to the stability of the job market.
It was then the general assumption that the Nation’s
labor supply consisted only of persons actually work­
ing or actively seeking work. The notion that many
persons outside the labor force might have wanted
work but were not seeking it because they believed
that their search would be fruitless was not widely
entertained.
This popular concept of the labor supply was
probably relevant in the 1930’s, when the ranks of
the unemployed contained an apparently inexhaus­
tible reservoir of. manpower. It had to be gradually
abandoned, however, as evidence accumulated dur­
ing the post-World War II period showed that mil­
lions of persons entered and left the labor force
each year, not only because of personal reasons but
also in apparent response to changing labor market
conditions.
Recognizing these facts, the President’s Committee
to Appraise Employment and Unemployment Sta­
tistics (more familiarly known as the Gordon Com­
mittee) stated in 1962 that “the relatively simple
dichotomy between those in and out of the labor
force . . . [no longer provides] . . . a satisfactory
measure of the labor supply.” The Committee went
on to recommend that special efforts be made,
through the Current Population Survey (CPS), to
collect detailed data on persons not in the labor
force, particularly on the so-called “discouraged
workers” or “hidden unemployed”— those persons
who want work but are not looking for a job be-

cause of a belief that their search would be in vain.
In so doing, it should be noted, the Committee also
recommended that these persons n o t be included in
the unemployment count.
In 1964-66, following the recommendation of
the Gordon Committee, the Bureau of Labor Sta­
tistics began to experiment with a special set of
survey questions designed to elicit detailed infor­
mation on the reasons persons outside the labor
force did not participate in the job market. In Janu­
ary 1967, these questions were incorporated into the
regular CPS questionnaire. The data which they have
yielded have been published quarterly since late
1969 in a special set of tables in the monthly BLS
periodical, E m p lo y m e n t a n d E arnings.
The earlier analyses of these findings were, by
necessity, limited to cross-sectional examinations
done in snapshot fashion. Obviously, no time-series
analysis could have been undertaken until a number
of years had elapsed. Moreover, the first 3 years of
data were collected in a period of very low unem­
ployment, so that one could hardly draw any con­
clusion about their cyclical sensitivity.
Since the data have now been accumulated for 6
years— the last 3 years being a period in which vast
cyclical changes took place in the Nation’s economy
— it is possible to determine, at least tentatively, to
what extent workers will refrain from entering the
job market or may be induced to leave it because of
rising unemployment. Two variables are of particu­
lar interest for this purpose: (1) the number of “dis­
couraged workers,” and (2) the number of workers
leaving the labor force because of “slack work,”
who may or may not wind up as “discouraged
workers” under current definitions.

Paul O. Flaim is an econom ist in the D ivision o f E m ploym ent and U nem ploym ent A nalysis, Bureau o f Labor Statis­
tics. This article is based on a paper presented at the
August 1972 m eeting o f the A m erican Statistical A ssociation in M ontreal, Canada.

Identifying those discouraged

From the R eview of March 1973



Determining the extent of discouragement over
job prospects is a very difficult task. It involves the
95

prevent them from taking a job.
It is also important to note that separate data are
collected and published, from the same survey, on
the reasons for leaving the last job for those persons
who have recently left the labor force. As will be
discussed later, these “flow” data are an important
adjunct to the figures on discouraged workers in
terms of understanding the dynamics of the labor
force under changing economic conditions.

measurement of what are essentially subjective phe­
nomena, specifically one’s desire for work and one’s
perception of his or her chances of obtaining a job.
The pinning down of these “states of mind” is ren­
dered particularly uncertain by the fact that the
housewife is typically the only person interviewed in
each CPS household, and she must answer for all
members of the household.
Even if interviewed individually, however, some
persons may still not always admit their “real”
reason for leaving the labor force. It is possible, for
example, that some, although having been unable to
find a job, may attribute their nonparticipation status
to ill health or other “socially acceptable reasons”
rather than admit that they have failed in the job
market. Conversely, there may be some who indicate
that they want a job and who then explain their fail­
ure to look for one in terms of unavailability, even
though their desire for work is actually of very
limited intensity. Given the subjective and elusive
nature of “discouragement,” the extent of its pos­
sible overstatement or understatement cannot be
measured.
In order to identify the discouraged workers, the
CPS interviewer asks first if the persons not in the
labor force “want a regular job now, either full time
or part time.” If the answer is yes, or even a tenta­
tive yes, there is a follow-up question as to the
reasons they are not looking for work. In order to
be classified as discouraged, a person’s principal
reasons for not looking for work must fall into one
of the following five categories:
1.

How many discouraged workers?

The first examinations of the data on persons not
in the labor force, based on 1967-68 findings,
showed that less than one-tenth of these persons pro­
fessed any desire to be holding a job.1 Among these,
only about 700,000 were classified as discouraged—
that is, as not looking for work because of a belief
that they could not find a job. As shown in table 1,
the other nonparticipants reported as wanting a job
turned out either to be in school, in poor physical
condition, or prevented from seeking work by house­
hold responsibilities. The ranks of the 700,000 dis­
couraged workers, furthermore, were found to con­
tain relatively few men of prime working age— less
than 200,000. The great majority of discouraged
consisted, instead, of teenagers, housewives, and
elderly persons. These findings seemed to run counter
to the contentions that there were virtually millions
of discouraged workers and that they included large
numbers of men.2
However, the data being analyzed in the late
1960’s had been collected in a period of unusually
low unemployment, when the jobless rate held below
4 percent. Not until 1970, when unemployment rose,
could the relationship between changes in the unem­
ployment rate and in the number of discouraged
workers be discerned.
That the rise in unemployment in 1970 produced

B e liev es n o w ork a v a ila b le in lin e o f w ork o r area;

2.

H a d tried but c o u ld n o t find w ork;

3.

L ack s n ecessa ry sc h o o lin g , train in g, skills, or e x ­
perience;

4.

E m p lo y ers th in k to o y o u n g or to o old;

5.

O th er p erson al h a n d ica p in fin d in g a job.

It may be argued that this screening process, par­
ticularly the requirement that a person must first be
reported as wanting a job in order to be questioned
about possible discouragement, yields a rather re­
strictive definition of hidden unemployment. What
about those persons, one might ask, who, upon losing
their job, may decide to return to school and who
would then not want a job “now”? Should they not
also be regarded as discouraged workers? To answer
this, it should be noted that if the discouraged work­
ers’ data are to be useful as a measure of underutili­
zation of manpower for policy purposes, they should
hardly include persons who do not want a job,
especially when their current activity may actually




‘Hidden unemployment’

In this issue, three articles and a bibliography
deal with “hidden unemployment,” a problem
that is also referred to as a “manpower gap,”
or “discouraged workers.” The authors recog­
nize that hidden unemployment may mean dif­
ferent things to different researchers. Conse­
quently, each sets forth what the concept means
within the context of his analysis.
96

Table 1.

Distribution of persons not in the labor force, by reason, 1967-72

[Numbers in thousands]

Labor force status

1967

1968

1969

1970

1971

1972

Civilian noninstitutional population.......... ........................................................................
In civilian labor force...... ............................................................................................
Not in the labor force........ ..........................................................................................

129,873
77,347
52,484

132,026
78,737
53,289

134,335
80,734
53,596

136,995
82,715
54,275

139,775
84,113
55,662

143,325
86,542
56,783

Do not want job now, total.................................................................................
1n school...... ..................................................................
Ill, disabled.....................................................................
Homemaker.....................................................................
Retired, old.....................................................................
Other................................................................................

47,787
5,641
3,741
31.239
5,313
1,853

48,810
5,892
3,684
31,667
5,540
2,027

49,137
5,958
3,826
31,384
5,795
2,174

50,396
6,051
3,869
32,162
5,918
2,396

51,259
6,373
4,077
32,203
6,160
2,446

52,321
6,301
4,313
32,384
6,691
2,632

Want a job now, to ta l.............. ..................................... ....................... ...........
Reason not looking: School attendance......... ....... ..................................
Ill health, disability.................................................
Home responsibilities...............................................
Think cannot get job............................. ..................
All other reasons......................................................

4,698
1,104
768
1,325
732
769

4,477
1,115
656
1,263
667
777

4,459
1,126
627
1,257
574
875

3,877
1,075
489
926
638
749

4,404
1,242
555
1,020
774
813

4,461
1,200
632
1,098
765
766

Current activity:

NOTE: Because of separate computation, the figures on the civilian labor force
and on persons not in the labor force may not in all cases add up precisely to the

civilian noninstitutional population.

at least a temporary slackening in labor force par­
ticipation is now a historical fact. The slackening
was most evident in the first half of 1971, when the
labor force hardly grew at all. The question is the
extent to which this slackening in participation can
be attributed to discouragement over job prospects
caused by the rise in unemployment.
As chart 1 shows, there is, indeed, a positive rela­
tionship between the unemployment rate and the
number of discouraged workers. Both series trended
downward, though in differing degrees, during the
1967-69 period; both rose substantially during 1970;
both showed little distinct movement during 1971;
and both moved downward during 1972. In terms
of the actual number of persons involved, however,
it should be noted that the 1969-71 increase in the
number of discouraged workers was relatively small
when compared with the rise in the number of job­
less persons. While the number of unemployed rose
by 2.2 million between 1969 and 1971 (on an annual
average basis), the number of discouraged workers
increased by only 200,000.
Despite the positive relationship between unem­
ployment and discouragement, the two series did not
correlate very highly with each other. The coefficient
of correlation between these two variables, derived
on the basis of seasonally adjusted monthly data for
the 1967-71 period,3 was only 0.53. Nor was the
coefficient raised when the relationship between the
two series was tested on the basis of data disaggre­
gated by age, sex, and race. (Correlation and re­
gression results are shown in appendix table 1.)
Since it may be reasonably assumed that changes

in the number of discouraged workers lag behind
the changes in the unemployment rate, some experi­
mentation with lags was also conducted. By lagging
the discouraged workers’ series by 3 and also by 6
months behind the unemployment rate, the coeffi­
cients of correlation were raised somewhat— to 0.61
in both cases—but were still far from indicating a
very close relationship between the two variables.




‘Cyclical' vs. 'structural' discouragement

A closer examination of the disaggregated data on
discouraged workers for the 1967-71 period re­
vealed a significant change in composition in terms
of the specific reason cited by these persons for their
belief that they could not obtain a job. Specifically,
there was an increase in the proportion of workers
whose discouragement appears to have been directly
related to the changing conditions of the job market.
Conversely, there was a decline in both the number
and proportion of persons attributing their discour­
agement to personal situations or deficiencies.
Table 2 groups discouraged workers into these two
broad categories. The first included the workers re­
ported as believing that there were no jobs in their
line of work or area and those who had tried unsuc­
cessfully to find a job before giving up the search.
The second category includes those workers reported
as thinking they could not get a job due to their very
young or advanced age, those who saw their lack of
education or training as the major obstacle, and those
who cited other personal handicaps, Such as lan­
guage difficulties.
97

It would appear, given the different nature of the
reasons for discouragement, that the first category
of discouraged workers should be much more cycli­
cally sensitive than the second. Discouragement
among the second category appears to be more of a
“structural” nature and thus not necessarily related
to the tightness, or looseness, of the job market. The
data in table 2 confirm this hypothesis. As shown, all
of the 200,000 increase in the number of discouraged
workers between 1969 and 1971 took place among
those blaming their situation on job-market weak­
nesses.

Table 2. Composition of discouraged workers by reason
for believing they cannot find a job, 1967-72
[Numbers in thousands]
Reason

Chart 1. Unemployment rate and number of discouraged
workers, 1967-72

In thousands

900 ---------

1967

1968

1969

1970

1971

1972

Total..............................

732
383

667
371

574

Job-market factors.....................
Had looked but could not
find job...........................
Thinks no job available.......
Personal factors.........................
Employers think too young
or too old........................
Lacks education, skills,
training...........................
Other personal handicap___

311

638
437

774
537

765
540

168
215
349

161
210
297

161
150
263

244
193
201

300
237
236

300
240
226

216

171

139

105

112

111

84
49

74
52

78
46

60
36

85
39

78
37

Percent distribution.......... 100.0
Job-market factors..................... 52.3
Had looked but could not
find job............................ 23.0
Thinks no job available....... 29.4
Personal factors......................... 47.7
Employers think too young
or too old........................ 29.5
Lacks education, skills,
training.................... ...... 11.5
Other personal handicap___
6.7

100.0
55.5

100.0
54.2

100.0
68.5

100.0
69.5

100.0
70.6

24.1
31.4
44.5

28.0 38.2
26.1 30.3
45.8 . 31.5

38.8
30.7
30.5

39.2
31.4
29.5

25.6

24.2

16.5

14.5

14.5

11.1
7.8

13.6
8.0

9.4
5.6

11.0
5.0

10.2
4.8

NOTE: Because of rounding, sums of individual items may not equal totals.

Correlation analysis also lent support to this hy­
pothesis. Whereas, as noted above, the total number of
discouraged workers did not correlate highly with the
overall unemployment rate, yielding a coefficient of
only 0.53, the number of workers discouraged be­
cause of job market reasons yielded a much higher
correlation coefficient — 0.79 — when regressed
against the unemployment rate.
On the other hand, when the number of persons
whose discouragement seemed to hinge on personal
factors was regressed against the unemployment rate,
the result was a negative coefficient of correlation—
—0.47. There is no ready explanation for this nega­
tive relationship, but some possibilities may be raised.
For example, the passage of legislation designed to
reduce job discrimination because of age may have
accounted for a downward trend in the number of
elderly workers who thought that they could not get a
job due to their advanced age. It may also be hypoth­
esized that when the unemployment rate rises, some
workers who had previously been attributing their
discouragement to personal reasons may then attrib­
ute their situation to the deteriorating job market.
It is clear, nevertheless, that if we limit our analy­
sis to the group of discouraged workers who link
their situation to the conditions of the job market,
we find that their number did increase and decrease

Number discouraged
because Qf personal
reasons 1

0 L I I I 1 I I I 1 I I I 1 I I I I I I 1 1.1 1,1
Percent




98

in line with the underlying movement of the unem­
ployment rate during the 1967-71 period.
Unexpected discontinuity

One of the findings from the 6 years of experience
in obtaining statistics on labor force nonparticipants
is that it apparently makes quite a bit of difference
whether the questions about current desire for work
and future jobseeking plans are asked in the first
month in which they are visited by the CPS inter­
viewer or in subsequent months.
Since a person’s reasons for nonparticipation in
the labor force are not likely to change from one
month to another, this information is asked in only
one of the four consecutive monthly interviews con­
ducted in households falling in the CPS sample.
From 1967 through 1969, the questions were asked
in the month in which a given household first entered
the CPS sample and then again 1 year later, when
the same household reentered the sample for the
second and final 4-month stint after an 8-month
hiatus. In January 1970, the questions were switched
from the first and fifth month-in-sample to the fourth
and eighth. In effect, instead of being asked when a
household enters or reenters the sample, they are
now being asked only when the household leaves the
sample.4
This switch turned out to have a noticeable effect
on the data on persons not in the labor force. Fol­
lowing the switch, proportionately fewer persons,
particularly among the housewives, were reported as
either wanting a job at present or as planning to look

for work in the near future. Evidently, having be­
come increasingly more at ease with the interviewer
with each passing month, a respondent is less likely
to exaggerate his (or her) attachment to the labor
force in the fourth monthly interview than in the
first one.
As far as the data on discouraged workers are
concerned, the switch appears to have caused a small
drop in the number and proportion of persons attrib­
uting their discouragement to personal reasons (a
factor which no doubt contributed to the negative
relationship between this variable and the unemploy­
ment rate). Although this discontinuity did not have
a great effect on the overall numbers, it is a good
illustration of the difficulties which arise in the meas­
urement of what are essentially attitudes on the part
of workers or potential workers.
Profile, 1972

As was the case during the first years for which
data on discouraged workers are available, the pro­
portion of men of prime working age among this
group is still relatively small. Of the 765,000 persons
classified as discouraged workers in 1972, only about
70,000, or less than one-tenth, were men 25 to 59
years of age. (See table 3.)
Blacks are even more overrepresented among the
discouraged workers than they are among the un­
employed. They made up only one-ninth of all the
persons of working age outside the labor force but
one-fourth of the discouraged workers in 1972.
(Blacks also make up one-ninth of the civilian labor

Table 3. Discouraged workers by time elapsed since last job and jobseeking intentions, 1972

Total
discouraged
(In thousands)

Percent distribution by time elapsed since last lob
Total

Less than
1 year

1 to 5
years

More than
5 years

Never
worked

Percent who
Intend to
seek work
within
12 months

Total 16 years and over................................................... ......

765

100.0

36.3

30.6

19.0

14.1

77.1

Male, 16 years and over........................................................ ..........
16-19 years...............................................................................
20-24 years...............................................................................
25-59 years............... ..............................................................
60 years and over.....................................
..............

240
64
33
67
75

100.0
100.0
100.0
100.0
100.0

45.0
40.0

30.0
12.3

9.6
1.5

15.4
46.2

77.1
84.4

0

55.2
37.3

0

32.8
42.7

0

9.0
18.7

0

3.0

83.6
61.3

Female, 16 years and over................................................................
16-19 years........................................
20-24 years...............................................................................
25-59 years...............................................................................
60 years and over......................................................................

525
68
79
299
79

100.0
100.0
100.0
100.0
100.0

32.4
35.3
44.3
29.8
28.8

30.9
10.3
32.9
33.4
37.5

23.2

13.5
54.4
20.3
5.4
3.8

77.0
82.4
89.9
77.9
55.7

White*.................
Negro and other races *

578
188

Sex, age, and color

* Breakdown of discouraged workers in terms of time elapsed since last job and
future job-seeking intentions is not available separately for whites and Negroes.

1 Percent distribution not shown where base is less than 50,000.




3.8
32.1
30.0

0

99

case, are persons other than those to whom the ques­
tions relate. The “flow” data which these questions
produce should thus be subject to less response error
than those on the actual number of discouraged
workers.
As shown in table 4, the volume of the gross flows
out of the labor force has not changed much in recent
years, averaging close to 10 million. This would
indicate that the cyclical changes in labor force
growth during this period have stemmed primarily
from fluctuations in the in-flow of new entrants and
reentrants into the job market. There have been,
however, some cyclical changes in the composition
of the out-flows by reasons for leaving last job.

force and account for about one-fifth of the unem­
ployed.)
In terms of previous work history, nearly (two-fifths
of the discouraged workers had been out df the job
market less than 1 year when interviewed. Only 14
percent had never worked before. These findings,
however, differ by age and sex.
Evidently, most discouraged workers regard their
status as only temporary. Although they do not deem
it worthwhile to look for a job at the time of the
interview, they are apparently more hopeful in terms
of their future prospects. Nearly 80 percent of the
total were reported as planning to actively seek work
within the next 12 months. It would thus be errone­
ous to assume that most discouraged workers have
permanently given up on the job market.

Chart 2. Unemployment rate and percent of persons
leaving labor force because of “ slack work," 1967-72

Examining out-flows

While the data on discouraged workers may not
fully explain how cyclical changes in the employment
situation affect the dynamics of the labor force, other
data gathered through the same survey shed addi­
tional light on this phenomenon. For example,
through the special set of questions asked of the non­
participants since 1967, it has been possible to group
those with recent work experience according to their
reasons for leaving their last job—regardless of
whether or not they are currently counted as dis­
couraged workers.
Unlike the questions designed to identify the dis­
couraged workers, those designed to determine when
and why a person left his last job deal with facts
which are of a more overt, observable nature. As
such, these questions should present fewer problems,
particularly when the respondents, as is often the

Percent

Percent

Table 4. Persons exiting from the labor force, by reason
for leaving last job, 1967-72
Reason for leaving
job during previous
12 months

1967

1968

1969

1970

1971

1972

Number (in thousands)..................
Percent........................

9,327
100.0

9,752
100.0

10,175
100.0

10,130
100.0

10,098
100.0

9,624
100.0

School, home responsi. bilities______________
III health, disability...........
Retirement, old age_____
Economic reasons............ .
End of seasonal job...
Slack work_______
End of temporary job.
All other reasons............

49.2
9.5
5.3
17.1
9.2
3.3
4.6
18.9

50.3
9.2
6.0
17.8
9.1
3.1
5.6
16.7

50.5
9.6
6.1
16.6
8.5
3.1
5.1
17.2

49.3
8.9
6.7
18.0
8.1
4.3
5.7
17.2

47.7
8.7
7.4
19.5
8.5
5.2
5.8
16.7

46.8
9.1
8.1
19.3
8.6
4.9
5.8
16.7




100

as a major brake against the lowering of the unem­
ployment rate.
□

Specifically, there was an increase between 1969
and 1971 in the proportion of persons attributing
their exit from the labor force to the fact that their
jobs had been terminated, either temporarily or per­
manently, due to economic reasons. Of the three
categories under the “economic” heading, “slack
work” appears to have been most cyclically sensi­
tive. As illustrated on chart 2, the changes in this
variable have been closely related to the changes in
the unemployment rate.

----------FOOTNOTES---------1 See Robert L. Stein, “Reasons for Nonparticipation in
Labor Fcvce," M onthly Labor Review, July 1967, pp. 22-7,
and Paul O. Flaim, “Persons not in the labor force: Who
they are and why they don’t work,” M onthly Labor Review,
July 1969, pp. 3-14.

This relationship was also tested through regres­
sion and correlation analysis. The coefficient of cor­
relation between the overall unemployment rate and
the number of persons reporting they had left the
labor force after having lost their jobs due to slack
work was 0.83 on the basis of monthly data for the
1967-71 period. The substitution of data on unem­
ployment due to job loss for the overall measure­
ments of unemployment yielded coefficients of
roughly similar magnitude. (See appendix table 2.)

Summary and conclusion
After 6 years of experience in the collection of
data on discouraged workers through the Current
Population Survey, it appears that this survey is,
indeed, a viable vehicle for such a purpose. Al­
though the definition of “discouragement” used for
the purposes of the survey might not be universally
agreed upon, the data gathered so far have shed
important light both on the discouraged-worker phe­
nomenon and other aspects of labor force dynamics.
While 6 years of data may not be sufficient to
enable us to establish with any certainty the rela­
tionship between two variables, the hypothesis that
changes in the number of discouraged workers are
closely related to changes in the unemployment rate
can now be verified at least tentatively. The same
can also be said for changes in the number of work­
ers leaving the labor force because of slack work.
To the extent that this is true, it would appear that
we should take into account these variables, as well
as the data on unemployment and underemployment,
when assessing the waste of manpower which ac­
companies an economic recession. In the most recent
slowdown, however, the increase in the number of
discouraged workers, as currently defined, was rela­
tively small when compared with the magnitude of
the changes in unemployment. That being the case,
it would be unreasonable to assume that the return
of these workers to the job market as economic con­
ditions improve could be of such magnitude as to act




101

2 These contentions were based largely on econometrically
derived estimates of hidden unemployment published in
various journals during the mid-1960’s. Among the first to
construct such estimates were Alfred Telia and Thomas
Dernbu.g and Kenneth Strand. See A. Telia, “The Relation
of Labor Force to Employment,” Industrial and Labor Rela­
tions Review, April 1964, pp. 454-69, and T. Dernburg and
K. Strand, “Cyclical Variation in Labor Force Participa­
tion,” Review of Economics and Statistics, November 1964.
Many other economists, using a variety of econometric tech­
niques, have since undertaken similar research. Essentially,
they have attempted to measure the elasticity of labor force
participation rates, especially for women and youth, in re­
sponse to the intensity o f the demand for labor as reflected
by the unemployment rate, the wage rate, and other varia­
bles. Optimal participation rates, those consistent with con­
ditions of “full employment,” were then applied to the
population to obtain a “full employment labor force.” To
the extent that the actual labor force, as measured through
the Current Population Survey, fails to match this theoretical
labor force, they ascribe the gap to the discouraged workers
phenomenon or hidden unemployment.
For an analytical discussion of the early econometric esti­
mates of “hidden unemployment” or “discouraged workers,”
see Jacob Mincer, “Labor Force Participation and Unem­
ployment: A Review of Recent Evidence” in R. A. Gordon
and M. S. Gordon, eds., Prosperity and Unemployment
(N ew York, Wiley, 1966), pp. 73-112. For a comparison
of the more recent econometric estimates with the survey
data presented in this article, see the discussions by Joseph

Gastwirth and Jacob Mincer elsewhere in this issue.
3 Although the not-in-the-labor-force data are published
only on a quarterly basis, they are being tabulated monthly.
They have also been seasonally adjusted experimentally,
although not yet regularly published in this form.
4 The switch was instituted in an attempt to determine
whether these added questions were having an effect on the
so-called “first-month bias” in the unemployment figures.
It had long been evident that the reported incidence of job­
lessness in households entering or reentering the CPS sample
was higher than in households which had been in the sample
for 2 consecutive months or more. This “first-month bias”
became even larger around 1967, and it was hypothesized
that the increase might have been related to the introduction
of the not-in-the-labor-force questions. The reduction in the
reported incidence o f unemployment for the first and fifth
month-in-sample groups and concomitant rise for the fourth
and eighth following the January 1970 switch of the not-inthe-labor-force questions seems to have amply confirmed
this hypothesis.

APPENDIX

In order to determine quantitatively to what ex­
tent workers may refrain from entering the labor
force or may be induced to leave it because of rising
unemployment, recourse was made to regression
analyses. A large number of simple, or two-variable,
regressions were run, with the variables consisting
in each case of 60 monthly observations covering
the 1967-71 period.
The independent variable (X ) consisted in all
cases of seasonally adjusted observations concerning
the unemployment situation, either at the aggregate
or disaggregated level. The dependent variable (Y )
consisted of observations concerning either the num­
ber of “discouraged workers” or the number of
workers having left the labor force after being laid
off because of economic factors affecting their jobs.
As was the case with the unemployment data used
for the independent variable (X ), all observations
on the number of discouraged workers had also been
seasonally adjusted. The data on the number and/or
Appendix table 1.
ployment

percent of persons who had left the labor force be­
cause of various economic reasons, on the other
hand, were not seasonally adjusted. Although these
latter data are obtained monthly, they refer to exits
from the labor force occurring over a 12-month span.
As such, these data are, in effect, 12-month moving
averages, which should be relatively devoid of sea­
sonal fluctuations.
The result of a selected number of regression
analyses focusing on the relationship between unem­
ployment and the number of discouraged workers
are shown in appendix table 1. As shown, it is only
when the number of discouraged workers is reduced
to include only those whose discouragement is di­
rectly attributable to job market reasons, and which
is thus of cyclical nature, that the regression yields a
reasonably good fit— as denoted by relatively high
values of the coefficient of correlation (r ), the co­
efficient of determination (r2), and the T-Ratio.
A smaller number of regressions were run spe-

Regression of selected categories of discouraged workers against various measurements of unem­

Variables
Independent(X)

Regression results
Dependent(Y)

Regression
equation

r

r*

s

T ratio

DurblnWatson

Unemployment rate, overall.........................

Discouraged, total.

Y=472
+ 47. OX
(43.7)
(10.0)

0.53

0.28

77.8

4.70

0.88

Unemployment rate, men age 20 or over...

Discouraged, men age 20 or over.

Y=145.6 +
(11.8)

6.8X
(4.1)

0.21

0.05

29.8

1.66

1.51

Unemployment rate, women age 20 or over

Discouraged, women age 20 or over

Y=257.4 + 31.7X
(44.1)
(9.9)

0.39

0.15 _

62.2

3.18

1.31

Unemployment rate, teens 16-19.................

Discouraged, teens 16-19.................

Y=-14.4 +
(31.9)

9.4X
(2.3)

0.47

0.22

35.8

4.10

1.90

Unemployment rate, overall.......................

Discouraged, total lagged 3 months.

Y=432.7 + 57.5X
(42.2)
(9.9)

0.61

0.36

72.7

5.80

1.02

Unemployment rate, overall.........................

Discouraged, total lagged 6 months .

Y=413.2 + 63.8X
(44.9)
(10.8)

0.61

0.37

72.3

5.90

1.08

Persons unemployed 15 weeks and over__

Discouraged, total.............................

Y= 568.1 +142.3X
(20.2)
(26.4)

0.58

0.34

74.5

5.41

0.98

Average duration of unemployment.......... .

Discouraged, total.............................

Y=285.6 + 43.IX
(65.6)
(7.2)

0.62

0.38

71.9

5.97

1.18

Unemployment rate, overall........................

Discouraged, job market reasons...

Y= 77.2 + 76.OX
(33.8)
(7.7)

0.79

0.62

60.3

9.81

1.54

Unemployment rate, overall.........................

Discouraged, personal reasons.........

Y= 387.8 - 27.IX
(31.2)
(7.1)

-0 .4 5

0.20

55.5

- 3 .8 0

0.67

Unemployment rate, overall.........................

Discouraged, job market reasons
lagged 3 months..........

Y= 58.6 + 82.4X
(37.7)
(9.1)

0.81

0.65

58.0

10.41

1.73

Unemployment rate, overall.........................

Discouraged, job market reasons
lagged 6 months..........

Y= 44.1 + 88. IX
(37.7)
(9.1)

0.79

0.62

60.7

9.70

1.61




102

Appendix table 2.
unemployment

Regression of selected categories of workers leaving labor force against various measurements ol

Variables
Independent(X)

Regression results
Regression
equation

r

r*

s

T ratio

DurblnWatson

Y=1271.4+113.5X
(68.7) (15.7)

0.69

0.47

112.3

7.2

1.17

Dependent (Y)

Unemployment rate................................................ Total leaving due to economic reasons___
Unemployment rate..............................................

Left due to slack work................................... Y=

2 .1 + 86.3X
(33.0)
(7.6)

0.83

0.69

58.8

11.4

1.92

Number of unemployed who lost last job............

Left due to slack work................................... Y= 140.7+ 0.15X
(21.8) (1.46)

0.82

0.67

60.6

10.9

1.88

Job-losers rate............................................ ...........

Left due to slack work as percent of total
leaving labor force.

0.82

0.67

0.6

11.0

2.04

cifically to determine the relationship between
changes in unemployment and in the flow of workers
out of the labor force following a job loss stemming
from economic factors in general and slack work in
particular. Results are shown in appendix table 2.




Y=

1 .4 +
(0.2)

1.3X
(0.1)

These equations indicate that there is, indeed, a
rather close and positive relationship between
changes in unemployment and in the number and
proportion of workers who drop out of the labor
force after losing their jobs.
□

103

Education
of workers:
projections
to 1990
T h e l a t e s t p r o j e c t io n of the educational attain­
ment of adult workers points to an accelerated rise
in the number of college graduates and a rapid de­
cline in the number of workers at the lower end of
the educational ladder.1 Starting with the March
1970-72 average, the number of workers 25 years
and over with different amounts of formal educa­
tion, as shown in table 1, is projected to change as
follows:
— W ith 4 years o f college or m ore— from 9 .6 m illion
to 14.3 m illion by 1980 and 21.8 m illion by 1990,
increasing from 14.6 percent to 23.8 percent o f the
civilian labor force.
— W ith 1 to 3 years o f college— from 7.9 m illion to
10.8 m illion by 1980 and 15.0 m illion by 1990,
increasing from 12 percent to 16.4 percent o f the
labor force.
— W ith 4 years o f high school— from 2 4 .6 m illion to
31.4 m illion by 1980 and 37 .7 m illion by 1990,
from 37.5 percent to 4 1 .2 percent o f the labor
force.
— W ith 1 to 3 years o f high school— from 11.1 m il­
lion to 11.7 m illion by 1980 and 11.4 m illion by
1990, from 16.9 percent to 12.5 percent o f the
labor force.
— W ith 8 years o f elem entary school or less— from
12.5 m illion to 9.1 m illion by 1980 and 5.6 m il­
lion by 1990, from 19 percent to 6.1 percent o f
the labor force.

The imbalance between the rates of growth of the
adult male civilian labor force and of the number of
college graduates among them has been especially
pronounced. Over the 13 years between 1958 and
1971, the labor force has increased by only 0.6 per­
cent a year, on average, while its component of men
with at least 4 years of college has grown at the av­
erage annual rate of 3.8 percent, or about six times
as rapidly. The corresponding disparity among adult

Denis F. Johnston is senior demographic statistician in the
Office of Manpower Structure and Trends, Bureau of
Labor Statistics.

104
From the Review of November 1973



Special Labor Force Report
shows rapid advances
in educational attainment
of workers during
the next two decades
DENIS F. JOHNSTON

working women has been much smaller— 2.5 and
4.8 percent.
According to the projections, this disparity in
growth rates may be expected to continue in the fu­
ture. Between the early 1970’s and 1990, the adult
male civilian labor force is estimated to increase at
an annual average rate of 1.6 percent, with the
holders of college degrees increasing at the average
rate of 4.0 percent. The corresponding increases
during that period for the adult female civilian labor
force are estimated at average rates of 1.9 and 5
percent a year. It is apparent that one of the major
challenges to be met by the economy, both during
the current decade and the 1980’s, is the continued
absorption of this rapidly growing supply of well ed­
ucated workers.
The prospective change in the number of less ed­
ucated workers is equally dramatic. In the late
1950’s, over one-third of the adult civilian labor
force— 19.3 million workers— had completed 8
years or less of formal education. By the early
1970’s, this group had been reduced to about 12.5
million, or less than one-fifth of the entire labor
force. It is projected to decrease to about one-eighth
of the labor force by 1980 and to about one-six­
teenth by 1990. This continuing drop in the number
of less educated adult workers implies an average
annual rate of decline of 3.9 percent throughout the
1958-90 period (chart 1).

Changes in the 1970’s
Between 1972 and 1980, the civilian labor force
16 years old and over is projected to increase by
nearly 13.3 million, reaching 99.8 million in 1980.2
Nearly 60 percent, or 7.6 million, of this increase is
expected to consist of workers aged 25 to 34 years.
The concentration of the projected increase in the
number of workers with at least 4 years of college
education is similar, with 55 percent of the 4.7-mil-

Table 1. Years of school completed by persons 25 years old and over in the civilian labor force, actual 1957-72,
projected to 1980, 1985, and 1990
[Percent distribution]
Total

Elementary school

High school

College

Sex and year
Number in Percent
thousanda

Less than
5 y e ars1

5 to 7
years

8 years

1 to 3
years

4 years

1 to 3
years

4 years
or more

4 years

5 years
or more

BOTH SEXES
1957-59 average1______
1964-65-66 average____
1967-68-69 average........
1970-71-72 average........

55,909
60,067
63,618
65,655

100.0
100.0
100.0
100.0

6.3
4.1
3.1
2.6

11.4
8.7
7.2
6.4

16.8
13.4
11.0
10.0

19.2
18.9
17.6
16.9

27.8
32.8
36.4
37.5

8.4
9.6
11.0
12.0

10.2
12.5
13.7
14.6

(>)
7.4
8.1
8.3

(*)
5.1
5.6
6.3

Projected: 1980.............
1985______
1990............

77,227
84,731
91,456

100.0
100.0
100.0

1.5

1.0
.6

3.9
2.7
1.9

6.4
4.8
3.6

15.1
13.8
12.5

40.7
41.3
41.2

14.0
15.2
16.4

18.5
21.2
23.8

10.4
11.6
12.7

8.1
9.6
11.1

1957-59 average1______
1964-65-66 average____
1967—
68—
69 average........
1970-71-72 average____

38,527
39,821
40,941
41,668

100.0
100.0
100.0
100.0

7.1
4.8
3.6
3.1

12.1
9.3
7.7
7.1

17.6
14.1
11.7
10.7

19.2
18.7
17.3
16.6

25.1
30.0
33.0
34.0

8.2
9.7
11.5
12.2

10.8
13.6
15.2
16.3

(3)
7.7
8.5
8.9

(*)
5.9
6.7
7.4

Projected: 1980______
1985.............
1990.............

48,283
52,772
56,815

100.0
100.0
100.0

1.8
1.2
.8

4.3
3.0
2.2

6.9
5.3
4.0

14.4
12.8
11.4

37.5
38.4
38.6

14.7
16.2
17.6

20.4
23.1
25.5

10.6
11.5
12.2

9.8
11.6
13.3

1957-59 average*______
1964-65-66 average____
1967-68-69 average.........
1970-71-72 average........

17,382
20,246
22,677
23,987

100.0
100.0
100.0
100.0

4.5
2.8
2.2
1.8

9.9
7.8
6.2
5.2

15.2
12.0
9.6
8.6

19.1
19.3
18.2
17.6

33.7
38.5
42.5
43.6

8.9
9.5
10.3
11.4

8.7
10.3
11.1
11.8

(•)
6.7
7.3
7.5

(*)
3.6
3.8
4.3

Projected: 1980.............
1985...........
1990.............

28,944
31,959
34,641

100.0
100.0
100.0

1.0

3.2
2.2
1.5

5.5
4.0
2.9

16.4
15.4
14.4

46.1
46.1
45.3

12.7
13.6
14.5

15.3
18.1
21.0

10.0
11.7
13.4

5.3
6.4
7.7

MEN

WOMEN

.6
.4

1 Includes persons reporting no formal education.

3 Data not available.

* Totals exclude persons whose educational attainment was not reported.

NOTE: Data for combined years are averages of March Current Population Survey
figures.

lion increase in their number to be found among
workers 25 to 34 years old. In 1972, about 32 per­
cent of the 11.8 million college graduates in the
labor force were in the 25-to-34 age group; by
1980, it is estimated that about 38 percent, or 6.3
million, of the projected 16.4 million college gradu­
ates in the labor force will be in this age group.
Among women workers, the sharpest increase in the
number of college graduates in the labor force is
projected to occur in the 25-to-34 age group. In
March 1972, this group of women college graduates
numbered about 1.2 million; by 1980, its number is
estimated to reach nearly 2.1 million— an average
annual gain of 6.8 percent. The corresponding rate
of increase among men college graduates in this age
group is 6.2 percent a year.
The current decade is also expected to witness
significant changes in the median age of workers
with different amounts of formal schooling. During a
period when the median age of the labor force is
rapidly declining (from about 39.5 years in 1970 to




105

about 35.6 years anticipated in 1980), the median
age of workers with 8 years education or less rises
slowly to about 50.4 years, while that of workers
with at least 4 years of college falls below 35 years.
The steady decline in the number of less educated
workers is la rg ely a ttrib u tab le to their growing con­
centration in the older age groups, whose labor force
participation rates have been dropping steadily.

Changes in the 1980’s
The anticipated slowdown in the growth rate of
the labor force during the 1980’s (1.0 percent a year,
as compared with 2.2 percent during the 1970’s),
will be accompanied by an accelerated rate of de­
cline in the number of less educated workers (8
years of school or less), and a more moderate rate
of growth among the more highly educated workers
(table 2 ). The differences between men and women
workers with respect to these projected rates of
change are primarily attributable to differences in

Chart 1. Educational attainment of the civilian labor
force 25 years old and over, 1957-59 average, 1970-72
average, and projected 1980 and 1990

gards college graduates: except for workers 55 years
old and over, the percent of women with college de­
grees is expected to rise somewhat faster than that
of men. By 1990, the gap between the educational
levels of both sexes will be considerably smaller
than it is today.
A major feature of the projected educational at­
tainment of workers over the next 17 years is the
growing educational attainment of the several age
groups of the labor force (table 3 ). For example,
the percentage of workers with at least 4 years of

Table 2. Civilian labor force 16 years old and over, by
sex and years of school completed, 1950, 1960, and 1970
censuses, projected to 1980 and 1990
Years of school completed
Sex and year
Total

8
or
le s s 1

9 to 11

12

13 to 15

16
or
more

57,141
67,545
80,393
99,809
110,576

23,671
20,832
14,431
10,002
6,139

11,222
15,016
17,157
17,262
15,683

13,593
18,623
28,168
40,302
44,771

4,545
6,855
10,556
15,844
19,960

4,110
6,219
10,081
16,399
24,023

41,051
45,339
49,634
60,630
66,947

18,607
15,315
10,034
6,933
4,316

8,049
10,044
10,688
10,301
8,955

8,584
11,161
15,647
22,568
25,468

2,950
4,373
6,424
9,895
12,615

2,861
4,446
■ 6,841
10,933
15,593

16,090
22,206
30,759
39,179
43,629

5,064
5,517
4,397
3,069
1,823

3,173
4,972
6,469
6,961
6,728

5,009
7,462
12,521
17,734
19,303

1,595
2,482
4,132
5,949
7,345

1,249
1,773
3,240
5,466
8,430

BOTH SEXES
1950 census........................
1960 census........................
1970 census........................
Projected 1980...................
Projected 1990...................
MEN
1950 census........................
1960 census........................
1970 census........................
Projected 1980..................
Projected 1990...................

their projected rates of labor force participation, and
the effect of the changing age distribution of work­
ers with different amounts of schooling.
By 1990, 4 out of 5 workers are projected to
have completed at least 4 years of high school, with
a range from just over 60 percent among workers
65 years and over to nearly 90 percent among
workers 25 to 34 years old. A somewhat wider
range is evident in the 1990 projection for those
with at least 4 years of college: they are expected to
make up over 20 percent of the labor force, ranging
from about 16 percent among those 65 years old
and over to nearly 30 percent among workers 25 to
34 years old (chart 2 ). It is also evident that the
differences between working men and women with
respect to high school graduation are expected to
narrow somewhat over the projection period. The
percentage of high school graduates is projected to
increase somewhat faster among working men than
among working women but the reverse is true as re­




106

WOMEN
1950 census........................
1960 census___________
1970 census......................
Projected 1980_________
Projected 1990_________

Average annual percent change
BOTH SEXES
1950-60...........................
1960-70...............................
1970-80...........................
1980-90..................... .........

1.7
1.7
2.2
1.0

-1 .3
-3 .7
-3 .7
-4 .9

2.9
1.3
.1
-1 .0

3.1
4.1
3.6
1.0

4.1
4.3
4.1
2.3

4.1
4.8
4.9
3.8

1.0
.9
2.0
1.0

-1 .9
-4 .2
- 3 .7
- 4 .7

2.2
.6
-.4
-1 .4

2.6
3.4
3.7
1.2

3.9
3.8
4.3
2.4

4.4
4.3
4.7
3.6

3.2
3.2
2.4
1.1

.8
-2 .3
-3 .6
-5 .2

4.5
2.6
.7
-.3

4.0
5.2
3.5
.8

4.4
5.1
3.6
2.1

3.5
6.0
5.2
4.3

MEN
1950-60..................... .........
1960-70....___________
1970-80..............................
1980-90............................
WOMEN
1950-60_______________
1960—70..............................
1970-80____________ _
1980-90______ ______ _

1 Includes persons reporting no formal education.

Table 3. Projected educational attainment of persons 16 years old and over in the civilian labor force, by age and sex,
1980, 1985, and 1990
[ Percent distribution)
1980

1985

1990

Age and years of school completed
Both
sexes

Men

Women

Both
sexes

Men

Women

Beth
sexes

Men

Women

99,809
100.0
27.3
72.7
1.3
3.3
5.4
17.3
40.4
15.9
16.4
9.7
6.7

60,630
100.0
28.4
71.6
1.6
3.8
6.1
17.0
37.2
16.3
18.1
9.8
8.3

39,179
100.0
25.6
74.4
.9
2.6
4.4
17.8
45.3
15.2
14.0
9.6
4.4

105,716
100.0
23.0
77.0
.9
2.4
4.2
15.5
40.7
17.1
19.2
11.0
8.2

64,057
100.0
23.5
76.5
1.1
2.8
4.7
14.9
37.9
17.8
20.8
10.8
10.0

41,659
100.0
22.2
77.8
.6
1.9
3.3
16.4
45.0
16.2
16.7
11.3
5.4

110,576
100.0
19.8
80.2
.6
1.8
3.2
14.2
40.5
18.0
21.7
12.0
9.7

66,947
100.0
19.8
80.1
.7
2.0
3.7
13.4
38.0
18.8
23.3
11.6
11.7

43,629
100.0
19.6
80.3
.4
1.3
2.5
15.4
44.2
16.8
19.3
12.7
6.6

3,295
100.0
97.9
2.1
.5
1.4
4.0
92.0
2.0
.1

1,868
100.0
98.7
1.3
.5
1.6
5.0
91.6
1.2
.1

1,427
100.0
96.9
3.1
.5
1.2
2.6
92.6
3.0
.1

2,831
100.0
98.0
2.0
.4
1.1
3.5
93.0
1.9
.1

1,584
100.0
98.9
1.2
.4
1.3
4.6
92.6
1.1
.1

1,247
100.0
97.0
3.1
.4
.9
2.2
93.5
3.0
.1

2,716
100.0
98.0
2.0
.3
.8
3.1
93.8
1.9
.1

1,511
100.0
98.8
1.2
.3
.9
4.0
93.6
1.1
.1

1,205
100.0
97.0
3.0
.3
.7
1.8
94.2
2.9
.1

4,803
100.0
31.0
69.0
.8
1.3
1.7
27.2
55.5
13.4
.1

2,569
100.0
37.3
62.6
.8
1.7
2.1
32.7
49.3
13.2
.1

2,234
100.0
23.7
76.3
.7
.9
1.1
21.0
62.6
13.6
.1

4,095
29.5
70.5
.6
1.1
1.4
26.4
56.8
13.6
.1

2,147
100.0
35.6
64.3
.6
1.4
1.8
31.8
50.7
13.5
.1

1,948
100.0
22.6
77.5
.5
.7
.9
20.5
63.6
13.8
.1

4,134
100.0
27.9
72.0
.4
.9
1.1
25.5
57.9
14.0
.1

2,159
100.0
33.7
66.3
.5
1.2
1.4
30.6
52.2
14.0
.1

1,975
100.0
21.6
78.4
.4
.5
.7
20.0
64.2
14.1

14,484
100.0
12.6
87.4
.6
1.5
1.9
8.6
42.3
•30.5
14.6
11.5
3.1

7,910
100.0
15.3
84.7
.7
1.9
2.3
10.4
40.2
31.0
13.5
10.0
3.5

6,574
100.0
9.4
90.6
.6
1.0
1.3
6.5
44.7
30.0
15.9
13.3
2.6

14,059
100.0
10.4
89.7
.5
1.3
1.6
7.0
40.1
33.1
16.5
12.8
3.7

7,554
100.0
12.8
87.3
.6
1.7
2.0
8.5
38.4
33.6
15.3
11.2
4.1

6.505
100.0
7.5
92.5
.4
.8
1.1
5.2
42.1
32.5
17.9
14.7
3.2

12,270
100.0
8.0
92.0
.4
1.0
1.3
5.3
38.0
35.7
18.3
14.1
4.2

6,462
100.0
10.1
89.8
.5
1.4
1.6
6.6
36.6
36.2
17.0
12.4
4.6

5,808
100.0
5.6
94.5
.3
.6
.9
3.8
39.6
35.1
19.8
16.0
3.8

26,299
100.0
16.0
83.9
.3
1.2
2.6
11.9
42.2
17.6
24.1
13.4
10.7

17,052
100.0
15.9
84.2
.4
1.4
3.1
11.0
40.7
18.5
25.0
12.5
12.5

9,247
100.0
16.7
83.4
.2
1.0
1.8
13.7
44.9

29,259
100.0
13.3
86.9
.2
.8
2.0
10.3
41.0
18.8
27.1
14.5
12.6

18,929
100.0
12.7
87.5
.2
.9
2.4
9.2
40.1
20.0
27.4
13.0
14.4

10,330
100.0
14.2
85.8
.1
.6
1.3
12.2
42.8
16.5
26.5
17.2
9.3

30,051
100.0
10.8
89.2
.2
.4
1.4
8.8
39.8
19.7
29.7
15.3
14.4

19,382
100.0
10.1
89.9
.2
.5
1.8
7.6
39.2
21.2
29.5
13.3
16.2

10,669
100.0
12.1
87.9
.1
.3
.8
10.9
40.8
17.0
30.1
18.8
11.3

16 YEARS AND OVER
Total: Number (in thousands)....................... .........
............
Percent................... ...................................
Less than 4 years of high school1..........................
4 years of high school or more.......................
Elementary:
Less than 5 years1..................
5 to 7 yeari............................
8 years...................................
High school:
1 to 3 years........... ..........
4 years....... ...................................
College:
1 to 3 years........................
4 years or more.......................................................
4 years...............................................................
5 years or more...................................................
16 AND 17 YEARS
Total: Number (in thousands)........................
Percent............................. .................
Less than 4 years of high school1............
4 years of high school or more................................................................
Elementary:
Less than 5 years1...........
5 to 7 years........................
8 years...................... .......
High school:
1 to 3 years...................................
4 years...................................
College:
1 to 3 years........................
4 years or more............
18 AND 19 YEARS
Total: Number (in thousands)...............................................................
Percent..................................................
Less than 4 years of high school1____________
4 years or high school or more.....................................................
Elementary:
Less than 5 years1...........................
5 to 7 years................................
8 years..................................
High school:
1 to 3 years..............................
4 years...........................
College:
1 to 3 years............................
4 years or m o re ..........................................................

ioo;o

.1

20 TO 24 YEARS
Total: Number (in thousands)...... ........................
........ .............
Percent________________________
_________
Less than 4 years of high school1..................................... ......................
4 years of high school or more........ .......................................................
Elementary:
Less than 5 years1....................
.........................
5 to 7 years______________
8 years.................................................... ........................
High school:
1 to 3 years___________ _______________ _______
4 years_____________________
____________
College:
1 to 3 years.....................................................................
4 years or m ore................................. ............. ...........
4 years........... ............................ ... ...................
5 years or more___ _____ __________________
25 TO 34 YEARS
Total: Number (in thousands).................. ................................ ...........
Percent____ ___________________ _______ ____________
Less than 4 years of high school1.................... .......................................
4 years of high school or more____________ ______ ____ ________
Elementary:
Less than 5 years1.............. ................ .........................
5 to 7 years.......................................... ............. ...........
8 years....................................................................... .
High school:
1 to 3 years...................... ............ ....................... .......
4 years_____ _____ ______________ ___________
College:
1 to 3 years...................................................................
4 years or more.......... ..................................................
4 years................................... ..............................
5 years or more.......................................................
See footnotes at end of table.




22.6
15.1
7.5

107

Table 3.

Continued—Projected educational attainment of persons 16 years old and over in the civilian labor force

[Percent distribution]
1980

1985

1990

Age and years of school completed
Both
sexes

Men

Women

Both
sexes

Men

Women

Both
sexes

Men

Women

18,450
100.0
24.4
75.6
.9
3.0
4.5
16.0
42.9
13.9
18.8
10.7
8.1

11,584
100.0
24.4
75.7
1.1
3.5
4.8
15.0
39.3
14.8
21.6
11.4
10.2

6,866
100.0
24.5
75.5
.5
2.3
3.9
17.8
48.9
12.4
14.2
9.6
4.6

22,907
100.0
19.6
80.3
.4
1.7
3.4
14.1
42.5
15.7
22.2
12.2
10.0

14,350
100.0
19.0
81.1
.5
2.0
3.8
12.7
39.6
16.8
24.7
12.4
12.3

8,557
100.0
20.8
79.1
.2
1.3
2.8
16.5
47.5
13.9
17.7
11.7
6.0

27,347
100.0
16.1
83.7
.2
1.0
2.5
12.4
41.8
17.0
24.9
13.2
11.7

17,131
100.0
15.1
84.9
.2
1.2
2.9
10.8
39.4
18.3
27.2
13.0
14.2

10,216
100.0
18.0
82.0
.1
.7
2.0
15.2
45.9
14.9
21.2
13.6
7.6

16,397
100.0
33.4
66.5
2.4
5.3
8.2
17.5
40.1
11.3
15.1
8.5
6.6

9,862
100.0
35.5
64.6
3.2
6.2
9.3
16.8
34.9
11.8
17.9
9.7
8.2

6,535
100.0
30.4
69.6
1.1
4.0
6.6
18.7
48.1
10.6
10.9
6.8
4.1

16,238
100.0
28.5
71.5
1.6
4.0
5.9
17.0
42.7
12.4
16.4
9.4
7.0

9,698
100.0
30.2
69.9
2.2
4.9
6.6
16.5
37.7
13.0
19*2
10.4
8.8

6,540
100.0
26.3
73.8
.7
2.9
4.9
17.8
50.1
11.6
12.1
7.9
4.2

18,225
100.0
23.3
76.7
.9
2.6
4.2
15.6
43.5
14.2
19.0
10.8
8.2

10,863
100.0
23.9
76.0
1.3
3.1
4.6
14.9
39.4
14.9
21.7
11.4
10.3

7,362
100.0
22.2
77.8
.4
1.7
3.5
16.6
49.6
13.2
15.0
10.0
5.0

12,784
100.0
37.4
62.6
2.5
6.4
11.1
17.4
39.4
11.1
12.1
7.0
5.1

7,727
100.0
39.9
60.2
3.0
7.1
12.1
17.7
34.8
11.7
13.7
7.5
6.2

5,057
100.0
33.8
66.2
1.8
5.5
9.6
16.9
46.3
10.3
9.6
6.0
3.6

12,926
100.0
34.0
66.1
2.0
5.6
9.4
17.0
40.5
11.5
14.1
8.0
6.1

7,713
100.0
36.2
63.9
2.5
6.2
10.6
16.9
35.4
12.1
16.4
8.9
7.5

5,213
100.0
30.8
69.2
1.4
4.6
7.7
17.1
48.0
10.6
10.6
6.6
4.0

12,307
100.0
30.5
69.6
1.6
4.7
'7 .7
16.5
41.6
11.9
16.1
9.1
7.0

7,304
100.0
32.5
67.5
2.0
5.4
8.9
16.2
36.4
12.4
18.7
10.0
8.7

5,003
100.0
27.7
72.3
1.0
3.8
6.0
16.9
49.2
11.1
12.0
7.6
4.4

3,297
100.0
51.9
48.1
5.4
12.8
19.2
14.5
25.6
9.0
13.5
6.7
6.8

2,058
100.0
54.9
45.1
5.8
14.1
20.4
14.6
23.3
8.0
13.8
6.5
7.3

1,239
100.0
47.0
53.0
4.8
10.6
17.2
14.4
29.4
10.7
12.9
7.0
5.9

3,401
100.0
43.9
56.0
4.3
9.4
15.3
14.9
31.4
9.7
14.9
7.3
7.6

2,082
100.0
45.9
54.1
4.7
9.8
15.9
15.5
29.4
8.8
15.9
7.4
8.5

1,319
100.0
40.8
59.0
3.6
8.8
14.4
14.0
34.6
11.1
13.3
7.2
6.1

3,526
100.0
38.3
61.8
3.0
7.9
12.4
15.0
34.8
10.6
16.4
8.0
8.4

2,135
100.0
40.1
59.9
3.1
8.5
13.2
15.3
31.8
10.1
18.0
8.4
9.6

1,391
100.0
35.2
64.8
2.7
6.8
11.3
14.4
39.4
11.4
14.0
7.5
6.5

35 TO 44 YEARS
Total: Number (in thousands)_________________ _____ ___
Percent____ ________________ _______ ______
Less than 4 years of high school1___________________________
4 years of high school or more________________ _______
Elementary:
Less than 5 years1_____
_____
5 to 7 years...............................................
8 years......... ........................
High school: 1 to 3 years______ ____
4 years_______________________
College:
1 to 3 years_____________________
4 years or more_______________ _______________
. .
4 years__________________
5 years or m ore.._______ _________________
45 TO 54 YEARS
Total: Number (in thousands)_______________________________
Percent____ _______ ______ _____
. . ___ . . .
Less than 4 years of high school1_______________ _____________
4 years of high school or more_______________________________
Elementary:
Less than 5 years1. . ______ __________________
5 to 7 years__________________________________
8 years___ _________________________________
High school:
1 to 3 years___ ______ _________________ _____
4 years_____________________________________
College:
1 to 3 years.._________ ______________________
4 years or more______________________________
4 years___________________________________
5 years or more________ _______ _____ ______
55 TO 64 YEARS
Total: Number (in thousands)_______________________________
Percent__________ ____
___ __________________
Less than 4 years of high school1____ __________________ _____
4 years of high school or m o re.._____________________________
Elementary:
Less than 5 years1_____ ______________________
5 to 7 years__________________________________
8 years_____________________________________
High school
1 to 3 years_____ ____________________________
4 years_____________________________________
College:
1 to 3 y e ars...___________________________ . . .
4 years or more___________________ ____ ______
4 years___________ _____ ___________________
5 years or more___ ____ _____________ _______
65 YEARS AND OVER
Total: Number (in thousands)_______________________________
Percent_____________ ______ _______________________
Less than 4 years of high school1___ _________________________
4 years of high school or more________________ ______ _________
Elementary:
Less than 5 years * . .. __________ ______________
5 to 7 years___________________ ______ ________
8 years___ _____ _______________ __________ _
High school: 1 to 3 y e a rs....--------------- ----- -----------------------4 years............................................................................
College:
1 to 3 years_______ _____________ _____ _____ _
4 years or more__________ _______ ________ ____
4 years................................. ....................... ............
5 years or more..........................................................
1

Includes persons reporting no formal education.

NOTE: Because of rounding, percentages may not add to exactly 100.

48.1 percent, respectively, and by 1990, at 89.2 and
61.8 percent, respectively. Thus the range narrows
from 37.1 percentage points in 1972 to 27.4 per­
centage points in 1990. However, the gap widens

high school in March 1972 ranged from 78.6 per­
cent among workers 25 to 34 years old to 41.5 per­
cent among those 65 years old and over. By 1980, the
corresponding percentages are projected at 83.9 and




108

between the projected percentages of college gradu­
ates in the two age groups. In March 1972, the
range was from 20.1 percent among workers 25 to
34 years old to 12.4 percent among those aged 65
years and over. By 1980, the corresponding percent­
ages are projected at 24.1 and 13.5 percent, respec­
tively, and by 1990, at 29.7 and 16.4 percent, re­
spectively— a range increase from 7.7 to 13.3
percentage points.
These trends suggest that while the projected
labor force at all ages will tend increasingly to have
achieved at least a high school education, the
younger workers will enjoy a growing advantage
over their elders with respect to the completion of
college studies.
The new projection also reveals the highly signifi­
cant proportion of the college graduates who will
have pursued some form of graduate education in
the future. Among the nearly 16.4 million college
graduates projected to be in the labor force in 1980,
about 6.7 million (41 percent) are expected to have
completed at least 1 year of graduate work. By
1990, the number of workers with at least 5 years
of college education rises to 10.7 million, or 45 per­
cent of the college graduates in the labor force. The
corresponding proportions among working men are
46 percent in 1980 and 50 percent in 1990; among
working women, 31 percent in 1980 and 34 percent
in 1990. Here also, the disparity between the
younger and older workers is pronounced. By 1980,
10.7 percent of the workers 25 to 34 years old are
estimated to have completed at least 5 years of col­
lege, compared with 6.8 percent of the workers 65
years and over; by 1990, the corresponding propor­
tions are 14.4 and 8.4 percent, respectively.
The educationally disadvantaged

In recognizing the growing preponderance in
the civilian labor force of persons with higher educa­
tion, one may overlook the plight of the less edu­
cated ones, whose competitive disadvantage grows
apace with the rising educational attainment of the
majority. Their problems are exacerbated by their
growing concentration in the older age groups of
the work force. As previously noted, the median
age of workers- with 8 years of formal education or
less increases over the projection period, while that
of the more highly educated declines. In March
1972, about 38 percent of the workers with 8 years
or less of formal schooling were aged 55 years and




109

Chart 2. Percent of civilian labor force with at least 4
years of high school and at least 4 years of college, by
age, 1970-72 actual and 1990 projected

over. By 1980, this proportion is expected to remain
unchanged, despite the decline in the median age of
all workers; and by 1990, it rises to 41 percent.
Thus, the employment and retraining problems to be
overcome in fitting the educationally disadvantaged
into our increasingly sophisticated economy will be
complicated by the relatively advanced age of these
less educated workers.
Workers who have attended, but failed to com­
plete, high school have experienced particularly se­
vere employment problems during the past 20 years.
If workers under age 18 (most of whom are still en­
rolled in high school) are excluded, the number of
workers with 1 to 3 years of high school is pro­
jected to rise slightly, from 13.7 million in March
1972 to 14.2 million in 1980. It declines steadily
thereafter, reaching 13.2 million by 1990. The pro­
portion of younger workers (18 to 24 years old)
among these high school dropouts has apparently
reached a peak and is expected to decline in the fu­

ture. In March 1962, 15.6 percent of workers 18
and over with 1 to 3 years of high school completed
were in the group 18 to 24 years old. By March
1972, the corresponding percentage had risen to
18.5 percent, reflecting the large inflow of new
young workers which has occurred since the mid1960’s. By 1980, the corresponding proportion is
projected at 17.9 percent, and by 1990, it is pro­
jected to have declined sharply to 13.0 percent. In
numerical terms, these young high school dropouts
increased from about 2 million in 1962 to 2.5 mil­
lion in 1972. They are projected to remain about
constant in number to 1980, and to decline to about
1.7 million by 1990. Thus, the demand for man­
power programs designed to assist these young but
educationally disadvantaged workers to improve
their employment qualifications will remain steady
throughout the 1970’s, but will decline thereafter.

therefore, require a movement by these women in
unprecedented numbers into traditionally male-dom­
inated professional and technical occupations.4
These highly qualified workers may also displace
increasing numbers of less educated workers in oc­
cupations which have formerly been the preserve of
those without college education, particularly if the
kinds of jobs which typically have been held by col­
lege graduates do not increase fast enough to absorb
the prospective growth of college graduate jobseek­
ers. The upgrading of job requirements already ob­
served suggests that the employers’ expectations
with respect to the educational qualifications of their
prospective employees tend to rise with increases in
such qualifications of the jobseekers themselves.
Thus, if college graduates are forced to seek jobs
which have not traditionally attracted them, they are
likely to be hired in preference to the less educated,
quite apart from the actual education needed to per­
form such jobs adequately. Should such displace­
ment take place on a large scale, the potential con­
sequences could be damaging both to the
college-educated workers and to the less educated
workers they displace. For the former, limited op­
portunity to utilize and develop the skills and
perspectives acquired in college could give rise to al­
ienation, frustration, and other problems associated
with this type of underemployment. For the latter,
the prospect of competition with the educationally
advantaged for jobs and promotions could also give
rise to serious strains.5 But it is also possible to en­
vision more favorable consequences, such as the im­
proved job performance which may be expected to
accompany the educational upgrading of workers in
different occupations. In addition, increased compe­
tition between workers with different amounts of
formal education might eventually lead to less exclu­
sive reliance upon academic credentials in hiring
and promotion, and the supplementing of such crite­
ria with more valid indicators of work-related ability
and potential.6

The college graduates: supply and demand

The performance of the U.S. economy in absorb­
ing the growing number of college graduates during
the past two decades warrants some optimism with
respect to the employment prospects of highly edu­
cated workers in the future. Between 1950 and
1970, the number of employed men 25 years and
over increased by 19 percent— 0.9 percent a year,
on average— while that of the college graduates
among them rose by 134 percent, or an average of
4.3 percent a year. Over the same 20-year period,
the number of employed women workers in the
same age group increased by 89 percent— an aver­
age gain of 3.2 percent a year— while the number of
college graduates among them increased by 147 per­
cent, or 4.5 percent a year, on average.
However, the recent employment experience of
new college graduates suggests that their short-term
employment prospects may be quite sensitive to
cyclical changes in the economy and to the changing
mix of demand for highly trained professional and
technical workers in particular fields.3 In particular,
a note of caution should be expressed in regard to
the employment needs of women college graduates
whose numbers are projected to grow so rapidly.
This growth must be considered in relation to the
limited job opportunities expected in primary and
secondary education, where a large proportion of
these women have been employed in the past. The
absorption of these women into the labor force may,




Comparison with earlier projection

In comparison with the 1970 projection it su­
persedes, the current projection of the male civilian
labor force 25 years old and over is smaller by
about 400,000 for 1980 and about 500,000 for
1985.7 As for the female civilian labor force of the
same age span, the current projection is larger by
about 1.3 million for 1980 and 1.6 million for

110

1985. In the aggregate, the current projection yields
an adult civilian labor force that is larger than the
earlier projection by about 900,000 for 1980 and
1.1 million for 1985 (table 4 ). In general, the cur­
rent projection shows a more rapid advance in the
educational attainment of the adult civilian labor
force than was indicated earlier. This advance is
more pronounced among men than among women
for both 1980 and 1985. It is also more pronounced
for 1985, for each sex, than for 1980.
The differences between the current and the pre­
vious projection in the size of the adult civilian
labor force are largely those resulting from changes
in the projected rates of labor force participation.
The recent changes in the size of the projected pop­
ulation 25 years old and over are generally minor.
The more rapid rate of increase in the projected ed­
ucational attainment of the labor force, on the other
hand, reflects primarily the higher levels of educa­
tional attainment projected for the adult population
as a whole by the Bureau of the Census, since the
projected educational attainment of the labor force
has been linked to that of the population. (See ex­
planation of the linkage in the following section.)

Table 4. Comparison of current projection of educa­
tional attainment of adult workers 25 years and over
with previous BLS projection, 1980 and 1985
1980
Sex and year* of
school completed

(1)

Civilian labor force:
Number (in thousands). 48,283
Percent........ ................
100.0
Less than 8 years1.............
6.1
8 to 11 years___________
21.3
12 years..............................
37.5
13 to 15 years__________
14.7
20.4
16 years or more_______

Differ­
ence
(4W 5)

(2)

(3)

(4)

(5)

(6)

48,665
100.0
6.4
23.2
39.7
12.1
18.6

-3 8 2

52,772
100.0
4.2
18.1
38.4
16.2
23.1

53,282
100.0
4.7
19.9
42.3
12.6
20.5

-5 1 0

31,959
100.0
2.8
19.4
46.1
13.6
18.1

30,362
100.0
3.1
19.8
48.2
12.9
16.0

-.3
-1 .9
-2 .2
2.6
1.8

-.5
-1 .8
-3 .2
3.6
2.6

WOMEN
Civilian labor force:
Number (in thousands). 28,944
Percent........ ..................
100.0
Less than 8 years*......... .
4.2
8 to 11 years............... .......
21.8
12 years......... ...................
46.1
13 to 15 years....................
12.7
16 years or more........ .......
15.3

27,662
100.0
4.5
22.5
47.2
12.0
14.0

1,282
-.3
-.7
- 1 .1
.7
1.3

1,597
-.3
-.4
- 2 .1
.7
2.1

>Denis F. Johnston, "Education of adult workers: projections to 1985,” Monthly
Labor Review, August 1970, pp. 43-56, reprinted as Special Labor Force Report 122.
* Includes persons reporting no formal education.

T hese differences reflect observed trends over the
period 1 9 5 7 -5 9 to 1 9 7 0 -7 2 , either toward increasing
or decreasing differences. Otherwise, the differences
w ere assum ed to rem ain constant.

These projections were developed by a method
which provides a systematic linkage with the latest
available projections of educational attainment of
the population, by age and sex, prepared by the Bu­
reau of the Census.8 The projections were devel­
oped in the following steps:

Step 3. T he differences— p ositive or negative—
were applied to the projected educational distribu­
tions o f the p opulation to obtain a first approxim a­
tion o f the educational attainm ent o f the labor force
in 1980, 1985, and 1990.

Civilian labor force 25 years old and over

Step 4. T hese educational distributions (in per­
centage terms) w ere then applied to the previously
projected civilian labor fo rce totals for each age-sex
group. T he resultant num bers w ere then divided by
the corresponding population to obtain a labor force
participation rate for the population in each age,
sex, and educational attainm ent category for the
target dates.

Step J. Percentage distributions by educational
attainm ent categories w ere obtained for m en and
w om en in the population and the civilian labor force
for age groups 25 to 34, 35 to 44, 45 to 54, 55 to
64, and 65 and over. T h e follow in g educational
attainm ent categories were used: less than 5 years
o f schooling (including no school years com pleted),
5 to 7 years, 8 years, 9 to 11 years, 12 years, 13 to
15 years (that is, 1 to 3 years o f college), 16 years,
and 17 years or m ore. T hese data were obtained
from the M arch Current Population Surveys for
four periods— an average o f 1957 and 1959; an
average o f 1964, 1965, and 1966; an average o f
1967, 1968, and 1969; and an average o f 1970,
1971, and 1972.9

Step 5. T he labor force participation rates ob­
tained in step 4 were then com pared with observed
trends in these rates over the period 1 9 5 7 -5 9 to
1 9 7 0 -7 2 . R elatively m inor adjustm ents in these rates
were then introduced w herever necessary to m aintain
consistency with the observed trends, w hile also pre­
serving consistency with the previous projection o f
the civilian labor force in each age-sex group.

Step 2. T he observed differences in the educa­
tional distribution o f the population and civilian
labor force (by age and sex) w ere projected to 1990.




Differ- C urrent Report
en ce projec­
1221
tion
(1 H 2 )

MEN

Assumptions and methodology

A.

C urrent Report
projec­
122 ‘
tion

1985

Step 6. T he adjusted rates o f labor force partici­
pation were then applied to the projected population

111

1957, 1959, and 1964-72.

to obtain the projected labor force by age, sex, and
years o f school com pleted for 1980, 1985, and 1990.
B.

For each of these groups, observed trends in the
percentage distribution of their educational attain­
ment were extrapolated to 1990, and the resultant
projections were applied directly to the projected
population and civilian labor force of each group to
obtain the numbers by years of school completed.
The labor force participation rates implicit in these
projections, by age, sex, and years of school com­
pleted were then computed for 1980, 1985, and
1990 by dividing the projected civilian labor force
by the projected population. These rates were thei*
compared and adjusted, as necessary, to ensure con­
sistency with trends in the reported participation
rates over the period for which actual data were
available. The adjusted rates were then applied to
the projected population to obtain the final pro­
jected labor force in each educational attainment
category.
□

C ivilian labor force 16 to 24 years old
F or som e planning purposes, it is useful to have
a projection o f the educational attainm ent o f workers
under the age o f 25 even though m any o f these
workers have not yet com pleted their form al sch ool­
ing. T he Bureau o f the Census does not develop pro­
jections o f the educational attainment o f the popula­
tion under 25; therefore, it was decided to develop
a projection o f the educational attainm ent o f both
the population and the civilian labor force 16 to 24
years old by m eans o f a direct extrapolation o f trends
in their reported educational attainment. T he re­
ported educational attainm ent distribution o f the
population and labor force 16 and 17 years old (by
sex) w as obtained from the 1950, 1960, and 1970
censuses and from th e M arch 1972 Current Popu­
lation Survey. C orresponding data for the population
and labor force age groups o f 18 and 19 and 20
to 24 (by sex) w ere obtained from the published
Current Population Survey reports for the years

-FOOTNOTES1 These projections supersede those presented in Denis F.
Johnston, “Education of adult workers: projections to 1985,”
M onthly Labor Review, August 1970, pp. 43-56, reprinted
as Special Labor Force Report 122. Information by color
or race, which was provided in the earlier report, is not yet
available. The size and age-sex composition of the civilian
labor force in this report are consistent with those o f the
total labor force as projected in Denis F. Johnston, “The
U.S. labor force: projections to 1990,” M onthly Labor R e­
view, July 1973, pp. 3-13. In addition, the projected edu­
cational attainment of the adult civilian labor force (25
years and over) is consistent with the latest projection of
the educational attainment of the adult population published
by the Bureau of the Census. For a description of the
assumptions and methodology of these projections, see
Demographic Projections for the United States, Current
Population Reports, Series P-25, No. 476 (Bureau of the
Census, February 1972).

February 1973, pp. 41-50, reprinted as Special Labor
Force Report 151. Also see Michael F. Crowley, “Pro­
fessional manpower: the job market turnaround,” M onthly
Labor Review, October 1972, pp. 9-15; and Persons in
Engineering, Scientific, and Technical Occupations: 1970
and 1972, Current Population Reports, Special Studies,
Series P-23, N o. 45 (Bureau of the Census, July 1973).
4
For additional perspective on the employment o f col­
lege-educated women, see Pamela Roby, “Women and
American Higher Education,” The Annals o f the American
Academy of Political and Social Science, November 1972,
pp. 118—39, and Alan L. Sorkin, “Occupational Status o f
Women, 1870-1970,” The American Journal of Economics
and Sociology, July 1973, pp. 235-43.
6
For additional information on the prospective occupa­
tional distribution o f college-educated workers, see the
forthcoming article by Neal H. Rosenthal in the Decem­
ber 1973 issue o f the M onthly Labor Review.

2 Unless otherwise specified, the projections in this article
relate to the entire civilian labor force, 16 years old and
over. The projected years of school completed refer to the
workers’ estimated attainment at the time they are in the
labor force, and not to their ultimate attainment upon
completion o f their formal education. Where data for 1972
are used for discussion of emerging changes during the cur­
rent decade, the 1972 data are taken from William V. Deutermann, “Educational attainment o f workers, March
1972,” M onthly Labor Review, November 1972, pp. 38-42,
reprinted as Special Labor Force Report 148.

6 “Years of school completed” is obviously a crude crite­
rion of either ability to perform well in a given job or o f
potential for further training and development; exclusive or
even primary reliance upon such a criteron for purposes of
job placement, promotion, or training is unlikely to provide
an adequate screening of the more capable individuals in
most work situations. Among the factors which are not
necessarily reflected by educational attainment are the qual­
ity of schooling received, the possible loss o f knowledge
and skills learned in the past, learning and experience ac­
quired outside of school, and such personal attributes as
discipline, interest, and motivation. On this general issue,
see Thomas F. Green, Work, Leisure, and the American

3 Information on the employment experience o f recent
college graduates is provided in Vera C. Perrella, “Employ­
ment o f recent college graduates,” M onthly Labor Review,




112

Schools (N ew York, Random House, 1968), and Creden­
tials and Common Sense: Jobs for People Without D iplo­
mas (U.S. Department of Labor, Manpower Administra­
tion, December 1968), Manpower Report 13.
7 These comparisons are restricted to the civilian labor
force 25 years old and over in 1980 and 1985 because the
1970 projection was limited to that age group and those
dates.
8 Demographic Projections for the United States, Current
Population Reports, Series P-25, No. 476 (Bureau of the
Census, February 1972), table 5. These projections re­
late to persons 25 years old and over, by age and sex. For
the age groups where two series of educational distributions
were developed (persons 25 to 34 years old in 1980 and




persons 25 to 44 years old in 1985 and 1990), an arith­
metic average of the two series was adopted.
9
Current Population Survey data on the educational at­
tainment of the population for the mentioned years are
published by the Bureau of the Census in Current Popula­
tion Reports, Series P-20, Numbers 77, 99, 138, 158, 169,
182, 194, 207, 229, and 243. Corresponding data for die
civilian labor force for these years are published by the
Bureau of Census in Current Population Reports, Series
P-50, No. 78 (for 1957) and by the Bureau of Labor Sta­
tistics in its Special Labor Force Reports, 1, 53, 65, 83,
92, 103, 125, 140, and 148 (covering March 1959 through
March 1972). The latter reports were reprinted, with addi­
tional tables, from the M onthly Labor Review.

113




Chapter III. Special Groups in the Labor Force




The economic
status of
families headed
by women

Nearly 5.6 million families
in the United States are headed by women;
despite employment growth of the 1960’s
about 2 million of these families
remain in poverty
ROBERT L. STEIN

O n e of the important domestic problems facing
the N ation in the 1970’s is how to improve the
economic status of families headed by women.
According to the latest estimates— for March
1970— 5.6 million families in the United States
are headed b y women, or more than 1 fam ily in 10.
The number has been increasing more rapidly
than the total of all families. Between 1960 and
1970, for example, it rose by 24 percent, whereas
total families increased by 14 percent.
Historically the em ployment and income sit­
uation of such families has generally been bleak.
M ost of the women are ill-equipped to earn an
adequate living. M any suffer from one handicap or
more to successful competition in the labor
market— lack of sufficient education or training,
irregular and unstable work histories, sex or racial
discrimination in hiring, ill health, and the diffi­
culty of arranging for satisfactory child care.
As a result, these women have not been able to
share fully in the N ation’s economic growth,
with its associated expansion in jobs and advances
in earnings. During the 1960’s, the income of
families headed by men remained more than
double the income of families headed by women.
While the number of families headed by men with
incomes below the poverty line ($3,700 for a family
of four in 1969) was reduced by one-half between
1959 and 1969, the number of poor families
headed by women remained virtually unchanged
at about 1.8 million. Em ploym ent growth, the
m ost powerful weapon in the antipoverty arsenal,
has not significantly reduced the number of poor
families headed by women.

Public assistance, a primary source of income
for many of the families headed by women, has
been expanding in coverage and in benefit levels,
but payments are still generally very low—in
most States below the poverty line.
The welfare system has been caught in a cross­
fire of public criticism. The target for m ost of the
hostility is the a f d c program—Aid to Families
with Dependent Children— designed to provide
income assistance to the families of children whose
fathers have died or deserted or are absent for a
variety of other reasons. On the one hand, welfare
programs are criticized because their paym ent
levels are considered too low to provide economic
security to families in need. On the other hand,
the programs are criticized on the grounds that
work, as well as need, should be a requirement for
eligibility. The welfare system has also been
faulted because of the widely disparate State
benefit levels, because it m ay discourage some
women from seeking employment, and because
it m ay induce some families to break up.
The attacks have become sharper in recent
years because of steady growth in the welfare
population during a period of rapid economic
growth and very low unemployment. B y March
1970, about three-fifths of the 3.4 million fam ilies
with children headed by women were already on
welfare and the rolls were still rising. These de­
velopments were placing a growing burden on the
already hardpressed taxpayer. One result of the
resistance to the rising welfare bill has been a
heightened interest in the possibility of em ploy­
ment for welfare mothers. One important aspect
of welfare reform involves the development of
training and job placement programs for ablebodied adult welfare recipients. The manpower
provisions of the Administration’s proposed Fam ily
Assistance Act of 1970 include a training and work
requirement for mothers of school-age children.

R obert L. Stein is an economist in the Office of Eco­
nomic and Social Research, Bureau of Labor Statistics.
Carol Milner of the sam e office assisted in the preparation
of the article.

From the Review of December 1970



116

less serious than that of younger families with
children since they had more freedom to accept
employment, they had more income from other
sources, and they had fewer dependents. H alf
had fully grown children in the household who
could contribute to the fam ily’s income. In 1969,
the median income of families headed by women
45 to 64 years of age who had no children under
18 in the household was $7,000, whereas the income
of families headed by women 24 to 44 years of
age who did have children was only $4,000.
Between 1960 and 1970, the number of women
heading families with children rose by 800,000.
Roughly one-third of this increase could be
attributed to general population growth. There has
been considerable speculation that rising welfare
benefits in the large industrial States of the North
have contributed to the breaking up of poor
families. However, it would be extremely difficult
to isolate this factor from the entire complex of
forces that leads to family disorganization. (Onethird lived in the South where welfare paym ents
are still com paratively low.)
The proportion of families headed by women is
highest among poorly educated and low income
groups, among minority groups, and among city
residents. On the other hand, the group is also more
heterogeneous than might be supposed. Among
women 25 and over, m ost of whom have completed
their formal schooling, one-third of the family
heads have no more than an elementary school
education (compared with one-fourth of other
women), but 13 percent have some college educa­
tion. Although one-third have incomes below the
poverty line, a small minority (nearly 300,000)
have incomes of $15,000 or more. These are m ainly
older white families without children.
Among the black urban poor, the proportion of
families headed by women was 66 percent in March
1970. Here, as in the N ation as a whole, the pro­
portion has been increasing; the trend is much more
pronounced among the urban poor.
Even among the 3.4 million families with child­
ren, the situation is uneven. About 65 percent
have only one or two children and their incomes
are somewhat higher than the incomes of larger
families. However, those with few or no children
tend to be at the extremes of the age scale. Among
women family heads age 25 to 44, presumably
the prime candidates for training and employment,
nearly half had three children or more. The prob­

Scope of th e problem

In March 1970, 5.6 million women were heads
of families (table 1); 2.4 million of these women
(43 percent; were widows and 2.6 million (46
percent) were divorced or separated from their
husbands. The remaining 600,000 had never
been married. About a third of these single women
had children under 18.
From the standpoint of society, foremost con­
cern is centered on the status of those families
with dependent children. The environment in
which these children are growing up is inevitably
affected by the stresses and strains on the mother
who m ust take over the responsibility for the
discipline, training, and guidance of the young as
well as their financial support. In March 1970,
there were 3.4 million such families, comprising 8
million children under 18 years of age (an average
of 2.4 per family) and 13 million persons altogether.
The remaining 2.2 million— women without
children under 18— were nearly all past the age of
45. Two-thirds were widows, and all but a few
were heads of small families consisting of only
two or three persons. These older family heads
were not without employment and income prob­
lems. B y and large, however, their situation was
Table 1.
women

Selected characteristics of families headed by

Thousands of
families

Percent of
families in each
category

Characteristic
March
1970

March
1960

March
1970

March
1960

Total, all families............................. . . . .............
With children........ ................ . ............

5,580
3,363

4,494
2,542

11
11

10
9

Below the poverty line................................
With children........................................

1,803
1,488

1,916
1,525

36
47

23
28

In central cities of metropolitan areas...............
Below the poverty level................. .............

2,269
738

1,764
585

15
50

12
29

Total, all families................................................
With children........................................

4,185
2,255

3,545
1,834

9
9

9
8

Below the poverty line.................................
With children........................................

1,063
831

1,233
948

30
40

20
25

In central cities of metropolitan areas...............
Below the poverty line................................

1,418
337

1,240
303

12
39

10
24

Total, all families................................................
With children........................................

1,395
1,108

949
708

27
31

22
25

Below the poverty line................. ..............
With children........................................

739
657

683
577

53
59

32
35

In central cities of metropolitan areas..............
Below the poverty line................................

851
402

524
282

29
66

23
28

ALL RACES

WHITE

NEGRO AND OTHER RACES




117

lems confronting women with many children are
compounded by the fact that they are also the
least educated and therefore the least equipped to
find employment.
F a m ily incom e

The relationship between income and family
stability is complex. When a breadwinner dies or
leaves his family, the loss or reduction of financial
support m ay be only partly offset by the wife’s
earnings and Social Security, private pensions or
insurance, welfare payments or other benefits.
Poverty or low income m ay itself create tensions
leading to fam ily breakup. Or the fact that a man
does not have a steady job at good pay m ay induce
him to leave so that his family can obtain public
assistance. These situations are not easily quanti­
fied. In any case, the data show a very strong cor­
relation between income and the presence or ab­
sence of fathers.
As table 2 shows, the percentage of families
headed by women moves down steadily as family
income rises. The proportion starts out at 63 per
100 families with incomes under $2,000, and then
moves down progressively to reach 2 per 100 fami­
lies with incomes of $10,000 and over.
Negro families with children are much more
likely than white families to be headed by a
woman— 1 in every 3 Negro families is in this
category, compared with 1 in every 10 white
families. The difference in family structure is one
reason for the lower average income of Negro
families. Although the proportion of black families
without husbands and fathers is higher than for
whites at every income level, it moves down
sharply and continuously from about 3 in 4 among
the lowest income families to about 1 in 20 among
the higher income families.
The median income of the families of 8 million
children who were being brought up by their
mothers— or other female relatives— was $4,000
in 1969. This contrasts with a median family in­
come of $11,600 for the 61 million children living
with both parents.
Only 38 percent of the families headed by women
had incomes over $5,000 and only 9 percent had
incomes over $10,000. B y contrast, 55 percent of
the husband-wife-children families had incomes
over $10,000. Although husband-wife families tend
to be larger than families headed by women, the




118

Table 2. Income in 1969 of families with children, headed
by women
Family income

All races

White

Negro
and other
races

Total: Number (in thousands)......... ....................

3,363

2,255

1,108

Percent................................................ .......
Under $2,000....................................................
$2,000 to $2,999____________ ___________
$3,000 to $3,999...............................................
$4,000 to $4,999____ ______ ______ ______
$5,000 and over.......................... .....................
$5,000 to $5,999................................................
$6,000 to $6,999_________________ ______
$7,000 to $7,999.......................................
$8,000 to $8,999........................................
$9,000 to $9,999... .......................................
$10,000 and over..................... ....... ................

100
21
15
14
12
38
10
8
5
3
3
9

100
18
13
12
12
45
10
9
6
4
5
11

100
26
18
18
11
27
10
6
3
3
1
4

Median income............ ...................... .....................

$4,008

$4,523

$3,327

Families headed by women as percent of all families
with children............... ........................................

11

9

31

Under $2,000...................................................
$2,000 to $2,999.._____ ___________ ____
$3,000 to $3,999.............................................
$4,000 to $4,999...............................................
$5,000 to $5,999............................. ..................
$6,000 to $6,999...................... .......................
$7,000 to $7,999......................... ____..............
$8,000 to $8,999___________________ ____
$9,000 to $9,999...................... .........................
$10,000 and over______ _____ _____ _____

63
54
40
28
20
14
8
5
r
2

57
48
33
26
17
12
7
4
5
2

74
67
56
36
31
22
12
14
6
5

differences in income between the two types of
families far exceed any differences in need.
Families headed b y women account for a large
and growing proportion of the remaining poverty
in the United States. In 1969, 47 of every 100 poor
families with children were headed by women. In
1959, the proportion was 28 out of 100.
The poverty line takes account of both family
income and family size. In 1969, the line was set
at $3,700 for a nonfarm family of four headed by
a woman. It goes up (or down) by roughly $700
for each additional person (or each person less) in
the family.
The poverty thresholds as used in this dis­
cussion 1 are not intended to provide a measure of
income adequacy; that is, it should not be inferred
that those with incomes above the poverty line
have necessarily achieved a minimally adequate
level of living. The cutoffs do provide a useful
device for measuring the prevalence of, and trends
in, very low income levels among various fam ilytype and family-size groups, and are more realistic
than are fixed dollar amounts of income (for ex­
ample, families with incomes under $3,000) because
they are graduated by family size. T hey are
varied over time to reflect annual changes in the
average price level as measured by the Consumer
Price Index.

The poverty statistics point up the importance
of fam ily size. If a fam ily headed by a woman has
on ly one or two children, it has about a 2 out of 3
chance of staying above the poverty line. H ow ­
ever, as the number of children increases, the
probability that the fam ily’s income is under the
poverty line rises sharply. Among those families
with four children or more, over two-thirds are
poor.
Additional children might have been economi­
cally helpful to poor families in an earlier era.
B ut in modern urban society with its complex
technology and its unrelenting emphasis on educa­
tion and skill, each additional child diminishes the
woman’s prospects for economic independence and
security through employment. The bearing and
rearing of children m ay interfere with the comple­
tion of her education, and most certainly will
interfere w ith the continuity of her employment.
Unless a woman can acquire at least a high school
education or can acquire meaningful job training
and job experience, and unless she can work full
time m ost of the year, it is unlikely that her annuaj
Table 3. Extent of poverty in 1969 among families headed
by women, by number of children
(Numbers in thousands]
Poor families
Total
number
of
families

Number

Percent
of
total

Median deficit
between total
income and
poverty line 2

Total...................................

5,580

1,803

32

$1,200

No children under 1 8 ..................
One child___________________
Two children________________
Three children_______________
Four children_____ __________
Five children or more_________

2,218
1,211
960
545
303
344

315
360
386
279
202
262

14
30
40
51
67
76

700
1,100
1,200
1,500
1,700
2,400

Total...................................

4,185

1,063

25

1,200

No children under 18....................
One child................................. . . .
Two children.................................
Three children.._____________
Four children...............................
Five children or more...................

1,931
906
702
353
163
130

232
227
258
163
97
86

12
25
37
46
60
66

700
1,100
1,300
1,700
1,700
2,400

1,395

739

53

1,400

286
306
258
191
140
214

83
133
128
116
105
174

29
43
50
61
75
81

700
1,100
1,100
1,500
1,600
2,400

Race and number of
children under 18 >

earnings alone would be sufficient to lift the in­
come of a family of four above the poverty line.
Additional children tend to reduce her earning
power, while raising family expenses. The extra
welfare allowance for each additional family mem­
ber is too small to prevent the gap from widening.
The situation is illustrated statistically in table 3.
On the average, poor families headed by women
had total incomes in 1969 which were $1,200
below the poverty threshold, but this income
deficit increased with each child added to the
family. The median difference between income
level and the poverty line (the “poverty gap”)
was $1,100 for those with one child, $1,500 for
those with three children, and $2,400 for those
with five children or more.
One-quarter of all families headed by a woman
arc black. For these families, the rate of poverty is
greater than for white families irrespective of the
number of children. Moreover, large families are
more common among blacks; one-third of the
Negro families headed by women has four children
or more compared with only one-eighth of the
white families.
Among families with children, nearly twothirds had only one child or two children. B ut when
the children themselves are considered by family
size, a different picture emerges— three-fifths lived
in families with three children or more. These are
the families where the poverty rate ranged from 51
to 76 percent and the poverty gap averaged from
$1,500 to $2,400.

ALL RACES

Extent of employment
The proportion of women holding paid jobs
outside the home has been climbing steadily for
25 years and by March 1970, 43 of every 100 women
16 years of age and over were in the labor force
(that is, either employed or seeking work).
The typical pattern has been for a woman to
enter the labor force after completion of her
education and prior to marriage, to leave after
starting a family, and to reenter the labor force
as family responsibilities diminish. During the
last 10 years, however, there has been some
modification of this pattern with the increasing
entry into the labor force of mothers with young
children. Their participation rate, although still
comparatively low, has increased much faster

WHITE

NEGRO AND OTHER RACES
Total............................. .
No children under 18............ .......
One child............................. .........
Two children......... .....................
Three children.................. .........
Four children................................
Five children or more..................
> Own or related.
> Based on data for 1968.




119

Table 4. Work experience of women1 heading families
and extent of poverty among these families in 1967

than the rate for other mothers. From 1960 to
1969, the rate for mothers with children under 6
years of age increased from 20 percent to 30
percent, while for mothers with children 6 to 17
years of age it increased from 43 to 51 percent.
The data indicate that the labor force partici­
pation of mothers responds to economic need. In
March 1969, divorced, separated, or widowed
women with young children under 6 had a par­
ticipation rate of 47 percent, compared with
29 percent for married women with children
under 6. The higher rate for women without
husbands reflects in part an insufficiency of in­
come from sources other than em ployment (ali­
mony, child support, welfare, and Social Security).
From the standpoint of developing programs
geared to assist women to earn their way off
welfare, these labor force trends appear somewhat
encouraging. However, the statistics on labor
force participation of women can be misleading
because they reveal nothing about the duration
of employment. I t is readily apparent that there
is a high rate of turnover in the female work
force. During 1968, an average of 28 million were
employed, but 37 million different women were
employed at some time during the year. For
insight into the duration of employment, it is
necessary to turn to data on work experience
during the entire calendar year rather than in an
average survey week. Because of concern with
the capacity of women not merely to hold jobs
but to support their families on the basis of their
earnings, it is particularly important to examine
the extent of full-time and part-time labor force
activity, and the extent of year-round work
compared with seasonal or temporary work.
Special tabulations of data on work experience
in 1967, compiled for the Manpower Adminis­
tration of the U.S. Department of Labor, were
summarized for female heads of families age 16
to 44 years. These are women who still have many
years of potential working life remaining and for
whom job training is a realistic possibility. They
are also the ones, however, who are m ost likely
to be prevented from working steadily by the
presence of children. Altogether, 70 percent worked
at some time during the year, but only 38 percent
worked throughout the year at full-time jobs.
As table 4 shows, working only part of the year
is not enough to enable many female family heads




(Numbers in thousands]
Poor families

Percent distribution

Total
number
of
families

Number

Percent
of
total

Total................................

2,263

1,029

45

100

100

Year round full time..................
All other workers............... .......
Part year full tim e ............
Part time.............................
No work at all..... .....................

862
728
439
289
673

135
373
217
156
521

16
51
49
54
77

38
32
19
13
30

13
36
21
15
51

Total................................

1,509

557

37

100

100

Year round full time..................
All other workers.......................
Part year full time..............
Part time.............................
No work at all............................

599
498
305
193
413

53
215
131
84
289

9
43
43
44
70

40
33
20
13
27

10
39
24
15
52

Total................................

752

470

63

100

100

Year round full time..................
All other workers......................
Part year full time..............
Part time.............................
No work at all...........................

262
231
134
97
260

82
157
85
72
232

31
68
63
74
89

35
31
18
13
35

17
33
18
15
49

Work experience
and race

Total

Poor

ALL RACES

WHITE

NEGRO AND OTHER RACES

1 16 to 44 years of age.

to support their families at a level of living above
the poverty line. Of the families headed by women
who were employed only part time or part year,
about half were poor. On the one hand, where
the mother was employed year round full time,
only 16 percent were poor. Of course, supple­
mentary income was a factor in some cases, but
the mother’s earnings were clearly the m ost de­
cisive factor. On the other hand, three-fourths of
the families headed by nonworkers were poor.
If a woman can hold a professional, managerial,
or clerical job, her chances of keeping her family
above the poverty line are very good (table 5);
only 16 percent of these families were poor. Over
two-fifths of the mothers who worked at all had a
job in one of these white-collar occupations.
Half of all female heads of poor families did not
work at all during the year so that any skills or
experience they might have were not being used.
Of those who did work, nearly half had low-paid
service jobs such as kitchen helpers, maids,
hospital attendants and aides, and laundry
workers. A fifth held semiskilled factory jobs.
Only one-fifth of those with any employment ex­
perience (one-tenth of the overall total) worked

120

at some time during the year in the better-paid
white-collar occupations.

Table 6. Educational attainment of women Heads 1 of
poor families and usual weekly earnings of full-time
women workers in May 1969
[Numbers in thousands]

W eekly earnings of women
Educational attainment by race

D ata on the usual weekly earnings of wage and
salary workers in full-time jobs reveal that in
general the median earnings of women full-time
workers are not very high. (See table 6.) The
overall median weekly earnings for all women full­
time workers in M ay 1969 were $87. Even among
white women with high school diplomas, who were
employed mainly in clerical jobs, usual weekly
earnings were only $88.
The data by educational attainm ent (years of
formal schooling completed) and occupation from
the M ay 1969 earnings survey are instructive.
They reveal that only among the college-educated
professional and managerial groups did a majority
of women working full time earn over $100 a
week. Among those with no college attendance
(three-fourths of the total), only 3 out of every 10
white women and 2 out of every 10 black women
Table 5. Occupation of women heads1 of families, by
poverty status in 1967
[Numbers in thousands]

Occupation, according to
longest job held

Total
number
of
family
heads

Poor family
heads

Percent
distribution

Num­
ber

Per­
cent
of
total

Total

Poor

1,584

504

32

100

100

185
483
77
354
109
331
45

21
84
32
106
73
159
29

11
17
42
30
67
48
C)

12
30
5
22
7
21
3

5
17
6
21
14
32
6

1,091

ALL RACES
Total with work experience........ .
Professional and managerial......................
Clerical.____ _______ ______________
Sales..........................................................
Operatives and other blue collar_______
Private household___________________
Other service workers________________
Farm workers______________________
WHITE
Total with work experience______

268

25

100

100

145
470
231
217,
28

17
83
53
101
14

12
18
23
47
0)

13
43
21
20
3

6
31
20
38
5

Total with work experience______

490

234

48

100

100

Professional and managerial__________
Clerical and sales___________________
Operatives and other blue collar_______
Private household and other services___
Farm workers______________________

39
89
123
222
17

5
33
52
130
14

(2)
37
42
59
<2)

8
18
25
45
3

2
14
22
56
6

Professional and managerial...................
Clerical and sales___________________
Operatives and other blue collar_______
Private household and other services___
Farm workers______________________
NEGRO AND OTHER RACES

1 16 to 44 years of age.
2 Percent not shown where base is less than 75,000.




121

Heads of pc or families2
Number

Percent

Usual weekly earnings
of full-time
workers (median)

Total...... .................. ...........

1,025

100

$87

White............................... ..............
8 years or less......... ..............
9-11 years________ ______
12 years..................................
13-15 years________ _____
16 years or more__________

556
140
188
163
53
12

54
14
18
16
5
1

88
70
76
88
100
138

Negro and other races_________
8 years or less____________
9-11 years_______________
12 years...... ...........................
13 years or more....................

469
147
212
97
13

46
14
21
9
1

74
54
66
80
115

1 16 to 44 years of age.
2 Poverty status as of 1967.

earned $100 a week or more.
The earnings potential of women heading poor
families is even more restricted because of limited
formal education. Nearly 70 percent of the 1
million in the 16- to 44-year age bracket never
completed high school; 300,000 never went beyond
elementary school. More than half of the least
educated are black. Negro women with less than
a high school education were earning only $60 a
week in the spring of 1969, even working at full­
time jobs. M any were working in domestic and
other service activities not covered by minimum
wage legislation and where hourly pa}' scales are
still comparatively low.
If all women heading poor families were to
become employed at jobs with weekly earnings
commensurate with their education levels, and
assuming that they would be subject to prevailing
practices of racial and sex discrimination in hiring
and pay scales, the}* would earn an average of
about $74 per week (as of the spring of 1969).
D ata from the Work Incentive Program show that
the average w i n graduate in a followup sample
was earning about $2 an hour or roughly $80 a
week. A woman who earned that much, and who
worked every week of the year, would make enough
to support herself and her family above the poverty
standard if she had no more than three children.
Women who can be trained to fill clerical,
technical, and lower grade professional jobs, and
who stay on those jobs on a regular year-round
basis, could expect to earn between $5,000 and

$7,500 a year, on the average. On the other hand,
average earnings are much lower in semiskilled
manual occupation and in service (excluding
domestic) occupations, where about two-fifths of
the female heads age 16 to 44 who work at all are
clustered. Year-round work in these occupations
would yield annual earnings of about $4,500 and
$3,500, respectively.

high, the woman’s earnings would have to be
considerably higher to equal welfare payments,
since State welfare benefits would not be reduced
under the proposal.
Of course, any increase in a woman’s earning
power would at least reduce her welfare subsidy.
It would be important, therefore, to take account
of trends in the average paym ent per fam ily, in
addition to the total number of beneficiaries, if
an integrated income support and em ployability
program were to go into effect.
The main issue in any em ployment strategy is
whether the incentives can be made strong enough
to induce welfare recipients to accept training and
jobs. In the recent controversy over the Fam ily
Assistance Program, proponents of the bill pointed
to the provisions for child care, training, job
counseling, and job placement, and to the flexi­
bility in program design to meet the individual
needs of each beneficiary. T hey stressed that the
poor in this country are imbued with a strong work
ethic, needing only the opportunity to exercise it.
T hey emphasized that the act was so designed that
the tax and benefit provisions would always make
it more profitable for a recipient to work than not
to work. For the small minority who might other­
wise reject the opportunity, the act includes a
provision requiring adults to register with the U.S.
Em ploym ent Service unless exempted because of
illness, age, or in the case of female fam ily heads,
the presence of children under 6. Opponents of the
act raised a number of questions about the appro­
priateness and effectiveness of the work require­
ment in the case of mothers. Skepticism was voiced
about the availability of jobs; about the costeffectiveness of child care and training; and, above
all, as to whether the monetary incentives would
be strong enough to offset the loss of welfare pay­
ments and in-kind benefits (food stamps, medicaid,
etc.) associated with increased earnings.
Perhaps some answers will be forthcoming from
experimentation with income maintenance pro­
grams which is now under way in several com­
munities. In the meantime, the data available on
the work experience, occupational and educational
backgrounds, and, particularly, the earnings of
women fam ily heads do give some useful per­
spective on the feasibility of providing em ploym ent
as a substitute for welfare.
□

Program s to upgrade em ployability

Paid work would appear to be a logical solution
to the income problems of many welfare mothers.
However, the data point up several constraints
operating against any employment strategy. If
em ployment is to be effective in raising family
standards, it must be full time and year round.
E ven for the mother of a small or average-sized
fam ily, the cost and difficulty of finding adequate
child care, and the lack of sufficient education and
job training, are formidable barriers to steady
work at good wages.2 For mothers of large families,
these problems are compounded because their
fam ily responsibilities are greater, and their
income needs are larger.
In an effort to overcome these barriers to
employment, Federal programs such as the Work
Incentive Program ( w i n ) and the proposed Fam ily
Assistance A ct ( f a p ) have been developed in
recent years.3 B oth of these programs have train­
ing, job placement, and child care provisions which
are designed to enable employable adult mem­
bers of poor families to find jobs and gain economic
independence.
The benefit and tax rate schedules under f a p
provide some idea of how much a mother would
have to earn to get off welfare completely. If a
four-person fam ily received $3,920 or more in
earned income, its Federal income supplement
would be eliminated entirely. The earnings equiva­
lent of that annual income would be roughly $2
an hour for 2,000 hours of work, or $80 a week for
at least 50 weeks. The head of a six-person family
would have to earn more than $2.50 an hour or
over $100 a week all year long before the income
supplement would phase out completely. In many
northern States (Connecticut, M assachusetts,
N ew Jersey, N ew York, Pennsylvania, Minnesota,
in particular), where a f d c payments are relatively




122

FOOTNOTESa n d a d iscu ssio n of th e re lia b ility of th e d a t a a re c o n ta in e d
in C u rre n t P o p u la tio n R e p o rts S eries P -6 0 , p u b lish e d b y
th e B u re a u of th e C en su s.

T h e d a ta in th e ta b le s a n d m u c h of th e d a ta u n d e rly in g
th e te x t fo r th is a rtic le w ere o b ta in e d fro m th e C u rre n t
P o p u la tio n S u rv e y ( c p s ) w h ic h is c o n d u c te d b y th e B u re a u
of th e C ensus, in p a r t fo r th e B u re a u of L a b o r S ta tis tic s .
T h e th r e e p rin c ip a l so u rces of in fo rm a tio n w ere th e s u p p le ­
m e n ta ry in q u irie s on fa m ily in co m e, on w o rk ex p erien ce,
a n d on w e e k ly e a rn in g s. D e ta ile d ta b u la tio n s on th e s e
s u b je c ts w ere m a d e a v a ila b le b y th e P o p u la tio n D iv isio n ,
B u re a u of th e C e n su s; th e Office of M a n p o w e r a n d E m ­
p lo y m e n t S ta tis tic s , B u re a u of L a b o r S ta tis tic s ; a n d th e
Office of R e se a rc h , M a n p o w e r A d m in is tra tio n , D e p a rtm e n t
of L a b o r. F o r a d e sc rip tio n of th e C u rre n t P o p u la tio n
S u rv e y , see b l s R e p o r t 313, “ C o n c e p ts a n d M e th o d s u sed
in M a n p o w e r S ta tis tic s fro m th e C u rre n t P o p u la tio n S u r­
v e y .” A n e x p la n a tio n of th e in c o m e a n d p o v e rty co n c e p ts




1 F o r a d iscu ssio n of th e uses a n d lim ita tio n s of p o v e rty
s ta tis tic s , sec M ollic O rsh a n sk y , “ H o w P o v e rty is M e a s­
u re d ,” M onthly Labor Review, F e b ru a r y 1969, p p . 3 7 -4 1 .
2 Sec G e n e v ie v e W . C a rte r, “ T h e E m p lo y m e n t P o te n tia l
of a f d c M o th e rs ,” Welfare in Review, J u ly - A u g u s t 1968,
p p . 1 -1 1 .
3 F o r a d e sc rip tio n of th e s e p ro g ra m s, sec th e Work
Incentive Program, F ir s t A n n u a l R e p o rt of th e U .S . D e ­
p a r tm e n t of L a b o r on T ra in in g a n d E m p lo y m e n t, 1970.
Also see T h e F a m ily A ssistan ce A ct of 1970, n o w p e n d in g
in C o n g re ss.

123

Women find jobs in the
fastest growing industries,
but remain clustered
in fewer occupation groups
than men
ELIZABETH WALDMAN AND
BEVERLY J. McEADDY

The last three decades have been years of
extraordinary economic and social change in the
status of women—the tremendous response of
married women to labor market demand; an
increasingly service-oriented economy, accompa­
nied by an increased need for white-collar workers;
changing attitudes toward careers for women
outside the home; the trend toward smaller fami­
lies; the increase in the number of households
headed by women; and landmark legislation pro­
hibiting employment discrimination based on sex.
Despite these changes, today’s figures on the
employment o f women in American industry bear
a striking resemblance to those o f yesterday. In
1970, just as in the three previous census years—
1940, 1950, and 1960—the service industry ranked
first in the employment o f women. Over this 30year span, about 60 percent of all em ployees in
the service industry were women— some 60 per­
cen t o f the w orkers in ed u cation al ser v ic es;
around 75 percent in the medical-health industry;
and about 75 percent in personal services, includ­
ing those in hotels and private hom es. Within
other major industrial categories, such as manu­
facturing and trade, certain subgroups remain as
fem ale-intensive today as they were yesterday.
Examples are the manufacture of clothing and gen­
eral merchandising, where at least 50 percent of all
em ployees are women. (See table 1.)
The recordbreaking growth achieved by Ameri­
can industry since 1940 was made possible, in
part, by the phenomenal increase in the number
and proportion o f w om en, esp ecially married
wom en, who were able and willing to join the
work force. From 1940 to 1970, nonagricultural
employment of all persons expanded from 32.1
million to 69.1 million, with women nearly half of
Elizabeth Waldman is an economist and Beverly J. McEaddy
a social science research analyst in the Division of Labor
Force Studies, Bureau of Labor Statistics.

From the R eview of May 1974



Where women
work— an analysis
by industry
and occupation
the increase (18 million). In 1940, women were 31
percent o f all workers in nonagricultural indus­
tries; by 1970, 40 p ercen t, as their num bers
alm ost tripled to 27Vi million. In 1940, alm ost
half the women in the labor force were single and
only 30 percent were married; in 1970, about 20
percent were single and 60 percent were married.
Over these three decades, the labor force partici­
pation rate o f married women rose from 15 to 41
percent and the rate o f mothers with children
under age 6 from 9 to 30 percent.
The enorm ous expansion in the labor force
participation o f women has sometimes been re­
ferred to as the response o f married women to the
tidal w ave o f paperwork that occurred in the
industrial world o f the 1950’s and 1960’s. The
population explosion o f post-World War II con ­
tributed to the need for expanding all types o f
services— among them, medical, educational, per­
sonal, and recreational—thus generating more jobs
o f the types considered to be traditionally female.
Many job s in the service industry can be de­
scrib ed as e x te n sio n s o f w hat w o m e n do as
homemakers— teach children and young adults,
nurse the sick, prepare food.
Another factor contributing to the concentration
o f women in the service industries is that parttime employment is more readily available there
than in other major industry categories (with the
exception o f retail trade). In recent years, about
one-fourth o f all employed women held part-time
jobs. A lso, many service industries employ full­
tim e w ork ers but op erate at oth er than the
standard 9-to-5 schedule—for example, hospitals,
schools, libraries, and hotels. Shift work or other
atyp ical h ours o f em p loym en t may be more
attractive to women who have children.
The following sections discuss in greater detail
the trends in w om en’s employment by industry
and occupation, and are based on (a) establish­
ment data from the monthly nationwide sample

124

survey o f nonagricultural payrolls; (b) the Current
Population Survey (CPS), a monthly nationwide
sample survey of households; and (c) the U .S.
Decennial Census o f Population. The establish­
ment series provides a count o f jo b s; the CPS and
Census provide a count of individuals. Despite
these differences, data on the industrial employ­
ment of women from one series complements and
confirms the trends indicated by the others.1
Table 1.

Establishment data

Payroll statistics from establishments in non­
agricultural industries provide one o f the most
detailed, up-to-date appraisals o f the employment
o f women in American industry. These data also
permit more precise industry identification than
that obtained through the household interviews of
either the decennial cen su ses or the monthly

Women employed as wage and salary workers in nonagricultural industries, 1940-70

[Numbers in thousands]

1940

1960

1950

Women

Women

1970

Women

Women

Industry
Total
employed

Total nonagricultural industries......................

Per­
Number cent of
total

Education__________________ ____ _
Legal2_____________________________
Other services, including recreation and amusement
Public administration 1_______________________
Postal service___________________________
Federal public administration____ __________
State and local____________________ _____
Local

..

.

. . __________

6,984
3,268
2,196
583
2,796
745
0
0
1,514
0
0
156
338
1,758
309
299
848
('1
0
800

Per­
Number cent of
total

Total
employed

Per­
Number cent of
total

31

43,478

14,113

32

54,579

19,449

36

69,115

27,496

40

12
33
2,323
604
1,719
130
522
139
197
32

1
2
23
12
33
60
68
38
19
15

880
2,752
14,053
7,460
6,478
189
1,036
371
1,328
267

22
86
3,594
1,219
2,337
119
754
167
311
55

2
3
26
16
36
63
73
45
23
21

626
3,062
17,142
9,621
7,464
195
1,131
345
1,757
311

30
130
4,354
1,707
2,627
132
858
173
417
76

5
4
25
18
35
68
76
50
24
24

616
3,976
19,566
11,596
7,970
962
1,201
282
1,364
0

50
244
5,623
2,483
3,140
445
939
160
356
0

8
6
29
21
39
46
78
57
26
0

32

36

140

57

41

202

85

42

0

36
76
345
210
5
17
1,669
174
1,495
468
385
181
435

51
18
12
54
22
5
30
17
33
65
47
51
34

73
642
4,138
637
0
545
8,122
1,687
6,434
873
1,266
408
1,670

37
129
666
389

51
20
16
61

42
3,013
362
2,651
595
719
266
739

8
37
21
41
68
57
65
44

75
850
4,268
818
87
776
9,653
1,943
7,710
1,211
1,429
442
2,417

38
163
757
426
22
61
3,835
422
3,413
821
912
317
1,181

51
19
18
52
25
8
40
22
44
68
64
72
49

0
980
5,039
1,071
131
992
13,810
2,907
10,903
2,005
2,061
667
3,610

0
222
1,106
522
33
106
5,871
699
5,172
1,392
1,271
437
1,870

0
23
22
49
25
ii
43
24
47
69
62
66
52

4,321
2,449
1,931
64
1,736
543
0
0
976
0
0
85
72
350
36
104
203

62
75
88
11
62
73

58
72
88
16
63
74
77
72
64
60
65
68
27
26
12
34
25
38
21
42

11,668
3,247
1,880
1,181
6,803
2,223
544
1,679
3,292
762
2,530
144
437
3,194
551
1,266
1,377
396
981
2,548

7,241
2,464
1,701
287
4,352
1,710
448
1,262
2,062
464
1,598
109
138
909
65
444
400
152
248
1,013

62
76
90
24
64
77
82
75
63
61
63
76
32
28
12
35
29
38
25
40

11,436
2,256
(1,124)
624
8,352
3,096
1,039
2,057
3,788
957
2,831
178
204
1,297
144
546
557
202
355

63
75
0
31
66
79
83
77
62
62
62
46
36
31
20
36
31
38
28

38

5,007
2,097
1,405
155
2,623
1,006
302
704
1,283
281
1,002
81
131
648
52
339
258
100
158
338

18,282
3,010
0
2,006
12,707
3,907
1,249
2,658
6,080
1,550
4,530
386
559
4,216
719
1,528
1,824
538
1,286

305

8,584
2,930
1,598
999
4,166
1,365
393
972
2,019
470
1,549
120
489
2,471
451
1,003
1,017
264
753
809

64
54
21
20
12
35
24

0

0

NOTE: Because some industries are not included in this table, subgroups do not
always add to total for major industrial division.

1 Data not available.
2 1940 figures include engineers and miscellaneous professionals.
1 1940 figures are for government instead of public administration.




Per­
Number cent of
total

Total
employed

9,794

32,058

Mining____________________________________
869
Construction__________ ______ ______________
1,603
Manufacturing............................. ..............................
10,317
Durable goods___________________ _______
5,162
Nondurable_____________________________
5,155
Knitting m ills___ ____ ______________
217
Apparel, etc________________________
768
Leather and leather products.....................
363
Food and Kindred products____ ________
1,054
Meat products......................................
207
Canned and preserved food and seafood.. _____________________
89
Confectionary and related products_____ ________ ________
70
Chemicals and allied products__________
433
Transportation and public utilities_____________ _
2,911
392
Telecommunications__________ . . . . . . .
Radio and TV______
23
Trucking and warehouse.______ ___________
330
Wholesale and retail trade____________________
5,522
Wholesale______*_______________________
1,009
Retail_________________________________
4,514
General merchandise.______________ _
718
817
Eating and drinking__________________
356 .
Apparel and accessory s to re s... ______ l
1,294
Finance, insurance and real estate........... ....... .........
Services_________ . . . ____ ________
___
Personal, including private household and hotels
Private households..
___ . . ___ ._
Business and repair_____________ ________
Professional service______________________
Medical and health________ _________
Medical and health, except hospitals..

Total
employed

SOURCE: Census of Population, Industrial Characteristics, 1940 (Vol. Ill),
1950 (P-E No. ID), 1960 (PC(2) 7F), 1970 (PC(2) 7C); (Bureau of the Census).

125

Current Population Surveys. Payroll data were
first collected in 1919, but until 1964 information
on women was available on a regular basis for
only a few selected industries. Today, detailed
tables on the employment o f women in more than
400 industrial categories are published quarterly by
the Bureau of Labor Statistics.2
The follow ing discussion uses establishm ent
data to review recent changes for women in major
industry divisions, with an eye on prospective
trends, and describes changes in the occupational
mix within industries.
From January 1964 to January 1973, the num­
ber o f w om en on p ayrolls in nonagricultural
industries expanded from 19.1 to 27.9 million.
(S e e tab le 2 .) M arried or form erly m arried
w o m en , m any re sp o n sib le for sch o o l-a g e or
younger children, accounted for the largest share
o f the increase in payroll employment o f women.
Most o f the 8.8 million labor force entrants or
reentrants found jobs in the four major industry
divisions that were the fastest growing:
Millions
o f workers
Services ................................................................ 2.5
Government ........................................................ 2.4
Wholesale and retail trade ................................
1.9
Manufacturing ....................................................
1.1
In the late 1960’s, the service industry main­
tained its position as a principal em ployer o f
wom en, and by 1973 had more female workers
(6.8 m illion) than any other industry. O f the
several industries within the service sector that
recorded a robust expansion, the most spectacular
was health care services. The forces that contrib­
uted to this industry’s growth—gains in the size o f
the population, a rising affluence that enabled
more persons to afford health care and to demand
improved services, and increases in the roles of
special programs covering medical and health
serv ices, such as medicare and medicaid— are
expected to continue and to bring similar rapid
employment increases in the near future. It seems
likely that this industry, in which 8 out o f 10
em p lo y ees are w om en, can continue to b e a
source o f jobs for women.
In January 1973, the trade industry was the
second largest employer o f women (6.3 million),
most o f whom held jobs in retail stores. Women
were only one-fourth (900,000) o f the em ployees
in wholesale trade, but nearly half (5.4 million) in




retail trade. Within retail trade, women made up
two-thirds o f the em ployees in department stores,
clothing and accessory shops, and drugstores, and
over h alf in restaurants and other eating and
drinking establishm ents. N ew job openings in
trade during the rest of the 1970-80 decade are
expected to be little more than half what they
were in the 1960’s, because of the greater use o f
such laborsavers as computers, automated equip­
ment, self-service stores, and vending m achines.3
Manufacturing still employs the largest share o f
the male work force, but since the mid-1960’s has
dropped to fourth place for women. In part, this
reflects the fact that in some nondurable goods
industries employing relatively larger proportions
o f w om en— textile mill products, apparel and
related item s, and food and kindred item s—
increased automation and other improved plant
processes have boosted output without any great
increase in employment. During the 1970’s, the
need for additional workers in manufacturing is ex­
pected to be largest in such durable goods industries
as machinery, rubber and plastic products, and in­
struments, all currently male-dominated.
In the same way that the phenomenal growth o f
the service industry in the private sector made
jobs available for women, services provided by
govern m en t agen cies were responsible for the
soaring em ploym ent o f wom en on governm ent
payrolls, especially at the local and State levels.
Nearly half (1.1 million) o f the entire 1964-73
increase in w om en’s jobs in government occurred
at the local level in one industry— education.
Two-thirds o f all em ployees in schools and related
educational activities supported by city, county,
and other local tax jurisdictions were women. In
January 1973, the local and State education indus­
try accounted for nearly 60 percent of the 6.1
million women on government payrolls. Demand
for workers in the education field is expected to
taper off considerably as a result of the decline in
birth rates that began in the late 1950’s. In
contrast, government health and welfare services,
industries in which women are also prominent,
are expected to increase at a rapid pace through
the remainder o f the 1970’s.
The industry division encom passing fin ance,
insurance, and real estate became predominantly
female during the 1960’s, and by January 1973
women were 52 percent o f the em ployees (the
1940 and 1950 censu ses reported much smaller
proportions, 34 and 44 percent, respectively). The

126

Table 2.

Women employees on nonagricultural payrolls, by selected industries, January 1964 and January 1973

[Numbers in thousands]
1964

1973

Industry group
Number of
women

Percent of total
employed

Number of
women

Percent of total
employed

Total nonagricultural industries____ ___________________ __________ _.

19,096

34

27,920

38

P rivate...___________________ ____________________ _______________
Mining_________ ____ __________________ ____ ______________________
Construction_______________________________________________________
Manufacturing_____________________________________________________
Durable goods____ . ___________________ _________ _______
Fabricated metal products________________
_____
Machinery, except electrical____
__________________ _______
Electrical equipment and supplies____ _______________________
Transportation equipment____________________________________
Instruments and related products...______ ____________________
Miscellaneous manufacturing_______ ______ _______________
Nondurable goods___ _ _____________ . . . . . . . . .
Food and kindred products___________ _____ . . . ___________
Meat products_________________________________ _________
Poultry dressing plants_______________________________
Canned, cured, and frozen foods_____________ ____________
____ _
Canned, cured, and frozen seafoods_________ _
Confectionary and related products____________ ____ ________
Tobacco manufacturers____________________ _______ _________
Textile mill products_____ ______________________ ____________
Knitting mills__________ ________________ _____ _________
Apparel and other textile products_________________ ___________
Printing and publishing_________________ _______ .
.
.
Periodicals... ________ _ . . . _____________ ______ . . .
Blankbooks and bookbinding______ _______________ _____ . .
Chemicals and allied products_________________________________
Leather and leather products.__ _ __________________ _______

15,421
34
143
4,385
1,717
192
201
571
168
123
145
2,668
387
79
35
85
20
39
40
373
134
994
270
33
21
160
179

33
6
6
26
18
17
13
37
10
34
40
37
23
25
53
42
58
51
46
43
67
79
29
48
45
19
53

21,854
37
193
5,464
2,357
264
297
781
199
183
179
3,107
420
94
52
89
21
41
30
467
174
1,062
366
34
29
208
175

37
6
6
28
21
19
15
41
11
38
43
39
25
28
55
39
56
51
42
46
65
81
34
50
51
21
60

Transportation and public utilities_____
.
...
. . _.
Communications.. . . . . . . . . . .
.
___
Telephone communication... . . . __________________
_ _____
Radio and television broadcasting________ ______ ____ ___________

706
410
380
22

18
50
56
22

949
542
493
34

21
47
51
25

Wholesale and retail tr a d e ___ _________________ . _____ _______ . . .
Wholesale trad e.. _ . . .
___________ _____ _________ _______
Retail trade.. . . . .
__________________
_______ _______ _
Retail general merchandise__________ __________ _____ _______
Food stores__________________________________ _____ ________
Apparel and accessory stores____ ___________________ ________
_______
Eating and drinking places__________________ ______ _
Miscellaneous retail stores____________________ ______ . . . . .
Drug stores and proprietary stores__________________________

4,404
686
3,718
1,163
451
387
969
427
222

37
22
43
70
32
65
56
42
58

6,338
912
5,426
1,708
694
505
1,431
620
295

40
23
46
68
37
66
55
46
62

Finance, insurance, and real estate.__________________ ____ ____ ____ ___
Banking.. ________
_ _____ . ______ . . . _________
Credit agencies other than banks___
______ ____________ _______
Security, commodity brokers, and services._ .
. ____ . . . . . . . .
Insurance carriers______ ________ ______________________ . . _____
Insurance agents, brokers, and service_______ . ___________________
Real estate____________________________________________________

1,445
454
167
38
435
124
190

50
60
54
31
49
56
36

2,070
721
234
68
578
172
250

52
64
57
35
52
59
34

Services__________ . . .
.
___________________________
Hotels, tourist courts, and motels_______________ __________________
Personal services__________ _____________ __ ________ _______
Miscellaneous business services________________ __________________
Advertising_____ ___________ ______________________________
Credit reporting and collection__ _______ _______________________
Services to buildings.__________________________ ____________
Medical and other health services_____________ _____________________
Hospitals._____ ________ _______ _____ ______________________
Legal services_____________________ _____ ___________ ______ ____
Educational services_____________________________________________
Elementary and secondary schools..----------- ------- ------------- ------ Colleges and universities............................................................. ...............

4,304
245
553
333
40
43
42
1,474
1,029
105
398
175
197

51
48
60
34
37
70
27
78
82
62
44
58
37

6,803
346
555
600
50
57
119
2,850
1,641
171
593
255
272

55
52
62
35
43
71
35
80
80
63
49
61
42

Government............. ............................................................... ............... .......................
Federal___________ ______ _______ ______________ _____ ___ _____ ____
State......................................... ............................................... .............................. .
State education............................................................................. ....................
Other State government.......... .............................................. ...........................
Local............................................................ ............................ .............. ..................
Local education............................................................... ................................
Other local government..---------------------------- ---------------------------------

3,675
520
692
245
448
2,463
1,831
633

39
22
38
40
37
46
63
26

6,066
767
1,248
535
713
4,050
2,956
1,095

45
29
43
43
43
50
63
32

NOTE; Because some industries are not included in this table, subgroups do not
always add to total for major industrial division.

SOURCE; Bureau of Labor Statistics,




127

1964-73 expansion in this industry’s job s for
women occurred primarily in banking and insur­
ance. These industries and credit agencies are
ex p ec ted to con tin ue to expand through the
remainder of this decade, providing new oppor­
tunities for women.

which stemmed from the extraordinary demands
for clerical support made by the education and
health service industries.
In the 30-year span, the proportion of women in
the service industry who perform service jobs
increased from 20 to 28 percent. Examples of
occupations still dominated by women are food
services, practical nurses, and dental assistants.
In p r o fe ssio n a l-te ch n ica l o cc u p a tio n s, 2
million women are teachers in elem entary and
high school. Women also predominate in regis­
tered nursing, social work, libraries, d ietetics,
physical therapy, and dental hygiene.
The 1940-70 redistribution shifted men into the
more p restigiou s, better paying p rofession altechnical group. In the education industry, about 70
percent o f the teachers in colleges and universities
are men; about 70 percent of the teachers in elem en­
tary and high sc h o o ls are w om en. D o cto rs,
lawyers, engineers, and many other professional-

Occupation by industry

W om en, like m en, find job s in the fa stest
growing industries. H ow ever, no matter what
industry women are in, they remain clustered in
fewer occupation groups than men. (See table 3.)
Service. As pointed out earlier, the service indus­
try employs the largest number of working women
and ranks third in the employment o f men. In
1970 as in 1940, most women in this industry were
employed in the same three occupation groups:
professional-technical, services, and clerical-sales
(chart 1). Yet there was a striking redistribution
Table 3.

Occupation group of employed wage and salary workers in nonagricultural industries, by sex, 1970
Percentage in each occupation group
Total
Profes­
sional,
techni­
cal and
kindred
workers

Managers
and
admini­
strators

Sales
workers

Clerical
and
kindred
workers

Craft
and
kindred
workers

Opera­
tives,
except
tran s­
port

100.0
100.0

16.6
14.6

3.2
10.5

7.4
6.9

37.5
8.6

1.8
22.2

15.0
15.2

0.5
6.4

0.9
7.0

17.1
8.8

Industry and sex
Number
(in thou­ Percent
sands)
Total: Women.......................................... 26,373
Men................................................ 41,619

Trans­
port
equip­
ment
opera­
tives

Laborers,
except
farm

Service
workers,
except
private
house­
hold!

Women__________
Men_____________

294
4,297

100 0
100 0

5.1
5.9

3.4
7.2

1.0
.7

72.1
2.8

8.5
53.2

3.7
9.8

.7
5.0

2.7
14.4

2.4
1.1

Manufacturing: Durable goods: Women_____
Men_______
Nondurable goods: W omen...
Men_____

2,483
9,113
3,140
4,831

100 0
100.0
100.0
100.0

3.8
14.0
3.9
10.0

1.1
5.8
1.0
7.8

.5
1.9
1.3
5.9

33.7
6.4
22.5
,6.8

4.8
26.7
4.4
23.8

52.6
34.0
63.9
30.8

2
3.3
.2
5.9

1.9
5.5
1.7
5.7

1.3
2.4
1.0
3.3

Transportation and public utilities: Women___
Men...........

1,106
3,934

100 0
100.0

4.2
8.8

3.2
8.1

1.3
1.3

74.6
11.0

2.0
28.6

1.1
3.9

7.7
25.3

.8
10.2

5.2
2.8

Wholesale trade: Women........................... .........
Men______ _______ ______

699
2,208

100 0
100.0

2.7
5.0

4.1
16.9

5.3
24.6

67.0
10.0

2.3
12.5

13.7
7.3

.7
14.2

2.6
7.9

1.6
1.5

Retail trade:

Women__________ ____ _____
Men.............................................

5,171
5,732

100 0
100.0

1.3
2.8

4.8
17.0

31.8
19.3

29.2
6.8

1.6
14.4

3.6
11.3

.2
5.5

1.3
9.3

26.2
13.5

Finance, insurance, and real estate: W omen...
Men..........

1,870
1,740

100.0
100.0

2.9
7.9

6 2
26.1

7.3
32.4

79.7
19.6

.3
3.4

.3
.5

.5

.2
2.3

3.2
7.5

Services: W om en............................................ 10,312
Men............................................. ....... 6,846
Medical and other health services: Women. 3,096
Men___
811
Educational services: Women...................... 3,788
Men.......................... 2,292

100.0
100 0
100 0
100.0
100.0
100.0

36.9
40.4
33.1
34.5
59.0
60.0

2.5
9.6
1.5
7.2
2.6
9.2

.7
1.4
.2
.2
.2
.3

28.0
6.0
20.4
5.4
22.7
5.2

.6
11.8
.4
10.0
.4
4.4

3.1
3.9
1.6
3.0
.4
1.0

.1
2.3
(l)
1.4
.2
.7

.4
3.8
.3
1.8
.1
1.8

27.7
20.8
42.6
36.4
14.5
17.5

Public administration: Women______ ______ _
Men...... .........................

100.0
100.0

11.3
18.9

6.3
12.5

.3
.2

72.6
25.3

.8
9.7

.8
2.0

.1
2.0

.6
4.0

7.2
25.5

Mining and construction:

1,297
2,919

1 Excludes all women, but includes a few men, who were private household workers.
* Less than 0.05 percent.

SOURCE: Census of Population, 1970: Occupation by Industry, Report (PC(2)
7C), (Bureau of the Census).

NOTE: Because of rounding, sums of individual items may not equal totals.




128

technical occupations have remained substantially
male-intensive.
Chart 1. Occupational distribution of wage and salary work­
ers, selected industries, 1940 and 1970

MEN

WOMEN
Percent
100

SERVICE
*

‘

100

Professional and
technical workers

■75

75-

, Managers - — .
Clerical and /
sales workers

/

50'

50
/C raft workers x
Operatives —

25'

■25

Service workers
(excl. domestic)
^
(2.4 mil.)

Laborers
(2.4 mil.)

(10.3 mil.:

(6.8 mil.)

Trade. Since 1940, very little occupational change
has taken place for women in trade. Nearly 9 out
of 10 women in this industry were in retail trade,
where sales, clerical, and service jobs predomi­
nated, in that order.
In general m erchandising, w om en held the
greatest proportion o f sales jobs, as well as such
clerical positions as bookkeepers, cashiers, office
machine operators, secretaries, and typists. A few
more were in managerial and administrative jobs
as buyers and sales managers, and a relatively
small number were in the operative group as
dressmakers and seam stresses. Women working
in eating and drinking establishments were mostly
waitresses, cooks, and clerical workers.
A smaller proportion of men (7 out of 10) were
in retail trade. Changes from 1940 to 1970 in their
occupational pattern reflect mainly the relative
increase in the need for managers and skilled craft
workers (carpenters, electricians, mechanics, and
repairers) and a decrease in the need for clerical
and sales workers.

TR AD E
„

...

Manufacturing. In both 1940 and 1970, approxi­
m ately 9 out o f 10 w om en w orking in the
manufacturing industry held semiskilled operative
or white-collar clerical jobs. Nearly three-fifths
were engaged in the production o f nondurable
goods. In this sector, most women work in the
production end as operatives (for example, assem ­
blers), as checkers, exam iners, and inspectors,
and as sewers and stitchers. About 11 percent of
all professionals in nondurable goods were women.

-^technical workers
^

/

M a n a g e rs /^

Clerical and
sales workers

/
Operatives----^ Service workers \
^ Laborers -—
(1.7 mil.)

(5.9 mil.)

(3,9 mil.)

M A N U FAC TU R IN G




(7.9 mil.)

G o v e rn m e n t. About 22 percent o f w om en on
nonfarm p ayrolls are governm ent em p loyees,
mostly (86 percent) in State and local govern­
ments working as teachers, administrators, cleri­
cal w orkers, maintenance w orkers, librarians,
nurses, and counselors.
A ccord in g to a su rvey by the U .S . C ivil
Service Commission4 o f women employed by the
Federal Government in 1970, m ost worked in
nonprofessional administrative, clerical, and office
service jobs. Roughly 77 out o f 100 (compared
with 44 o f 100 men) were in grades GS-1 through
- 6 in 1970. A n oth er 20 out o f 100 w om en
employees (compared with 32 of 100 men) were in
grades G S-7 through -1 1 , and only 3 out o f 100
w om en (but 23 out o f 100 men) w ere in the
highest paying grades, G S-12 and above.
W omen’s share o f full-time white-collar Federal
em ploym ent by individual grade reveals even

129

more dramatically their concentration in the non­
professional grade series. (See table 4.) Grades in
w hich w om en predom inated, G S-1 to - 4 , are
entry-level clerical and support positions in the
nonprofessional job series. G S-5 is the entry
grade for professional employment. Positions in
G S-5, -7 , -9 , and -11 are primarily professional,
technical, or administrative jobs, requiring a bac­
calaureate or higher degree or equivalent profes­
sional, technical, or administrative experience.
O ther industries. Most of the occupations women
hold in the finance, insurance, and real estate
industry are clerical— about 80 percent. Approxi­
mately 9 out o f 10 bank clerks and tellers are
women, but very few bank officers are. In the
transportation and public utilities industry, about
20 percent o f the em ployees are women; over half
o f these women work in communications, where 9
out o f 10 are telephone operators. O f wom en
employed in the construction and mining indus­
tries, most are in clerical occupations.

Occupational shifts
Overall, the 1940-70 changes in the occupa­
tional pattern o f women, as well as of men, mirror
the shift from the predominantly goods-producing
econom y prior to World War II to the serviceproducing econom y o f the 1970’s. For women,
the occupational pattern changed from half blue,
half white collar to one that is decidedly white col­
lar. Today, over 60 percent o f women and 40 per­
cent o f men employed in nonagricultural industries
are in white-collar work. (See chart 2.)
In 1940, w om en working in nonagricultural
industries were concentrated in three broad occu­
pational groups. Roughly half held service and
blue-collar operative jobs (30 and 21 percent), and
a third w ere in the w hite-collar clerical-sales
group. For those in services, prominent occupa­
tions were waitress, hairdresser, cook, and practi­
cal nurse. Operatives were mostly engaged in the
manufacture o f clothing (sewing machine opera­
tors). The clerical-sales group were mainly ste­
nographers, typ ists, secretaries, bookk eepers,
cashiers, and retail saleswomen. Of the 14 percent
in white-collar professional-technical jobs, 2 out
o f 4 were teachers and 1 out o f 4 were nurses.
Thirty years later, working women were still
highly con cen trated in the same three broad
occupational groups, but a much larger share




were in the clerical-sales field, the professionaltechnical proportion had edged up, and the service
and operative proportions had declined.
Do the changes represent an improvement in
the lot o f em ployed wom en? The shift out o f
service jobs as dom estics and into white-collar
jobs is a profound improvement, especially among
N egro w om en. And w om en now are a larger
share o f em ployees in a few o f the more presti­
gious, better paying occupations. In 1940 women
were only 1 out o f 20 physicians, compared with 1
of 8 today (1973 average). From 1940 to 1973, the
proportion o f real estate agents and brokers who
were women grew from 9 to 36 percent.
In the area o f job discrimination by sex, there
are a few “ breakthroughs” into typically mascu­
line jobs. For example, today 30 percent of all
bartenders are women, compared with
per­
cent in 1940, and about 37 percent o f the busdrivers are women, a rarity in 1940 when they were
less than 1 percent. H owever, many o f today’s
female busdrivers may operate school buses part
time, part o f the year, and for low pay. Thus,
what appears to some persons to be an occupa­
tional improvement— the movement o f women out
o f their homes with its unpaid housework and into
the paid labor force— to others represents no gain.

Self-employed women
Women have made considerable inroads into
the traditionally male-intensive province o f selfTable 4. Women as full-time white-collar employees in
Federal Government agencies,1 October 31, 1970
General schedule
(GS)
grade
1..........................................
2..........................................
3..........................................
4..........................................
5...................................... 6..........................................
7..........................................
8..........................................
9..........................................
10........................................
11........................................
12........................................
13........................................
14........................................
15..............- ........... ...........
16 and higher...... ...............

Salary1

$4,125
4,621
5,212
5,853
6,548
7,294
8,098
8,956
9,881
10,869
11,905
14,192
16,760
19,643
22,885
26,547+

Number of
employed
women

Women as
percent of
total employed

2,913
18,576
86,274
139,664
191,678
65,089
54,037
12,431
43,441
3,890
19,325
9,870
4,622
1,817
942
158

68
76
78
63
32
48
38
26
24
12
12
7
5
4
3
2

1 Excludes employees of Central Intelligence Agency, National Security Agency,
Board of Governors of Federal Reserve System, and foreign nationals overseas.
1 The rate for basic pay for employees is step 1 of the grade.

SOURCE: "Study of Employment of Women in the Federal Government, 1970,"
(Washington, U.S. Civil Service Commission, Bureau of Manpower Information Sys­
tems, 1971), pp. 17, 235.

130

em ploym ent, w here their share rose from 17
percent in 1940 to 26 percent in 1973. A total of
1.4 million women were self-employed in nonagricultural industries in 1973, nearly 600,000 more
than in 1940. Over this period, the number of selfemployed men rose only slightly to 4 million— a
minor increase, especially when compared with
the doubled employment of men in nonfarm wage
and salary work. The shift from small owneroperated businesses to corporate ownership con­
tributed to the lack of increase for men. Increased
demands in the more female-intensive service and
trade industries drew more women into entrepre­
neurship.
N early all o f the self-em p loyed w om en in
nonagricultural industries in 1973 were in service
and retail trade. In the service industry, over 6
out o f 10 were in personal services (operating
beauty shops, laundries, dressmaking shops, and
child care facilities) and 3 out o f 10 in professional
services (m edical enterprises, such as nursing
homes, and educational services).
The occupational distribution of self-employed
women differs from that o f women in wage and
salary jo b s. Proportionately more o f the selfemployed were managers or proprietors (24 ver­
sus 4 percent) and fewer were in clerical-sales (18
versus 43 percent), where self-employed women
are found in such fields as court stenography or
real estate sales. Self-employed women are generChart 2. Occupational distribution of wage and salary work­
ers, nonagricultural industries, 1940 and 1970

WOMEN

MEN
Professional and




ally older than wage and salary women, in part a
reflection o f the greater experience and maturity
necessary to run their own businesses or careers.
Median ages in 1973 were about 46 and 36 years.

Earnings
Further evid en ce that w om en have not yet
penetrated the high-skill, high-paying jobs is found
in the payroll data on weekly and hourly earnings
in nonagricultural industries. In January 1973,
most industries paying average weekly earnings of
less than $100 were fem ale-in ten sive. Several
were paying under $90 a week, while the weekly
paycheck for all industries averaged $138.5 Fig­
ures on hourly earnings, which exclude the effect
o f part-time and overtime work, support conclu­
sions based on weekly earnings. (See table 5.)
In 1972, w om en w ere 28 p ercen t (or 572
million) o f the total workers in the manufacturing
industry, yet most o f these women were concen­
trated in the lower paid and less-skilled jobs. The
average salary for all manufacturing workers was
$159 a w eek in January 1973. For th o se in
manufacturing industries that were female-inten­
sive, the average was much lower—for example,
the apparel industry, in which 81 percent o f
em ployees were w om en, paid average w eekly
salaries o f only $93.
The service industry— the most female-inten­
sive o f the major industry groups, with 55 percent
o f its w orkers w om en— em ployed 6.8 m illion
women in January 1973; earnings averaged $111 a
week. About 1.6 million women worked in hospi­
tals, where weekly earnings averaged $108. An­
other 600,000 women worked in hotels and laundries-drycleaners, where average weekly wages
were $76 and $87, respectively.
Another low-paying female-intensive industry is
retail gen eral m erch an d ise, w ith an average
w eekly wage o f $82. Although part-time work
undoubtedly accounts for some o f the low earn­
ings, most jobs in department stores and restau­
rants are known to be low paying.
M ale-intensive industries are on the higher
rungs o f the wage ladder: Construction—6 percent
female, paying $223 average a week; transporta­
tion and public utilities— 21 percent female, pay­
ing an average o f $196 a w eek (sw itchboard
operators averaged $126, line construction em ­
ployees $228); m anufacturing— here the list is
extremely long. In transportation equipment, 10

131

percent fem ale, average earnings were $210 a
week; in food products, the malt liquor industry
em ployed 7 percent w om en and the average
worker earned $229 a week. Among retail trade
industries, the most fem ale intensive were the
lowest paying. Yet em ployees on the payrolls of
motor vehicle dealers, only 11 percent of whom
were women, were among the highest paid work­
ers in retail trade— $152 a week.
Male-female earnings differentials have always
existed, but in recent years with the increase in
w om en’s labor force activity these differentials
have becom e the focus o f great concern. One
form of discrimination was the barring o f women
from the type o f jobs men held, such as skilled
workers and executives. A typical example was
cited in a standard college-level econom ics text;
. . . in a large electrical-goods plant, jo b evalu a­
tion experts divide all factory work into tw o parts:
w o m en ’s job s and m en ’s jo b s. The lo w est pay o f the
men begins about w here the w o m en ’s highest pay
lea v es off; yet both m anagem ent and the union will
adm it, o ff the record, that in many borderline job s
the productivity o f the w om en is greater than that o f
the m en .6

Many researchers believe the earnings differen­
tial results partially from the role that society has
traditionally arrogated to women. In a study on
wage differentials by sex, using 1959 and 1960
Census of Population and Housing data, Victor
Fuchs suggested that the percentage point differ­
ence in male-female wages “ can be explained by
the different roles assigned to men and women.
Role differentiation, which begins in the cradle,
a ffects the ch o ice o f occup ation , labor force
attachment, location o f work, postschool invest­
ment, hours o f work, and other variables that
influence earnings.” 7 He believes a reduction in
this role discrimination would eventually result in
a narrowing o f the m ale-fem ale differential in
earnings. Professor Fuchs also stated that con­
sumer discrimination (such as the preference of
custom ers in expensive restaurants for waiters
rather than waitresses) may be more significant
than employer discrimination.
Area wage surveys covering six industry divi­
sions recently published by the Bureau of Labor
Statistics indicate that on a nationwide basis, pay
levels were consistently higher for men than for
women working in the same occupation.8 In a
study o f pay differences between men and women
in the same job , John Buckley— while neither




denying nor confirm ing that wage discrim ina­
tion by sex existed—acknowledged that “ experi­
ence in implementing the Equal Pay Act indicates
that some discriminatory practices do ex ist.” 9
Differences in female-male earnings in the Fed­
eral Government occur in part because women
remain clustered in the lowest paid grades.
Table 5. Gross hours and earnings of production or non*
supervisory workers 1 on private nonagricultural payrolls,
selected industries, January 1973
Average
earnings

Average
hours

Industry
Weekly Hourly

Weekly

Over­
time

Total private................................................

$138

$3 . 7 7

3 6.6

-

Mining.....................................................................

190

4 .60

4 1.3

-

Contract construction................. ...........................

223

6 .42

3 4 .8

_

Manufacturing___________________________
Durable goods________________________
Fabricated metal products......................
Machinery, except electrical...................
Electrical equipment and supplies____
Transportation equipment___________
Miscellaneous manufacturing________
Nondurable goods_____________________
Food and kindred products__________
Canned, cured, and frozen foods...
Confectionery and related products.
Textile mill products_______________
Knitting mills_________________
Apparel and other textile products____
Chemical and allied products...............
Leather and leather products________

159

3.98

4 0.0

173

4 .23

4 1.0

169

4 .13

4 1 .0

188

4 .44

4 2.4

1 53

3.80

4 0.3

210

5 .00

4 1.9

124

3.24
3 .61

3 8.4

140
149

3 .75

3 9.8

119

3 .15

3 7 .8

125

3 .29

3 7.9

1 12

2 .8 7

39.1

99

2 .76

3 5.7

93

2 .72

3 4 .1

181

4 .36

41.5

103

2 .77

37.2

Transportation and public utilities___________
Telephone communication______________
Switchboard operating employees2___
Line construction employees3........ .......

3 8.7

3.6
3.9
3.9
4.5
3.0
4.8
2.4
3.2
3.8
3.0
2.2
3.9
2.4
1.2
3.5
1.9

196

4.87

4 0.2

—

175

4 .47

3 9 .1

126

3 .65

3 4.4

228

5 .23

4 3.6

—
—
—

2 .08

30.6
2 9.7

152

3.77

4 0 .2

82

2.67

3 0.7

—
—
—
—
—
—
—
—
—

Finance, insurance, and real estate......................
Banking..........................................................

131

3 .54

3 7 .0

—

114

3 .07

3 7 .0

—

Services...................................................................
Hotels, tourist courts, and motels4...............
Laundries and drycleaning plants..................
Hospitals..........................................................

111

3 .27

3 3 .9

—

76

2 .35

3 2.3

87

2 .50

3 4.7

—
—

108

3 .15

3 4.3

Wholesale and retail trade______ ____ ______
Wholesale trade___ __________________
Retail trade__________________________
Retail general merchandise__________
Food stores_______________________
Apparel and accessory stores............... .
Eating and drinking places4 _________
Motor vehicle dealers______________
Drug stores and proprietary stores........

107

3.11

3 4 .5

158
91

3 .99

3 9.5

2 .78
2 .61

32.9

82
102

3.20

3 2 .0

78

2 .54

62

3 1.3

1 Data relate to production workers in mining and manufacturing; to construction
workers in contract construction; and to nonsupervisory workers in wholesale and re­
tail trade; finance, insurance, and real estate; transportation and public utilities; and
services.
1 Data relate to employees in such occupations in the telephone industry as switch­
board operators; service assistants; operating room instructors; and pay station
attendants. In 1971, such employees made up 29 percent of the total number of
nonsupervisory employees in establishments reporting hours and earnings data.
3 Data relate to employees in such occupations in the telephone industry as central
office craft workers; installation and exchange repair craft workers; line, cable, and
conduit craft workers; and laborers.
4 Money payments only; tips not included.
SOURCE: Bureau of Labor Statistics.

132

It is not within the scop e o f this article to
explore in depth the various reasons for malefem ale d ifferen tials in p ay, but research has
continued in this area.10 Data presented here takes
the differential into account as an important factor
in wom en’s industrial characteristics.
Education

For today’s working woman to achieve profes­
sional status in the higher paying, traditionally
m ale-intensive occupations in many industries,
she must acquire more years o f formal higher
education. To illustrate, doctors, dentists, airline
pilots, metallurgists, architects, and certified pub­
lic accountants must have more years of schooling
than are needed for most occupations. For men,
the returns on the investment in education are
usually high in terms o f m oney and prestige.
Women, even with the required years of schooling,
often do not obtain returns equal to men’s.
The 1972 amendments to the Equal Pay Act
and the Civil Rights Act have outlawed many

11.2
30.3

17.4
37.2

18.3
36.1

13.9
32.5

11.9
33.6

7.9
25.5

4.0
19.4

9.0
20,7

10.3
15.5

9.5
18.4

8.8
25.6

10.7
21.7

9.7
24.7

3.9
12.8

Toward tomorrow

4.4
18.2
6.3

2.5
14.0
3.1

2.0
13.0
2.8

1.8
13.3
4.1

3.5
11.2
7.4

7.1
16.8
8.3

8.7
43.6
7.6

[Percent distribution]
Years of school completed
Total
Less
than
8

8

9
to
n

12

13
to
15

16
or
more

WOMEN
Total: Number (thousands). 29,968 1,153 1,547 4,987 13,995 4,376 3,910
Percent______ ____ 100.0 100.0 100.0 100.0 100.0 100.0 100.0
Mining and construction.......... .
Manufacturing________ _____ _
Transportation and public
utilities__________________
Trade......................... ..................
Finance, insurance, and real
estate_______ ___ _____ _
Service, except private household........................................ .
Public administration...................

1.2
19.4

0.4
41.1

0.8
35.0

0.5
28.3

1.4
19.9

2 0
9.7

0.6
4.1

3.7
22.9

1.0
21.9

1.0
27.1

2.3
33.7

4.9
24,6

4.2
19.1

2.1
6.0

7.5

2.2

2.3

3.3

10.3

10.3

3.5

40.9
4.4

31.5
2.0

32.2
1.6

29.5
2.3

33.2
5.6

48.8
5.8

80.2
3.5

MEN
Total: Number (thousands). 48,152 3,344 3,399 8,575 17,585 7,223 8,026
Percent....................... 100.0 100.0 100.0 100.0 100.0 100.0 100.0
Mining and construction..............
Manufacturing....... ....... ..........
Transportation and public
utilities................. ....................
Trade...........................................
Finance, insurance, and real
estate....................... ................
Service.................................... .
Public administration...................

SOURCE: Bureau of Labor Statistics.




U n em p lo ym en t
Labor
ra te
fo r c e rate
11 years or less .........................
8.6
32
12 years ......................................
51
5.3
4 years o f co lleg e or more . .
61
2.7

Y ea rs o f sc h o o l c o m p le te d

Women who have graduated from college earn
over twice as much annually as women at the
low est end o f the education scale. For women
em ployed year round in full-time job s, median
earnings in 1972 w ere $8,925 for the college
graduates, $5,770 for high school graduates, and
$4,305 for those who did not complete elementary
school. A similar education-earnings relationship
was evident for employed men, but their yearround, full-time earnings were substantially above
w om en’s—$14,660, $10,075, and $7,575, respec­
tively, for the corresponding groups.
Earnings in the different educational categories
are also a reflection o f the differences in the
industrial distributions o f em ploym ent. O f the
women in nonagricultural jobs who were college
graduates, 8 out of 10 were in service industries,
mostly professional services, in March 1973.11
(S ee table 6.) W omen who w ere high school
graduates, but had no college education, were more
widely distributed: (1) one-third in the service in­
dustry, largely professional services; (2) one-fourth
in trade, largely retail stores; and (3) one-fifth in
manufacturing. Among working women at the bot­
tom of the educational scale, mostly older women
who had either not completed or never attended
elementary school, two-fifths were in manufactur­
ing, largely semiskilled em ployees.

Table 6. Women and men employed in nonagricultural
industries, by educational attainment, March 1973

Sex and Industry

barriers to employment, among them job quotas
by sex and unequal male-female wage scales for
the same job. While educational attainment alone
is not a cure-all for working women (legislation
helps), the statistics on w om en’s education and
labor force participation indicate that more years
o f formal schooling would assist in equalizing
wom en’s position with that of men.
Annual surveys on educational attainment show
that, for both men and women, participation in the
labor force is lowest and unemployment rates high­
est for those who complete the fewest years of
school. The March 1973 survey shows:

Today’s working women are in the throes of
obtaining equal consideration with men in the job
market through the legislation that prohibits dis­

133

legislation prohibiting discrimination, provision of
child care services for mothers on industry pay­
rolls, and the ex ten t o f form al ed u cation or
technical training o f women for traditionally male
occupations.
CH

crimination in employment. Tomorrow’s working
women will be affected not only by the measure
of success achieved today, but also by econom ic
con d ition s, changes in lifestyle (for exam ple,
smaller fam ilies), the mode o f enforcem ent o f

-FOOTNOTES
1
Establishment data are based on payroll records compiled
monthly from mail questionnaires by the Bureau of Labor
Statistics in cooperation with State agencies. The payroll
surveys provide detailed industry information on wage and
salary employees in nonagricultural establishments. The Cur­
rent Population Survey (CPS) is conducted each month for the
Bureau of Labor Statistics by the Bureau of the Census. It is
based on household interviews obtained from a sample of the
population 16 years old and over
The CPS definition of nonagricultural employment com­
prises persons in nonagricultural industries who were wage
and salary workers (including domestics and other private
household workers), self-employed, or unpaid and working 15
hours or more during the survey week in family-operated
enterprises. The payroll survey covers only wage and salary
employees on the payrolls of nonagricultural establishments.
In the household approach, employed persons holding more
than one job are counted only once and are classified
according to the job at which they worked the greatest number
of hours during the survey week. In the payroll series, persons
who worked in more than one establishment during the
reporting period are counted each time their names appear on
payrolls. For example, workers may be counted more than
once if they hold two jobs concurrently or leave one job for
another during the same reference period and thus appear on
the payrolls of both employers.
The household survey includes among the employed all
persons who had jobs but were not at work during the survey
week—that is, were not working but had jobs from which they
were temporarily absent because of illness, bad weather,
vacation, labor-management dispute, or because they were
taking time off for various other reasons, even if they were not
paid by their employers for the time off. In the payroll series,
persons on leave paid for by the company are included, but
not those on leave without pay for the entire payroll period.
For a detailed description of these series and differences
between them, the following publications are available from
the Bureau of Labor Statistics: Concepts and Methods Used
in Manpower Statistics from the Current Population Survey,
Report 313 (Bureau of Labor Statistics, 1967); Handbook o f
Methods for Surveys and Studies, Bulletin 1711 (Bureau of
Labor Statistics, 1971), ch. 2; and Gloria P. Green, “ Compar­
ing employment estimates from household and payroll sur­
v ey s,” Monthly Labor Review, December 1969, pp. 9-20,
reprinted as BLS Reprint 2651.
The Decennial Census o f Population is conducted by the
Bureau of the Census to obtain a house-to-house enumeration
of each person by questionnaire. For an employed person,
1970 census information pertains to one specific job held
during the reference week. If employed at two jobs or more,
the job at which the person worked the greatest number of
hours during the reference week was to be represented.
Census reports on specific subjects, such as occupation by




134

industry, are frequently based on representative samples of the
total population. The intercensal statistics provided by the
Current Population Survey are generally designed to be
comparable to decennial census statistics. For a detailed
description of the 1970 decennial census and comparability
with earlier censuses and other data, see Census o f Popula­
tion: 1970, General, Social and Economic Characteristics,
Final R ep o rt P C (1 )-C 1 , U. S. Sum m ary (Bureau of the
Census).
2 Detailed establishment data on women employees were
first published in the May 1963 Employment and Earnings.
Since 1967, the data have been published in Employment and
Earnings once each quarter (February, May, August, and
November) as table B-3, “ Women Employees on Nonagricul­
tural Payrolls, by Industry.” In addition, annual and monthly
data appear in Employment and Earnings, United States,
1909-72, Bulletin 1312-9 (Bureau of Labor Statistics, 1973),
and the Handbook o f Labor Statistics, 1973, Bulletin 1790
(Bureau of Labor Statistics, 1974).
3 See Occupational Outlook Handbook, 1972-73 Edition,
Bulletin 1700 (Bureau of Labor Statistics, 1973).
4 See Study o f E m ploym ent o f Women in the F ederal
Government, 1970, Pamphlet SM 62-06 (Washington, U.S.
Civil Service Commission, Bureau of Manpower Information
Systems, 1971), table B.
5 See Employment and Earnings, May 1973, table B -3,
“ Women Employees on Nonagricultural Payrolls, by Indus­
try,” and Employment and Earnings, Aprif T973v~4able^ Cr2,
“ Gross Hours and Earnings of Production or Nonsupervisory'
Workers on Private Nonagricultural Payrolls, by Industry.”
6 Paul R. Samuelson, Economics (New York, McGraw-Hill
Book Co., 1967), p. 120.
7 Victor R. Fuchs, “Differences in hourly earnings between
men and women,” Monthly Labor Review, May 1971, pp. 915. See also pp. 23-26, this issue.
8 Area Wage Surveys: Metropolitan Areas, United States
and Regional Summaries, 1969-70, Bulletin 1660-92 (Bureau
of Labor Statistics, 1971).
9 John E. Buckley, “ Pay differences between men and
women in the same job,” Monthly Labor Review, November
1971, pp. 36-39.
10 See, for example, Paul O. Flaim and Nicholas I. Peters,
“ Usual weekly earnings of American w orkers,” M onthly
Labor Review, March 1972, pp. 28-38. Reprinted as Special
Labor Force Report 143.
11 See W illiam V. Deuterm ann, “ E ducational attain­
ment of workers, March 1973,” M onthly L abor R eview ,
January 1974, pp. 58-62. Reprinted as Special Labor Force
Report 161.

Special Labor Force Report reviews
employment gains of veterans
during the year ending in June 1972,
and new'data on occupations,
industry, and residence
KOPP MICHELOTTI AND KATHRYN R. GOVER

Job prospects brightened for veterans during the
year ending in June 1972, as young, newly separated
servicemen returned to an economy in which em­
ployment was generally on the rise while unemploy­
ment remained stable. The number of veterans with
jobs increased steadily during this period, and the
unemployment rate for Vietnam Era veterans1 in
ages 20 to 29 dropped a full percentage point to 8.0
percent (seasonally adjusted) in the second quarter
of 1972. Subsequently, the rate fell even further to
7.2 percent in the third quarter.
The civilian economy had to absorb fewer new
veterans, as military discharges declined. In fiscal
1972, discharges numbered 880,000, down from an
average of one million in each of the 3 preceding
years, reflecting in part the drop in Armed Forces
inductions that began about 3 years earlier.
At the close of fiscal 1972, the United States had
been engaged in the war in Southeast Asia for 8
years, and 5.7 million men were Vietnam Era veter­
ans. About 80 percent of the veterans were in their
twenties and another 12 percent were 30 to 34 years
old. The older group has been increasing in size as
the men separated several years ago move out of
their twenties. In the second quarter of 1972, there
were about 660,000 in this age group compared with
420,000 a year earlier. About 97 percent were in the
labor force and their unemployment rate (not sea­
sonally adjusted) was 2.7 percent, not materially dif­
ferent from the 3.0 percent rate for nonveterans 30
to 34 years old.
The number of veterans in ages 30 to 34 is still
too small to permit either reliable adjustment for
recurring seasonal patterns in their employment or
detailed tabulations for such basic characteristics as
race and duration of unemployment. Since the job­
finding problems of veterans 30 to 34 years old are
Kopp M ichelotti and Kathryn R. G over are Social Science
Research Analysts in the D ivision o f Labor Force Studies,
Bureau o f Labor Statistics.

135
From the Review of December 1972



The employment
situation of
Vietnam Era
veterans
much less serious than for the group under age 30,
this analysis will continue to focus on those 20 to 29
years old.
This annual review of the employment situation of
male Vietnam era veterans includes, for the first
time, information on occupation and industry of em­
ployment, residence, household relationship, and rea­
sons for being unemployed or out of the labor force.
Employment

During fiscal 1972 all of the net growth in the
veterans’ labor force was in employment, as the
number of 20- to 29-year-old veterans with jobs rose
by 550,000 to average 3.9 million. Similar patterns
of increase occurred with respect to the nonveteran
labor force and employment levels. (See table 1.) A
year earlier employment had accounted for only
three-fourths of the labor force increase for veterans
and two-thirds for nonveterans.
Occupation. The occupational distribution of em­
ployed veterans and nonveterans 20 to 29 years old
is generally the same, with the exception of profes­
sional and technical workers and craftsmen. (See
table 2.) In the second quarter of 1972, about onefourth of the veterans were craftsmen (such as skilled
construction workers and mechanics), compared
with one-fifth of the nonveterans. A smaller propor­
tion of veterans than nonveterans were in profes­
sional and technical jobs (11 and 17 percent, re­
spectively). For the 20- to 24-year-olds, the propor­
tion of veterans in these occupations was less than
half that of nonveterans. This gap reflects the lower
percentage of college graduates among the veterans.
Younger veterans (age 20-24) were more con­
centrated in jobs which generally require less educa­
tion, training, and experience. In the second quarter
of 1972, about two-thirds of the employed younger
veterans but only half of the veterans 25 to 29 years
old were blue-collar workers— craftsmen, operatives,

than white veterans in the less skilled laborer and
service occupations. (See chart 1.) These differences
result from a combination of several factors, such as
job discrimination and the somewhat larger propor­
tion of employed Negro veterans who were in the
less experienced age group 20 to 24 years— 50 per­
cent, compared with 41 percent of the young whites.

and nonfarm laborers. On the other hand, less
than a third of the younger veterans were in whitecollar jobs, compared with 40 percent of the older
veterans. Only 6 percent of the younger group but
14 percent of the older were in professional and
technical occupations.
Negro12 veterans were more heavily concentrated

Table 1. Employment status of male Vietnam Era veterans and nonveterans 20 to 29 years old, quarterly averages,
1971 and 1972
[Numbers in thousands]
Seasonally adjusted
1971

1972

Veteran status and employment status

1971

1972

1

II

III

IV

■

II

1

II

III

IV

1

II

Total, 20 to 29 years:
Civilian noninstitutional population 2_____
Civilian labor force_________ _____
Percent of population_____________
Employed___________________
Unemployed________________
Unemployment rate_______

3,809
3,459
90.8
3,087
372
10.8

3,981
3,623
91.0
3,314
309
8.5

4,145
3,844
92.7
3,525
319
8.3

4,293
3,931
91.6
3,626
304
7.8

4,429
4,058
91.6
3,658
400
9.8

4,515
4,174
92.4
3,862
312
7.5

3,809
3,470
91.1
3,160
310
8.9

3,981
3,632
91.2
3,302
330
9.1

4,145
3,814
92.0
3,463
351
9.2

4,293
3,951
92.0
3,623
328
8.3

4,429
4,076
92.0
3,743
332
8.2

4,515
4,180
92.6
3,848
332
8.0

20 to 24 years:
Civilian noninstitutional population 2..........
Civilian labor force_______________
Percent of population_____________
Employed_____ _____________
Unemployed_________________
Unemployment rate_______

1,902
1,668
87.7
1,424
244
14.6

1,947
1,711
87.9
1,499
212
12.4

1,974
1,782
90.3
1,583
199
11.2

1,990
1,782
89.5
1,587
195
11.0

2,000
1,788
89.4
1,544
244
13.6

1,967
1,788
90.9
1,606
182
10.2

1,902
1,676
88.1
1,471
205
12.2

1,947
1,719
88.3
1,490
229
13.3

1,974
1,768
89.6
1,551
217
12.3

1,990
1,783
89.6
1,579
204
11.4

2,000
1,801
90.0
1,596
206
11.4

1,967
1,792
91.1
1,596
196
10.9

25 to 29 years:
Civilian noninstitutional population2_____
Civilian labor force_______________
Percent of population_____________
Employed____ ______________
Unemployed_________________
Unemployment rate........... .

1,907
1,791
93.9
1,663
128
7.2

2,035
1,912
94.0
1,815
97
5.1

2,171
2,062
95.0
1,942
120
5.8

2,303
2,149
93.3
2,039
109
5.1

2,429
2,270
93.5
2,114
156
6.9

2,549
2,387
93.6
2,256
130
5.5

1,907
1,794
94.1
1,689
105
5.8

2,035
1,912
94.0
1,811
101
5.3

2,171
2,046
94.2
1,912
134
6.5

2,303
2,168
94.1
2,044
124
5.7

2,429
2,274
93.6
2,148
127
5.6

2,549
2,388
93.7
2,251
136
5.7

Total, 20 to 29 years:
Civilian noninstitutional population 2____
Civilian labor force_______________
Percent of population_____________
Employed____________ ______
Unemployed_________________
Unemployment rate...... .........

9,209
7,844
85.2
7,188
656
8.4

9,334
8,093
86.7
7,524
569
7.0

9,454
8,436
89.2
7,852
584
6.9

9,567
8,200
85.7
7,633
567
6.9

9,716
8,264
85.1
7,566
698
8.4

9,930
8,604
86.6
8,006
598
7.0

9,209
7,997
86.8
7,419
578
7.2

9,334
8,076
86.5
7,502
574
7.1

9,454
8,136
86.1
7,544
592
7.3

9,567
8,371
87.5
7,727
644
7.7

9,716
8,435
86.8
7,816
619
7.3

9,930
8,586
86.5
7,978
608
7.1

20 to 24 years:
Civilian noninstitutional population 2_____
Civilian labor force_______________
Percent of population_____________
Employed___________________
Unemployed____________ ___
Unemployment rate_______

5,327
4,158
78.1
3,709
449
10.8

5,468
4,439
81.2
4,016
423
9.5

5,582
4,741
84.9
4,321
420
8.9

5,620
4,456
79.3
4,061
394
8.8

5,825
4,573
78.5
4,072
501
10.9

5,980
4,860
81.3
4,421
439
9.0

5,327
4,321
81.1
3,911
410
9.5

5,468
4,421
80.9
4,004
417
9.4

5,582
4,448
79.7
4,028
420
9.4

5,620
4,610
82.0
4,162
448
9.7

5,825
4,753
81.6
4,293
460
9.7

5,980
4,842
81.0
4,404
437
9.0

25 to 29 years:
Civilian noninstitutional population 2_____
Civilian labor force_______________
Percent of population_____________
Employed___________________
Unemployed_________________
Unemployment rate_______

3,882
3,686
95.0
3,479
207
5.6

3,866
3,654
94.5
3,508
146
4.0

3,872
3,695
95.4
3,531
164
4.4

3,947
3,744
94.9
3,572
172
4.6

3,891
3,691
94.9
3,494
197
5.3

3,950
3,744
94.8
3,585
159
4.2

3,882
3,676
94.7
3,508
168
4.6

3,866
3,654
94.5
3,497
157
4.3

3,872
3,687
95.2
3,516
171
4.6

3,947
3,762
95.3
3,566
196
5.2

3,891
3,682
94.6
3,523
159
4.3

3,950
3,745
94.8
3,574
171
4.6

VETERANS1

NONVETERANS

1 Vietnam Era veterans are those who served after August 4, 1964; they are all classi­
fied as war veterans. About 80 percent of the Vietnam Era veterans of all ages are
20 fo 29 years old. Post-Korean peacetime veterans are not included in this table.
2 Since seasonal variations are not present in the population figures, identical num­
bers appear in the unadjusted and seasonally adjusted columns.
NOTE: Because of rounding, sums of individual items may not equal totals. Rates




are based on unrounded numbers. Data are subject to sampling variability which
may be relatively large in cases where numbers are small. Therefore, differences
between numbers or percents based on them may not be significant. Eor a detailed
explanation of the reliability of estimates, including standard error tables, see the
Technical Note in the October 1972 issue of Employment and Earnings.

136

Industry. The distribution by industry of employed
veterans 20 to 29 years old was virtually the same as
that of employed nonveterans the same ages— nearly
a third held jobs in manufacturing, primarily in the
durable goods industries, and a fifth were in trade,
mostly in retail trade.
Among the veterans, Negro men, to a greater ex­
tent than white men, seem to take advantage of pref­
erential hiring programs in the public sector. In the
second quarter of 1972, 20 percent of the employed
Negro veterans 20 to 29 years old worked for Fed­
eral, State, or local governments, compared with 12
percent of the white veterans. (See chart 2.)

Table 2.
Major occupation and industry group of
employed male Vietnam Era veterans and nonveterans
20 to 29 years old, second quarter averages, 1972
[Percent distribution)
Veterans

Non veterans

Major occupation and
industry group
20 to 29 20to24 25 to 29 20 to 29 20 to 24 25 to 29
years years years years years years
Total employed (in thous a n d s )...___________

3,862

1,606

2,256

8,006

4,421

3,585

100.0

100.0

100.0

100.0

100.0

100.0

10.6

5.7

14.1

17.4

13.0

22.8

7.8
9.7
6.7
23.5
23.6
7.7

6.2
10.0
5.7
22.7
27.9
7.5

8.9
9.6
7.4
24.1
20.5
7.9

8.3
7.4
5.6
18.5
21.8
7.6

5.8
8.6
5.5
17.8
23.4
9.4

11.0
5.8
5.9
19.4
19.8
5.4

1.9

2.6

1.4

3.8

4.0

3.5

8.4

11.5

6.1

9.8

12.5

6.5

100.0

100.0

100.0

100.0

100.0

100.0

2.4
97.6
94.8
9.2
31.6
20.1
11.5

3.3
96.7
94.4
10.6
31.6
19.9
11.8

1.8
98.2
95.0
8.1
31.5
20.3
11.2

4.6
95.4
91.9
8.5
28.1
18.0
10.1

5.1
94.9
92.6
9.1
27.4
17.2
10.2

4.1
95.9
91.0
7.8
28.9
18.9
10.9

9.0
19.1

8.0
21.1

9.7
17.7

6.0
19.0

5.7
21.9

6.3
15.6

3.8
8.1
12.8

3.2
8.3
10.0

4.3
8.0
14.8

3.8
11.4
14.0

3.4
11.7
12.4

4.4
11.1
15.9

2.8

2.3

3.1

3.5

2.4

4.8

OCCUPATION
Total_____ ___________
Professional and technical
workers__________________
Managers and administrators,
except farm_______________
Clerical workers______________
Sales workers_______________
Craftsmen and kindred workers..
Operatives and kindred workers..
Service workers______________
Farmers and farm laborers,
foremen_______ ____ ______
Laborers, excluding farm and
mine________ : ___________
INDUSTRY
Total____ ____________
Agriculture__________________
Nonagricultural industries______
Wage and salary workers___
Construction_________
Manufacturing________
Durable goods.........
Nondurable goods..
Transportation, communication, and
public utilities______
Trade_______________
Finance, insurance, and
real estate......... .........
Service............................
Government__________
Self-employed and unpaid
family workers_____ _______

NOTE: For definitions and notes on data limitations, see table 1.




Unemployment
The unemployment rate of veterans 20 to 29 fell
from 9.1 percent to 8.0 percent (seasonally ad­
justed) in the year ended in the second quarter
1972, while the rate of nonveterans remained the
same at 7.1 percent (seasonally adjusted). All of the
improvement in the veterans’ unemployment rate oc­
curred among the veterans in ages 20 to 24, whose
average rate dropped to 10.9 percent in second
quarter 1972 from 13.3 percent a year earlier. At
5.7 percent, the unemployment rate of veterans 25 to
29 was roughly the same as in second quarter 1971.
The gap between the unemployment rate of veter­
ans and nonveterans narrowed substantially between
mid-1971 and 1972. For the second quarter of 1972,
the difference was 0.9 percentage points compared
with 2.0 percentage points a year before. Although
most of the narrowing reflects an improved job situa­
tion for veterans, some reflects a shift in the age
composition of the veterans compared to the nonvet­
erans. Very little of the increase in the 20- to 29year-old veteran population and labor force was in
ages 20 to 24, where unemployment problems are
more severe than for older veterans. With fewer men
going into military service (draft calls fell from
152,000 in fiscal 1971 to 25,000 in fiscal 1972), the
nonveteran population and labor force increased pri­
marily in ages 20 to 24. Regardless of veteran status,
the jobless rate for men 25 to 29 is lower than that
for men 20 to 24, for such reasons as greater work
experience, more familiarity with the job market, and
higher seniority.
By the third quarter of 1972, the unemployment
rate for veterans 20 to 29 years old had dropped
to 7.2 percent, and in October the veterans’ rate of
6.4 percent was little different from the 6.6 percent
rate for nonveterans the same ages.
Duration. Following the economic downturn of
1970, the duration of unemployment for both veter­
ans and nonveterans lengthened. The percentage of
unemployed veterans looking for work for 15 weeks
or more increased from an annual average of 9 per­
cent in 1969 to 15 percent in 1970 to 25 percent in
1971. The comparable statistic for nonveterans has
increased in a similar fashion. In the second quarter
of 1972, about 30 percent of the jobless veterans and
nonveterans had been unemployed for at least 15
weeks, the same proportions as in the second quarter
a year earlier. (See table 3.)

137

Reasons for unemployment. Some persons become
unemployed by losing or quitting a job, while others
are unemployed as a consequence of coming into the
labor force and starting to look for work. As the
following percentages for the second quarter of 1972
indicate, veterans and nonveterans differed slightly in
their reasons for unemployment:

Table 3. Duration of unemployment of male Vietnam
Era veterans and nonveterans 20 to 29 years old,
quarterly averages, 1971 and 1972
[Percent distribution]

1

II

III

IV

-

II

Total unemployed:
Number (in thousands).........
Percent__________ ______

372
100.0

309
100.0

319
100.0

304
100.0

400
100.0

312
100.0

Less than 5 weeks............. .
5 to 14 weeks................. .......
15 weeks or more_________

38.4
37.9
23.7

40.8
29.8
29.4

42.9
33.5
23.5

42.6
33.6
23.9

41.1
33.9
25.0

40.6
28.6
30.8

Total unemployed:
Number (in thousands).........
Percent........................... .......

656
100.0

569
100.0

584
100.0

567
100.0

698
100.0

598
100.0

Less than 5 weeks.................
5 to 14 weeks_____ ____
15 weeks or more.............. .

39.3
38.9
21.8

45.1
25.2
29.8

44.0
35.3
20.7

39.8
36.4
23.9

35.7
34.7
29.6

46.0
24.7
29.3

VETERANS

V eterans N on vetera n s
Total unem ployed (in thousands)
Percent ...............................................
Job losers ................................
On layoff .......................
Other job l o s e r s ..........
Job le a v e r s ..............................
Labor force e n t r a n t s ..........
Reentrants ....................
N ew workers ...............

312
100.0
4 5 .2
11.5
33.7
12.5
42.3
34.6
7.7

598
100.0
50.2
9.4
40 .8
14.2
35.6
30.1
5.5

In the second quarter of 1972, veterans and non­
veterans were about equally likely to be on layoff,
but the veterans were less likely to have lost their
jobs for such reasons as dismissal, expiradon of a
temporary job, or plant closing. A somewhat greater
percentage of the veterans than of the nonveterans
were either reentrants to the labor force or had never
worked before.
Younger veterans were more likely than older vetChart 1. Occupational distribution of male Vietnam
Era veterans 20 to 29 years old, by race, 2d quarter
averages, 1972

1972

1971
Veteran status and duration
of unemployment

NONVETERANS

NOTE: For definitions and notes on data limitations, see table 1.

erans to be labor force entrants, because more of
them had only recently left the Armed Forces. Job­
finding problems for newcomers to the labor force,
whether veterans or nonveterans, tend to be exacer­
bated by the fact that they may not be as familiar
with the intricacies of the job market as those who
left or lost a job.

Men not in the labor force
Percent

Professional,
technical

100

Managerial
Clerical

75 -

Sales
Craftsmen

50
Operatives

_ Laborers,
excluding farm

25 -

Farm workers
Service
White




Negro and
other races

In the second quarter of 1972, about 8 percent
(340,000) of the veterans were neither working nor
looking for work, compared with 13 percent (1.3
million) of the non veterans. The proportion not in
the labor force was smaller for veterans because of a
combination of demographic and social factors.
Among these is the larger proportion of veterans in
their late twenties, an age group in which the labor
force participation rate is higher than for those 20 to
24 years old. Another factor is the larger proportion
of veterans than nonveterans who head households.
In the April-June quarter of 1972, about two-thirds
of the veterans but only half of the nonveterans were
household heads, with a wife and, perhaps, young
children to support. All of this difference was ac­
counted for by the 20- to 24-year-old men, among
whom about half of the veterans compared with
about a third of the nonveterans had these family
responsibilities.
Attendance at school was by far the most impor-

138

Chart 2.

Employment of male Vietnam Era veterans 20 to 29 years old, by race and major industry group, 2d quarter

White

Negro and other races

1 Excluding government.

increase in the educational attainment of the popu­
lation. For all the servicemen discharged from
August 1964 through the end of 1971, the median
years of schooling completed at time of separation
was 12.5 years. This compares with 12.3 years for
Korean Conflict veterans and 11.5 years for World
War II veterans. Educational attainment at separa­
tion was highest for Vietnam Era veterans 25 to 29
years old (12.9 years). Nearly half (46 percent) of
the men in this age group had completed at least
1 year of college.3
Since the midsixties, the median educational at­
tainment of veterans at time of separation has in­
creased gradually from 12.4 to 12.6 years. In fiscal
1965 through 1967, about 17 percent of the separa­
tees had completed a year of college or more. This
proportion reached 27 percent in the first half of
fiscal 1972, including 13 percent who had graduated
from college.4
Roughly 10 percent of the veteran population 20
to 24 years old in the year ending in June 1972,
reported school as their major activity.5 The propor­
tion for nonveterans of the same age was twice as
high. Among 25- to 29-year-olds, about 6 percent of
the veterans and 3 percent of the nonveterans were
in school.

tant reason given for not being in the labor force. In
the second quarter of 1972, about two-thirds of the
veterans and three-fourths of the nonveterans not in
the labor force gave school as their reason for non­
participation. The next most frequently given reason
was not wanting a job. (See chart 3.)
On an annual average basis for 1971, veterans 20
to 24 years old and those 25 to 29 years old differed
little in their reasons for nonparticipation in the
labor force. In contrast, nonveterans exhibited large
differences by age. Younger nonveterans were almost
twice as likely as older nonveterans to be in school,
while they were less likely than older nonveterans to
mention ill health or disability as a reason for non­
participation. Few of the veterans in either their
early or late twenties gave this reason. Relatively few
(2 to 3 percent) of the veterans and nonveterans not
in the labor force in the first half of 1972 gave as
their reason the belief that they could not find a job.

Education
Vietnam Era veterans are better educated when
they leave the service than were World War II or
Korean Conflict veterans at the time of their separa­
tion from military service, reflecting in part a general




139

Students generally have a much lower labor force
participation rate than those whose major activity is
something else. In the second quarter of 1972, 30
percent of the veterans 20 to 29 years old in school
were in the labor force in contrast to 97 percent of
the veterans out of school. The labor force participa­
tion rate of students was the same for veterans and
nonveterans, but among nonstudents veterans had a
slightly higher rate.

Chart 3. Reasons for nonparticipation in the labor force
of male Vietnam Era veterans and nonveterans 20 to 29
years old, 2d quarter averages, 1972

About a tenth of the unemployed veterans and
nonveterans were students, and most of these were
seeking part-time work. In the second quarter of the
year, however, the proportion seeking part-time jobs
usually decreases, probably because students begin
working or looking for full-time summer jobs before
the end of the school year, as shown by the following
tabulation for veterans in the first and second quarters
of 1972:
I
Total unem ployed (in th o u sa n d s). 4 0 0
P e r c e n t ....................................................... 100.0

II
312
100.0

Major activity: s c h o o l ...................................
Looking for full-tim e w o r k .............
L ooking for part-tim e w o r k .............

13.0
4.3
8.7

9 .0
4.8
4 .2

M ajority activity: other ..............................
L ooking for full-tim e w o r k .............
L ooking for part-tim e w o r k .............

87.0
84.3
2.7

9 1 .0
88.8
2.2

Race and residence

tively more eligible Negroes than whites reenlist
when their enlistments expire.7
The employment situation of veterans of Negro
and other minority races can be discussed only in
general terms because the data are based on small
numbers of sample cases and sampling variability is
high. The unemployment rates of Negro veterans
have not been significantly different statistically from
those of Negro nonveterans, but have been substan­
tially higher than the unemployment rates of white
veterans. (See table 4.) During 1971 and the first
half of 1972, the quarterly average unemployment
rate of Negro veterans was in the range of 12 to 15
percent, compared with 7 to 10 percent for white
veterans.

Race. Negroes constitute a smaller proportion of
Vietnam Era veterans than of nonveterans. In the
first half of 1972, they made up about 9 percent of
the 20- to 29-year-old veteran population and labor
force but almost 13 percent of the nonveteran popu­
lation and labor force. The smaller proportion of
Negroes among veterans is primarily due to two rea­
sons. Relatively more Negroes than whites are dis­
qualified from entering the Armed Forces,6 and rela­

Residence. Following the national pattern,8 more
Vietnam Era veterans and nonveterans 20 to 29
years old reside in the Southern and North Cen­
tral regions of the United States than in the North­
east and West. The unemployment rates for veterans
and nonveterans in the Southern and North Central
regions are considerably lower than comparable rates
the Northeast and West. (See table 5.) In the second
quarter of 1972, the jobless rates for veterans 20 to

In contrast, the overwhelming majority of unem­
ployed nonstudents look for full-time jobs the year
round.
The unemployment rate of men 20 to 29 years old
in school is far higher than that of those not in
school. For veterans, in the second quarter of 1972,
the unemployment rate was 29 percent for students
in contrast to 7 percent for nonstudents. The corre­
sponding unemployment rates for nonveterans were
13 percent for students and 7 percent for nonstu­
dents.




140

29 years old were 5.6 and 6.7 percent, respectively,
in the Southern and North Central regions, compared
with 8.9 and 9.9 percent in the West and Northeast.
About half the Negro veterans 20 to 29 years old
live in the South, in contrast to about one-quarter of
the white veterans; this is comparable to the distri­
bution of the total population by race. The unem­
ployment rate for the Negro veterans in the South,
at 13.8 percent, was about three times as high as for
white veterans, though not significantly higher than

for Negro veterans living elsewhere (11.7 percent).
Outside the South, the unemployment rate of Negro
veterans was only one and a half times as high as for
white veterans.

Special programs
Among the continuing programs and benefits for
veterans are the longstanding GI Bill administered by
the Veterans Administration, Project Transition,

Table 4. Employment status of male Vietnam Era veterans and nonveterans 20 to 29 years old, by race, quarterly
averages, 1971 and 1972
[Numbers in thousands]
White
Employment status

Negro and other races

1971

1972

1971

1972

•

II

III

IV

1

II

1

II

III

IV

Total, 20 to 29 years:
Civilian noninstitutional population, _........
Civilian labor force_______________
Percent of population_____________
Employed______ ______ ______
Unemployed_________________
Unemployment rate..............

3,446
3,135
91.0
2,812
323
10.3

3,596
3,274
91.0
3,008
266
8.1

3,721
3,456
92.9
3,191
265
7.7

3,878
3,558
91.7
3,306
252
7.1

4,028
3,708
92.1
3,361
347
9.3

4,102
3,799
92.6
3,535
264
7.0

363
324
89.3
275
49
15.1

386
350
90.7
308
42
12.1

425
388
91.3
334
54
14.0

415
373
89.9
322
52
13.8

401
350
87.3
297
53
15.3

413
375
90.8
327
48
12.7

20 to 24 years:
Civilian noninstitutional population______
Civilian labor force_______________
Percent of population_____________
Employed....................................
Unemployed____________ ____
Unemployment rate....... .......

1,699
1,489
87.6
1,282
207
13.9

1,737
1,527
87.9
1,347
180
11.8

1,761
1,593
90.5
1,424
169
10.6

1,798
1,615
89.8
1,447
168
10.4

1,800
1,617
89.8
1,411
206
12.7

1,748
1,595
91.2
1,442
153
9.6

203
179
88.2
142
37
20.9

210
184
87.6
153
31
17.0

214
189
88.3
159
30
16.0

192
167
87.0
140
27
15.9

200
171
85.5
133
38
22.4

219
193
88.1
164
2.9
15.1

1,747
1,646
94.2
1,529
117
7.1

1,859
1,747
94.0
1,661
86
4.9

1,961
1,863
95.0
1,767
96
5.2

2,080
1,943
93.4
1,859
84
4.3

2,228
2,091
93.8
1,950
141
6.7

2,354 ’
2,205
93.7
2,093
112
5.1

160
145
90.6
133
12
8.0

176
165
93.8
154
11
6.7

211
199
94.3
175
24
12.0

223
206
92.4
181
25
12.0

201
179
89.1
164
15
8.6

195
182
93.3
164
2.9
10.2

Total, 20 to 29 years:
Civilian noninstitutional population........ .
Civilian labor force_______________
Percent of population_____________
Employed_____ _____________
Unemployed________________
Unemployment rate......... .

7,964
6,798
85.4
6,277
521
7.7

8,072
7,020
87.0
6,567
453
6.5

8,183
7,338
89.7
6,888
450
6.1

8,260
7,116
86.2
6,679
437
6.1

8,463
7,232
85.5
6,678
553
7.6

8,652
7,539
87.1
7,053
486
6.4

1,245
1,045
83.9
910
135
12.9

1,262
1,073
85.0
958
115
10.7

1,271
1,098
86.4
963
135
12.3

1,307
1,084
82.9
955
129
11.9

1,253
1,032
82.4
888
145
14.0

1,278
1,065
83.3
953
112
10.5

20 to 24 years:
Civilian noninstitutional population______
Civilian labor force______ _________
Percent of population_____________
Employed____ ______________
Unemployed_________________
Unemployment rate_______

4,616
3,604
78.1
3,252
352
9.8

4,739
3,850
81.2
3,519
331
8.6

4,834
4,119
85.2
3,795
324
7.9

4,838
3,853
79.6
3,549
304
7.9

5,066
3,994
78.8
3,596
397
9.9

5,220
4,263
81.7
3,913
350
8.2

711
554
77.9
457
97
17.4

729
589
80.8
497
92
15.6

748
621
83.0
525
96
15.5

782
603
77.1
513
90
15.0

759
579
76.3
476
104
17.9

761
597
78.4
508
89
14.9

20 to 29 years:
Civilian noninstitutional population........ .
Civilian labor force____________
Percent of population_____________
Employed___________________
Unemployed.
................... __
Unemployment ra te ,............

3,348
3,195
95.4
3,026
169
5.3

3,333
3,170
95.1
3,048
122
3.8

3,349
3,219
96.1
3,093
126
3.9

3,422
3,263
95.4
3,130
133
4.1

3,397
3,238
95.3
3,082
156
4.8

3,433
3,277
95.5
3,140
136
4.2

534
491
91.9
453
38
7.8

533
484
90.8
460
24
4.9

523
477
91.2
438
39
8.1

525
481
91.6
442
39
8.0

494
453
91.7
412
41
9.0

517
467
90.3
444
23
4.9

II

VETERANS

25 to 29 years:
Civilian noninstitutional population_____
Civilian labor force........ ...... .............
Percent of population..........................
Employed____ ______________
Unemployed__________ ____
Unem ploym ent rate . _

__

NONVETERANS

NOTE: For de'mitions and notes on data (imitations, see table 1.




141

Table 5. Employment status of male Vietnam Era veterans and nonveterans 20 to 29 years old, by region and race,
second quarter averages, 1972
[Numbers in thousands]
Veterans

Nonveterans

Labor force status and race
Total

North­
east

North
Central

South

West

Total

North­
east

North
Central

South

West

4,515
4,174
92.4
3,862
312
7.5
341

997
921
92.4
830
91
9.9
76

1,293
1,209
93.5
1,128
81
6.7
84

1,344
1,242
92.4
1,173
69
5.6
102

881
802
91.0
731
71
8.9
79

9,931
8,603
86.6
8,005
599
7.0
1,328

2,427
2,034
83.8
1,856
178
8.8
393

2,675
2,370
88.6
2,210
161
6.8
305

3,091
2,699
87.3
2,570
129
4.8
392

1,738
1,500
86.3
1,369
131
8.7
238

4,102
3,799
92.6
3,535
264
6.9
303

927
859
92.7
775
84
9.8
68

1,214
1,134
93.4
1,061
73
6.4
80

1,139
1,055
92.6
1,012
43
4.1
84

822
751
91.4
687
64
8.5
71

8,653
7,539
87.1
7,054
486
6.4
1,114

2,190
1,851
84.5
1,697
154
8.3
339

2,429
2,164
89.1
2,034
131
6.1
265

2,466
2,162
87.8
2,072
90
4.2
304

1,568
1,362
86.9
1,251
111
8.1
206

413
375
91.8
327
48
12.8
38

70
62
(l)
55
7
0
8

79
75
95.0
67
8
10.7
4

205
187
91.2
161
26
13.8
18

59
51

1,278
1,064
83.3
951
113
10.6
214

237
183
77.2
159
24
13.1
54

246
206
83.7
176
30
14.6
40

625
537
85.9
498
39
7.3
88

170
138
81.2
118
20
14.5
32

ALL MEN
Civilian noninstitutional population___________________________
Civilian labor force.........................................................................
Percent of population__________________________________
Employed_______________ _____ _____ _______ ______
Unemployed__________ ___________________________
Unemployment rate._______________ ____ _______
Not in labor force........................................ ..................................
WHITE
Civilian noninstitutional population......................................................
Civilian labor force________________ _____ ______________
Percent of population_________ ____ ____________________
Employed________________________________________
Unemployed______________ _______ ___________ ____
Unemployment rate____________________________
Not in labor force........................................................................
NEGRO AND OTHER RACES
Civilian noninstitutional population....... ...............................................
Civilian labor force________ ____ ___________ ______ _____
Percent of population___________ ______________________
Employed_______ ______ __________________________
Unemployed.____ _____ _____________ _____ ________
Unemployment rate..........................................................
Not in labor force____ ____ _______ ____ ________________

44
7
0
8

NOTE: For definitions and notes on data limitations, see table 1.

1 Percent not shewn where base is less than 75.CC0.

under the Department of Defense, and Employment
Services, Unemployment Compensation for Ex-Serv­
icemen, and Reemployment Rights, all under the De­
partment of Labor. In the past year, many of these
programs have been expanded and new ones added.
The President’s 6-point veterans program spurred
substantial increases in veterans’ job counseling,
placement, and training benefits, and also prompted
increased job opportunities in private industry
through such organizations as the National Alliance
of Businessmen.
Through June 1972, 41 percent of all Vietnam
Era veterans have participated in educational pro­
grams under the current GI Bill (effective June
1966). The comparable proportions of veterans par­
ticipating under previous GI Bills after a similar
length of time were 40 percent of Korean Conflict
servicemen and 46 percent of World War II veter­
ans. On October 24, 1972, amendments were signed
into law raising the amount of the current GI Bill
educational benefits for full-time students from the
$175 per month for a single veteran to $220, with
corresponding increases for veterans with depend­
ents.




0

142

By the end of June 1972, about 1.5 million serv­
icemen had received some type of job counseling for
civilian jobs under Project Transition which began in
January 1968. In addition, some 258,000 had partic­
ipated in a job-training program, frequently run on
or near military bases by private industry. Although
Project Transition is primarily for those servicemen
most in need of vocational training and education for
civilian life, it recently incorporated many other spe­
cial programs. One such program is Military Experi­
ence Directed Into Health Careers (M EDIHC), a
joint program of the Departments of Defense and
Health, Education, and Welfare, in which servicemen
who received military training in the health or medi­
cal fields are assisted in obtaining jobs in the civilian
health fields. The placement rate in mid-1972 ranged
from 40 to 70 percent depending upon the State. A
companion program, the Veterans Construction Jobs
Clearinghouse, assists servicemen who have been
trained as construction mechanics. The program is
supported by Department of Labor funds and
manned by representatives of the construction indus­
try.
Other examples of new or amplified benefits for

Vietnam Era veterans were additional payments to
eligible veterans (as well as others) under the Tem­
porary Unemployment Compensation Program and
the employment of veterans (and others) under the
Public Employment Program.

These are only a few examples of the nationwide
efforts which helped Vietnam Era veterans get edu­
cational and vocational training and contributed to
the improvement in their employment situation dur­
ing fiscal year 1972.
□

-FOOTNOTES1
About 83,000 women veterans of the Vietnam Era are
not included in this report because employment data are not
available for them. In this report, Vietnam Era veterans are
those who served in the Armed Forces after Aug. 4, 1964,
have been separated from active duty, and are now in the
civilian noninstitutional population. Korean Conflict veterans
served during the period June 27, 1950, to Jan. 31, 1955.
World War II veterans served at any time from Sept.
16, 1940, to July 25, 1947. Nonveterans include those who
have never served in the Armed Forces or who served only
in peacetime prior to June 27, 1950. Post-Korean Conflict
veterans— men who served between Feb. 1, 1955, and Aug.
4, 1964— are not included in this report.
Unless otherwise indicated, data on the civilian noninstitu­
tional population, labor force, employment status, and edu­
cational attainment are derived from the nationwide Current
Population Survey (CPS) sample of about 50,000 house­
holds. The CPS, conducted each month by the Bureau of the
Census for the Bureau of Labor Statistics (BLS), is the
source of special tabulation by veteran status prepared for
the Veterans Administration and BLS. The data are subject
to sampling variability, which may be relatively large for the
smaller figures and for small differences between figures.
Standard errors of monthly sample estimates are published by
BLS in Employment and Earnings. These standard errors
must be reduced by a factor of .7070 for quarterly averages,
and .4472 for annual averages. Details about basic labor
force concepts, sample design, and estimating methods are
decribed in Concepts and M ethods Used in Manpower Sta­
tistics From the Current Population Survey (BLS Report
313, 1967).




143

The latest in this series of annual reports on the employ­
ment situation of Vietnam Era veterans was published in the
Monthly Labor Review, September 1971, pp. 3-11, and
reprinted as Special Labor Force Report 137.
2 Data for all persons other than white are used in this re­
port to represent data for Negroes, since the latter constitute
about 92 percent of all persons other than white persons in
the United States.
3 See Data on Vietnam Era Veterans, December 1971
(Veterans Administration, 1972), p. 8.
4 Ibid., p. 7.
5 Respondents in the Current Population Survey were
asked, “What were you doing most of last week?” On the
basis of their replies, persons were classified into two groups
— Major activity: going to school and Major activity: other.
In this report, those whose major activity was going to school
are referred to as “students” and those whose major activity
was something else are classified as “not in school.”
8 Data on disqualifications on the basis of medical, mental,
and trainability tests were provided by the Medical Statistics
Agency, Office of the Surgeon General, Department of the
Army. In these data, statistics for Negroes refer to Negroes
only and statistics for whites refer to all others (non-Negro).
7 Data on reenlistment rates and ineligibility to reenlist
were provided by the Director of Procurement Policy, Office
of the Assistant Secretary of Defense for Manpower and
Reserve Affairs. In these data, statistics for Negroes refer to
Negroes only and statistics for whites refer to whites only.
8 See Geographic Profile of Employment and Unemploy­
ment, 1971 (BLS Report 402, 1972).

Occupational
characteristics
of urban
workers

Special Labor Force Report shows workers
in metropolitan areas to be highly skilled,
but with substantial differences
between residents of the central cities
and those living in the suburbs
CHRISTOPHER G. GELLNER

T w o - t h i r d s o f a l l w o r k e r s in the United States
now reside in metropolitan areas— the centers of
economic activity and growth for the Nation.1 The
economic importance of these areas is reflected in the
high proportion of highly skilled workers within their
populations. Professional, technical, and managerial
occupations are more common in such areas (particu­
larly the 20 largest) than in the Nation as a whole.
This article is based on occupational data for
Standard Metropolitan Statistical Areas (SMSA’s)
that have recently become available on an annual
average basis from the Current Population Survey
(CPS). It explores the major differences in the
occupational distribution of employment among our
large metropolitan areas and between their central
cities and suburban rings. It also attempts to deter­
mine whether such skill differences have any direct
bearing on the disparity between central city and
suburban unemployment rates.
Data on the occupational distribution of the labor
force are essential in order to study the purported
skill gap between central city workers and suburban
workers.2 In the absence of such information, it has
been widely assumed that the great majority of
central city workers are concentrated in semiskilled
and low skilled jobs and that when unemployed they
seek work in similar fields.

and managerial occupations. Nonmetropolitan area
workers are more likely to be employed in bluecollar work. The proportion of workers in the service
occupations is roughly the same in metropolitan and
nonmetropolitan areas.
The higher skill level of the metropolitan area
labor force is apparent among both Negro and white
workers. One-third of the Negro workers residing
in metropolitan areas were engaged in white-collar
work in 1970, with 14 percent working as skilled
professionals and managers. Outside these areas,
only 14 percent were in white-collar work and 8
percent in the professional and managerial occupa­
tions. Among whites, the percentage employed in
white-collar occupations is also significantly higher
in metropolitan areas (56 percent) than in non­
metropolitan areas (41 percent). Moreover, the
proportion of whites employed in the professional
and managerial occupations in metropolitan areas,
at 28 percent, is also significantly higher than the
22 percent in nonmetropolitan areas. Outside metro­
politan areas, about one-tenth of the whites and
one-eighth of the Negroes were in farming occupa­
tions in 1970.
Generally speaking, the larger a metropolitan area,
the larger its proportion of higher skilled workers.
This is confirmed by data on the 20 largest SMSA’s,
which contain about half of the workers of all metro­
politan areas.3 Approximately 56 percent of the
workers residing in these areas are in white-collar
occupations, a slightly greater proportion than in
all metropolitan areas. Moreover, these large urban
areas have a slightly larger proportion of professional
and technical workers than do all metropolitan areas.
Conversely, the proportion of workers in both bluecollar and service occupations is somewhat lower
in these large SMSA’s than in smaller urban areas.
Differences in occupational distribution between
the labor force in the 20 SMSA’s and that for all
metropolitan areas are evident both for Negroes and
for whites. In the 20 largest areas, approximately

Skill pattern by nature of area
As table 1 shows, over one-half of all workers
residing in metropolitan areas are employed in whitecollar work and 16 percent are in professional and
managerial occupations. In nonmetropolitan areas,
slightly less than two-fifths of the workers are in
white-collar work and 11 percent in professional

Christopher G. Gellner is a labor economist in the Division
of Employment and Unemployment Analysis, Bureau of
Labor Statistics.


From the Review of October 1971


144

politan areas contain approximately three-tenths of
all U.S. employment. This is less than half, however,
of total metropolitan employment. Over the past two
decades, the net number of employed persons living
in central cities has remained almost the same, and
occupational upgrading has proceeded slowly. In
contrast, in the surrounding suburban rings the resi­
dent labor force has grown rapidly in size and has
experienced substantial occupational upgrading. The
relative change in the skill levels of the suburban and
central city residents has occurred principally be­
cause higher skilled workers have moved to the
suburbs, while the central cities have received large
numbers of less skilled workers from smaller towns
and rural areas.
Half the workers living in the central cities were
employed in white-collar jobs in 1970. This is slightly
lower than the suburban proportion. However, the
highly skilled professional and technical fields ac­
counted for 14 percent of all central city workers—
the same as the U.S. average, but over 2 percentage
points below the suburban average. (See table 2.)
In comparison to their suburban counterparts, cen­
tral city residents were generally less represented in
all of the skilled white-collar occupations and in the
craftsmen trades, but were more represented in
the lower skilled occupations (operative, nonfarm
laborer, and service).
Among central city residents, about one-third of
the employed Negroes were working in white-collar
jobs. The same proportion was found among Negroes
living in the suburbs. While about one-half the

37 percent of Negro employment and 59 percent of
white was in white-collar fields— both larger percent­
ages than for all SMA’s combined. The relatively
large proportion of white-collar workers in the Negro
labor force of the 20 largest areas compared with
that of all SMSA’s, however, results almost entirely
from a greater representation in clerical jobs, gen­
erally occupied by women. White employment, on
the other hand, stems from a greater representation
in professional and technical as well as clerical
occupations. In blue-collar and service occupations,
the percentages of both Negro and white workers
are slightly lower in the 20 largest areas than in all
metropolitan areas combined. The smaller propor­
tion of Negro workers in these two occupational
groups in the 20 areas is due chiefly to the fact that
a smaller proportion hold nonfarm laborer and pri­
vate household jobs in the large urban areas than
in smaller areas.

Central city versus suburb
As metropolitan areas have grown in size and
importance, the socioeconomic dichotomy between
the central city and its surrounding surburban ring
has increased. Each component of the SMSA is
dependent upon the other for economic survival.
Central cities, however, have experienced a dispro­
portionate amount of the economic hardship in
metropolitan areas, as reflected by their higher un­
employment rates and less skilled work forces.
Today, the central cities of the Nation’s metro­
Table 1.

Employed persons in the United States, by major occupation group and color, 1970 annual averages

[Percent distribution]
Total

White

Negro and other races

United
States

All non­
metro­
politan
areas

All
metro­
politan
areas

20 largest
metro­
politan
areas

United
States

All non­
metro­
politan
areas

All
metro­
politan
areas

20 largest
metro­
politan
areas

United
States

All non­
metro­
politan
areas

All
metro­
politan
areas

20 largest
metro­
politan
areas

Total employed (thousands)____
Percent............ .............. .......

78,627
100.0

27,011
100.0

51,616
100.0

26,180
100.0

70,182
100.0

24,798
100.0

45,384
100.0

22,643
100.0

8,445
100.0

2,213
100.0

6,232
100.0

3,537
100.0

White-collar workers._________
Professional and technical...
Managers, officials, and
proprietors..................... .
Clerical workers......... ..........
Sales workers....... ................

48.3
14.2

38.8
11.4

53.3
15.6

55.7
16.5

50.8
14.8

41.1
11.9

56.1
16.4

58.6
17.4

27.9
9.1

13.6
5.6

33.0
10.3

37.2
10.6

10.5
17.4
6.2

9.7
12.8
5.0

11.0
19.9
6.8

11.1
21.4
6.7

11.4
18.0
6.7

10.3
13.5
5.4

12.0
20.4
7.4

12.1
21.7
7.4

3.5
13.2
2.1

2.3
4.4
1.3

3.9
16.3
2.4

4.3
19.7
2.6

Blue-collar workers_____ _____
Craftsmen and foremen........
Operatives...................... .......
Nonfarm laborers......... .........

35.3
12.9
17.7
4.7

38.4
13.0
20.0
5.4

33.7
12.9
16.5
4.4

32.3
12.5
15.8
4.0

34.5
13.5
17.0
4.1

37.7
13.5
19.5
4.8

32.8
13.5
15.6
3.7

31.2
13.1
14.8
3.4

42.2
8.2
23.7
10.3

45.9
7.0
26.2
12.9

40.8
8.6
22.9
9.3

39.4
8.3
22.5
8.1

Service workers............... ............
Private household workers...
Other service workers...........

12.4
2.0
10.4

12.8
2.4
10.4

12.1
1.8
10.3

11.7
1.5
10.2

10.7
1.3
9.4

11.5
1.6
9.9

10.3
1.1
9.2

9.9

8.9

1.0

26.0
7.7
18.3

27.7
11.5
16.2

25.4
6.4
19.0

23.1
4.8
18.3

Farm workers............. .................

4.0

9.9

.9

.4

4.0

9.7

.9

.4

3.9

12.7

.7

.3

Occupation group




145

Table 2. Employed persons in the central cities and suburban rings of all SMSA's and the 20 largest SMSA’s, by major
occupation group and color, 1970 annual averages
[Percent distribution]
Total

White

Negro and other races

Occupation group
Central city

Suburban ring

Central city

Suburb: n ring

Central city

Suburban ring

Total employed (thousands)......................................
Percent................................................................

23,234
100.0

28,382
100.0

18,471
100.0

26,913
100.0

4,764
100.0

1,469
100.0

White-collar workers...........................................
Prof essional and technical..........................
Managers, officials, and proprietors............
Clerical workers...........................................
Sales workers..............................................

516
U .4
9.6
21 2
6.3

54.7
16.6
12.1
18.8
7.2

56.4
15.6
11.2
22 3
7.3

55.9
16.9
12.5
9.1
7.4

32.8
19.7
3.7
17.1
2.3

33.6
12.3
4.8
13.8
2.7

Blue-collar workers.............................................
Craftsmen and foremen..............................
Operatives ...............................................
Nonfarm laborers........................................

34.0
17.6
11.5
4.9

33.5
14.0
15.6
3.9

32.1
12.2
16.1
3.8

33.2
14.4
15.2
3.6

41.3
8.9
23.2
9.3

39.2
7.8
21.9
9.5

Service workers...................................................
Private household workers.........................
Other service workers.................................

14.2
2.1
12.1

10.4
1.5
8.9

11.2
1.0
10.2

9.6
1.2
8.4

25.7
6.3
19.4

24.6
6.5
18.1

Farm workers......................................................

.2

1.4

.2

1.3

.2

2.6

Total employed (thousands)......................................
Percent................................................................

11,223
100.0

14,957
100.0

8,399
100.0

14,244
100.0

2,823
100.0

714
100.0

White-collar workers...........................................
Professional and technical.......................
Managers, officials, and proprietors...........
Clerical workers...........................................
Salesworkers...................................... .........

52.8
14.4
9.2
23.4
5.7

57.9
18.0
12.5
19.9
7.5

58.3
16.0
11.0
24.4
6.8

58.7
18.2
12.8
20.0
7.7

36.3
9.6
4.0
20.3
2.4

40.7
14.9
5.8
17.0
3.0

Blue-collar workers.......................... .................
Craftsmen and foremen........ ......................
Operatives............. .......................................
Nonfarm laborers........................................

33.3
11.0
17.8
4.6

31.5
13.7
14.3
3.5

31.1
11.6
16.1
3.4

31.2
13.9
14.0
3.3

39.9
9.0
22.8
8.2

37.1
8.2
21.4
7.5

Service workers.................................................
Private household workers..........................
Other service workers.................................

13.8
1.8
12.0

10.1
1.3
8.8

10.5
.8
9.7

9.5
1.1
8.4

23.6
4.8
18.8

21.1
5.1
16.0

Farm workers......................................................

.1

.6

.1

.6

.2

1.0

ALL METROPOLITAN AREAS

20 LARGEST METROPOLITAN AREAS

white-collar Negro suburbanites were in professional,
technical, managerial, or official jobs, less than half
the Negro white-collar workers living in central
cities were in this highly skilled group. This indi­
cates that a considerable proportion of Negroes with
high skilled— and thus high paying— jobs have
moved to the suburbs. Such a selective process does
not appear to have been at work among Negroes
outside the white-collar sector, however, as the skill
distribution of Negro blue-collar and service workers
living in the suburbs is not measurably different from
that of central city Negroes. In contrast, white
workers living in the suburbs are in higher skilled
jobs than their central city counterparts, within both
the white-collar and blue-collar categories. The pro­
portion of suburban Negroes working in farm jobs
(2.6 percent) is twice as large as the proportion of
suburban whites.
Workers living in the central cities of the 20 largest
metropolitan areas have essentially the same array
of jobs as those in the central city of all SMSA’s




combined. The suburbanites in the largest areas, on
the other hand, hold higher skilled jobs than sub­
urbanites in general. Especially, they are more con­
centrated in professional and technical occupations.
As a result, the skill gap between central city and
suburban residents is wider in the 20 largest areas
than in all SMSA’s combined.
The relatively wide skill gap is evident among
both whites and Negroes. In the 20 largest metro­
politan areas, suburban workers— both white and
Negro— hold a relatively larger proportion of pro­
fessional, managerial, and sales jobs than do their
city counterparts (table 2 ). In addition, a larger
proportion of the white suburban labor force than
of the white central city labor force are skilled
craftsmen. The proportion of professional and tech­
nical workers is IV2 times as large among Negroes
living in the suburban rings of the 20 largest
SMSA’s as among Negroes in the central cities of the
same areas, and equals the proportion of these
highly skilled workers in the total U.S. labor force.

146

There has, however, been a great deal of worker
movement into and from the city over the past two
decades. Occupational upgrading of the city labor
force has been hindered not only by outmigration
of highly skilled workers (largely whites) to the
more affluent suburbs, but also by a substantial
inmigration of unskilled, untrained Negroes (many
of them coming from rural areas). Because it is
mainly the younger whites who have been moving to
the suburbs, white workers in the city tend to be
older than white workers in the suburbs. In con­
trast, Negro workers in the city are relatively young.
However, they often lack appropriate skills or edu-

The majority of Negro white-collar workers living
in the central cities of these areas are in clerical
occupations.

Reason for the skill gap
General differences in occupational levels be­
tween residents of central cities and those of suburbs
are explained in part by the differences in the racial
composition of the labor force in the central cities
and the suburbs. Negroes hold a disproportionate
share of the lower skilled, less desirable jobs, and
their dense concentration in central cities tends to
skew the occupational distribution of city workers
in the low skilled direction. This is especially the
case in the 20 largest SMSA’s, where four-fifths of
the Negro labor force resides in the central cities.
The nature and geographic location of the indus­
trial growth within or in the vicinity of a metropolitan
area may have some effect on the occupational dis­
tribution of its central city and suburban labor forces.
For years most new metropolitan industry and busi­
ness has been placed in the suburbs. The majority
of building permits for office buildings and stores
issued in the early 1960’s were for suburban sites.4
If the new higher skilled better paying jobs are
available only in one section of the metropolitan
area— that is, the suburban ring— workers with the
appropriate qualifications for these jobs may prefer
to live in this section in order to be close to the
expanding employment opportunities.
The nature and location of industry growth, how­
ever, has probably had a greater effect on the differ­
ences in occupationrl distribution among the labor
forces of individual SMSA’s than on the differences
in occupational distribution between the labor forces
of a particular city and its surrounding suburb.
Another factor that has affected the occupational
gap between the central cities and suburbs is patterns
of population growth and migration. From 1950 to
1970, there has been virtually no growth in the
number of workers residing in the central cities,
if annexations are excluded. Extensive growth in
the number of workers residing in the suburban
rings has, in the meantime, pushed the number of
suburban workers past the number of city workers.
Many workers, when they attain a sufficient level of
education and skill to obtain a more remunerative
job, move to the suburbs. Continuance of this trend
is a serious obstacle to closing the gap between the
suburban labor force and that in the cities.




A note concerning data
The labor force data discussed in this article
were collected and tabulated for the Bureau of
Labor Statistics by the Bureau of the Census as
part of the Current Population Survey (CPS) pro­
gram. The CPS is a national survey conducted
monthly in about 50,000 households. The data for
Standard Metropolitan Statistical Areas and their
central cities and suburban rings have larger sam­
pling errors than national data collected through
the CPS, even when averaged over 12 months. For
this and other reasons, the metropolitan area esti­
mates in this article may differ somewhat from
1970 Census estimates that are scheduled to be
released in 1972. This should be taken into account
when making further use of the data.
Standard Metropolitan Statistical Areas as de­
fined by the Office of Management and Budget
consist of large cities and their adjacent suburban
counties. Central cities include the corporate limits
of the city or cities named in the SMSA title, while
the suburban rings include all SMSA territory out­
side the central city or cities.
The metropolitan areas used in the report refer
to the 212 SMSA’s as defined and ranked in 1960.
This means that for the purposes of this report the
geographic boundaries of the 212 SMSA’s are
those which were in effect in 1960 even though,
subsequently, the boundaries of some of these
areas have been redefined to include additional
counties or exclude counties. Since 1960, the num­
ber of areas defined to be metropolitan in character
has been expanded to over 240. SMSA’s added
since 1960 are not included in the data in this
report.
It should also be noted that the data in this
report have been tabulated according to the place
of residence of workers rather than their place of
work.

147

Table 3. Employed persons in the 20 largest SMSA’s,
their central cities, and their suburban rings, by occupa­
tion, 1960 1 and 19702

the blue-collar sector. The proportion of service
workers remained the same. (See table 3.)
Workers residing in the suburbs accounted for all
the employment growth shown by the 20 largest
SMSA’s during the 1960’s. Their number increased
by about two-fifths and was accompanied by a
general occupational upgrading of the labor force.
Today, workers residing in the suburbs have a much
higher representation in the major white-collar
occupations (except as sales personnel) than they
did in 1960. Today’s suburban workers also have a
smaller representation in all blue-collar jobs (par­
ticularly as craftsmen and operatives) than they
had a decade ago. The proportion of private house­
hold workers and the proportion of farm workers
have also declined in the suburbs since 1960.
Over the same period, the central city labor force
declined slightly in size and exhibited a somewhat
slower rate of occupational upgrading. The number
of employed persons residing in the central cities
of the 20 largest SMSA’s declined by 400,000 (or
nearly 4 percent) between 1960 and 1970. Although
these workers have achieved a higher representation
in professional, technical, and managerial occupa­
tions, the disparity in skill level between city and
suburban residents is slightly greater today than a
decade ago.

[Percent distribution]

Occupation group

Central cities
of 20 largest
SMSA’s

20 largest
SMSA's

Suburban rings
of 20 largest
SMSA’s

1960

1970

1960

1970

1960

1970

22.287
100.0

26.180
100.0

11.628
100.0

11.223
100.0

10.659
100.0

14.957
100.0

White-collar workers.........
Professional and
technical.................
Managers, officials,
ana proprietors___
Clerical workers.........
Sales workers........... .

50.4

55.7

48.6

52.8

52.4

57.9

13.6

16.5

11.9

14.4

15.4

18.0

9.2
19.3
8.3

11.1
21.4
6.7

8.0
21.0
7.6

9.2
23.4
5.7

10.5
17.5
9.0

12.5
19.9
7.5

Blue-collar workers..........
Craftsmen and
foremen................. .
Operatives................. .
Nonfarm laborers___

37.6

32.3

37.9

33.3

37.2

31.5

14.5
18.7
4.3

12.5
15.8
4.0

12.7
20.4
4.8

11.0
17.8
4.6

16.5
16.9
3.8

13.7
14.3
3.5

Service workers............... .
Private household___
Other service workers.

11.3
2.2
9.1

11.7
1.5
10.2

13.4
2.5
10.9

13.8
1.8
12.0

9.1
1.9
7.2

10.1
1.3
8.8

Farm workers................... .

.7

.4

.2

.1

1.2

.6

Total employed
(thousands).................. .
Percent5............................

1 1960 Decennial Census data, collected in April 1960. Persons 14 and 15 years old
are included (unlike the 1970 data collected by the CPS). However, the number of
employed 14- and 15-year-olds is small and should have only minor effect on the distri­
bution of employment.
* 1970 annual averages collected by the Current Population Survey.
5 Percentage distribution of 1960 Census data is the distribution of those persons
who reported an occupation.

cation for the many available jobs that call for
managerial, professional, or technical personnel.

Occupation and joblessness

1960-70 changes in 20 SMSA’s

For several years, the unemployment rates in the
central cities of metropolitan areas have been signifi­
cantly higher than the jobless rates in suburban
rings.5 In 1970, for example, the jobless rate in all
central cities combined was 5.6 percent, in all
suburban rings combined 4.7 percent. Several
hypotheses have been offered to explain the central
city v. suburban differences in joblessness— a mis­
match between skills and jobs, life style, and differen­
tial occupational status.
The mismatch hypothesis argues that the main
cause of the urban unemployment problem is not
a shortage of jobs, but a mismatch between the
skill requirements of the jobs available in the
central city and the actual skills of the resident labor
force.6 It argues that jobs in the central cities have
been growing very slowly and those jobs that have
been created are of a highly skilled, white-collar,
“professional” character, for which central city
residents do not have the training to compete success­
fully. It further maintains that jobs in the suburbs

Since 1960, the work force in our 20 largest
metropolitan areas has grown numerically. Its qual­
ity, measured by its occupational distribution, has
also increased. A decade ago, just half the employed
in these areas were working in white-collar occupa­
tions; today, 56 percent of employment in these
areas is white collar.
Within the white-collar sector, there has been a
marked increase in the proportion of professional
and managerial workers. The proportion of clerical
workers has also increased, while the proportion
of sales workers has decreased. This decline does
not stem from a lack of proportionate growth in
the number of sales jobs relative to other jobs.
Instead, the decline can probably be attributed to
the greater use by retail establishments of part-time
sales personnel whose primary job is in another field.
While the percentage of white-collar workers
increased during this period, there was a com­
mensurate decline in the percentage of workers in




148

have been growing extensively because of the reloca­
tion of manufacturing, retail trade, and services out­
side the city, and that many of these suburban jobs
require the low skilled or semiskilled labor which
city residents could provide.
A test of the validity of this hypothesis requires
reliable data both on the skills of city workers and
on the location and quality of job growth. A recent
study undertaken with limited data concluded that
a so-called job-worker mismatch in the city is largely
imaginary.7 It found that low skilled jobs had con­
tinued to grow in the central cities, though not as
fast as in the suburbs. According to this study,
based on data for 1965-67, almost enough jobs
were being created in the cities studied to eliminate
all unemployment even if all the jobless were semi­
skilled or low skilled workers. In light of this fact,
the persistence of an unemployment gap between
the central city and the suburbs was attributed
largely to discriminatory employment practices.
The life style hypothesis is in direct opposition to
the mismatch hypothesis. It argues that there is an
abundance of unfilled low skilled job vacancies in
or near city areas. It also holds, however, that most
of these job vacancies are for jobs with low wages
or bad working conditions. The availability of a
large number of unfilled low skilled jobs has allegedly
created excess labor demand and tends to make
workers very independent of their employers, thus
creating a high rate of voluntary termination.8 This
hypothesis also argues that many of the jobs that
are concentrated in the cities (warehouses, main­

tenance services, and so on) are compatible with
a high rate of worker turnover, which, in turn, is
considered a norm in city slum areas.
According to this rationale, high unemployment
in the inner city has been caused not only by the
low skill level of the workers who live there, but
mainly by their cultural norms and life style. As with
the mismatch hypothesis, adequate testing of this
argument cannot begin until data on the quality of
the central city job growth become available.
Under the occupational hypothesis, the higher
jobless rates in the city compared with the suburbs
stem from the fact that the city has greater propor­
tions of workers in those occupations with tradi­
tionally high unemployment rates (operative, non­
farm laborer, service, and so forth). Even when
comparisons are made by broad occupational cate­
gories, central city unemployment rates are higher
than suburban unemployment rates. (See table 4.)
The disparity between city and suburban un­
employment rates by occupation must be attributed
at least partly to the higher proportion of Negroes
in the city labor force. This can be seen if we look
at the central city and suburban occupational jobless
rates by race. The absolute differences between
city and suburban unemployment rates by race for
most occupations are smaller than the differences
for all races combined. Since Negroes generally
have substantially higher unemployment rates than
whites for the same occupation, their concentration
in the city tends to increase the overall gap between
the city and suburban occupational jobless rates.

Table 4. Unemployment rates by occupation for all SMSA’s, their central cities, and their suburban rings, by occupa­
tion and color, 1970 annual averages

Total
Occupational group

White

Negro and other races

All
SMSA’s

All
central
cities

All
suburban
rings

All
SMSA’s

All
central
cities

All
suburban
rings

All
SMSA’s

All
central
cities

All
suburban
rings

1970
All workers........................................................................ .
White-collar workers_______________ ___________
Professional and technical......................................
Managers and officials..............................................
Clerical workers......................................... ..............
Sales workers......... .................................................

5.1
3.0
2.2
1.5
4.1
4.0

5.6
3.4
2.5
1.9
4.3
4.5

4.7
2.7
1.9
1.3
3.9
3.6

4.7
2.8
2.2
1.5
3.8
3.7

4.9
3.1
2.5
1.9
3.8
4.0

4.5
2.7
1.9
1.3
3.8
3.5

8.1
5.0
2.1
2.0
6.9
8.9

8.3
5.3
2.4
(l)
7.0
9.5

7.4
4.2
0
0
6.3
0

Blue-collar workers......... ............................... ...............
Craftsmen and foremen-------------------------------Operatives....... .......................................................
Nonfarm laborers....... ................. ...........................

6.4
3.8
7.3
10.2.

6.9
4.4
7.4
10.8

5.9
3.5
7.1
9.5

6.0
3.7
6.9
9.9

6.3
4.3
6.8
10.3

5.7
3.4
6.9
9.6

8.7
5.2
9.0
11.0

8.8
4.8
9.0
11.6

8.5
6.8
8.9
9.0

Service workers............... ................................................
Farm workers................................................ - ...............

5.4

5.6

5.2

0

0

0

5.0
(2)

4.3
(2)

5.1
(2)

6.7
(2)

6.8
(2)

0

1 Not shown where unemployment estimate is less than 5,000 or where labor force
is less than 50,000.




6.6

1 Rates for farm workers are not shown since these workers constitute a very minute
percent of metropolitan employment.

149

This effect can most clearly be seen in the clerical,
operative, and service occupations where Negroes
are most concentrated.
A factor which may be influencing the disparity
between central city and suburban jobless rates for
the same occupation is that central city and sub­
urban workers of the same broad occupational
grouping may not have the same level of skills.
Within the same broad occupational grouping the
city workers may be in lower skilled suboccupations,
with relatively higher unemployment rates, than sub­
urban workers. This effect cannot be quantified with
the limited occupational data available.
Notwithstanding that, for the same occupation,
urban residents have higher unemployment rates
than suburban residents, the occupational hypothesis
seems to be supported by the data in tables 2 and 4.
It would be spurious reasoning, nevertheless, to
conclude that the central city-suburban jobless dif­
ferential is attributable entirely to differences in the
occupational levels of the respective labor forces.

Eastern seaboard and in the Midwest) may contain
relatively large cities that share most of the problems
of the urban cores.
The boundaries of the metropolitan areas listed
in table 5 correspond to 1960 definitions. Since
then, approximately half of these SMSA’s have been
redefined either to include additional suburban coun­
ties or to exclude counties. However, the effect on
most areas is probably very slight. Table 6 shows the
additions and deletions of territory since 1960 to
the areas affected and the proportion of the 1970
SMSA population attributed to the change in
definition.

Areas where white collars predominate
As table 5 shows, in eight metropolitan areas
(New York, Los Angeles-Long Beach, San Francisco-Oakland, Washington, D.C., Minneapolis-St.
Paul, Boston, Cincinnati, and Dallas), relatively
large proportions of the work force— over 56 per­
cent— are employed in white-collar jobs. In virtually
all, no more than 30 percent of the workers are
employed in blue-collar occupations.

Individual area highlights
Different political, social, and economic circum­
stances have contributed to the nature of the occupa­
tional distribution within each metropolitan area.
Among these are:
1. The racial composition of the area’s labor
force. If the area (specifically the central city) houses
a large proportion of minority workers, the occupa­
tional distribution of the labor force will gravitate
toward the low skilled occupations.
2. The nature of the industries most important
to the area’s economy. Workers living in or near an
area will generally be in occupations associated with
the industries that dominate its economy.
3. The rate of labor force growth. In a period
when the number of workers living in a particular
area is expanding rapidly, the labor force tends to
be relatively skilled, because usually only workers
with high paying jobs can afford the housing and
other economic amenities common to areas of this
nature.
4. Delineation of the areas in question. Bound­
aries between central city and suburbs are drawn
according to political criteria and not according
to economic differentiation. Some central cities may
be so defined as to contain large neighborhoods of
“suburban” character, while the suburban rings of
some SMSA’s (especially the older ones on the




New York. Because of the local concentration of
corporate headquarters and other public and private
offices in New York, both central city and suburban
workers are primarily white collar. However, nearly
half the central city white-collar workers are in
clerical jobs, while two-thirds of the white-collar
suburbanites are in professional, managerial, and
sales occupations.
Since 1960, the number of workers living in New
York’s suburbs has grown extensively, by about 30
percent, while the number living in the city has not
increased. However, work forces in both the central
city and the suburbs have been occupationally up­
graded during this period, at a fairly even rate, thus
maintaining the relative skill relationships between
residents of the two areas.
Los Angeles-Long Beach and San FranciscoOakland. The relatively high proportion of profes­
sional, technical, and managerial workers in the
Los Angeles and San Francisco SMSA’s is a reflec­
tion of the industrial mix in the two areas. Los
Angeles has many aerospace and research-related
industries, while San Francisco has a heavy concen­
tration of service-producing industries (transporta­
tion and public utilities, trade, finance, insurance,
150

that the suburban ring contains areas and cities 10
of an urban nature.
In both California metropolitan areas, suburban
and central city workers have been occupationally
upgraded fairly evenly since 1960. In San FranciscoOakland, about the same proportion of city workers
are in professional and technical occupations (19
percent) as in the suburbs. However, proportions of

and real estate, and government) that require whitecollar workers. Both areas also have relatively large
educational institutions, which employ large numbers
of professional workers.
Los Angeles-Long Beach is an anomaly in that
workers living in the city hold proportionately more
highly skilled jobs than workers living in the
suburbs.9 This can be attributed in part to the fact

Table 5. Total employment by occupation for the 20 largest SMSA’s, their central cities, and their suburban rings,
1970 annual averages
(Percent distribution]

SMSA

Occupation group

Cen­
tral
city

Sub­
urban
ring

SMSA

NEW YORK

Cen­
tral
city

Sub­
urban
ring

SMSA

Cen­
tral
city

Sub­
urban
ring

SMSA

CHICAGO

LOS ANGELESLONG BEACH

Cen­
tral
city

Sub­
urban
ring

PHILADELPHIA

Total employed (thousands)......................... .
Percent............................................................

4,517
100.0

3,132
100.0

1,385
100.0

3,364
100.0

1,281
100.0

2,083
100.0

2,865
100.0

1,366
100.0

1,499
100.0

1,876
100.0

777
100.0

1,099
100.0

White-collar workers...............................
Professional and technical...............
Managers, officials, and proprietors.
Clerical workers...............................
Sales workers...................................

60.1
16.7
11.9
25.2
6.3

59.1
15.2
10.8
27.7
5.5

62.3
20.0
14.6
19.5
8.2

56.1
17.0
12.4
19.5
7.3

59.0
18.4
12.9
20.6
7.1

54.3
16.1
12.1
18.7
7.4

53.9
15.9
10.5
21.2
6.3

48.9
13.5
7.2
23.2
4.9

58.6
18.2
13.5
19.2
7.7

52.6
15.7
9.8
19.9
7.2

46.8
10.8
7.1
23.1
5.7

56.6
19.1
11.8
17.5
8.3

Blue-collar workers.................................
Craftsmen and foremen...................
Operatives........................................
Nonfarm laborers.............................

27.5
10.7
13.9
2.9

28.0
9.5
15.6
2.9

26.5
13.4
10.1
3.0

32.3
12.5
16.1
3.8

29.3
10.7
15.1
3.6

34.2
13.6
16.7
3.8

35.5
13.2
18.0
4.2

39.8
12.8
21.5
5.5

31.4
13.6
14.8
3.0

36.0
13.1
18.9
4.1

39.8
12.3
22.2
5.3

33.4
13.6
16.6
3.2

Service workers.......................................
Private household workers..............
Other service workers......................

12.3
1.3
11.0

12.9
1.3
11.6

10.8
1.2
9.6

11.0
2.0
9.0

11.3
2.3
9.0

10.8
1.8
9.1

10.3
.8
9.5

11.2
.7
10.5

9.4
.9
8.4

10.8
1.5
9.3

13.4
1.7
11.7

9.0
1.5
7.5

Farm workers................. .........................

.1

.3

.6

.4

.7

.3

.6

.6

0

1.0

0

SAN FRANCISCO—
OAKLAND

DETROIT

Total employed (thousands)......................... .
Percent............................................................

1,571
100.0

582
100.0

989
100.0

1,371
100.0

White-collar workers...............................
Professional and technical...............
Managers, officials, and proprietors.
Clerical workers...............................
Sales workers...................................

47.4
13.5
8.2
20.1
5.6

39.8
10.2
4.6
20.4
4.5

52.0
15.5
10.3
19.9
6.3

60.9
19.1
12.2
22.9
6.7

Blue-collar workers.................................
Craftsmen and foremen...................
Operatives........................................
Nonfarm laborers.............................

40.5
14.7
21.8
4.0

45.4
12.7
28.0
4.7

37.5
15.9
18.1
3.5

Service workers.......................................
Private household workers..............
Other service workers......................

12.0
1.6
10.4

14.9
1.8
13.1

10.2
1.5
8.7

Farm workers...........................................

(l)

0

0

BOSTON

1,165
100.0

239
100.0

926
100.0

876
100.0

174
100.0

702
100.0

60,6
19.7 '
10.9
24.3
5.8

61.0
18.9
12.8
22.3
7.1

59.9
18.3
10.8
23.6
7.1

58.1
15.5
7.7
30.0
4.9

60.4
19.1
11.6
21.9
7.6

48.0
15.5
8.7
16.8
7.0

47.8
13.5
8.0
21.1
5.2

48.2
16.1
8.9
15.6
7.5

26.4
11.2
10.9
4.2

25.6
8.7
12.0
4.8

26.7
12.4
10.3
4.0

28.0
11.5
12.7
3.7

28.7
10.8
12.2
5.7

27.8
11.7
12.9
3.2

38.7
16.8
15.3
6.6

33.1
13.0
12.7
7.4

40.1
17.8
15.9
6.3

12.0
1.9
10.1

13.7
2.5
11.2

11.2
1.6
9.6

11.9
.9
11.0

13.1

11.6
.8
10.8

12.7
1.6
11.1

19.0

0
0

0
0

11.1
1.6
9.6

0

1.1

0

0

0

0

.9

WASHINGTON, D.C.

ST. LOUIS

PITTSBURGH

920
100.0

.7

451
100.0

0

.6

CLEVELAND

BALTIMORE

Total employed (thousands)......................... .
Percent............................................................

909
100.0

228
100.0

681
100.0

1,140
100.0

343
100.0

797
100.0

770
100.0

203
100.0

567
100.0

766
100.0

348
100.0

418
100.0

White-collar workers...............................
Professional and technical...............
Managers, officials, and proprietors.
Clerical workers.............................. .
Sales workers..................................

52.2
14.1
10.5
19.6
8.0

41.0
8.0
6.6
21.1
5.4

55.9
16.2
11.9
19.1
9.0

69.5
25.3
11.4
26.9
5.9

54.0
15.0
5.9
28.7
4.3

76.3
29.7
13.8
26.2
6.5

53.2
12.8
12.3
21.4
6.8

35.2
7.9
5.1
17.3
4.9

59.6
14.4
14.9
22.8
7.6

49.8
15.0
8.8
20.2
5.8

37.9
10.1
5.8
18.0
4.1

59.6
19.1
11.5
21.8
7.4

Blue-collar workers.................................
Craftsmen and foremen...................
Operatives........................................
Nonfarm laborers.............................

35.2
13.1
17.7
4.4

39.9
10.9
23.3
5.7

33.6
13.8
15.9
4.0

17.8
8.6
5.6
3.6

23.8
7.0
9.5
7.3

15.2
9.3
4.0
2.0

35.5
14.0
17.5
4.0

47.7
13.7
26.7
7.3

31.1
14.1
14.2
2.8

36.6
13.9
16.1
6.5

43.8
13.2
21.1
9.5

30.5
14.4
12.0
4.1

Service workers.......................................
Private household workers..............
Other service workers......................

12.2
1.5
10.7

19.1
3.3
15.8

9.8

12.5
2.4
10.1

22,1
4.5
17.6

8.4
1.5
6.9

11.2
.7
10.5

17.1

9.2

8.8

0
0

0
0

13.4
2.2
11.2

18.3
3.3
15.0

9.2
1.2
7.9

Farm workers..........................................

0

0

0

0

0

0

0

0

0

0

0




1.0

151

0

Table 5. Continued—Total employment by occupation for the 20 largest SMSA’s, their central cities, and their
suburban rings, 1970 annual averages
Percent distribution]

SMSA

Cen­
tral
city

Sub­
urban
ring

SMSA

Cen­
tral
city

Sub­
urban
ring

SMSA

Cen­
tral
city

Sub­
urban
ring

SMSA

Cen­
tral
city

Sub­
urban
ring

Occupation group
NEWARK

MINNEAPOLISST. PAUL

BUFFALO

HOUSTON

Total employed (thousands)........................... ............. ..........
Percent.....................................................................................

752
100.0

92
100.0

660
100.0

759
100.0

291
100.0

468
100.0

509
100.0

167
100.0

342
100.0

758
100.0

550
100.0

208
100.0

White-collar workers........................................................
Professional and technical.........................................
Managers', officials, and proprietors________ ____
Clerical workers.........................................................
Sales workers.............................................................

52.0
16.0
11.6
18.7
5.6

27.4
4.7
(l)
15.0

0

55.6
17.7
13.0
19.2
5.8

56.3
18.1
11.0
20.8
6.4

52.5
16.6
6.9
24.3
4.7

58.7
19.2
13.6
18.6
7.3

53.7
18.0
9.8
18.0
7.8

48.0
15.4
5.6
20.4
6.7

56.0
19.2
11.7
17.0
8.4

50.2
12.5
11.0
19.2
7.5

54.8
14.3
12.3
20.2
8.0

38.8
7.9
7.5
16.8
6.5

Blue-collar workers...........................................................
Craftsmen and foremen.............................................
Operatives...................................................................
Nonfarm laborers.......................................................

38.1
13.0
21.1
4.0

55.9
13.1
34.3
8.5

35.6
13.0
19.2
3.4

28.9
10.9
14.0
4.0

30.0
10.8
14.6
4.5

28.1
11.0
13.6
3.7

34.5
13.7
16.1
4.8

36.8
11.8
19.1
5.9

33.6
14.5
14.8
4.2

36.0
14.7
16.3
5.0

30.7
11.8
13.5
5.4

49.5
22.0
23.4
4.2

Service workers.................................................................
Private household.......................................................
Other service workers................................................

9.8
1.4
8.4

16.6

0
0

8.9
1.4
7.6

13.7
1.5
12.2

17.5

11.2
1.7
9.5

11.4
.9
10.5

15.2

9.7

13.6
2.7
10.9

14.3
3.3
11.0

11.7

Farm workers.....................................................................

0

(l)

0

1.7

0

0

0

1.0

0
0
0

PATERSONCLIFTON-PASSAIC

MILWAUKEE

0
0
0

0
0
0

CINCINNATI

0
0
0

DALLAS

Total employed (thousands).....................................................
Percent......................................................................................

526
100.0

285
100.0

241
100.0

556
100.0

132
100.0

424
100.0

435
100.0

197
100.0

238
100.0

696
100.0

385
100.0

311
100.0

White-collar workers..........................................................
Professional and technical.........................................
Managers, officials, and proprietors..........................
Clerical workers..........................................................
Sales workers.............................................................

49.4
13.4
11.8
16.8
7.5

43.9
12.1
8.8
16.7
6.3

56.2
15.0
15.5
16.7
8.6

53.5
13.9
11.3
20.9
7.3

39.2
12.8
7.2
14.4
4.7

58.2
14.4
12.5
22.9
8.3

56.7
18.7
11.5
19.3
7.2

58.8
22.6
9.7
20.4
6.2

55.3
15.4
13.0
18.3
7.7

59.3
16.0
13.1
21.7
8.4

59.1
14.7
13.1
22.0
9.2

59.6
18.2
13.0
21.2
7.4

Blue-collar workers...........................................................
Craftsmen and foremen.............................................
Operatives...................................................................
Nonfarm laborers.......................................................

36.6
14.0
18.7
3.9

40.4
14.8
21.7
3.9

32.2
12.9
14.6
4.3

37.1
14.2
19.0
3.9

46.6
11.8
29.4
5.4

34.2
14.9
15.8
3.5

32.7
12.3
16.6
3.8

29.4
10.2
15.7
3.5

35.0
13.8
17.1
4.1

28.6
11.4
13.3
3.9

27.2
8.8
13.6
4.8

30.9
15.2
13.0
■2.6

Service workers.................................................................
Private household.......... .........................................
Other service workers................................................

13.9
1.4
12.5

15.7

11.6

9.4
1.3
8.1

14.2

7.8
1.2
6.6

9.5
2.0
7.5

11.8
2.7
9.1

8.1

11.5
1.9
9.6

13.5
2.0
11.5

Farm workers.....................................................................

0

0

0

0

0

0
0
0

0
0
0

0

1 Percent not shown where employment estimate is less than 5,000.

0
0
0

0

0
0
0

8.2

0
0
0

Percent not shown where private household employment estimate is less than 5,000.

managers, salesmen, and craftsmen are larger in
the suburbs, as was the case in 1960.

metropolitan area have also experienced nearly
equal occupational upgrading during this period.

Boston, Minneapolis-St. Paul, and Dallas. The high
proportion of white-collar workers in Boston,
Minneapolis-St. Paul, and Dallas is also attributable
largely to the relative importance of the serviceproducing industries— especially trade and finance,
insurance, and real estate. The high proportion of
white-collar workers in these three areas (particularly
in Minneapolis-St. Paul) may be explained in part
by the great majority of workers residing in the
central city being white. However, central citysuburban skill gaps of average magnitude are still
evident in Boston and Minneapolis-St. Paul.
In Dallas, the number of workers has increased
substantially in both the city and suburbs over the
decade. The central city and suburban ring of the

Cincinnati. The occupational distribution in the cen­
tral city and the suburbs of Cincinnati is atypical,
in that the proportion of workers in the professional
and technical field is one-third larger in the city
than in the suburban ring 11. However, larger pro­
portions of managers and craftsmen live in the
suburbs. In both the city and suburbs, workers have
been occupationally upgraded extensively since
1960, and the number of workers has also increased.




Washington, D.C. In the metropolitan area that in­
cludes the Nation’s capital, seven-tenths of all
workers are white-collar, with one-fourth in pro­
fessional and technical occupations. This is, of
course, a reflection of the dominant position of the
152

Federal Government in the area. The skill level of
workers living in the Washington, D.C., suburban
ring is the highest of all 20 suburbs: three-fifths of
the workers are either professionals, technicians,
managers, officials, or craftsmen.
Central city residents, though holding more skilled
jobs in comparison with workers in many central
cities, have a relatively small share of the higher
skilled jobs available in the metropolitan area. In
the District of Columbia itself (the central city),
seven-tenths of the resident workers are Negro.
Because of extensive migration during the past
decade, both in and out, the present city residents
show virtually no occupational upgrading over the
1960 residents.

Moderate proportions of both blue and white
Chicago, Philadelphia, St. Louis, Cleveland,
Buffalo, and Houston have relatively moderate pro­
portions of both blue-collar workers (around 35
percent) and white-collar workers (between 50 and
54 percent). No particular industry predominates
in these areas, although, with the exception of
Houston, manufacturing is relatively strong.
Chicago, Philadelphia, and Buffalo. In these three
areas, about 55 percent of the workers living in the
Table 6.

suburbs are employed in professional and technical,
managerial, sales, and craftsmen occupations. In
comparison, only about 37 percent of the workers
who live in the central cities are in these occupations.
In all three areas, the number of workers living
in the central city has not grown since 1960. In
Philadelphia and Chicago, occupational upgrading
of the work force has been slight; in Buffalo, more
substantial. In comparison, the suburban work force
in these areas has grown, both numerically and in
terms of the proportion of higher skilled jobs.
Cleveland. Here the difference in occupational array
between those living in the central city and in the
suburbs is wide. Three-tenths of suburban workers
are in professional and managerial occupations, but
only one-eighth of all city workers. The majority of
city workers (about two-thirds) are employed in
blue-collar and service jobs.
Sf. Louis. The central city-suburban occupational
gap is also wide in St. Louis. Suburban residents
hold a disproportionate number of the skilled jobs
available in the metropolitan area. The central city
labor force has decreased both numerically and in
terms of the proportion of skilled jobs since 1960.
Migration of workers helps to explain the slight
occupational downgrading of the inner city. A large

Definitional changes in the 20 largest Standard Metropolitan Statistical Areas, 1960-70

SMSA

No change

Additions

Deletions

Percent to which use of 1960 boundary
definitions overestimate (underesti­
mate) 1970 population 1
SMSA

Suburban ring

Orange Co.

+20

+37

Solano Co.

+ 5
—less than 1

+ 8
—less than 1

Loudoun Co., Va., Prince
William Co., Va.

- 5

- 7

Franklin Co., Mo.

- 2

- 3

Geauga Co., Medina Co...............

- 7

-1 1

Hartford Co..................................

- 6

-1 0

Brazoria Co., Fort Bend Co.,
Liberty Co., Montgomery Co.

-1 2

-3 2

Kaufman Co., Rockwall Co.........

- 3

- 6

Ozaukee Co., Washington, Co__
Clermont Co., Ohio, Warren Co.
Ohio, Dearborn Co., Ind.
Boone Co., Ky.

- 8
-1 8

-1 7
-2 6

No change
No change
No change
No change
Sherborn Town, Middlesex Co.
Millis Town, Norfolk Co.

No change
No change

No change

No change
No change

1 Positive percent denotes an overestimation and a negative percent denotes an
underestimation.




153

SOURCE: 1970 Census of Population, Preliminary Report PC (P3)-3 (U.S. Depart*
ment of Commerce, Bureau of the Census, 1971).

outmigration of white workers decreased the city
labor force by one-fifth. Consequently, the city
labor force became increasingly Negro, due both
to the shift of white workers and inmigration of
Negroes. Today, two-fifths of the city’s workers
are black.

except Baltimore, employment is very heavy in the
durable goods industries.
Detroit and Pittsburgh. About two-fifths of all metro­
politan workers in Detroit and Pittsburgh are in
blue-collar jobs. The influence of the automobile
industry is strong in Detroit, with about 22 percent
of the labor force being operatives. Close to one-half
of the workers who live in Detroit’s central city are
in blue-collar jobs (mostly as operatives), and
15 percent are in service occupations.
Over half the workers living in Pittsburgh’s central
city and the same proportion in its surrounding
suburban ring are in blue-collar and service occupa­
tions. Service workers living in the city account for
one-fifth of total city employment. Pittsburgh has
a higher proportion of blue-collar workers in the
suburbs (40 percent) than in the city (33 percent).
The suburban blue-collar workers, however, are
more likely to be craftsmen and foremen and less
likely to be nonfarm laborers.

Houston. Today, one-half of Houston’s metropolitan
work force is white collar. Workers living in the city
are in white-collar and service jobs. About onehalf of the workers in the suburban ring (as defined
in 1960 12) are blue collar— the converse of the
situation in most metropolitan areas. One reason
for the high proportion of craftsmen and operatives
in the suburban ring may be that over the last
decade employment in contract construction and
the oil industry has grown extensively in the
Houston metropolitan area.
Newark and Paterson-Clifton-Passaic. In New
Jersey’s two largest metropolitan areas— Newark
and Paterson-Clifton-Passaic— the occupational gap
between the city and suburban work forces is wide.
In both areas, employment is moderately heavy in
both durable and nondurable goods manufacturing.
The suburbs of both areas have fairly high propor­
tions of professional and managerial workers, while
the central cities are peopled mainly by blue-collar
workers, especially operatives.
In the Paterson area, workers residing in either
the city and suburbs have been occupationally up­
graded only slightly since a decade ago. Most of
the occupational change has been in movement of
workers among the lower skilled occupations.
Workers in the Newark and Paterson metropolitan
areas (especially in the city of Newark) are strongly
represented in operative occupations. Moreover,
about seven-tenths of the workers living in Newark
are employed in blue-collar and service occupations,
a slightly larger proportion than a decade ago.
The relatively low skill level of workers in Newark’s
central city results in part from the inmigration of
untrained Negroes and Puerto Ricans.

Baltimore. In the Baltimore metropolitan area, onehalf the labor force is in white-collar occupations.
In the central city, this proportion is 38 percent
and in the suburban ring 60 percent. About 30
percent of the suburban work force is in professional
and managerial occupations— twice as large a pro­
portion as that for the central city work force.
Since 1960, the labor force living in the Baltimore
suburbs has grown extensively (passing the number
of city workers) and has become increasingly white
collar. On the other hand, the labor force living in
the city has not grown nor shown any significant
upgrading. Today, half the workers who live in
Baltimore city are Negroes.
Milwaukee. In the Milwaukee SMSA, one-half of
the labor force is in white-collar occupations. Its
central city-suburban skill gap is not of the magni­
tude of Baltimore’s, however. Because of the
importance of durable goods manufacturing in this
area, approximately one-third of the workers are
employed as craftsmen and operatives.

Areas with strong blue-collar orientation
Conclusions
In Detroit, Pittsburgh, Baltimore, and Milwaukee,
one-half or less of the work force is employed in
white-collar occupations. Stated conversely, at least
half of the workers in these areas are in blue-collar
and service occupations. In all of these SMSA’s




Reflecting their important and growing role in
the Nation’s economy, metropolitan areas have
larger proportions of professional and technical,
managerial and official, and clerical workers than
154

The same factors that have affected the occupa­
tional gap in the past will probably continue to
do so. Two interrelated factors— the racial com­
position of the city labor force and the rate of
city-to-suburb migration— may have the most effect
on the future occupational distribution of city
workers and the chance of closing the gap in skills
between them and their suburban counterparts.
At present, the proportion of Negro workers in the
central cities of the 20 largest metropolitan areas
is about one-fourth. If the central city-suburban
skill gap is to be narrowed, greater effort must be
made to make city living attractive to skilled
workers, to equip relatively unskilled city residents
with sufficient skills to enable them to compete for
good jobs, and to remove discriminatory hiring and
housing practices.
□

other areas of the country. These areas (especially
the 20 largest) are the Nation’s centers of industrial
and business activity, and their labor forces can
be expected to lead the path of occupational evolu­
tion in the future as they have in the past.
So long as workers tend to move from the central
city as their occupational level rises, the central
city labor force will remain below that of the
suburbs in terms of skill levels. This situation is
unlikely to change until there are substantial changes
in housing patterns and in socioeconomic conditions.
Earlier, it was concluded that in the 20 largest
SMSA’s the gap in skills between central city
residents and those in the suburbs is only slightly
larger today than a decade ago, notwithstanding
significant upgrading in both SMS A components.
However, in many individual SMSA’s the gap has
substantially increased over the decade.

FOOTNOTEST he U rb a n Institute, W ashington, D .C ., Ja n u a ry 1970.

1 D a ta in this article on m etro p o litan areas re fe r to the
212 S tan d ard M etro p o litan S tatistical A reas as defined in
1960. It should be n o ted th a t the d a ta in this re p o rt re p re ­
sent the place o f residence o f w orkers ra th e r th a n th eir
place o f w ork.

8 See P eter B. D oeringer, “L a b o r M ark et R ep o rt fro m the
B oston G h e tto ,” M onthly Labor Review, M arch 1969,
pp. 55-56.
9 D a ta fro m the 1960 C ensus show th a t this relationship
w as also true a decade ago.

2 U ntil recently, d a ta on the occupational distrib u tio n of
m e tro p o litan w orkers have been available only fro m the
decennial census and to a lim ited extent fro m the U rb an
E m ploym ent Survey (U E S ), w hich obtained som e industry
and occupational d a ta on residents living in C on cen tra ted
E m ploym ent F ro g ram (C E P ) areas o f six m ajo r U.S.
cities betw een July 19 6 8 -Ju n e 1969. See BLS R ep o rt 370,
O ctober 1969.

10 T he su b u rb an ring o f the L os A n g eles-L o n g B each
SM SA as defined in 1960 contained cities in O range C ounty,
such as A naheim , F u lle rto n , G a rd e n G rove, S anta A na,
and so fo rth , w ith som e u rb a n ch aracteristics. In 1963,
O range C ou n ty w as deleted fro m the L os A ngeles SM SA
and m ade a separate SM SA (A n a h e im -S a n ta A n a -G a rd e n
G ro v e ). Since the d a ta in this article are based on 1960
definitions, these u rb a n areas are included in the L os
A ngeles-L ong B each su b u rb an boundaries.

3 These are the 20 largest SM SA ’s as defined and ranked
in 1960. D a ta fro m the 1970 decennial census show that
d uring the 1960’s th ree S M SA ’s m oved out o f the top 20
(C in cin n ati, P a te rso n -C lifto n -P a ssa ic , and B uffalo) and
w ere replaced by A n a h e im -S a n ta A n a -G a rd e n G rove,
S eattle -E v e re tt, and A tlan ta.

11 O ne reason fo r this m ay be th e fact th a t the fo u r
counties added to the C incinnati SM SA since 1960 are not
included in the d a ta in this article. T hese counties in 1970
housed o n e-fo u rth o f the C incinnati su b u rb an popu latio n
(tab le 6 ). Because o f the absence o f these counties fro m
the su b u rb an figures, the su b u rb an p ro p o rtio n o f p ro fes­
sional and technical w orkers is no d o u b t u nderstated.

4 See D o ro th y K. N ew m an, “D ecentralization o f Jo b s,”

Monthly Labor Review, M ay 1968, table 1, p. 8.
5 See P au l O. F laim , “Jobless T ren d s in 20 L arge M e tro ­
po litan A re a s,” Monthly Labor Review, M ay 1968,
pp. 16-28.

12 T he H o u sto n su b u rb an rin g as defined in 1960 did not
contain fo u r counties subsequently added to the SM SA.
These counties in 1970 contained three-tenths o f the H o u s­
to n su b u rb an p o p u latio n (ta b le 6 ). T h e absence o f these
counties fro m the su b u rb an figures show n here has no
d o u b t affected the occu p atio n al distrib u tio n o f the H o u sto n
su b u rb an w ork force.

8 N ew m an, op. cit.
7 Study by C h a rlo tte F rem o n , “T he O ccupational P attern s
in U rb a n E m ploym ent C hange, 19 6 5 -6 7 ” (w orking p a p e r)




155

Employment and
unemployment
among Americans
of Spanish origin

Quarterly publication of new series
begins this month; data for 1973
show persons of Spanish origin
had an unemployment rate of 7.5 percent, and
were more likely to be jobless than white workers
ROBERTA V. McKAY

O f t h e 6 m i l l i o n Americans age 16 and over
who identified them selves as being o f Spanish
origin or d esce n t in 1973, an average o f 3.6
million were in the labor force, and they had an
unem ploym ent rate o f 7.5 percent. T hese are
summary findings o f a new Bureau o f Labor
Statistics data series on the employment status of
Americans o f Spanish origin, now available for
the first time on a regular basis.
In the recen t p a st, data on A m erican s o f
Spanish origin have been collected only once a
year. Moreover, very little detail had been availa­
ble on a consistent and continuous basis. Since
March 1973, monthly information on labor force
characteristics o f the civilian noninstitutional pop­
ulation o f Spanish origin 16 years of age and over
have been tabulated separately by the Bureau o f
the Census as a part o f the ongoing monthly
survey o f the N a tio n ’s labor fo r c e .1 U nder a
program sponsored by the U .S . Department o f
L abor’s M anpower Adm inistration, these data
will be published quarterly by the Bureau o f Labor
Statistics, beginning in April 1974.
This article introduces the continuous labor
force data series for Americans o f Spanish origin.
It first traces the evolution o f the self-identifica­
tion method for classifying persons o f Spanish
origin and discusses a few o f the major technical
caveats. Its main focus, however, is on analyzing
initial findings from the survey based on 1973 an­
nual averages.2
Data com parability

Before 1973, the two major sources o f pub­
lished data on the labor force characteristics of
persons o f Spanish origin have been the decennial
R o b e rta V . M c K a y is an e c o n o m is t in th e D iv isio n o f
E m ploym ent and U nem ploym ent A nalysis, B ureau o f L abor
S tatistics.

From the Review of April 1974



156

cen su ses and once-a-year supplem ents to the
Current Population Survey (CPS) in 1969, 1971,
and 1972.3 In addition to labor force data, both
the census and Current Population Survey supple­
mental series included data on a wide range of
characteristics o f the population.
The socioeconom ic data collected during the
decennial census were derived from a population
universe based on a changing characterization o f
Spanish background and ethnicity. The earliest
published social and econom ic data on Spanish
ethnicity from decennial censuses were derived
from questions on the country o f birth o f the
individual (1850) and subsequently, birthplace of
parents (1890). Still later, when the first direct
question on Spanish ethnicity was introduced
(1930), Mexican Americans were identified from a
“ race” question (a one-time question in which
M exican A m ericans were considered to be a
racial group). In subsequent decennial years, the
characterization of Spanish Americans was pro­
gressively expanded by including questions on
Spanish mother tongue (1940) and (1950) classifi­
cation by Spanish surname in the five Southwest­
ern States where many Am ericans o f Spanish
origin reside. In 1970, Spanish Americans were
identified by use o f four identifiers: origin or
descent, mother tongue, surname, and place of
birth or parent’s birth. Three o f these definitions
are utilized in defining Spanish heritage, a term
used in many 1970 census reports.4 The Spanish
origin or descent definitions, used in many of the
Census o f Population Subject Reports, relies on
self-identification o f Spanish ethnicity.
In the annual ethnic origin supplement to the
Current Population Survey, initiated in Novem bei
1969 and then conducted in March o f 1971 and
1972, estimates have been based on the respond­
ent’s identifying him self as of Spanish origin or
descent. Self-identification essentially consists of

asking all survey respondents, “ What is . . . ’s
origin or descent?’’, with the enumerator coding
according to seven Spanish categories.5
With changing terms and differing collection
methods, the population counts resulting from the
1970 census and the estim ates from the annual
ethnic supplem ents to the Current Population
Survey have varied widely. A lack o f comparability
between the census and the Current Population
Survey and, over the past 5 years, within the Cur­
rent P op ulation S u rvey series has th erefore
emerged. The Current Population Survey estimates
were found to yield differing population growth
counts because o f technical factors such as sample
redesign (in 1972-73), reclassification o f the origin
of children under age 14, revision of Mexican-origin
categories, and sampling variability.6

Population and labor force
The 6 million persons 16 years and over o f
Spanish origin in 1973 accounted for 4 percent o f
the N ation’s civilian noninstitutional population of
th is age group. T he se x co m p o sitio n o f the
population was about the same as that o f both
racial groups, but persons of Spanish origin, like
blacks, had a lower median age than the white
population. Teenagers o f Spanish origin com ­
prised a proportion o f their working age popula­
tion nearly 1Yi times that o f their white counter­
parts.
Table 1. Employment status of persons of Spanish origin,
whites, and blacks, by sex and age, annual averages, 1973
[In thousands]

The new series

The new Spanish origin labor force estimates
are derived from the ongoing monthly Current
Population Survey, which collects information on
labor force activities o f all persons age 16 and
over in the United States. The data on persons o f
Spanish origin in this article are somewhat limited
in detail. The relatively small size o f this group— less than 5 percen t o f the total population
—particularly subjects these estimates to a high
degree o f sampling variability, that is, the variation
that might occur by chance because only a sample
o f the population has been surveyed. C on se­
quently, only the larger estimates can be reliably
reported. However, the availability and use o f an­
nual averages serve to enhance data reliability, thus
permitting a number o f comparisons between the
labor force experience o f Spanish origin workers
and other workers.7
Persons classified as o f Spanish origin also are
counted as white or black. For purposes of labor
force a n alysis, persons o f Spanish origin are
tabulated separately, without regard to race. Ac­
cording to the 1970 cen su s, approxim ately 98
percent o f the population group is white. The size o f
the Spanish ethnicity component within each color
group is not large enough to either affect the m ove­
ment of the entire group or substantially bias any
one estim ate.8
(The comparisons in this article also introduce
specially tabulated data for black workers from
the Current Population Survey. Previously, only
the “ Negro and other races’’ classification, of




which blacks comprise 89 percent, has been used
in racial comparisons.)

157

Employment status

Spanish
origin

White

Black

145,936
88,714
60.8
84,409
3,452
80,957
4,304
4.9
57,222

5,997
3,603
60.1
3,333
222
3,111
280
7.5
2,394

129,302
78,689
60.9
75,278
3,144
72,134
3,411
4.3
50,613

14,788
8,890
60.1
8,061
258
7,803
829
9.3
5,898

60,943
49,539
81.3
47,946
2,500
45,445
1,594

2,425
2,084
85.9
1,973
167
1,806
111
5.3
341

54,503
44,490
81.6
43,183
2,269
40,915
1,307
2.9
10,013

5,662
4,430
78.2
4,170
193
3,977
260
5.9
1,232

69,249
30,713
44.4
29,228
550
28,678
1,485
4.8
38,536

2,718

61,319
26,647
43.5
25,494
506
24,988
1,153
4.3
34,672

7,050
3,635
51.6
3,325
37
3,288
310
8.5
3,415

15,744
8,461
53.7
7,236
402
6,834
1,225
14.5
7,283

855
401
46.9
321
27
294

13,481
7,552
56.0
6,602
370
6,232
950
12.6
5,929

2,076
824
39.7
566
28
537
259
31.4
1,251

Total

TOTAL, 16 YEARS OLD AND OVER
Civilian noninstitutional population.............
Civilian labor force................................
Percent of population.............
Employment....................................
Agriculture___ _____ _____
Nonagricultural industries__
Unemployment...............................
Unemployment rate........
Not in the labor force................ ..........

MALES, 20 YEARS OLD AND OVER
Civilian noninstitutional population.............
Civilian labor force............. ..................
Percent of population__
Employment....................................
Agriculture..............................
Nonagricultural industries...
Unemployment_____________ _
Unemployment rate........
Not in the labor force............................

3.2

11,404

FEMALES, 20 YEARS OLD AND OVER
Civilian noninstitutional population--------Civilian labor force-----------------------Percent of population__
Employment..................................
Agriculture........... .................
Nonagricultural industries...
Unemployment.......... ...................
Unemployment rate........
Not in the labor force............................

1,118

41.1
1,038
28
1,010

81
7.2
1,599

BOTH SEXES, 16 TO 19 YEARS OLD
Civilian noninstitutional population.............
Civilian labor force.........................—
Percent of population...
Employment...............................
Agriculture..............................
Nonagricultural industries...
Unemployment...............................
Unemployment rate____
Not in the labor force.................. .........

79

19.8
454

NOTE: Since persons of Spanish origin are also counted as white or black, 3 groups
shown will not sum to total.

The Spanish origin civilian labor force averaged
3.6 m illion persons in 1973, com posed o f 2.1
million adult men, 1.1 million adult women, and
400,000 teenagers (table 1). In percentage terms,
this age-sex distribution o f the labor force was
similar to that for whites. Compared with blacks,
however, the work force o f Spanish origin had a
larger proportion o f adult men and a smaller
proportion o f adult women.
On an overall basis, the labor force participa­
tion rates— the civilian labor force as a percent of
population— o f Spanish origin persons did not
differ much from those of their white and black
counterparts. In 1973, 60.1 percent o f the Spanish
origin were in the labor force, identical with the
black proportion but slightly lower than that o f
white workers.
The major a ctivity o f persons in the three
groups who were not participating in the labor
force were likewise similar. (See table 2.) At least
7 out o f 10 workers in each o f the population
groups cited home responsibilities or school as
reasons for not working or looking for work.
The labor force participation rate o f adult men
of Spanish origin, at 85.9 percent in 1973, was
Table 2. Major activity of persons not in the labor force,
Spanish origin, whites, and blacks, by age and sex, annual
average, 1973
Spanish origin

Black

White

Major activity
Age
16 to
19

Age
20 and
over

Age
16 to
19

Age
20 and
over

Age
16 to
19

Age
20 and
over

above the 81.6-percent rate for white men and
considerably higher than the 78.2 percent for
black men. (See table 3.) A t every age lev el,
except for those age 65 and over, adult Spanish
men were considerably more likely to participate
in the labor force than black adult men. H owever,
adult Spanish men participated at a higher rate
than white men only in the age group 20 to 24
years— 88.5 compared to 85.8 percent— reflecting
the fact that a smaller proportion o f Spanish men
are still in school at that a g e.9 The dominant
influence on the differing overall labor force
participation rates of adult men, Spanish versus
white, has been the dissimilar age distributions of
their populations and labor forces. Young adult
men, who have higher participation rates than
those age 55 years and over, comprise a relatively
greater proportion o f the Spanish origin than
white population and labor force groups. The
proportionately older age o f white men serves to
lower their overall participation rate.
The small proportion o f adult men of Spanish
origin who did not work or look for work was
similar to that o f whites and blacks. The main
reasons were inability to work and voluntary idle­
ness, retirement, waiting to enter school or Armed
Forces, and discouragement over job prospects.
The labor force participation rate o f Spanish
teenage boys was 55.2 percent, less than that for
whites but greater than that for blacks. When not
in the labor force, Spanish teenage boys were
as likely to be in school as other teenage boys.
Table 3. Civilian labor force participation rates of persons
of Spanish origin, whites, and blacks, by age and sex, annual
average, 1973

TOTAL
Not in the labor force (in
thousands)__________ ____
Percent distribution--------Home responsibilities, _
School_____________
Unable to w ork,..........
Other....................... .

454

1,940

5,929

76.1
3.9

17.3

14.1

12 7
71.6
5
15 2

44,685
100 0
69 9
3.8
5 0
21 3

1,251
100 0
13 5
70.4
5
15.7

4.647

19 6
62.9

185

341
100 0
2.3
11.7
23 8
62.2

2,551

10,013

542

1,232

100 0 100 0 100.0
2

6.0

100.0

Spanish origin

61 9
5.3
11.5
21 2

Black

White

Age
Males

Fe­
males

Males

Fe­
males

Males

Fe­
males

81.5

40.9

79.5

44 1

73.3

49.3

55.2
85.9

39.1
41.1

62.0
81.6

50.1
43.5

45.6
78.2

34 2
51.6

88.5

48.6

85.8

61.6

83.6

58.0

94,3

44.0

96.3

48.5

91.8

62.7

94.6

45.5

96.8

52.2

90.9

61.6

91.5

45.2

93.5

53.4

87.5

56.1

74.8

29.4

79.0

40 8

69.4

44.7

21.5

7.1

22.8

8.7

22.4

11.4

MALES
Not in the labor force (in
thousands)...............................
Percent distribution. ____
Home responsibilities..
School_____________

100.0
1.1

Other_______ ______

22.1

76 8

100.0
6
81 3
.7
17.4

100.0 100.0 100.0
9
3.2
1.8
9.6
9.8
78 8
14.1
.6 25.2
74.3

19.7

62.0

FEMALES
Not in the labor force (in
thousands)_______________
Percent distribution_____
Home responsibilities..
School_____________
Unable to work______
O ther...........................




269

1,599

3,377

100 0 100.0 100.0
91.9
32.1
21 8
2.2 64 3
53.7
4
2.2
.4
13.8

3.8

13.5

34,672
100 0
89.6

2.0
6.0

2 4

709

3,415

23.0
64.0
4

83 1
3.8

100.0 100.0
12.6

6.6
6.5

Total, IS years old and
over________ _______
Both sexes, 16 to 19
years old........ .........
20 years old and over.
20 to 24 years
old.................. .
25 to 34 years
old__________
35 to 44 years
old__________
45 to 54 years
old............ .......
55 to 64 years
old__________
65 years old and
over..................

158

At 41 percent, the labor force participation rate
o f adult women o f Spanish origin was slightly
lower than the participation rate of white women
and considerably lower than that of black women.
It was even below that o f white teenage girls, a
group whose participation has been among the
lowest o f all race-sex groups. The high degree of
nonparticipation o f Spanish women undoubtedly
reflects the traditional role o f women in the Spanish
hom e.10 Household responsibilities were the reason
92 percent of these women not in the labor force
were neither working nor looking for work.
Although the major reason that teenage girls of
Spanish origin were not in the labor force was
“ sch ool,” the proportion citing school was lower
than that for w hite or black teenage fem ales.
Moreover, they were more likely to cite household
responsibilities as keeping them from working or
looking for work than did other teenage girls.

Employment
An average of 3.3 million persons o f Spanish
origin were em ployed in 1973. Adult men ac­
counted for nearly 2.0 million o f this number,
adult women, 1.0 million, and teenagers, 320,000.
The occupational distribution of employed per­
sons o f Spanish origin was essentially similar to
that o f blacks, except that a smaller proportion of
Americans o f Spanish origin were employed in
service occupations and slightly larger proportions
were blue-collar and farm workers. The differ­
ences between the occupational distribution o f
white and Spanish workers are striking. Whereas
two-thirds o f Spanish workers were employed in
blue-collar and service occupations and fewer
than one-third held jobs in white-collar occupa­
tions, half o f the whites held jobs in white-collar
occupations. The proportion o f Spanish workers in
the professional and managerial occupations was
less than half the proportion for white workers in
such occupations. (See table 4.)
Underlying the occupational distribution o f em­
ployed persons o f Spanish origin w ere som e
important sex differences. Six out o f 10 men of
Spanish origin held jobs in blue-collar occupa­
tions, a proportion comparable to that of blacks
but higher than the 46-percent figure for white
men. W hite-collar em ploym ent among Spanish
males was again roughly the same as for black men,
but less than half the proportion of white men.




159

Women o f Spanish origin were em ployed in
blue-collar occupations to a greater extent than
either white or black women. Thirty-five percent
of women o f Spanish origin compared with 16
percent o f white and 19 percent o f black women
were so employed. A lesser proportion held jobs
in white-collar occupations than white women.
Forty percent o f women of Spanish origin had
white-collar jobs; 63 percent o f white women did.
But Spanish women were only half as likely as
black wom en to work in service occupations,
including dom estic job s. These sex differences
help to explain the more than proportional con­
centration o f Spanish workers in blue-collar occu­
pations, white workers in white-collar occupa­
tions, and black workers in service occupations.

Unemployment
Unemployment is a severe problem for persons
of Spanish origin. In 1973, an average of 270,000
were jobless. At 7.5 percent, their unemployment
rate was more than halfway betw een the 4.3percent rate for white workers and the 9.3 per­
centage for blacks. Workers of Spanish origin
accounted for over 6 percent o f total unemploy­
ment but only 4 percent of the labor force.
Table 4. Employment and unemployment rates, experi­
enced workers only of Spanish origin, whites, and blacks, by
occupation, annual average, 1973
.Employment
(Numbers in thousands)

Unemployment rate
(P ercent of labor
force)

Occupation
Spanish
origin

White

Black

Spanish White Black
origin

Total experienced workers__
Percent distribution__

3,333
100 0

75,278
100 0

8,061
100 0

6.6

3.7

7.8

White-collar workers.........
Professional and technical________________
Managers and administrators, except farm__
Sales workers________
Clerical workers...............

28.9

49 8

28.6

4.3

2.7

6.7

6.5

14.4

8.5

3.3

2.0

4.5

5 5
3.7
13.2

10 0

3.5

6.9
17.5

1.4
5.9
5.5

1.4
3.4
3.8

11.5

14.5

Blue-collar workers______
Crafts and kindred
workers____________
Operative, except transport_______ ________
Transport equipment
operatives__________
Nonfarm laborers.............

49.8

34.7

42.3

7.7

5.0

2.2
8.2
8.0

13.0

13.9

8.8

6.4

3.6

5.3

24.3

12.5

17.5

8.3

5.6

9.4

4.5
80

3.7
4.6

5.8

4.4
9.5

3.9

8.1

5.1
9.5

Service workers_____ _____
Private household______
Other service workers___

15.8

11.7

1.1
10.6

26.4
6.3

6.2

14.0

Farm workers____ ______

5.6

3.7

1.8

2.1

10.2

8.7

20.1

6.6

5.0
2.9
5.2

6.8

2.7

8.7

2.2

6.0

3.2'

9.2

As with the total population, men, women, and
teenagers o f Spanish origin are not equally af­
fected by unem ploym ent. The job less rate for
adult men averaged 5.3 percent, while the rates
w ere 7.2 p ercen t for adult w om en and 19.8
percent for teenagers. These unemployment rates
were higher than those for whites for each age-sex
group, but lower than those for blacks (table 5).
The burden o f joblessness among persons of

Table 5. Unemployment rates and ratios, persons of
Spanish origin, whites and blacks, by age and sex, annual
average, 1973
Ratio

Unemployment rate
Age and sex
Spanish
origin

White

Black

Spanishwhite

Blackwhite

Total, 16 years and over..............

7.5

4.3

9.3

1.7:1

2.2:1

Both sexes, 16 to 19 years____

19.8

12.6

31.4

5.3

5.9

4.6
5.2

2.9
6.5
2.3
2.5

4.6
3.2

1.6:1
1.8:1
1.3:1
2.0:1
2.1:1

2.5:1

Males, 20 years and over.......
20 to 24 years____ ______
25 to 54 years____ ______
55 years and over________
Females, 20 years and over...
20 to 24 years__________
25 to 54 years.............. .......
55 years and over_______

7.2
9.0
7.0
4.5

4.3
7.0
4.0
2.7

8.5
18.3
7.0
3.4

1.7:1
1.3:1
1.8:1
1.7:1

8.2

12.8

2.0:1
2.0:1
2.0:1
1.3:1

2.0:1
2.6:1
1.8:1
1.3:1

Spanish origin can also be demonstrated by the
use o f a Spanish-white unemployment rate ratio,
similar to that com monly used to exam ine the
relationship between black and white unemploy­
ment. The ratio of Spanish to white unemploy­
ment rates of 1.7:1 indicates that relative to the
sizes o f their respective labor forces, for every 10
white workers unemployed there were 17 unem­
p loyed w orkers o f Spanish origin. T his w as
considerably less than the 2.2:1 black-white ratio
in 1973. H owever, for adult men in prime working
years— ages 25 to 54— Spanish-white ratio was
2.0:1, about the same as the black-white ratio.
The new current labor force data available for
1973 tend to confirm the results o f earlier surveys
on the labor force characteristics of workers of
Spanish origin. Joblessness affects a significantly
higher proportion o f Spanish than white workers,
but their unem ploym ent rates are low er than
those for black workers. Adult men of Spanish
origin are more likely than black men to partici­
pate in the labor force, but with the exception of
th o se 20 to 24 years o ld , their labor force
participation rates are lower than those o f their
white counterparts in every age group. Moreover,
adult Spanish w om en participate to a lesser
degree than both black and white workers.
□

-FOOTNOTES
1 F o r a d e ta ile d d e sc rip tio n o f th e C u rre n t P o p u la tio n
S u rv e y , see C on cepts and M ethods U sed in M anpow er
Statistics from the Current Population Survey, R eport 313
(B ureau o f L abor S tatistics, 1967). T his re p o rt is available
from the Bureau on request.
2 A lthough collection o f the data began in M arch of 1973, a
12-month series is provided w hich includes estim ated levels for
January and February.
3 See Persons o f Spanish Origin in the United States:
N ovem ber 1969, Current Population R eports, Population
Characteristics. Series P-20, N o. 213 and N o. 249 (Bureau of
the C ensus, 1971). See also C ensus o f Population: 1970, Subject
Reports, Persons o f Spanish Origin, Series PC (2)-1C (Bureau of
the C ensus, 1973).
4 S p a n ish h e rita g e in c lu d e s p e rs o n s w ith th e follow ing
characteristics: persons o f Spanish language (Spanish m other
tongue and all other persons in fam ilies in which the head o r
the wife reported Spanish spoken in the home as a child) and
persons of Spanish surnam e in the five S outhw estern S tates of
A riz o n a , C alifo rn ia, C o lo rad o , N ew M exico, and T e x a s;
persons of P uerto Rican birth o r parentage in the three Middle
A tlantic S tates o f N ew Jersey , New Y ork, and Pennsylvania:
persons of Spanish language in the rem aining 42 S tates and the
D istrict of Colum bia.
5 T he seven Spanish origin categories are M exican A m eri­
can , C hicano, M exican (M exicano), P uerto R ican, C uban,




C entral or South A m erican, and “ O ther S panish.”
6 T he m ost significant o f these technical adjustm ents took
place betw een M arch 1972 and M arch 1973, resulting in a
gross population increase o f 1.4 million. The basis for these
changes and th eir im pact on the population estim ates have
been detailed in a recent C ensus Bureau report. See Persons
o f Spanish Origin in the United States: March 1973 (A dvance
report), Current Population Reports, Population Characteris­
tics, Series P-20, No. 259 (Bureau of the C ensus, 1974).
7 Sam pling e rro rs fo r S panish estim ates may be obtained
from the au th o r upon request.
8 In the 1970 D ecennial C ensus of Population, it w as found
th at the p ro p o rtio n o f N egroes am ong persons o f Spanish
origin at the national level is probably in the range of IV 2 to 2
percent. See Census o f Population: 1970, Subject Reports,
Persons o f Spanish Origin, Series PC (2)-1C (B ureau of the
C ensus, 1973) and its addendum issued A ugust 1973.
9 See Anne M. Y oung, “ The high school class of 1972:
more at w ork, few er in college” Monthly Labor Review, June
1973, pp. 26-32.
10 F o r a d isc u ssio n o f lab o r force p a rtic ip a tio n ra te s of
A m erican w om en o f S panish origin in term s o f n u m b er of
children, see Paul M. R yscavage and Earl F. M ellor, “ The
econom ic situation o f Spanish A m erican s,” Monthly Labor
Review, April 1973, pp.3-9.

160

Multiple
jobholding
in 1970
and 1971

Survey definitions

For purposes of this survey, multiple job­
holders are defined as those employed persons
who, during the survey week, (1 ) had jobs as
wage or salary workers with two employers or
more, (2 ) were self-employed and also held
wage or salary jobs, or (3 ) worked as unpaid
family workers but also Led secondary wage
or salary jobs. The primary job is the one at
which the greatest number of hours were
worked. Also included as multiple jobholders
are persons who had two jobs during the sur­
vey week only because they were changing
from one job to another. This group is very
small— only 1 percent of all multiple jobholders
in May 1969.

Unemployment and moonlighting

The increase in unemployment during the past 2
years again has focused interest on the relationship
between unemployment and multiple jobholding
rates. The unemployment rate increased from 2.9
percent in May 1969 to 4.1 percent a year later and
reached 5.3 percent in May 1971. In each of these
months, however, the multiple jobholding rate was
either 5.1 or 5.2 percent.
Based on the limited number of over-the-year
comparisons available for the same month and one
Howard V. Hayghe and Kopp Michelotti are economists in
the Division of Labor Force Studies, Bureau of Labor
Statistics.




HOWARD V. HAYGHE AND KOPP MICHELOTTI

comparison over a 3-year period, changes in multi­
ple jobholding rates appear to be unrelated to
changes in unemployment rates. (See chart 1.) Of
the 10 pairs of observations covering a span of 15
years, in only three periods did significant changes
in the same direction occur in both multiple jobholding and unemployment rates (in years ending
in 1963, 1964, and 1966). In two periods (ending

T h e n u m b e r a n d p r o p o r t io n of American workers
who held two jobs or more in May 1971 were vir­
tually unchanged from May of 1969 and 1970, even
though the unemployment rate increased sharply
over that period. In May 1971, 4 million or 5.1
percent of all employed workers had two jobs or
more.
The multiple jobholding rate for men has re­
mained more than double that for women and the
rate for whites continues to be higher than for
workers of Negro and other races. Between May of
1970 and 1971, the number with two wage or salary
jobs in the nonagricultural sector was unchanged.
A small decline in the number with one job in agri­
culture was offset by an increase in the number
combining wage or salary jobs with self-employment.
(See box.)
This report on multiple jobholders, frequently
called moonlighters, includes a discussion of the
relationship between the unemployment rate and
multiple jobholding, earnings of multiple jobholders
on their second jobs, and seasonality and trends in
multiple jobholding.1

From the Review of October 1971

Special Labor Force Report,
based on May 1971 survey,
shows moonlighters earned
an average of $30 a week
on second jobs

161

Persons employed only in private households
(as a maid, laundress, gardener, babysitter, and
so on) who worked for two employers or more
during the survey week were not counted as
multiple jobholders. Working for several em­
ployers was considered an inherent character­
istic of private household work rather than an
indication of multiple jobholding. Also ex­
cluded were self-employed persons with addi­
tional farms or businesses, and persons with
second jobs as unpaid family workers.

Chart 1.

Comparison of changes in multiple jobholding and unemployment rates

Percentage points

9 out of 10 of all employed men in the central
age group were married and most had children under
age 18. Thus, for many, the need for additional
income was strong enough to induce them to seek
second jobs. A recent survey showed that half the
male moonlighters gave meeting household needs or
paying off debts as reasons for holding more than
one job.2
Married women who work generally have family
responsibilities which preclude holding more than
one job. Thus, the multiple jobholding rate for mar­
ried women was lower than that for other women.
Also, a smaller proportion of female than male
moonlighters, about 50 and 90 percent respectively,
were married, reflecting in part the household re­
sponsibilities of married female workers.
The likelihood of the husband working at two
jobs does not appear to be affected by the presence
of the wife in the labor force. The multiple jobhold­
ing rate for married men whose wives were in the
work force was about the same as for those whose
wives were not.
There is a sharp difference in the proportions of
male and female multiple jobholders who combine
a full-time job with a part-time job. The following
tabulation shows that in May 1971 the largest pro­
portion of men who moonlighted worked full time

in 1958 and 1969), the unemployment rates and
multiple jobholding rates moved in opposite direc­
tions. In the remaining five periods, in which the
unemployment rate increased in three and decreased
in two, the multiple jobholding rates remained essen­
tially unchanged.

Personal characteristics
Multiple jobholding is almost entirely a male
phenomenon. In May 1971, only 765,000 out of a
total of 4 million multiple jobholders were women.
Although there was no overall change in the number
of multiple jobholders between May 1970 and May
1971, the number of women with two jobs or more
rose by about 130,000 while the number of men
with two jobs declined correspondingly. The multi­
ple jobholding rate for women, at 2.6 percent, was
higher than in May 1970, as shown in table 1. The
rate for men, on the other hand, decreased slightly
over the year to 6.7 percent from 7.0 percent.
Multiple jobholding was most prevalent among
employed men 25 to 44 years old; 7.9 percent of
these men, representing over half of all male multiple
jobholders, held more than one job in May 1971.
It is not surprising that the rate for men in these
ages is higher than for older or younger men. Almost



162

on their first job and part time on their second:
Total .....................................................
Worked at two part-time j o b s ......................
Worked full time on first job
and part time on s e c o n d ..........................
Worked at two full-tim e j o b s .....................

Men

Women

100
19

100
52

75
6

47
1

It should be noted that, among the women, about
the same proportions worked at two part-time jobs
or combined a full-time with a part-time job.

Industry and occupation
One-half of all multiple jobholders worked in
manufacturing or in service and finance on their first
jobs. This reflected the fact that more persons work
in these two industries than in any others, rather
than high proportions of the workers in these indus­
tries holding second jobs. (Multiple jobholding rates
in these two industries were about average.) Rates
significantly above average are found among workers
whose first jobs are in public administration, and in
transportation and public utilities.
The multiple jobholding rate of men was highest
in May 1971 for those whose primary jobs were in
State and local government, 15 percent. The multiple
jobholding rates of postal service workers and those
in educational services were about as high. In com­
parison, those who worked in mining had a rate of
only 4 percent.
Table 1.

By occupation, men who were employed on their
primary jobs as teachers or as protective service
workers (guards, policemen, and firemen) had the
highest multiple jobholding rates— IS and 16 percent,
respectively. Laborers and waiters, cooks, or bar­
tenders had the lowest multiple jobholding rates, at
about 4 percent.
The opportunity to work at a second job may be
just as important a factor in multiple jobholding
as the need for extra income. Many of the workers
in the industries and occupations where multiple jobholding rates are high have flexible hours on their
principal jobs as among farmers, or working hours
are different from the usual ones, as for many postal
and protective service workers.
About one-third of the multiple jobholders worked
at wage or salary jobs in service and finance on their
second jobs. An additional 15 percent took second
jobs in retail trade. It is not surprising that large
proportions found jobs in these industries, because
they typically provide the part-time or off-hours jobs
that multiple jobholders seek. About one-third of
the multiple jobholders were self-employed in farm
and nonfarm industries on their second jobs. Be­
tween May 1970 and 1971, the number of selfemployed in nonagricultural industries rose by 166,000 to 728,000 (table 2 ). Almost all of the increase
was among those who had their own businesses in
trade and service industries. It may be that some
persons who wanted but could not find a second

Employed persons with two jobs or more, by sex, race, and unemployment rates, 1955-70

[In p e rc e n t]

D a te

M u lt ip le jo b h o ld in g r a t e 1
N um ber
(th o u sa n d s)
B o th se x e s

M en

W om en

W h it e

N e g ro and
o th e r ra ces

U n e m p lo y e d
as p e rce n t
o f c iv ilia n
la b o r f o r c e 1

J u l y 1 9 5 6 . . . . ............................................................................................................
J u ly 1 9 5 7 .............................................................................. ......................................
J u ly 1 9 5 8 ......................... ............................................................................................

3 ,6 5 3
3 ,5 7 0
3 .0 9 9

5 .5
5 .3
4 .8

6 .9
6 .6
6 .0

2 .5
2 .5
2 .2

( s)
( s)
(*)

(»)
(*)
( J)

4 .4
4 .1
7 .4

D e c e m b e r 1 9 5 9 ..........................................................................................................
D e c e m b e r I 9 6 0 4 .......................................................................................................

2 ,9 6 6
3 ,0 1 2

4 .5
4 .6

5 .8
5 .9

2 .0
2 .0

4 .6
4 .6

4 .2
4 .1

5 .1
6 .4

M ay
M ay
M ay
M ay
M ay
M ay
M ay

3 ,3 4 2
3 ,9 2 1
3 ,7 2 6
3 ,7 5 6
3 ,6 3 6
4 ,0 0 8
4 ,0 4 8

4 .9
5 .7
5 .2
5 .2
4 .9
5 .2
5 .2

6 .4
7 .4
6 .9
6 .7
6 .4
6 .9
7 .0

2 .0
2 .4
2 .1
2 .3
2 .2
2 .3
2 .2

4 .9
5 .7
5 .3
5 .3
5 .0
5 .3
5 .3

4 .6
5 .2
4 .7
4 .0
4 .3
4 .5
4 .4

5 .1
5 .5
4 .8
4 .3
3 .7
2 .9
4 .1

N o v e m b e r 1 97 0 ..........................................................................................................

3 ,8 3 2

4 .9

6 .3

2 .5

5 .0

3 .4

5 .5

M a y 1 9 7 1 .....................................................................................................................

4 ,0 3 5

5 .1

6 .7

2 .6

5 .3

3 .8

5 .3

1 9 6 2 . . .................................... .......................................................................... 1 9 6 3 .....................................................................................................................
1 9 6 4 ................ ................... .................................................................................
1 9 6 5 . . . . ............................................................ ......................................... —
1 9 6 6 ...................................... ............... ...............................................................
1 9 6 9 ................................................................. ....................................................
1 9 7 0 ..................................................................... ...............................................

1 M u lt ip le jo b h o ld e r s a s p e r c e n t o f a l l e m p lo y e d p e rs o n s.

* D ata n o t a v a ila b le .

1 N o t s e a s o n a lly a d ju s t e d .

4 D a ta f o r A la s k a a n d H a w a ii in c lu d e d b e g in n in g 1960.




163

Table 2. Type of industry and class of worker of primary and secondary jobs for persons holding two jobs or more, May
and November 1970 and May 1971
[ N u m b e r s in th o u s a n d s ]

Date, type of Industry, and
class of worker of primary job

Total
employed

Type of industry and class of worker of secondary job

Persons holding two
jobs or more

Agriculture

Nonagricultural industries

Percent
of total
employed

Total

Wage and
salary
workers

Selfemployed
workers

Total

Wage and
salary
workers

Selfemployed
workers

4 ,0 4 8

5 .2

738

122

616

3 ,3 1 0

2 ,7 4 8

562

71
44
27

47
20
27

24
24
0)
0

205
45
127
33

1%
36
127
33

9
9

Number

M A Y 1970

Total.....................................

7 8 ,3 5 8

A g r ic u lt u r e ............................................. ..
W a g e a n d s a la r y w o r k e r s . . . ..........
S e lf - e m p lo y e d w o r k e r s ............. ..

3 ,7 2 5
1 ,2 0 0
1 ,9 2 7
598

276
89
154
33

7 .4
7 .4
8 .0
5 .5

N o n a g r ic u lt u r a l in d u s t r ie s ............ ...........
W a g e a n d s a la r y w o r k e r s . . . ..........
S e lf - e m p lo y e d w o r k e r s ....................

7 4 ,6 3 3
6 8 ,9 0 5
5 ,2 2 6
502

3 ,7 7 2
3 ,5 7 0
194
8

5 .1
5 .2
3 .7
1 6

667
661
6

75
69
6

592
592

3 ,1 0 5
2 ,9 0 9
188
8

2 ,5 5 2
2 ,3 5 6
188
8

5 53
553

0
( 2)

Total.....................................

7 8 ,7 4 0

3 ,8 3 2

4 .9

614

88

526

3 ,2 1 8

2 ,6 4 8

5 70

A g r ic u lt u r e ....................................................
W a g e a n d s a la r y w o r k e r s ................
S e lf - e m p lo y e d w o r k e r s ....................

3 ,2 2 8
1 ,0 4 2
1 ,7 6 2
424

198
51
129
18

6 .1
4 .9
7 .3
4 .2

44
26
18

26
8
18

18
18
(l)
(»)

154
25
111
18

149
20
111
18

N o n a g r ic u lt u r a l in d u s t r ie s .......................
W a g e a n d s a la r y w o r k e r s ................
S e lf - e m p lo y e d w o r k e r s ........... ........

7 5 ,5 1 2
6 9 ,6 1 3
5 ,3 5 5
544

3 ,6 3 4
3 ,4 4 3
180
11

4 .8
4 .9
3 .4
2 .0

570
563
7

62
55
7

508
508

3 ,0 6 4
2 ,8 8 0
173
11

2 ,4 9 9
2 ,3 1 5
173
11

565
565

0
0

Total.....................................

7 8 ,7 0 8

4 ,0 3 5

5 .1

700

96

6 04

3 ,3 3 5

2 ,6 0 7

728

A g r ic u lt u r e ....................................................
W a g e a n d s a la r y w o r k e r s ................
S e lf - e m p lo y e d w o r k e r s ....................
U n p a id f a m ily w o r k e r s .....................

3 ,5 9 8
1 ,2 4 5
1 ,8 1 2
541

217
65
129
23

6 .0
5 .2
7 .1
4 .2

66
38
20
8

41
13
20
8

25
25

151
27
109
15

147
23
109
15

4
4

N o n a g r ic u lt u r a l in d u s t r ie s .......................
W a g e a n d s a la r y w o r k e r s ................
S e lf - e m p lo y e d w o r k e r s ....................
U n p a id f a m ily w o r k e r s .....................

7 5 ,1 1 0
6 9 ,1 5 0
5 ,4 2 9
531

3 ,8 1 8
3 ,6 4 1
167
10

5 .1
5 .3
3 .1
1 .9

634
629
4
1

55
50
4
1

3 ,1 8 4
3 ,0 1 2
163
9

2 ,4 6 0
2 ,2 8 8
163
9

0
0

0
0

N O V E M B E R 1970

5
5
0
0

0
0

M A Y 1971

1 S e lf - e m p lo y e d p e r s o n s w ith a s e c o n d a r y b u s in e s s o r fa r m , b u t no w a g e o r s a la r y

0
0
579
5 79
0
0

0
0
724
724
0
0

* P e r s o n s w h o s e p r im a r y jo b w a s a s a n u n p a id f a m ily w o r k e r w e re c o u n te d a s m u l­

jo b , w e r e n o t c o u n te d a s m u lt ip le jo b h o ld e r s .

t ip le jo b h o ld e r s o n ly if th e y a ls o h e ld a w a g e o r s a la r y jo b .

wage or salary job turned to self-employment in
order to earn additional income.
Half the women who were multiple jobholders
were employed in service and finance on their sec­
ondary jobs, compared with a quarter of the men,
and a much larger proportion of women than men
worked in trade. In contrast to the men, about a
fifth of whom were employed in agriculture on their
secondary jobs, nearly all of the women had second
jobs in nonagricultural industries.
Greater proportions of Negro 3 than white moon­
lighters have second jobs in service industries, and
smaller proportions are self-employed. In May 1971,
39 percent of the Negroes, but about 29 percent of
the white multiple jobholders worked in the service
industries. Only 24 percent of the Negroes were

self-employed on the second job, compared with 34
percent of the whites.
Multiple jobholders do not tend to work in the
same major occupational groups on their secondary
jobs as on their primary jobs, with one exception.
A majority of the professional and technical workers
who held at least two jobs were employed in the
same occupational group on both jobs. About a third
of the service workers (except private household)
worked in the same occupational groups on their
second jobs. In none of the other occupation groups
did more than one-fourth of the workers do the
same kind of work on both jobs.
Blue-collar workers on their first jobs were more
often farmers or farm managers on their second jobs
than persons in other occupations. Nearly a quarter




164

of the blue-collar workers— compared with fewer
than 10 percent of the professional, clerical, and
service workers— operated farms on their secondary
jobs. Similarly, 35 percent of those who were farmers
and farm managers on their first jobs worked as
craftsmen or operatives on their second jobs. Many
of the construction and maintenance skills that are
needed on the farm can be used off the farm as well.
The difference by race in the occupational distri­
bution of multiple jobholders on their second jobs
corresponds roughly to the differences in occupations
by race of all workers. Thus, white dual jobholders
are considerably more likely to work in white-collar
occupations, especially as managers, than their Negro
counterparts. About 46 percent of the whites worked
at white-collar jobs on their second jobs, compared
to 31 percent of the Negroes.
At the same time, proportionately over twice as
many Negroes as whites worked in service occupa­
tions on their second jobs. The same proportions of
whites and Negroes were blue-collar workers on their
second jobs.

Earnings on secondary jobs
Before the May 1970 survey, there was no infor­
mation on a nationwide basis on how much moon­
lighters earned on their second jobs. Data obtained
in May of 1970 and 1971 indicate that multiple
jobholders who were wage or salary workers on their
second jobs had median earnings of $30 on those
jobs during the survey week. (See table 3.)
On average, men earn more than women in their
primary jobs; this tendency also prevailed among
multiple jobholders. Thirty-four percent of the men,
but only 10 percent of the women, added at last $50
to their weekly income through moonlighting in May
1971. Earnings of under $20 were reported by onefourth of the men and one-half of the women. Median
earnings on the second job were $35 for men and
$19 for women. These sharp differences reflect the
variation in the distributions of occupations and in­
dustries in which men and women find secondary
jobs, and the difference in the number of hours they
work at these jobs.
Earnings depend not only on the rate of pay but
also on the number of hours worked. During the
May 1971 survey week, men worked a median of
13 hours on their second wage or salary jobs, in
contrast to 9 hours for women. Only 7 percent of
the women worked as many as 22 hours, while one-




fifth of the men worked at least this long at secondary
wage or salary jobs.
For both men and women, secondary job earnings
increased as hours worked increased. The following
tabulation shows that median weekly earnings for
persons who worked 22 to 34 hours on their second
jobs were about four times as high as earnings of
those who worked 1 to 7 hours.
Both sexes
1 to 7 h o u r s .................
8 to 14 h o u r s ...............
15 to 21 h o u r s ............
22 to 34 h o u r s ............

...............
...............
...............
...............

$15
26
43
61

Men

Women

$17
28
46
62

$12
20
34
0)

1Median not shown where base is less than 75,000.

By age and marital status. Wage or salary earnings
on the second jobs of men in the central age groups
tended to be higher than those of younger men.
About 24 percent of the men 25 to 44 years old
earned $70 or more on their second jobs during the
survey week, and a similar proportion of men. 45 to
64 years old earned as much (table 3 ). On the other
hand, much smaller proportions of the men 20 to 24
years old and of male teenagers earned as much
as $70. About three-fourths of the teenagers earned
under $20, reflecting the relatively small number of
hours that they worked on their second jobs.
Married men earned twice as much as single
men— $37 and $18, respectively. This is not sur­
prising since men in the central age groups had the
highest earnings and most of these men were married.
By industry and occupation. Men with secondary
jobs in educational services and in manufacturing
earned substantially more than the average moon­
lighter. One-third of the men working in these in­
dustries earned $70 or more on their second jobs
during the survey week, compared with only one-fifth
of those in service (other than educational) and
one-tenth of those in trade. The longer-than-average
hours worked by men in manufacturing help to ac­
count for their higher than average weekly earnings.
On the other hand, the high earnings of men in
educational services, many of them teachers, reflect
high hourly rates of pay since they worked fewer
hours than average. Men with secondary jobs in
finance and in agriculture had the lowest earnings.
Almost half of those in finance and two-thirds of
those in agriculture earned less than $20 during the
survey week.
Relatively high earnings were most frequent among
men who were professional and managerial workers
on the second job, A third of each earned at least

Table 3. Wage or salary earnings on second job for persons holding two jobs or more, by age, sex, and marital status,
May 1971
[Percent distribution]
W e e k ly e a r n in g s o n s e c o n d j o b
T o ta l

A g e , s e x , a n d m a r it a l s t a t u s

M e d ia n
U n d e r $20

$ 20 t o $29

$ 30 t o $39

3 2 .0

1 7 .7

1 3 .0

$ 40 t o $49

$ 50 t o $69

$ 70 o r m o r e

8 .7

1 2 .7

1 5 .9

$ 30

BOTH SEXES
T o t a l.....................................................................................................

1 0 0 .0

M EN
T o ta l, 16 y e a r s o ld a n d o v e r ........................................................

1 0 0 .0

2 5 .6

1 7 .2

1 3 .6

9 .1

1 5 .0

1 9 .5

$35

y e a r s _____ ___________________ _______ ____________________
y e a r s ....................................... ........................................... ............
y e a r s ________________________________________________ ____
y e a r s ............ ...............................................................................
y e a r s .............................. .................................................. ...............
a n d o v e r ...................
.
...........................................

1 0 0 .0
1 0 0 .0
1 0 0 .0
1 0 0 .0
1 0 0 .0
(»)

7 4 .0
3 2 .1
2 1 .8
1 4 .8
1 9 .4

1 5 .9
2 0 .6
1 7 .0
1 3 .5
1 9 .0

3 .1
1 3 .4
1 3 .T
1 8 .2
1 4 .9

1 .3
1 0 .1
9 .3
1 3 .7
4 .4

4 .0
1 1 .9
1 4 .7
1 8 .9
2 3 .8

1 .7
1 1 .9
2 4 .1
2 0 .9
1 8 .5

$14
$28
$38
$42
$37

S in g le .................................................. .............................................................
M a r r ie d , s p o u s e p r e s e n t ....................................... ................... .................
O th e r m a r it a l s t a t u s .....................................................................................

1 0 0 .0
1 0 0 .0

5 3 .9
2 1 .5

1 6 .5
1 7 .1

6 .8
1 4 .9

5 .0
9 .7

6 .8
1 6 .0

1 1 .0
2 0 .8

$ 18
$37

16
20
25
45
55
65

t o 19
to 2 4
to 44
to 54
to 64
y e a rs

(*)

(l )

C1)

W O M EN
T o t a l, 16 y e a r s o ld a n d o v e r ........................................................

1 0 0 .0

5 2 .2

1 9 .5

1 1 .1

7 .4

5 .3

4 .5

$19

y e a r s . . . .......... ............................................................. .................
y e a r s ................................................................................ ...............
y e a r s ................................................................................................
a n d o v e r .........................................................................................

1 0 0 .0
1 0 0 .0
1 0 0 .0
(l)

6 4 .5
4 2 .1
5 1 .7

2 2 .7
1 5 .7
2 2 .1

5 .7
1 0 .3
1 6 .8

2 .8
1 4 .4
3 .4

3 .5
9 .2
2 .7

.7
8 .3
3 .4

$15
$25
$19
( i)

S in g le ................................................................................................................
M a r r ie d , s p o u s e p r e s e n t . . . .......... ............................ ..............................
O th e r m a r it a l s ta tu s ........................................... .........................................

1 0 0 .0
1 0 0 .0
1 0 0 .0

6 2 .2
4 8 .0
4 6 .9

1 8 .9
2 0 .2
1 9 .0

6 .7
1 1 .9
1 5 .4

2 .9
9 .3
9 .8

5 .6
5 .6
4 .0

3 .7
4 .9
4 .9

$16
$20
$21

16
25
45
65

to 24
to 44
to 64
y e a rs

NOTE: Because of rounding, sums of individual items may not equal totals.

1 Percent and median not shown where base is less than 75.CC0.

$70 on their extra jobs during the survey week. The
high earnings of the professionals reflect their high
wage rates since they worked fewer hours than
average. In contrast, only one-fifth of the sales work­
ers earned as much as $50; the median number of
hours worked by sales workers was about equal to
the average on the second job for all workers.

on their second job, in contrast to only 6 percent of
those who primary jobs were less remunerative.
The tendency for the high earners on first job
also to earn the most on the second job is to be
expected since they may be assumed to have highly
valued skills. They are also less likely than others
to be moonlighting to meet regular household ex­
penses and therefore may be more selective in their

By earnings on primary job. The multiple jobholders
with the highest weekly wage or salary earnings on
the primary job tended to have the highest earnings
on their second jobs. Median weekly earnings on
the second job in May 1971 increased from $14 for
men who earned under $60 a week on their first
job to $52 for those who earned $200 or more, as
shown in the following tabulation:

c h o ic e s o f seco n d a ry job s.

Weekly earnings on
primary job

Median weekly
earnings on
secondary job

Percent of
earnings on
primary job

Less than $60 . . .
$60-$99 ...........
$100-$124 .............
$125-$149 ...........
$150—$199 ...........
$200 or m o r e ____

$14
33
27
36
38
52

50
40
24
26

Although low earners on the first job also earned
comparatively little on the second job, the supple­
mental earnings were a much larger proportion of
their basic earnings than they were for the much
better paid men. Men who earned between $60 and
$99 on their primary jobs averaged an additional 40
percent on their second jobs; for those who earned
$150 or more, the average earnings on the second
job were less than one fourth of their primary job
earnings.
Trends

22
23

In most of the years since 1962, data on multiple
jobholding have been obtained for the month of May.
Over this period, the number of moonlighters gen­

Among the men earning $200 a week or more on
their primary jobs, 22 percent made at least $100




166

erally fluctuated within a narrow range. It reached
the decade’s high of 4 million in May 1969 and
remained at that level in May 1970 and May 1971,
about 700,000 more than in 1962. During this
period, the multiple jobholding rates remained rela­
tively constant, at about 5 percent.
Most of the increase in the total number of multi­
ple jobholders between May 1962 and May 1971 was
among persons who were wage and salary workers
in nonfarm industries on both their first and second
jobs (table 4 ). Nearly 2.3 million multiple jobholders
held two such jobs in May 1971; they represented
57 percent of all multiple jobholders, up from 52
percent in May 1962.
About 900,000 workers combine a wage or salary
job with self-employment in nonfarm industries.
Examples of persons in this group are policemen,
firemen, postal workers, and teachers who on their
second jobs drive their own taxis, do home main­
tenance or repair, or are free lance writers or artists.
That number was relatively stable between 1962 and
1970 but in 1971, as previously indicated, it in­
creased because of the rise in the number selfemployed on one job. About four-fifths of all who
combine a nonfarm wage or salary job with selfemployment are self-employed on the second job.
A third group of moonlighters, about 850,000 or
roughly one-fifth of the total (somewhat lower than
Table 4. Agricultural and nonagricultural employment of
persons holding two jobs or more, 1953-70
[Numbers in thousands]
A t le a s t o n e jo b
in a g r i c u l t u r e

T w o j o b s in n o n ­
a g r ic u lt u r a l in d u s t r ie s

W age
and
s a la r y
jo b
and
s e lf em p lo y m ent

T o ta l

N um ber

P e r­
cent
of dual
jo b ­
h o ld e r s

J u l y 1 9 5 6 _____________
J u l y 1 9 5 7 .................. ..
J u ly 1 9 5 8 .........................

3 ,6 5 3
3 ,5 7 0
3 ,0 9 9

1 ,5 0 3
1 ,4 1 4
1 ,1 2 2

4 1 .1
3 9 .6
3 6 .2

2 ,1 5 0
2 ,1 5 6
1 ,9 7 7

1 ,6 1 1
1 ,5 5 8
1 ,4 2 7

539
598
550

D e c e m b e r 1 9 5 9 .............
D e c e m b e r 1960 1..........

2 ,9 9 6
3 ,0 1 2

829
781

2 8 .0
2 5 .9

2 ,1 3 7
2 ,2 3 1

1 ,5 3 3
1 ,6 4 7

604
584

M ay
M ay
M ay
M ay
M ay
M ay
M ay

1 9 6 2 _____________
1 9 6 3 ...................... ..
1 9 6 4 ........................
1 9 6 5 ........................
1 9 6 6 ........................
1 9 6 9 ______ _______
1 9 7 0 ..................... ..

3 ,3 4 2
3 ,9 2 1
3 ,7 2 6
3 ,7 5 6
3 ,6 3 6
4 ,0 0 8
4 ,0 4 8

868
1 ,0 7 1
1 ,0 6 9
1 ,0 6 5
936
939
943

2 6 .0
2 7 .3
2 8 .7
2 8 .4
2 5 .7
2 3 .4
2 3 .3

2 ,4 7 4
2 ,8 5 0
2 ,6 5 7
2 ,6 9 1
2 ,7 0 0
3 ,0 6 9
3 ,1 0 5

1 ,7 4 9
2 ,0 7 3
1 ,9 2 8
1 ,9 1 4
1 ,9 3 4
2 ,3 2 6
2 ,3 5 6

725
777
729
777
766
743
749

N o v e m b e r 1 9 7 0 ............

3 ,8 3 2

768

2 0 .0

3 ,0 6 4

2 ,3 1 5

749

M a y 1 9 7 1 .........................

4 ,0 3 5

851

2 1 .1

3 ,1 8 4

2 ,2 8 8

8%

D a te

T o ta l

Tw o
w age
and
s a la r y
jo b s

Most of the persons who hold at least one job in
agriculture are self-employed farmers on the second
job. Typically, they are wage and salary workers in
nonagricultural industries on their first jobs who work
the family farm in their free time. Many of these
persons may have taken nonfarm jobs in nearby
towns because they could not earn an adequate
living on their marginal or submarginal farms.

Seasonality and multiple jobholding
In an effort to ascertain whether multiple jobholding varies from one season to another, informa­
tion on multiple jobholding was obtained in Novem­
ber 1970 as well as in May of 1970 and 1971. Both
the number and rate of multiple jobholding in
November were somewhat lower than in either spring
month. (See table 2.) About 3.8 million persons
had two jobs or more in November 1970— 200,000
fewer than in May 1970— and the multiple jobholding rate, at 4.9 percent, was also somewhat
below the May 1970 level of 5.2 percent. By May
1971, both the number of moonlighters and the rate
had returned to their prior levels. The drop from May
to November 1970 appeared to be in line with the
seasonal decline in agriculture: the number of multi­
ple job holders with at least one job in agriculture
fell by 175,000 to 770,000. However, the upswing
in this group between November 1970 and the
following May was not as large as the decline had
been, so that it is difficult to assess how much of
either the drop or increase was seasonal, although it
is reasonable to assume that a substantial part was
seasonal.
As indicated earlier, the multiple jobholding rate
for Negroes generally has been slightly lower than
for whites in the past decade. Between May and
November of 1970 the difference widened because
the rate for Negroes declined more sharply than for
whites. The reduction for Negro multiple jobholders
occurred in both agricultural and nonagricultural
industries, but for whites the decline was only among
those in agriculture.

1 Data for Alaska and Hawaii included beginning 1960.




in May 1970), have one or both jobs in agriculture.
It is remarkable that this number has not changed
materially over the decade, since the total number of
persons employed in agriculture declined by a third
between 1962 and 1971.

167

Although the overall multiple jobholding rates
declined between May and November, the changes
in rates by industry and by occupation were generally
not significant. Also, the median number of hours
worked in May of each year and in November 1970
was the same (13), with little variation by industry.
The last time a multiple jobholding survey was
made during a winter month was in December 1960.
Between that time and November 1970, the number
of multiple jobholders with two nonfarm jobs has
increased by 800,000 or nearly 40 percent, but the
number who had at least one job in agriculture has
remained about the same.
The stability over the decade in the number of
persons with at least one job in agriculture in a
winter month is the balancing off of two divergent
trends. Over this period, the number of farm workers
holding a second job in a winter month decreased,
primarily because of a decline in the total number
employed in agriculture, rather than a material




change in the multiple jobholding rate. On the other
hand, the number of nonfarm workers holding a
second job in agriculture increased.
□
----------FOOTNOTES---------1 D ata in this report are based prim arily on inform ation
from supplementary questions to monthly survey o f the
labor force, conducted for the Bureau o f Labor Statistics
by the Bureau o f the Census through its Current Popu­
lation Survey. The data for the three surveys relate to
the weeks o f M ay 11-1 7 and N ovem ber 1 5 -2 1 , 1970, and
M ay 16 -2 2 , 1971. The m ost recent report in this series was
published in the Monthly Labor Review, A ugust 1970, pp.
5 7 -6 4 , and reprinted w ith additional tabular data and
explanatory notes as Special Labor F orce Report N o . 123.
! See Vera C. Perrella, “M ultiple jobholders in M ay 1969,”

Monthly Labor Review, A ugust 1970, pp. 5 8 -5 9 .
*D ata for all persons other than w hite persons are used
in this report to represent data for N egroes, since the latter
constitute about 92 percent o f all persons other than white
persons in the U nited States.

168

Chapter IV. Price Measurement and Price Trends




How a general price index
could be constructed,
what it should accomplish,
and virtues and limitations
of various approaches
ALLAN D. SEARLE

T h er e is no adequate comprehensive measure of

price change in the U.S. economy. Such a system
is needed. It would measure price change at inter­
mediate steps of production and at the level of
final distribution would lend itself to a variety of
analytical and statistical uses. It would provide a
check on what is happening to prices in any
important segment of the economy and, at the
same time, gage price m ovem ent for the economy
as a whole.
This article’s principal purpose is to discuss a
number of possible general price indexes and an
underlying system of price indexes for primary and
intermediate production and distribution which
would be consistent with the general indexes. An
attem pt will be made to show how interindustry
(input-output) measures and industrial sector
measures could be fitted into a total structure,
which itself would be consistent with N ational
Accounting (Gross National Product) concepts.1
A vailable price m easures

In the absence of a comprehensive measure,
analysts turn to a number of important but
limited sources of information about price trends.
These are the Consumer Price Index ( c p i ) of the
Bureau of Labor Statistics, which measures price
change in goods and services purchased by urban
wage earner and clerical workers’ families; the
Wholesale Price Index ( w p i ) of the b l s , which
measures price changes for commodities sold on
primary m arkets; the Indexes of Prices Paid and
Received by Farmers of the U.S. Departm ent of
Agriculture; and the Implicit Price Deflator
( i p d ) of the Office of Business Economics of the
U.S. Departm ent of Commerce, which measures
price change in the Gross National Product.
A llan D . Searle is an eco n o m ist in th e Office of P rices
an d L ivin g C on d ition s, B ureau of Labor S ta tistics.

From the Review of March 1971




T oward
comprehensive
measurement
of prices
All of these sources have drawbacks as general
measures of price change. The c p i is not intended
as a measure of price change for the goods and
services bought by consumers other than wage
earners and clerical workers and omits purchases
by Government and business. While the w p i
covers many industrial commodities, it does
not cover construction, transportation, communi­
cations, Government purchases and financing, or
industrial services. (However, individual com­
m odity series in both indexes would be used in
constructing comprehensive indexes.) The agri­
cultural indexes are confined to price changes that
affect farmers. The i p d depends for its coverage
of prices very largely upon the detailed price
information collected in the w p i and c p i programs
and is similarly lim ited.2 Moreover, the i p d ’ s
weighting scheme causes it to reflect changes in
composition of goods and services as well as
changes in prices. In 1969, the Office of Business
Economics began publishing a fixed-weight index
quarterly starting with 1965: This index has the
same lim itations as to commodity coverage as
the i p d .
An adequate, tim ely measure of price change,
broad enough to assess accurately inflationary
and deflationary forces and detailed enough to
provide insight into the interplay of economic
forces in the economy, should do at least two
th ings: It should provide measures of price change
for all goods and services purchased by consumers,
business, Government, and separately for foreign
countries, and it should provide sufficient detail
by industry to promote understanding of price
behavior at intermediate levels of demand (both
for transactions that enter final demand and those
that do not).
The measures of price change described in this
article, as well as the underlying industry sector
structure, are consistent with an industrial classi­
fication scheme rather than a com m odity group-

ing, such as that used in the Bureau’s Wholesale
Price Index. The general indexes are designed to
deal with current price developments in terms of
analysis of increasing and decreasing price trends.
Thus, the pricing of existing capital assets is not
discussed, although this would be particularly
important if a census of wealth were undertaken.
For the present, however, it is considered more
urgent to m ove pricing concepts and development
toward a comprehensive system generally con­
sistent with the framework of the current accounts
portion of the N ational Accounts (consumer
purchases, business investm ent, Government pur­
chases, and net foreign sales), modified only to
the extent necessary for current price analysis.
The first part of the article will deal with
approaches to a general price index and the
remainder, with approaches to price indexes for
industrial sectors, concluding with a detailed
consideration of problems involved and the
relationship among the indexes.

for each industry. These two sets of indexes would
be applied to data on the dollar value of output
and input to derive constant-dollar output and
input, and constant-dollar value added (the
difference between output and input), as described
in the section on double deflation (p.15)
These three sets of indexes are especially val­
uable in industrial economic analysis, permitting
consistency in comparisons with series of wage
rates, productivity, unit labor cost, employment,
average weekly hours, and hourly earnings in a
particular industry. However, these indexes do
not provide for a complete assessment of the
impact of changes in prices of materials purchased
and in the prices of the output of the industry.
Final dem and

The other approach to constructing a general
price index— directly pricing goods and services
in final markets— can be accomplished by follow­
ing the interindustry framework (with its ad­
vantages of an industry classification scheme) in
such a way that the spreads between prices at
producers’ and purchasers’ levels are evident
industry by industry at each point in final demand.
The general outline of the approach can be
developed within the lim its of the concepts cur­
rently used in constructing the Gross National
Product, or the national income concepts can
(for this purpose) be modified in order to present
a price index more responsive to measuring price
changes as they occur in ac.tual markets.

A general price index

One way to visualize a general price index is as
the end product of all interindustry purchases
and sales, unduplicated and totaled at the level of
final demand. Such an index would measure the
average price change of goods and services enter­
ing final markets in proportion to their values.
A general index can be constructed in two
principal ways: (1) indirectly through the concept
of “ Gross Product Originating in Industry” ;
that is, the aggregation of indexes developed
industry by industry for the value added per unit
at each stage of production, or (2) directly by
pricing the goods and services in final demand.
The latter scheme can, in turn, rely on one of two
routes: producers’ prices or purchasers’ prices.
Each approach— value added or final demand—
has its own virtues and limitations.

a n d m o d i f i e d g n p . Adherence to g n p
concepts is desirable because the g n p is a pro­
duction-oriented measure and a system of price
measurement consistent with production values,
input values, em ployment, man-hours, and so
on, is essential to the understanding of output,
inputs (in real and monetary terms), and pro­
ductivity. However, it m ay be advantageous to
make certain modifications in those concepts in
order to restrict measurement to those price
changes which are continually occurring in the
market p lace:
1.
Because the general price index would focus
on the prices at which goods are sold in domestic
markets, the measure would differ from the g n p
deflator by eliminating prices related to inven­
tory change which o b e must take into account
so that g n p measures the total quantity and price

Standard

Gross output method

Construction of a general price index by the
gross-product-originating route emphasizes indus­
trial classification of the N ation’s sales or produc­
tion. Price indexes would be developed for the
output of each industry (similar to those b l s pub­
lishes for about 100 for manufacturing and mining
industries). In addition, price indexes of purchases
made by producers (inputs) would be developed




171

of production. Also, export sales would be
beyond the modified domestic-market scope of
the general price index conceived here, and the
foreign trade part of final demand would require
different handling.
2 . The g n p deflator includes prices imputed to
nonmarket transactions, such as use by homeowners of their own dwellings and the consump­
tion of homegrown food. A General Price Index
would measure only prices of goods and services
actually sold.
3 . The production concept underlying the g n p
requires that used cars and resale of homes be
measured to include only that production which
is involved in their resale. Assigning proper weights
and pricing such item s in a General Price Index
would reflect the total transaction viewed from
the purchaser’s point of view.
If a general price measure is viewed as the end
product of the multitude of interindustry trans­
actions (many of which are not passed on except
as components or constituent parts of final sales),

Table 1.

then the interindustry structure can be made to
serve as a reference framework for the develop­
ment of a generalized price measure.3
Table 1 shows how the structure of the Bureau’s
currently published Stage of Processing Indexes
relate to the interindustry structure. The signifi­
cance of these indexes in price analysis is set
forth in a later section.
Input-output data show the dollar value of
transactions among the various industries (in­
cluding sales among establishments within the
same industry) for the reference year. Each row
in the table would show how the output of goods
and services of each industry would be distributed
among other industries, and to final users. For
example, part of the output of agriculture is
sold to agriculture (for example, livestock feeds)
with some output going to manufacturers for
further processing and the renainder going to
consumers as unprocessed foods. Were numerical
values shown, the columns would show the value
of each industry’s intake (input) of raw materials,

Input-output flows, classified by the codes used in the Stage of Processing Indexes, showing coverage gaps
Industry

Con­
struction

Manu­
facturing

Trans­
portation,
other
public
utilities

1220

1100
1210
1310

1320

Industry
Agri­
culture

Agriculture...................................
Mining..........................................
Construction.................................
Manufacturing..............................

Transportation, other public
utilities.
Trade......................... .................
Finances.......................................
Services........................................
Government..................................
Other.............................................

Mining

2622

2621
2622

2500

2200
2420

1210
2110
2120
2130
2140
2410
2500
2610

Final demand

Trade

Personal
con­
sumption
Finance Services Govern­
ment Other expendi­
tures

Gross
private
capital
forma­
tion

Govern­
ment
purchases
Net
exports (Federal,
State, and
local)

3111

2420
2622

2500
2622

3112
3120
3130

3210
3220

1210

NOTE: The codes in the BLS Stage of Processing Price Indexes are as follows:
1000 Crude materials for further processing
1100 Crude foodstuffs and feedstuffs
1200 Crude nonfood materials except fuel
1210 Crude nonfood materials except fuel, for manufacturing
1220 Crude nonfood materials except fuel, for construction
1300 Crude fuel
1310 Crude fuel for manufacturing industries
1320 Crude fuel for nonmanufacturing industries
2000 Intermediate materials, supplies and components
2100 Intermediate materials and components for manufacturing
2110 Intermediate materials for food manufacturing
2120 intermediate materials for nondurable manufacturing
2130 intermediate materials for durable manufacturing
2140 Components for manufacturing
2200 Materials and components for construction
2400 Processed fuels and lubricants
2410 Processed fuels and lubricants for manufacturing
2420 Processed fuels and lubricants for nonmanufacturing




2500 Containers
2600 Supplies
2610 Supplies for manufacturing industries
2620 Supplies for nonmanufacturing industries
2621 Manufactured animal feeds
2622 Other supplies
3000 Finished goods
3100 Consumer finished goods
3110 Consumer foods
3111 Consumer crude foods
3112 Consumer processed foods
3120 Consumer other nondurable goods
3130 Consumer durable goods
3200 Producer finished goods
3210 Producer finished goods for manufacturing industries
3220 Producer finished goods for nonmanufacturing industries

172

semifinished products and services used in pro­
ducing output for final sale.
Pricing for this part of the input-output table
calls for development of price indexes for each
cell of the table (or at least to represent each cell)
in order to promote understanding of how price
changes of supplying industries are related to
price indexes of the recipients. The prices which
enter this part of the table would be producers’
(sellers’) prices.
The final demand part of table 1 is aggregated
by consumers’ purchases (personal consumption
expenditures), business capital purchases (gross
private capital formation and net inventory
change), foreign purchases (net exports), govern­
ment purchases (Federal, State, and local), and
total. In this approach to a general index, the
indexes would be presented at two levels of pricing:
producers’ and purchasers’.
The spread between the two price indexes
would represent the “margin” added to values of
crude and intermediate goods exchanged and
channeled from the producers’ segment of industry
by wholesale and retail trade. The General Price
Index would be the index representing the allindustry average in both the producers’ and
purchasers’ columns of total final demand.
While indexes of producers’ and purchasers’
prices for goods and services entering final demand
would generally differ at the industry level, the
indexes for producers’ and purchasers’ prices should
be equal at the total level and are in essence in­
dexes of price change for the Gross National
Product regardless of how g n p is structured. (How­
ever, indexes based upon the traditional g n p
structure would differ from those based on the
modified g n p structure described in this article.)
The reason for this equality in the totals (but
inequality elsewhere) lies in the fact that the same
totals for the transportation and trade margins
are treated as separate purchases in the producers’
price column and as part of the value of each good
in the purchasers’ price column. That is, the pur­
chasers’ price column would include only that part
of transportation (either passenger or freight) in
personal consumption expenditures which is pur­
chased by the consumer. (A third portion of trans­
portation values are intermediate sales and do not
appear in either the producers’ or purchasers’ price
columns.)
In reality, the bulk of transactions of industries
in the producing sector are made with wholesale




173

trade and to a lesser extent with retailers. The
developers of the interindustry tables, however,
felt that much more economic insight could be
gained if the system of input-output accounts
could be set up to show the flow of product from
producer directly to final purchasers by industry
of origin. Consequently, each sector of final de­
mand is viewed as purchasing most goods and
services directly from producers, purchasing from
trade only “the margin”— operating expense and
profit. In the concept outlined here, then, the
total final demand at producers’ prices would
include separately a substantial value for trade.
The total at purchasers’ prices, however, would
include no value for trade because the remainder
would have been distributed among the various
other segments.
Relation to broad price indicators

A General Price Index represents only one
approach to measuring general price trends.
Another avenue is the Implicit Price Deflator of
the Office of Business Economics. This series is a
measure of price change for final demand, but
differs from the measure described in at least two
(and possibly three) principal respects: (1) The
o b e measure is built upon a system of classification
which varies from category to category of final
demand and does not conform to an industry or
commodity group structure, so that it is not
possible to trace effects of price change by in­
dustry. (2) The o b e measure is of the Paasche
type; that is, the weighting of prices changes as the
composition of output changes or, more precisely,
the weights for each year or quarter are used to
average the relative price change between that
year or quarter and the base year. The index
proposed in this article would be of the fixedweight, Laspeyres type,4 organized along indus­
trial lines. A General Price Index would deal only
with market sales, excluding the imputations in
the Implicit Price Deflator. (3) It would also be
useful to construct an index corresponding identi­
cally with g n p concepts.
The price deflator implied by another o b e
program ( g n p by Major Industries) has an
industrial structure but again the index is of the
Paasche type. The industrial breakdown at present
goes to no further detail than the 2-digit sic level
and is presented as a value-added price measure
as described in the section on “Standard and
modified g n p . ”

In comparison, a General Price Index would
have the following attributes:
1. The industrial structure of the index permits
interindustry analysis of price change and provides
insight into the effect on prices of the flow of goods
and services among industries at stages of produc­
tion prior to final sale.
2. The index’s industrial structure will alloAv
analyses of price trends at the individual industry
level for both outputs and (ultimately) inputs of
materials, adding to the store of understanding of
the functioning of the economy.
3. Basing the index on an industrial structure
of price measures will enhance the construction
and understanding of the Inter-Industry Program
by providing the means for interim extension of
the input-output data and contributing to im­
provement of the constant-value data themselves.
4. Because the index would be prepared with
fixed weights, its form will be comparable to that
of other leading price indicators, such as the
Consumer Price Index, the Wholesale Price Index,
and the U.S. Department of Agriculture’s indexes
of Prices Paid and Received by Farmers.
5. In one version, a modification of the g n p
concept would permit a better tie-in with other
price indexes and with market sales.
6. In another version, a more conventional
form, the tie-in with g n p (in common with other
implicit deflators) could unify the concept of the
entire price structure into a cohesive whole.
Another measure sometimes used in place of a
general price measure is the Wholesale Price Index.
The w p i , however, is limited in scope compared
with the proposed general index and is classified
along different lines. The w p i is aimed at measuring
price change of commodities at the first level of
transactions (primary m arkets); the general index
would include prices of both commodities and
services at the level of final sale (final demand)
with detail by industry representing all inter­
industry transactions. The classification structure
of the w p i is principally based on groupings of
com petitive commodities; the industrial portion
of the structure underlying the general index
would be based on an industry structure. The
w p i excludes from coverage retail and wholesale
trade while the general index includes these sectors.
Other areas included in the industrial substructure
of the general index, which are excluded from the
w p i , are interplant transfers and sales of military
items to the Government.




The structure and coverage of the Consumer
Price Index is even further removed from an
industrial concept. While much of the data used
in its construction are obtained from retail outlets,
the index is not a retail trade index. It is a
purchase price index for a broadly defined but
specific group of consumers— urban wage earners
and clerical workers. Hence, it excludes some
sales at retail (for example, qualities and types of
goods and services purchased by higher income or
lower income consumers) and is classified in line
with consumption categories (for example, housing,
medical care) rather than by industry. Because,
pricing for the c p i has been oriented toward the
urban wage earner and clerical purchaser, pricing
for retailing as a whole needs to be augmented to
include higher- and lower-priced stores, farm and
nonfarm, and types of business not covered or
only lightly covered.5 Also, the c p i provides only
spotty coverage for certain types of stores
(examples: lumber yards, retail bakeries, hard­
ware stores, and farm equipment dealers). Further­
more, even where a given product is represented,
it does not represent price differences for the
same product sold in particular types of retail
industry—shoes for example, are purchased in
shoe stores and in department stores.
Industrial sector price indexes

Where a prime purpose of a General Price
Index is to measure price change of goods entering
final demand, the purpose of a set of industrial
sector price indexes is to measure, separately,
price change of products and services moving
between industries at all stages of production,
processing, and distribution including final de­
mand. Thus, industrial sector price indexes can
provide increased insight into the interrelation­
ships of costs and prices at all stages of the
productive process and add to the understanding
of the resultant general price index.
In this context, the General Price Index may
be viewed as a summary indicator of price m ove­
ments within the entire industry structure
(including imports). Properly formulated, price
changes in intermediate industries could be
traced through the structure to final demand and
price change effected by demand changes could be
traced back through the structure in instances
when “demand pull” is the dominant force upon
prices.

174

An Industrial Sector Price Index ( ispi) is, es­
sentially, a composite index made up from several
series of prices that closely match the economic
activity of a defined industry or industry sector.
Industrial sector price indexes may be subdivided
into indexes of output or input prices based upon
either the products and services sold or the prod­
ucts and services purchased by an industry sector.
For example, an output price index for a given
industry represents a set of individual price indexes
for all the important products of the industry,
averaged together according to the relative im­
portance of each product to the industry. An input
price index for an industry consists of an aggre­
gation of price indexes for the commodities and
services purchased by the industry, weighted to­
gether according to the relative magnitude of the
purchases.

H istory. Early work in the field of industrial
price measurement was devoted largely to sup­
plying data for specific statistical purposes (de­
flation) but a consistent, continuous program was
not developed. In the early 1950’s, a set of annual
industry sector price indexes covering the years
1947-53 was prepared as part of a Bureau of Labor
Statistics project on interindustry economics.
These indexes were designed to revalue industry
outputs. Later, in 1959, a similar set of indexes
was compiled in connection with the need of the
Bureau of the Census to construct the 1958 pro­
duction index benchmark. Then, in 1961, the Price
Statistics Review Committee of the National
Bureau of Economic Research recommended to
the Bureau of the Budget that bls begin a
permanent program to develop price indexes or­
ganized along an industrial structure, and in 1962
the Bureau initiated its present modest program.
A t present, the program consists of m onthly
output price indexes for about 100 manufactur­
ing, mining, and agricultural industries (at the
4-digit sic level of detail). Price indexes for com­
modities primary to an industry are weighted
with values covering only the production within
the industry (that is, excluding the portion which
constitutes secondary production in other indus­
tries). Secondary production values of the industry
are used to weight the price movements of items
of types primarily made elsewhere but made as
secondary output of the industry. Thus, a given
price index for a com modity m ay be used in




175

several places in the ispi system : where output is
primary and in the industry or industries where it
is secondary. The inclusion of price series for
secondary output represents a departure from
the m ethodology of the earlier indexes, which
represented primary output only. Ultim ately,
input indexes (pricing industry purchases) will
also be constructed.6

T he ispi in deflation. Price measures m ay be
used as deflators to estim ate change in physical
quantities. Among the m ost noteworthy govern­
ment statistical programs using price indexes for
deflation are those connected with the National
Accounts: Gross National Product in constant
dollars, the final demand accounts and gross
product originating in industry. Other programs
include the bls productivity measures where
value data are deflated to obtain constant dollar
output per man-hour, and the Federal Reserve
Board’s production index which uses price data
to supplement physical output data, as does the
Census Bureau every 5 years.
M eaning of quantity measure. One of the pri­
mary purposes of the ispi is for use in deflation of
value figures in order to estimate changes in physi­
cal volume— value or value-added in “constant
dollars.” The deflator which is developed for this
purpose m ust be constructed (a t least in concept)
in such a way that the price index is comparable
in all respects with the value data being deflated—
or at least with the value concept which the com­
pilers of the value data had in mind. For example,
pricing must be timed to coincide with the value
data. Timing may be crucial where there are large
inventory holdings or sudden large changes in in­
ventory so that shipments and production may
have to be valued at different price levels and in
situations where the production cycle is very long
(for example, shipbuilding or construction).
The price index also must be constructed ac­
cording to a concept which will lead to a quantity
( “constant dollar”) figure which carries the mean­
ing which the user requires. For example, a price
index in which a quality adjustment for a new
machine is based on the measurable improved
performance of the machine results in a deflated
index in which quantity is measured in units of
performance. This is a different measure from one
in which the quality adjustment is made on the
basis of additional resources required to provide

the additional performance, except for the one
(unlikely) situation in which the resources used
are proportional to the added performance.
Another measure would result if the quality ad­
justm ent is based on the additional amount buyers
are willing to pay for the added performance. This
m atter is treated further, later in the paper. The
important point here is that it is essential not only
that the price measure represent the same scope
or coverage as the dollar values to be deflated, but
also that the meaning of the quantity measure be
constantly borne in mind, for any change in the
definition of price automatically changes the defi­
nition of quantity.

example, the value of steel in automobiles would
be counted in the automobile data and again in
the steel data. In contrast, the value-added data
derived in the value-added approach are additive,
industry-to-industry without duplication.
d e f l a t i o n . Because direct measurement
of value added per unit (value-added pricing) is
not feasible, the deflation of value-added data to
obtain the real net value of output must be done
in stages. The output price indexes and the input
price indexes are used to deflate total value of
production and of purchases, respectively. The
difference between the undeflated output and
input values (value added), when divided by the
difference between deflated output and input
(real value added), yields an implicit index of
value-added price or of unit value-added.

D ouble

R eal o u t p u t . Real output can be viewed as
either the value of production at constant prices
(gross output) or as “value-added at constant
values-added-per u nit.” The first approach is
akin to physical output measures in which output
is expressed in tons, car-miles, dozens, and so on.
It can also be represented in index form as value
at constant prices,
SP 0Qt/2 P 0Q0,

or

Another approach to
price measurement is that suggested by the Price
Statistics Review Com m ittee.8 In this proposal,
at each stage of aggregation, the weights assigned
represent only sales or output, which moves to
buyers outside that sector or stage of aggregation.
For example, as indexes are prepared for progres­
sively larger industry groupings (3-digit, 2-digit,
total industry division) in the Standard Industrial
Classification system , intrasector values and prices
are discarded as the sector definition becomes
broader. Weights are not additive to the total,
as they are in the double deflation approach.
However, this alternate method provides a concept
similar to that resulting from the double deflation
process. Furthermore, at the level of final demand
the results are consistent with other general price
indexes, but this system does not provide for
tracing and analyzing price movem ents as readily
as the i s p i system .
N et sector

SP.Q./SP.Qo

The first expression results from deflating a value
series by the Paasche-style formula for an index
of prices; the second, from use of a Laspeyresstyle (fixed market basket) price index.7
The second approach has no counterpart in
physical terms (for example, a quantum of valueadded is not readily visualized) but nonetheless
has an economic meaning as the utility added to
the materials and other purchased inputs in the
course of production. Deflated value-added can
be represented in index form as value added at
constant output and input prices:
SP qQ. Sp0q t
2 P 0Qo— 2p 0qo

°P

SP,Q ,— 2pt qj
S P ,Q 0— 2 p ,q 0

where “P ” and “Q” are output price and quantity,
“p” and “q” input price and quantity, respec­
tively, and “o ” and “i” signify data in the base
and current periods. In practice, however, the
values of output and input are deflated separately
by indexes of output prices and of input prices,
respectively, of either the current-weight or
fixed-weight form.

The

.

and price analysis

Price indexes constructed by industry have an
important use in economic analysis. They would
complement indexes which are classified by market
structure or similarity of use, such as the w p i .
Industry Sector Price Indexes allow comparison
of price trends for an industry with other com­
parable economic series for the industry.
For example, a consistent set of measures for
the basic steel industry under the industrial price
structure would cover price change of steel pro­

Gross output values (current or deflated)— and
the weights for the corresponding price indexes—
are not additive from industry to industry. For




ispi

m e a su r e m e n t

176

duced and sold, price change for goods and services
purchased and, through the technique of double
deflation, a value-added pi ice index summariz­
ing or netting out the input- and output-steel
price changes, so that an assessment of changing
industrial price spreads can be made. These trends
cOuld then be assessed in relation to changes in
production, average hourly earnings, productivity,
and unit labor cost, all of which are also compiled
by industry.
Price indexes developed according to an indus­
trial structure can also serve as tools in the analy­
sis of relationships between prices and wages,
materials costs, other costs, and profits. Experi­
mental work in this field is now underway to test
the feasibility of relating various aspects of chang­
ing costs and profits to price change in selected
industries.
While development and analysis of indexes
along industrial lines are proceeding, the improve­
m ent of data and analysis of price trends according
to stages of the production process m ust not be
neglected. The Bureau’s currently published Stage
of Processing ( s o p ) indexes show price movements
for commodities at various stages of production:
crude, interm ediate, and finished. In this set of
indexes each product is classified according to the
amount of processing, manufacturing, or assem­
bling it undergoes before entering the market.
Commodities m ay fall into more than one category.
For example, some fresh fruit is sold as a crude
material for further processing by canneries and
some directly to final demand. Because the s o p
indexes are currently tied into the w p i structure,
with its incomplete universe, the attem pt to recon­
cile the s o p and i s p i has results similar to that of
Procrustes and his bed. (See table 1.)
One of the principal limitations of the s o p
indexes is failure actually to price the various
different markets for the same product. This
should be largely corrected as a concomitant of
increased pricing for the i s p i .
In the long run, improvement of these indexes
will provide an additional facet to price analysis.
The s o p indexes will provide the connections
between prices at farm and mine through manu­
facturing and trade to the final consumer, to
supplement the industrial analyses provided by
the i s p i and the analysis of final demand price
movements provided by the general price indexes.
Table 1 not only shows the relationship of the




177

sop structure to the ispi, but also presents some
of the gaps which must be filled.

A fa m ily of indexes

The i s p i should be viewed, not as a single index,
as the w p i or c p i is, but as a system , or family,
of indexes which are flexible enough to serve as
deflators for a number of the more important sets
of economic statistics that relate to the National
Accounts as well as for assessment of the infla­
tionary and deflationary forces at work in the
economy. The dual organization of the General
Price Index by producers’ and purchasers’ prices
for final demand is but one example of the utility
of alternate sets of price indexes. To m eet a variety
of needs, the i s p i component series must be col­
lected in sufficient detail to perm it regrouping
(for example, imports and domestic prices should
be in separate series so that price changes for the
domestic industry can be separately analyzed).
Also, the i s p i components should be shown both
with and without taxes and possibly with and
without transport charges as discussed under the
sections on taxes and transportation.
All sets of alternative indexes will have certain
attributes in common. Comprehensive industry
coverage (or at least representation) would be
required. The i s p i should represent all industries
in the economy including importers and exporters.
The indexes should represent price movem ent for
goods not only as produced b y the primary in­
dustry (where the product takes its final form)
but also as exchanged, transported, marketed,
and further processed by the intermediate in­
dustries (wholesalers and jobbers) through which
the product passes and as sold at retail.9 In
addition, pricing for an industry would have to
cover not only the industry’s primary output but
price m ovements of its secondary products—goods
of a type normally made in other industries.
Thus, a product m ight be represented in the i s p i
system at any one stage of production in the in­
dustry where it is produced as a primary product
and as m any times as it appears as, secondary
output in other industries, or as it moves through
channels of distribution. Pricing is thus m ulti­
dimensional— horizontal throughout industries and
vertical along the lines of progress from raw
materials to retail or other final distribution.
N o t only m ust industrial coverage be compre­

hensive but all activities of the industries must be
represented— not sim ply the price movements
for commodities sold in the marketplace, as in the
wpi. Specifically, pricing should encompass pro­
duction and distribution of commodities, provision
of industrial services, interplant transfers between
establishm ents owned by the same company,
production of items for direct sale to “ultimate
consumers,” m ilitary item s (which are excluded
from wpi coverage) sold to the Government, and
purchase of commodities and services by industry,
including imports.
If the pricing system is comprehensive, the
variety of price data collected at the m ost detailed
level would be suitable to combine in a variety of
classification structures to m eet m ost major needs.
Price data could then be classified and group in­
dexes computed for the Standard Industrial Classi­
fication System covering agriculture, forestry, and
fisheries; mining; construction; manufacturing;
transportation, communication, electric, gas, and
sanitary services; wholesale and retail trade;
finance, insurance, and real estate; services; and
Government; the interindustry (input-output)
structure; and the N ational Income and Product
Accounts.

P roduct accounts. These consist of purchases of
(or sales to) individual consumers, Government,
business investors, and foreign countries and the
net “sales” to business inventory. Price indexes
developed for this set of data should be set at the
purchaser’s level rather than at the producers’
level, because that is the level at which the ac­
counts presently are aggregated. Separate indexes
would be compiled to show price change for per­
sonal consumption expenditures, gross private
domestic investm ent, net exports, and Govern­
m ent purchases of goods and services.
Pricing of inputs in the government sector is
im portant for a number of reasons. Federal, State,
and local purchases amounted to about one-fifth
of gnp in 1969. The largest single consumer
was the Federal Government. Because Federal
spending is used as a fiscal counterbalance to
inflationary and deflationary forces in the economy,
information on price trends of Government pur­
chases complement the price picture in the private
sector.
Indexes of Government purchase prices would
serve a number of budgetary purposes. Such




indexes could provide (through their use as de­
flators) estim ates of actual quantity purchased
(deflated value), and by this means answer the
question whether additional expenditure results
from increased quantity or higher prices. As an
example of detailed application, they could also
permit more accurate escalator clauses to be
written into government purchase agreements.

I ncome accounts. The income account side of the
national accounts consists of the returns to the
factors of production (wages, profits, rent, and so
on) plus nonfactor charges, such as direct business
taxes and depreciation. The gross product of an
industry measured from the income side is not
convertible to constant dollars according to the
same concept as measured from the product side.
This difference arises because of the definition of
quantity. The quantities implied in constantdollar product are the usual “final” outputs or
inputs measured in terms of tons, cubic feet, or
service rendered. The quantities implied in
constant-dollar income (when current-dollar in­
com e is deflated by indexes of wage rates, rents,
and so on) are measured in such “physical” units
as man-hours, use of a building for a year, and so
forth. The two approaches should be reconcilable
in terms of the total if suitable weights or inputoutput ratios (productivity ratios) could be
developed. In the foreseeable future, hQwever,
deflation will be confined to the product accounts
and to Gross Product Originating in industry
to which we now turn.
Gross product originating. This approach
focuses on the industrial origin of gross product.
While in the product account approach gnp is the
total of all final purchases, and in the income
approach the total of all factor incomes, plus
nonfactor charges such as indirect business taxes
and depreciation, in the gross product originating
approach gnp is the sum of the gross product of
all industries, representing each industry’s con­
tribution to the total output of goods and services^
Price indexes required for the constant-dollar
estimate of Gross Product Originating in each
industry consist of industrial output prices for
each industry and purchase price indexes. Output
price indexes should measure not only the primary
products but also the secondary products of in­
dustry. The “double-deflation” method described

earlier is used to obtain implicit “value-addedprice” indexes for the industry. Output prices for
this approach are at the ‘p roducers' level rather
than at the purchasers' level, in contrast to pricing
in the Product Accounts of the g n p . Since excise
taxes are included in the output data, the price
indexes should include them. Also, purchase of
services should be included in the input index.

possible decisions regarding them are presented in
the following sections:
I n t e r p l a n t t r a n s f e r s . Transfers to other plants
of the same company can be either included or
excluded from industry data on value of produc­
tion. (Census data carry the totals both ways.) In
the currently published industry price indexes of
the b l s , the decision was made to include the inter­
plant transfers in the concept because they are
part of total output. Their inclusion makes the
value and price data consistent with data on
man-hours, employment, and payrolls, all of which
im plicitly include them. (In practice, however,
only the weights are included because price m ove­
ments of interplant transfers are assumed to
parallel those of marketed products.) Moreover,
value added, excluding transfers, would be difficult
to estimate because data on cost of materials are
not collected according to whether the finished
product will be an interplant transfer. (In contrast,
such indexes as the w p i , which is market-oriented,
exclude interplant transfers.)

Pricing and concepts

Pricing for the various system s of Industry
Sector Price Indexes must be consistent with the
precepts of pricing— that the object priced be
standardized with respect to some highly specified
set of attributes— but within the general concept
of quantity (output or input) relating to the struc­
ture of the index of which the price series is a part.
If the price series is to be used as a deflator to con­
vert current-dollar data to constant-dollar data,
every conscious decision which is made concerning
the price series automatically results in an implicit
decision concerning deflated value, and, as a result,
production, quantity, and the unit of quantity.
Through its effect on quantity, the pricing
decision may, in turn, influence the measure
of productivity. For this reason, it is essential
that all decisions concerning the commodity or
service to be priced, the specification, level of
pricing, and the unit be determined with the
specific purpose in mind. After pricing is estab­
lished, it is essential that decisions concerning
adjustments for quality, timing of discounts,
changes in industrial vertical integration, treat­
m ent of transportation charges, and taxes should
be made with the larger goal in mind.
Sometimes several different or alternate deci­
sions might be required concerning particular
problems of pricing, because of the various uses
to which the i s p i would be put. For example, the
question whether the transportation charge for a
commodity should be included in the purchase
price would be answered in the affirmative in the
case of a deflator for materials inputs in a manu­
facturing industry whose input values, as reported
in the Census of Manufactures, include transporta­
tion implicitly. It would be answered in the nega­
tive for the interindustry chart, where transporta­
tion is treated as a separately purchased input and
materials are priced f . o . b .
Some problems that might be encountered in
constructing an Industrial Sector Price Index and




c h a n g e . It is not our purpose to deal
exhaustively with the quality change problem,
but only to indicate how decisions on this question
relate to the production measurement problem.
Unfortunately, there has never been agreement
among the various agencies of Government as to
the purpose to be served by any particular type of
quality adjustments— a statem ent which implies
that output itself has not been clearly defined.
Two aspects of the subject warrant attention in
particular: the nature of the measured quality
change and its incidence.
It is ap p a ren t th a t not all ch a n g es in products
or services will be greeted by purchasers as
quality improvements. Some will find new styling
appealing; others will object or at least be un­
moved. These are the subjective features of
quality change with which the psychologist, not
the economist, might deal. However, there are a
great number of changes in product which receive
a consensus—improved safety, contribution to
better health, performance, or economy of opera­
tion represent improvements while moves in the
opposite direction represent deterioration.
The nature of a change may be generally
accepted, but the nature of the measurement may
not be, largely because the question is approached
from opposite ends of the production-consump­

Q u a l it y

179

tion cycle. The production-oriented view would
recognize as quality changes only those specific
additions or deletions which require the use (or
removal) of productive resources in their creation.
The consumer-oriented view accepts changes
which contribute (positively or negatively) to the
utility, enjoyment, and so on of products or
services without regard to resource use. These
divergent attitudes lead to different types of
measures or adjustments for quality change, even
when there is agreement that quality has in fact
changed in a specific manner.
Some examples m ay serve to illustrate the
choices. Should adjustment for an improvement
in electric light bulbs, which extends the length of
life, be made on the basis of the additional lumens
provided or on the basis of the costed-price of the
added feature which made the improved perform­
ance possible? Should an improvement in the
ability of earthmoving equipment be based on the
additional tonnage-per-hour of earth moved or on
the cost-price of the improvement? Should changes
in motors be measured by changes in horsepower
ratings or by cost? Should quality change of a
new type, thinner, tin plate which allows more
beer cans to be produced per ton be measured by
the additional cans which m ay be produced or by
the incremental cost of the improvement? 10
The criterion used will determine the method
employed in the quality adjustment, which, in
turn, affects the magnitude of the measure because
the increment in resource inputs is rarely pro­
portional to the increment of usefulness or of
performance.
A third method— “let the consumer decide”—
shifts the decision to the market place. When the
product before improvement (or debasement) and
the original product are selling on the same market
at the same time, the price differential is taken
as the measure of quality change. This has con­
siderable appeal, especially where subjective m at­
ters, such as style, are involved, in that consumer
taste dictates and willingness to pay provides the
key to the adjustment. In some technical areas,
however, the universal application of this principle
m ay be less satisfactory because it assumes a high
degree of consumer sophistication. Rather than
an adjustment based on the increased tire miles
provided by an improved tire (consumer-oriented)
or the cost of providing more miles (produceroriented), the decision is based on the consumers’
belief or faith that the product performs better.




180

It is perhaps valid to observe that even some
sophisticated consumers m ay not be willing to pay
for increased safety either on the basis of the
amount of safety or the cost of it but m ay value
risk, dash, and adventure more. In this case, the
third method would be better. In any case, this
approach does not seem to m atch either the
consumer-oriented or market-oriented approach,
but m ay lie closer to the former.
Economists of the Office of Business Economics
have expressed the view that the appropriate
quality-change measure should be based on the
resource-use approach for their purpose.11 The
reason is that this method of measurement pro­
vides a production measure which can be used to
gage capital productivity change. In the earthmoving equipment example previously cited, the
percentage increase in performance would exceed
the cost increase and the difference would repre­
sent capital productivity gain. If the price index
were adjusted by the full amount of the perform­
ance increase (using the consumer-oriented
method) the quantity-of-machinery purchased or
used (input) would show a larger increase and
offset the increase in work done (output), in the
numerator of the productivity measure. For the
General Price Indexes and the i s p i , it seems that
the appropriate type of quality adjustment would
be cost-oriented.
For the c p i (an index outside the i s p i system)
it appears that the consumer-utility or perform­
ance adjustment has some merit. This method
would, however, recognize the c p i as primarily a
consumer welfare-type index. On the other hand,
adoption of this approach for the c p i would lim it
that measure’s usefulness as a contributor of
building blocks for the retail-trade-industry seg­
m ent of the i s p i system, and would lim it the
usefulness of comparisons between industrial
prices and their c p i counterparts.
A cost of living index would go even further
than the c p i in the direction of a consumeroriented approach to the quality problem. This
index would “take into account the fact that, for
m ost commodities, there is a rate at which the
consumer could substitute one for the other in
response to changes in relative prices and still
remain on the same plane or level of satisfaction,”
and a forced substitution—replacement by pro­
ducers of a low-priced item with one of higher
quality and price— would be treated as a price
increase.12

In connection with the Federal requirement
that smog-control devices be installed on new
automobiles, an interesting question has arisen:
Should the price indexes be adjusted at all for
these devices? The argument against adjustment
(and thus for treating the list price increase upon
introduction as a genuine price increase) holds
that the buyer does not benefit from the device—
others do. Also, the argument runs, acceptance
of such devices as improvement in quality (and
not price increases) overcompensates because of
failure to penalize for the environmental deteri­
oration which necessitated the device in the
first place. The opposing view is that the buyer
does benefit in a social sense. His paym ent “in
consideration of the paym ent by others’’ is of
value to him, hence a quality improvement. In
addition, this view states that the antismog
device does represent additional production and
a car with a device represents more than one
without (as does one with a radio or heater),
and failure to adjust ignores the additional
output— a consideration important for price series
used in deflation. This view of the index as a
deflator results in attaching a production-oriented
definition to the cpi, and the resulting index
measures social cost by the market cost of the
device.
If environmental improvement (or cessation of
deterioration) is paid for by price increases rather
than by taxes, this question will arise more fre­
quently in the future. The problems of quality
change associated with environmental improve­
m ent illustrate what has been said earlier con­
cerning pricing to meet concept requirements:
that all decisions on pricing, unit of measure,
quality adjustments, and so on, should be made
with the particular goal of measurement in mind
and, in case of deflators, with the particular effect
on the production measure in mind. Owing to the
variety of needs for price data, it seems reasonable
to assume that m any questions of this kind will
be solved by the construction of alternate meas­
ures, each for a specific purpose.
Turning to the problem of the incidence of
quality change, it is obvious that a change in
quality of the product of a given industry m ay
affect the performance and quality of goods and
services produced by other industries. In an
example already cited, the steel industry devel­
oped a better electrolytic tin plate that boosted




the number of cans manufacturers could produce
per ton. A decision to adjust the price index for
steel quality improvements by the full amount
of the increased performance “credits” the pro­
ducing industry with the full amount of the
improvement. Use of the cost-price adjustment,
on the other hand, can be shown to result in a
sharing of the effect between the producing and
using industries. This comes about because the
cost-price adjustment is usually of smaller mag­
nitude than the full-performance adjustment,
and the price of steel falls less and steel produc­
tion rises less than in the performance approach.
This in turn credits the using industry with less
steel consumption per can, and both industries
show a gain in productiveness.

T axes. If taxes are viewed as payment to govern­
ment for either specific or general service, then it
follows that they should be converted into some
sort of price paid for the general or specific serv­
ice and used in deflating government output. As
a consequence of considering taxes pa}unent for
government service, value figures and price in­
dexes for the private sector would exclude taxes
to avoid double counting. In addition, price series
needed for analysis of the interrelationship of
price, production, and productivity trends must
be comparable—hence, with taxes excluded.
However, there are practical necessities to con­
sider. Value-of-output-data used by obe in
deflation, for example, often contain the excise tax
both on the items in the output total and the hidden
taxes in the materials and components. In this
instance, the deflator should include excise and
sales taxes. The preference for inclusion of taxes
because of their inclusion in the value data is based
on the view that a tax increase should be reflected
as a price increase so that the deflated value series
will remain constant. This throws any concomitant
production increase which may flow from the tax
increase (new schools, roads, and so on) into the
government sector.
Because of variations in the value data with
respect to inclusion or exclusion of excise and sales
taxes, it seems likely that series should be available
with and without taxes in both the retail and nonretail price programs.
Income taxes and other taxes which are not
directly applicable to the transaction (sale or pur­
chase) or use of a good or service should be con­

181

sidered as payment to government and converted
to price indexes for the government output price
index, at least in concept.

does not apply when transport is treated as a
separate item. The handling of transportation
data is a special case of the general problem of
reconciling input pricing with output pricing.
B oth types of pricing are affected by productivity
changes or, in other terms, changes in input-output
ratios.
Several other categories of pricing m ay have to
be covered to m eet various needs. Inventory
pricing and goods in process pricing, while perhaps
not generally necessary, would be important in
industries such as shipbuilding where production
cycles are long or varying, and where it is essential
to adjust real value of shipments to represent
real output. Changes in vertical integration would
have to be watched carefully in dynamic situations
to assure continuous sample adjustment, as
today’s onsite production (for example, housing)
becomes tomorrow’s purchased component. Pricing
of large scale output (the purchase of the entire
crop by a cannery, for instance), and of long-term
purchase contracts are all part of the total picture
and m ust be taken into account.

T ransport charges. A s in the case of taxes, the
decision whether to include these charges is
based partly on the concept used, partly on the
nature of the data to be deflated. Values for
purchased materials reported to the Census of
M anufactures include the transportation charges,
so purchase price indexes would include transport.
As indicated, the interindustry concept views
purchasers as purchasing transportation sep­
arately, so purchase prices of goods would exclude
transportation charges for this purpose. Transport
would be separately deflated by indexes of
freight or passenger rates, as appropriate.
A t this point, it is important to note that the
different treatments do not lead to the same
results. W hen transportation is included in the
price, the series im plicitly prices the transport
charge per unit of product; when separately
priced, transport charge is standardized and
expressed as the charge for a given product
hauled a given distance, or a fixed number of
ton-miles. Thus, a change in the distance hauled
(change in the amount of transportation pur­
chased) becomes an integral part of price when
transportation is included— a situation which

Table 2.

Present program coverage

A t present, the Bureau’s Industry Sector Price
Indexes are published for only 100 4-digit manu­
facturing and mining industries out of a total of

Coverage of Gross National Product (by sector) and of industries by available price indexes
Percent of sector GNP (value of
shipments) covered2

Sector

Total...............................................................................................................

GNP
accounted
for*
(percent)

Information not
published

Information
pub­
lished

Good
coverage

No to
fair
coverage

16

*59

100.0

100

13

3.8
2.4
3.5
30.7
10.1
7.2
9.7
13.4
9.3
9.9

100
100
100
100
100
100
100
100
100
100

0
74
0
28
0
0
0
0
0
0

Agriculture, forestry, fisheries.................................................................................
Mining___I . .............................................................................................................
Contract construction................................................................................................
Manufacturing...........................................................................................................
Transportation, communication, electric, gas, and sanitary....................................
Wholesale trade............................................................... ......................................
Retail trade..................................................................... .........................................
Finance, insurance, and real estate.........................................................................
Services.....................................................................................................................
Government...............................................................................................................

Information not
published
Total

88
0
37
13
100
0
55

12
26
63
59
0
100
45

Information
published

Good
coverage

No to
fair
coverage

382

101

79

702

37
50
22
421
72
48
67
79
86

0
10
0
91
0
0
0
0
0
0

10
1
0
33
4
0
16
0
15

27
39
22
297
68
48
51
79
71

2 These industries are defined in the Standard Industrial Classification Manual, 1967
(Bureau of the Budget, Office of Statistical Standards).
* The remaining 12 percent is accounted for by industries for which detail on value
coverage does not permit adequate evaluation.

* The Office of Business Economics of the U.S. Department of Commerce is the source
of these data.
2 Percent of shipments values covered is derived from the following: For agricultural
sectors, from unpublished material of the U.S. Department of Agriculture; for mining and
manufacturing, the Censuses of Minerals Industries, and Manufactures, 1963; contract
construction estimated from residential construction as proportion of all construction;
for transportation and warehousing, communications, and so forth, and parts of services,
from unpublished data of the Office of Business Economics; and for retail trade from the
1963 Census of Business.




Total

Number of 4-digit industries covered2

NOTE: Dashes indicate information on availability of data is not known.

182

Some data are available for transportation from
the Interstate Commerce Commission and other
regulatory and ratemaking agencies, and the
Bureau of the Census has published a price index
for new single-family dwellings. Exploratory work
in other aspects of construction is continuing in
both b l s and the Census Bureau. Research on
import and export pricing is also underway.
Table 2 shows in detail the coverage available
in terms of the interindustry classification struc­
ture. It is evident that much remains to be done.
Coverage is low as a whole. Even in manufacturing,
the gaps are considerable and are characterized by
nonhomogeneous or rapidly changing products,
such as aircraft, electronics, and shipbuilding. □

a b o u t 500 in th e s e tw o d iv is io n s a n d o f a t o ta l of
a b o u t 9 0 0 in all d iv is io n s . T h e s e 100 in d u s tr ie s
c o v e r a b o u t 13 p e r c e n t o f th e to ta l d o m e s tic v a lu e
o f U .S . o u t p u t . C o v e ra g e a c c o u n te d f o r b y
p u b lis h e d in d e x e s is h ig h e s t fo r m in in g — a b o u t
75 p e r c e n t— fo llo w ed b y m a n u f a c tu r in g w ith 28
p e r c e n t. F o r th e la t t e r , a d d itio n a l p r o d u c t- c la s s
( 5 -d ig it) in d e x e s a r e a v a ila b le w h ic h w o u ld b r in g
c o v e ra g e to a b o u t 45 p e r c e n t. (S ee ta b le 2.)

Coverage in agriculture is reasonably good,
although there is some question as to the level of
pricing (whether close enough to the farm). There
is fairly good coverage in some utilities and in
retail trade (largely from the c p i ) , but insufficient
for publication of industry indexes.

7 2 P o0i _ Z P jQ, . Z P i Q j
2 PoQo ZPoQo ‘ 2 P 0Qi

1 See J a c k A lte rm a n a n d M a r tin L . M a rim o n t, Prices
find Price A nalysis in the Framework of the N ational A c­
counts, a p a p e r p re s e n te d a t th e 1 1 th G e n e ra l C o n fe re n c e

the Price Indexes (U.S. Congress, Joint
89th Cong., 2d sess., 1966). In term s of
Im plicit Price D eflator relies on the CPI
coverage, the WPI for 12 percent, agri­
and other prices and nonprice estim ates

3 See th e s ta te m e n t
S ta tis tic s , G eoffrey H .
on E c o n o m ic S ta tis tic s
C o n g ress of th e U n ite d

2 P iQ i_ Z P iQ o
2 P 0Q0 ■ 2 P qQ0

8 The Price Statistics of the Federal Government, a re p o r t
pf th e P ric e S ta tis tic s R e v ie w C o m m itte e of th e N a tio n a l
B u re a u of E c o n o m ic R e s e a rc h to th e B u re a u of th e B u d g e t,
h e a rin g s b efo re th e S u b c o m m itte e on E c o n o m ic S ta tis tic s
of th e J o in t E c o n o m ic C o m m itte e , J a n u a r y 24, 1961.

of th e I n te r n a tio n a l A sso c ia tio n fo r R e s e a rc h in In c o m e
a n d W e a lth , A u g u st 2 4 -3 1 , 1969, N a th a n y a , Is ra e l.
2 See Inflation and
E conom ic C om m ittee,
base year weights the
for 46 percent o f its
culture for 7 percent,
for the rem ainder.

ZPiQi
~LPiQ0

9 See fo o tn o te 5.
10 D iscu ssio n of th e s e issues can b e fo u n d in E d w a r d F .
D e n iso n , P ro b le m s of C a p ita l F o rm a tio n , S tu d ie s in I n ­
co m e a n d W e a lth , V o lu m e 19 (N ew Y o rk , N a tio n a l
B u re a u of E c o n o m ic R e se a rc h , 1957), p p . 2 1 7 -2 3 4 ; M ilto n
G ilb e rt, “ Q u a lity C h a n g e s a n d In d e x N u m b e rs ,” Economic
Development and Cultural Change, A p ril 1961, p p . 2 8 7 -2 9 4 ;
a n d Z v i G rilich es, “ Q u a lity C h a n g e a n d In d e x N u m b e rs : A
C ritiq u e ,” a n d M ilto n G ilb e rt, “ A R e p ly ,” b o th in
M onthly Labor Review, M a y 1962, p p . 5 4 2 -5 4 5 , a n d in th e
m in u te s of th e C o m m itte e on C o n su m e r a n d W h o lesale
P ric e s, B u sin ess R e s e a rc h A d v iso ry C o u n cil to th e B u re a u
of L a b o r S ta tis tic s , F e b r u a r y 18, 1964.

b y th e C o m m issio n e r of L a b o r
M o o re, b e fo re th e S u b c o m m itte e
of th e J o in t E c o n o m ic C o m m itte e ,
S ta te s , M a y 15, 1969.

4 T h e Im p lic it P ric e D e fla to r is d e riv e d b y d iv id in g to t a l
e x p e n d itu re s v a lu e d in c u rr e n t p ric e s b y to t a l e x p e n d itu re s
v a lu e d in b a se p e rio d p rices. In c o n s tru c tin g th e D e fla to r,
v a lu e d a t a a re d e fla te d b y a p p ro p r ia te p rice in d ex es (w h ic h
m a y be fix ed -w eig h t g ro u p indexes) a t th e fin e st d eg ree of
d e ta il feasib le a n d su m m e d to o b ta in th e to t a l c o n s ta n td o lla r figure. I t c a n b e sh o w n t h a t th e to t a l p ric e in d ex
d e riv e d is of th e P a a s c h e fo rm in so fa r as w e ig h ts b e tw e e n
th e m o s t d e ta ile d lev el of a g g re g a tio n is co n cern ed .

11 See G eo rg e Ja s z i, R o b e r t C. W asso n , a n d L a w re n c e
G ro se, “ E x p a n sio n of F ix e d B u sin e ss C a p ita l in th e U n ite d
S ta te s ,” Survey of Current Business, N o v e m b e r 1962, p. 10;
a n d L a w re n c e G rose, Ir v in g R o tte n b e rg , a n d R o b e r t C.
W asson, “ N ew E s tim a te s of F ix e d B u sin ess C a p ita l in th e
U n ite d S ta te s , 1 9 2 5 -6 5 ,” Survey of Current Business,
D e c e m b e r 1966, p p . 3 7 -3 8 .

5 F o r a m o re d e ta ile d a n a ly s is of th e n e e d fo r re ta il
p ric in g , see A llan D . S earle a n d M a ry E . L a w re n c e , “ R e ­
ta il P ric e s a n d th e C o n su m e r P ric e In d e x ,” F e b ru a r y 1969,
m im e o g ra p h e d .

12 F o r a d isc u ssio n of d ifferen ces b e tw e e n a co st-o f-liv in g
9
F o r a m o re c o m p le te d e s c rip tio n of th e B u re a u of in d e x a n d o th e r c o n su m e r p ric e in d ex es, see Jo e l P o p k in
“ T h e P ro g ra m fo r th e 1975 R e v isio n of th e c p i , ” a p a p e r
L a b o r S ta tis tic s c u rr e n t p ro g ra m , see “ In d u s try - S e c to r
p re s e n te d b e fo re th e N a tio n a l P la n n in g A s s o c ia tio n
In d e x e s ,” Handbook of Methods for Surveys and Studies
O c to b e r 1970.
(B L S B u lle tin 1458, 1966).




183

Updating the
Consumer Price
Index—
an overview

The place of the
Consumer Price Index
in today’s economy
and some of the problems
in keeping it up to date
JULIUS SHISKIN

T h e m o n t h l y Consumer Price Index is the only

is eroded by price increases, and serves as a major
economic indicator.

index compiled by the U.S. Government that is de­
signed to measure changes in the purchasing power
of the urban consumer’s dollar. It serves two major
functions:

A s an escalator. It is estimated that there are more
than 5.1 million workers covered by collective bar­
gaining contracts which provide for increases in wage
rates when the CPI rises. The number and applica­
tion of these escalator clauses is increasing, and
escalator clauses based on the index show signs of
changing. In the spring 1974 settlements in the
aluminum industry, for instance, a new step was
taken when aluminum producers agreed to provide
for annual automatic cost-of-living escalator adjust­
ment in pension benefit levels, so that pension pay­
ments to retired workers will rise partially (65 per­
cent) along with a rise in the Consumer Price Index.
In addition to workers whose wages or pensions
are adjusted according to changes in the CPI, some
44 million persons now find their incomes affected
by the index, largely as a result of statutory action:
almost 29 million social security beneficiaries, about
2 million retired military and Federal Civil Service
employees and survivors, 600,000 postal workers,
and about 13 million food stamp recipients. When
dependents are taken into account, the incomes of
somewhere in the neighborhood of one-half the pop­
ulation already are or soon will be pegged to the Con­
sumer Price Index.
Another group whose living standard is affected by
changes in the Consumer Price Index are the 24 mil­
lion children who eat lunch or breakfast at school,
under the National School Lunch Act and the Child
Nutrition Act of 1966. Under Public Law 93-150,
national average rates for these lunches and break­
fasts are adjusted semiannually by the Secretary of
Agriculture on the basis of the change in the CPI
series, “Food away from home.”
Also, the poverty threshold estimate, which is the
basis of eligibility in many health and welfare pro­
grams of both Federal and State and local govera-

• It is a yardstick for revising wages, salaries, and
other income payments to keep in step with rising
prices; and
• It is an indicator of the rate of inflation in the
American economy.

Because of changes in consumer buying patterns,
it is necessary to update and revise the Consumer
Price Index periodically. The Bureau of Labor Sta­
tistics is now in the midst of a major revision, sched­
uled for completion in 1977. The index will be
tested beginning in 1976. Starting in April 1977,
BLS will publish two Consumer Price Indexes:
an improved index for urban wage earners and cleri­
cal workers to meet the requirements of collective
bargaining, and an index for all urban households,
which will provide a new comprehensive measure of
price change for the economy.
This article briefly describes uses of the Consumer
Price Index, defines what it measures and describes
its limitations as a cost-of-living index, reviews
earlier revision programs,1 reports on some of the
problems encountered in the current revision, and
describes the additional data that will be avail­
able after the revision has been completed and the
new indexes are published in 1977.
Uses of the Consumer Price Index

Today, as in earlier years, the Consumer Price
Index plays an important role in consumers’ attempts
to assess the degree to which their purchasing power

Julius Shiskin is Commissioner of Labor Statistics.

From the Review of July 1974



184

ments, is updated periodically to keep in step with
the Consumer Price Index. Under the Comprehen­
sive Employment and Training Act of 1973, the “low
income” standard specified as one of the criteria for
distribution of manpower revenue-sharing funds is
kept current by reference to the index.
In addition, escalator clauses in an increasing num­
ber of rental, royalty, and child support agreements
automatically increase payments to an undetermined
number of people. The CPI is also used as a guide in
drawing up contracts and in wage negotiations.
A s an economic indicator. Beyond its application to
wages and other income payments to individual
Americans, the index has direct impact on the formu­
lation and evaluation of government economic
policy that affects virtually everyone. The Consumer
Price Index is, in fact, a major yardstick by which
the success or failure of government economic pol­
icies is judged.
As an indicator of cyclical change in the economy,
the index itself has typically lagged behind other
measures of economic performance, such as real
GNP (output) and unemployment. The Wholesale
Price Index also tends to lead the CPI, although the
leads are quite variable. On the other hand, the
Consumer Price Index seems to lead the GNP im­
plicit price deflator, although the lead is less clear
when comparisons are made with the GNP price
deflator computed with fixed weights.
In the light of these important issues involving the
CPI, it is clear that an accurate Consumer Price
Index is of the utmost importance. At present, a 1percent change in the index triggers at least a $1

Changes in nomenclature
Never a static instrument, the Consumer Price
Index has been responsive to changes in expenditures
and earnings patterns, as well as in its uses and in the
economy. Changes in use and in concept brought
changes in nomenclature as well.
Between 1913 and 1945, the Bureau of Labor Sta­
tistics referred to The Cost-of-Living Index for the
United States. In 1945, the name was changed to
Consumers’ Price Index for Moderate Income Fami­
lies in Large Cities. In 1964, the current title, Con­
sumer Price Index for Urban Wage Earners and
Clerical Workers, was adopted.
Beginning in 1977, a new Consumer Price Index
for A ll Urban Households will be published, in addi­
tion to an updated Consumer Price Index for Urban
Wage Earners and Clerical Workers.




185

billion increase in income under escalation provi­
sions. An error of only 0.1 percent can thus lead to
the misallocation of over $100 million.
Furthermore, while it is difficult to estimate the
effects of an error in the Consumer Price Index on
economic policy decisions, it is also clear that— with
inflation the major economic problem of the day—
the stakes involved in an accurate Consumer Price
Index are very great relative to the costs.
What the CPI measures— and doesn’t

The Consumer Price Index compares what the
“market basket” of goods and services cost this
month against what it cost a month ago, or a year
ago, or 10 years ago, or in 1967 (the base year for
the current index). Say that in 1967 the prescribed
market basket could have been purchased for $100.
In February 1974 the CPI was 141.5 and in March
1974 it reached 143.1. That means that the same
combination of goods and services that could have
been obtained for $100 in 1967 cost $141.50 in
February 1974 and $143.10 in March.
This does not necessarily mean the average con­
sumer actually spent $143.10 in March 1974 as
against $100 in 1967. Consumers tend to adjust their
shopping practices to the prices they encounter in the
marketplace and to substitute less costly items, or do
without, in order to hold their spending within their
means. For example, if the price of certain cuts of
beef rises rapidly, the purchasers may shift to poultry
or less expensive meat. If the charge for repair serv­
ices increases more than the customer believes is ac­
ceptable, householders tend to postpone having the
repairs made or to “do it themselves.”
The index does not take this sort of substitution
into account, but rather is predicated on the purchase
of the same market basket, in the same proportions
(or weight), month after month. This is one reason
why it is called a price index and not a cost-of-living
index— although the public often refers to it as a
cost-of-living index, and it is often used in that way.
There are other major differences between the two
types of indexes. For instance, the CPI does not in­
clude income and social security taxes since (unlike
sales taxes) these costs are not directly associated
with retail prices of specific goods and services,
whereas a true cost-of-living index would explicitly
include them.
The CPI does not immediately reflect changes in
expenditure patterns, nor can it immediately adjust
to the introduction into the economy of new products
or services. For example, the increased use of con­

ment agencies since the late 19th century,2 the Bu­
reau of Labor Statistics first consumer price index,
called a cost-of-living index,3 grew out of a decision
by the Shipbuilding Labor Adjustment Board during
World War I. In arriving at a “fair wage scale,” the
Board determined in November 1917 that readjust­
ment of wages in the shipbuilding yards was war­
ranted when there had been a general and material
increase in the cost of living.4 During 1918-19, in
cooperation with the Board, the Bureau investigated
the cost of living in a number of shipbuilding and
other industrial centers. Details of expenditures on
goods in the family market basket were obtained
from each of 12,000 wage-earner families in 92
cities, and records of retail establishments in 32
cities provided prices for a large number of articles.
Regular price collection was initiated after 1917 in
these 32 cities, with price information collected 1 to
4 times a year for about 145 commodities and serv­
ices. In 1919, the Bureau began the publication of
complete “cost of living” indexes at semiannual in­
tervals for 32 large shipbuilding and industrial cen­
ters, using a weighting structure based on expendi­
tures of wage-earner and clerical-worker families in
1917-19.5 In February 1921, regular, periodic pub­
lication of the U.S. index in roughly its present form
was established. Although over the years there have
been many changes in scope, coverage, frequency,
and publication format, the index has remained a
measure of change in the cost of a fixed market
basket of goods and services. Quarterly indexes were
initiated in 1935, and monthly indexes began in

venience foods as more and more women entered the
labor market and the rise in “fast-food” eating places
— these social and economic phenomena were in
place for some time before they could be adequately
reflected in the index. Similarly, a product which has
fallen from public favor— either because its place is
usurped by a better product, or simply because of a
change in fashion or consumer preference— may con­
tinue for a time to carry a disproportionate weight in
the index until it can appropriately be phased out.
But even within the fixed market basket concept of
the CPI, provision is made for some changes in prod­
ucts between the main decennial revisions.
The Consumer Price Index does not attempt to
report these changes in the style of living. It simply
measures the changes in prices for a scientifically se­
lected market basket based on the average experience
of certain population groups. Items in the market
basket for which the CPI measures price changes
run the gamut from bread and butter to television
and bowling fees, from prenatal and obstetrics serv­
ices to funeral expenditures, from popular paper­
backs to college textbooks. The CPI never has been
limited to price changes of so-called necessities.
Expenditures by a cross section of consumers liv­
ing in a representative selection of urban places, as
disclosed by Consumer Expenditure Surveys, provide
the basis both for the selection of items to be priced
and the importance of each of these items in the
index structure. The relative importance (or weight)
given to each item also is derived from the Consumer
Expenditure Survey. The weights reflect the experi­
ence of renters and of homeowners; of car owners
and of earless families and individuals; of families
with many children, childless families, and single
consumers.
Since the CPI is based on expenditures, it does not
reflect noncash consumption, such as food grown at
home, fringe benefits received as part of a job, serv­
ices supplied by government agencies without pay­
ment of a special tax fee, and so on. When the rela­
tive importance of such an item changes over time—
as with medical care, for which employers and the
government have in recent years assumed an in­
creased proportion of the expense— these changes
must be taken into account in interpretation of the
index.

Forthcoming articles on the CPI revision
Revision o f the Consumer Price Index, as this ar­
ticle indicates, is a long and complex endeavor, in­
volving the work of many staff members. The present
article is an overview of the revision process, and
interim progress report on some of the actions taken
and plans underway. Additional articles are planned
describing both methodology and survey results.
The principal BLS staff contributors to this over­
view article were Kenneth Dalton, Chief of the Divi­
sion of Consumer Prices and Price Indexes, Office of
Prices and Living Conditions and Robert Gillingham,
Economist, CPI Revision Group. Georgena Potts of
the Office of Publications assisted in bringing the mate­
rials for this article together and in writing the text.
Much of the work described above on the current re­
vision was carried out under the direction of Janet L.
Norwood, now Deputy Commissioner of Labor Sta­
tistics.

Origin of the CPI

Although studies of prices and living conditions
in the United States had been conducted by govern­




186

October 1940 at the request of the National Defense
Advisory Commission.6
First major revision— 1940

In 1933 the Secretary of Labor requested that
the American Statistical Association appoint a com­
mittee to advise the Department on its statistical pro­
grams. The Advisory Committee paid particular
attention to cost-of-living indexes, and on its recom­
mendation the Bureau of Labor Statistics initiated
steps leading to a comprehensive revision of its
indexes.
In 1934-36, the Bureau undertook a comprehen­
sive survey of “Money Disbursements of Wage Earn­
ers and Clerical Workers,” which covered 14,500
families of two persons or more in 42 cities with
50,000 inhabitants or more. Price collection proce­
dures were altered and methodological changes in
index calculation were made, modifying both the
weights used in combining group indexes to obtain
the “all-items” index and the population weights for
combining cities.7 The system of weights was re­
vised,8 with specific weights based on city food
expenditure patterns replacing the regional weights
formerly used. New commodities were added, and
food indexes were constructed on the new basis back

to March 1919. Also, the Bureau adopted the prin­
ciple of imputation— that is, ascribing to a sample
item that could not be priced the price change for
groups of items presumed to have price movements
similar to the sample item.
The comprehensive revision of the index was com­
pleted in 1940.9 At the same time, the reference
base period was shifted to 1935-39 = 100, on
advice of the Central Statistical Board (predecessor
of the present Statistical Policy Division of the
Office of Management and Budget).
Post-World War II revision

During World War II, temporary adjustments in
data collection procedures and in weights for foods,
fuels, transportation, and other selected items had
been made to take account of rationing and wartime
shortages.10 These adjustments were necessarily im­
perfect. In 1946, when wartime restrictions were re­
moved, prewar weight patterns were restored, with
adjustments to validate the actual change.
In 1946, also, a number of important changes
were made in the calculation of food prices. Sepa­
rate average prices were computed for chain and
independent stores, and these averages were com­
bined using fixed weights. Food outlet samples were

Publication of the Consumer Price Index
more in 1960, based on the pricing of full samples of
items. These indexes are computed monthly for five areas:
Chicago, Ill.-Northwestern Indiana; Detroit, Mich.; Los
Angeles-Long Beach, Calif.; New York, N.Y.-Northeast­
ern N.J., and the Philadelphia metropolitan area, and
once every 3 months on a rotating cycle, for the other
published “city” areas. Indexes are published monthly
for the food component for published “city” areas.
Because many users misinterpret the city indexes as
measures of intercity differences in prices, each report
cautions the user of these indexes that comparisons of
indexes for individual SMSA’s show only that prices in
one location changed more or less than in another. The
metropolitan area indexes cannot be used to measure dif­
ferences in price levels or in living costs between areas.
Besides publication of city indexes in a national press
release, statements are issued from the Bureau’s regional
offices on the same day as the national release. These
contain indexes and analyses of price movements in the
individual areas.
Starting in 1973, indexes have been published for
cities in five population-size groups, and in 1974 regional
indexes were added.

The national Consumer Price Index is compiled by the
Bureau of Labor Statistics and published about 3 weeks
following the month to which the data refer. The index
refers to the entire month, not any specific day of the
month. Prices are collected early in the month for foods,
around mid-month for rents and utilities, and over the
entire month for other goods and services. Approximately
15,000 retail stores and other retail outlets (bowling alleys,
doctors’ offices, and so on) are visited each month and
approximately 400 items are priced. A press release
contains a brief analysis of prices movements during the
month, as well as the latest available indexes and percent
changes over selected periods. A more detailed report
is published subsequently in the Monthly Labor Review
(table 25, pp. 103-08) and in a special periodical, The
Consumer Price Index. U.S. average indexes are published
monthly for “all items” and for groups, subgroups, and
selected items.
Individual “city” indexes for Standard Metropolitan
Statistical Areas, identified by the names of their central
cities, are published in the monthly press release, in
the Monthly Labor Review (table 27, p. 110) and in a
detailed report for individual Standard Metropolitan Sta­
tistical Areas (SMSA’s) having 1 million inhabitants or




187

revised, taking into account type of store, sales
volume and location.

new products (such as television sets and frozen
foods) and items that had not been previously cov­
ered, such as restaurant meals and owned homes.13
The new index was linked to the adjusted index in
December 1952 to form a continuous series.14

The 1953 revision

Expenditure surveys conducted in a few cities in
1947-49 showed significant post-war changes in con­
sumption patterns of wage-earner and clerical-worker
families, indicating a serious need for revision of the
index weights used and the market basket items.11
In 1949, the Congress authorized a large-scale 3-year
program for modernization of the index. By this time,
the postwar rise in prices, which followed elimina­
tion of price controls in mid-1946, appeared to have
run its course; prices had begun to decline from their
postwar peaks, and the period 1951-52 was expected
to be characterized by relatively stable economic
conditions.
The outbreak of hostilities in Korea, however, was
accompanied by sharp and diverse price increases
in the United States. These price changes, coupled
with widespread use of the index in wage escalation
clauses, made adjustment of the index weights to
post-World War II spending patterns extremely
urgent. In an interim revision,12 using data from
1947-49 expenditure surveys in seven cities, group
weights were adjusted, and 25 additional items were
selected for pricing. Both the “old series” index and
the adjusted index were published simultaneously
from 1950 through 1952, when the old series was
discontinued.
The comprehensive revision which was begun in
1949, was completed in 1953. Surveys of consumer
expenditures were conducted in 91 cities, the index
concepts were reexamined completely, and the
index reference base was changed from 1935-39
to 1947-49. The general concept of the index as
a measure of price change for a fixed market basket
of goods and services was retained, but a major
change was made by including the purchase of a
home in the weighting pattern. The classification
of goods and services into groups and subgroups
was revised, and indexes were computed retroactively
on the new base period (1 9 4 7 -4 9 ) for the new
major groups. The revision introduced a new sample
of 46 index cities out of the 91 cities in the Consumer
Expenditure Survey, including for the first time small
urban places (including areas with as few as 2,500
inhabitants) as well as large cities; revised weights
reflected the 1950 spending pattern of wage-earner
and clerical-worker families adjusted to 1952; and
the list of items priced was expanded to include




188

The 1964 revision

By the late 1950’s, it became apparent that the
index weights should not go unrevised for more than
a decade. The Bureau of Labor Statistics asked for
and received authorization for a revision program, to
take 5 years, which was begun in 1959. It included a
Consumer Expenditure Survey conducted in 196061 that provided information on the entire popula­
tion. These data were basic in selecting a new market
basket, new weights to reflect the distribution of con­
sumers’ expenditures, and a new and larger sample
of cities and retail stores. Chart 1 indicates how the
weighting pattern has changed over the years.
Since the 1950’s the population had mushroomed,
but, more important, it presented a composition
markedly different from that in 1950. The proportion
of persons at each end of the life cycle had increased.
Major changes had occurred in its geographic dis­
tribution. About 1 in 5 family units was moving
each year, many to the South and West, from farm
to city, from the central city to the suburbs, and to
peripheral areas in the process of urbanization.
Personal disposable income had moved upward
since 1950— about 37 percent between 1950 and
1956— and more than two-thirds of the rise was re­
flected in increased real income. Shifts in consumer
spending patterns were apparent. Further extension
of credit on easy terms made the consumer less and
less willing to defer purchasing a house, major appli­
ances, an automobile, and other large-ticket items.
Also, the decline of price maintenance laws and the
rise of the discount house had altered retail distribu­
tion patterns. Many new products or qualities had
come into being, ranging from deep freezers to new
household items made of plastics. Greater use was
being made of frozen foods, and there were import­
ant changes in housing, including a large number of
new units and a continuing shift from rental to owner
occupancy. Particularly significant was the increas­
ing share of consumer services in the economy as
a whole.
The basic objective of the index continued to be
the measurement of change in the price of a fixed
market basket of goods and services for urban
wage earners and clerical workers. A number of

Chart 1.

The consumer market basket, selected periods

1935-39

1952

1963

1 Includes personal finance charges other than automobile financing and mortgage interest. Imputed, not directly priced.

important changes were introduced, however: (1 )
the population coverage of the index was expanded
to include persons living alone; (2 ) the definition
of an urban wage-earner or clerical-worker family
was modified, so that a family was considered within
the scope of the survey if 50 percent or more of its
income came from wage and clerical occupations
and if at least one member of the family worked
for a minimum of 37 weeks of the year (in the old
series, this working member had to be the head of
the household; the change was made because of the
increasing importance of families with two workers
or more and of family units whose household head
was retired, but which had other working members);
and (3 ) the income limitation was dropped.15
These changes raised the population coverage to
about 55 percent of the urban population and under
45 percent of total population. At the time of the
1964 revision it was estimated that single workers
living alone represented about 10 percent of all urban
wage-earner and clerical-worker consumer units to
which the index applied, and family units 90 per­
cent. (On an expenditure weights basis, however,
the importance of single consumer units is only 6
percent of the composite wage-earner and clericalworker index.)
The average income of the population covered
was $5,963 in 1960-61 on the basis of the revised
definition, compared with $6,230 prior to revision.
This decline resulted from inclusion of single work­




ers, whose average income of $3,560 was consider­
ably below that of the family groups.
A new and expanded sample of metropolitan
areas and small urban places was introduced, based
on the 1960 Census of Population, and pricing was
extended to suburban areas. The sample of retail
stores was also revised and expanded. Probability
sampling techniques were used for the first time in
the selection of items for pricing. A system for
measuring sampling error was developed, and im­
provements were made in price collection methods.
The revision was completed in 1964. As before,
the new series was linked to the old in order to
maintain continuity. To provide an opportunity for
examining differences in price movements and to
allow persons using the index in contractual agree­
ments (such as labor contracts) to shift to the
revised index, both the old and new series indexes
were published for the period January to June 1964.
The two indexes did not diverge substantially.
Current revision program

The current revision of the Consumer Price Index
has been a major project of the Bureau of Labor
Statistics over the past 4 years. As in the past, the
revision program involves the development of a
greater amount of data and a review of the economic
and statistical concepts and operational procedures
used in constructing the index. Exhibit 1 shows the
189

progression of various steps in the process. Major
elements of the current revision, simply stated, are:

sumer Price Index is collected from a series of
sample surveys. The most important of these is the
Consumer Expenditure Survey, which collects in­
formation on what people buy. The latest such sur­
vey, conducted for the current revision, covers the
years 1972 and 1973. It differs from previous sur­
veys in several aspects of design and collection
methods, notably in combining the resources of the
Nation’s two major economic statistical agencies, the
Bureau of the Census and the Bureau of Labor
Statistics. BLS developed the questionnaire content
and specified the output. Census selected the house­
hold sample, spread throughout 216 areas of the
country, and conducted the interviews. Most of the
information was obtained in a series of quarterly
interviews involving about 20,000 families.
The remaining information was obtained from
another sample of about 20,000 families, who were
asked to complete a 2-week diary, in order to ob­
tain data on small, frequent purchases, such as food
and personal care items, which are typically difficult
to recall. The diary collection program started and
ended 6 months later than the quarterly survey.

• On the basis o f a survey o f consum er expendi­
tures, to determ ine
a. the proportion o f spending for food , shelter,
m edical care, and so on (to be used in the index
w eigh ts), and
b. the specific goods and services to be included
in the market basket.
• T o obtain a new sam ple o f stores w here people
buy, reflecting shifts in retail purchases, such as from
central cities to suburbs and from retail stores to m ail
order houses;
• T o m odernize the conceptual fram ew ork to m ake
the index more relevant to current and prospective
econom ic conditions, and to im prove statistical m eth­
odology, particularly sam pling techniques.

Present goals are that the revised CPI will have
sampling errors that are substantially lower than
those of the current CPI.
Surveys

Consumer Expenditure Survey— what people buy.
Information for the decennial revision of the Con­
Exhibit 1.
1972

CPI revision calendar
1973

1975

1974.

1976

1977

197S

Consumer Expenditure Survey—1st year quarterly data collection.
Consumer Expenditure Survey—1st year diary data collection.
Bureau of Census—processing and delivery of 1st year diary data.
Consumer Expenditure Survey—2d year quarterly data collection.
Bureau of Census—processing and delivery of 1st year quarterly data.
Consumer Expenditure Survey—2d year diary data collection.
Preparation for Point-of-Purchase Survey.
Bureau of Census—processing and delivery of 2d year diary data.
Preparation for rent survey.
Bureau of Census—processing and delivery of 2d year quarterly data.
Point-of-Purchase Survey data collection.
Screening, listing, and initiation of Rent Survey.
Item sample selection.
Rent test index.
Bureau of Census—delivery of Point-of-Purchase results.
Outlet sample selection.
Initiation of revised item and outlet samples.
Test indexes collected and compiled.
Publication of two indexes.
Evaluation and comparison of two indexes.
1972

1973

1974

1975

NOTE: Dates above refer to start of projects but not necessarily to their completion.




190

1976

1977

1979

substitution of nearby units is permitted. Six-month
rent changes are obtained in each city by part-time
agents every 2 or 3 months, by personal visit or tele­
phone inquiry to tenants of specified units in differ­
ent samples. The agent uses a detailed checklist cov­
ering fuels, gas, and electricity, telephone, garage,
furniture, water, maid service, switchboard service,
and so forth. In most cities, two subsamples of up to
500 rental units each are drawn, with each sample
priced semiannually in different months. In the five
largest cities, three subsamples of about 500 each are
priced semiannually in different calendar months,
providing data for one of the subsamples every 2
months.
In the 1974 survey, as a first step, BLS data col­
lectors in specified areas first list housing unit
addresses in the sample neighborhoods, including
both single and multifamily dwelling units. In the
second stage, the data collectors visit randomly
selected dwelling units and interview the occupants
to obtain information on whether the units are
owner- or renter-occupied, the type and amount of
rent, type of occupancy (year-round, transient, or
seasonal), age and type of structure and whether the
unit has complete kitchen facilities. In the final stage
of the survey, respondents will be asked to provide
information on the amount of rent paid and the
kinds of equipment and services included in the
rent. Thereafter, contact will be made semiannually
with samples of those dwelling units which meet the
specifications for inclusion in the Consumer Price
Index. The samples will be rotated and information
will be obtained each month on changes in the
amount of rent paid and the services and facilities
provided in the current and the previous month.
The new sample design will improve the timeliness
of the rent index, as well as its accuracy. Rent for
the current month will be compared with rent for
the immediately preceding month, rather than at 6month intervals as at present. The measurement of
short-term changes is a critical requirement for the
Consumer Price Index. The current rent system does
not provide an adequate measure of monthly change,
nor does it provide for time intervals of a few months
between changes. The revised system will yield
accurate short-term changes while allowing for close
to a 50-percent reduction in sample size.
In addition, attention is being given to the de­
velopment of better methods for adjusting for
changes in the quality of the rental units priced. Cur­
rent plans call for incorporating the new rent sample
and collection techniques into the ongoing Con­

Substantial gains in the accuracy of the index
should result from significant improvements in both
the sample and survey design of the Consumer Ex­
penditure Survey, such as improved stratification,
lower nonresponse rate, reduced length of recall
period, and improved estimating procedures. Also, a
substantially larger sample of items will be selected
for pricing. (Approximately 400 items are priced
in compiling the CPI each month.)
The Consumer Expenditure Survey is expected to
provide more accurate and more complete data than
have previously been available, and thus a sounder
basis for the selection and weighting of items in the
market basket. As with the last revision, data from
the Consumer Expenditure Survey will be used to
select a stratified probability sample of items (the
market basket) within the universe of items classified
into expenditure classes.
Information on such items as clothing, utilities,
and small household appliances is expected to be
more accurate because of the shorter recall period
resulting from the change to a quarterly survey.
The diary survey will provide greater detail on items
ordinarily purchased on a daily or weekly basis,
such as food and beverages, and on personal care
products, not otherwise covered in the survey.
Point-of-Purchase Survey— where people buy. Pric­
ing of items included in the Consumer Price Index
takes place in outlets selected to be as representative
as feasible of types and sizes of places where urban
wage earners and clerical workers shop. The Pointof-Purchase Survey, now underway, will provide
additional data on the retail stores, mail order
houses, bowling alleys, doctors’ offices, and other
places where goods and services are bought. Ap­
proximately 20,000 families are being asked where
they purchased various types of goods and services.
From the survey results, for the first time, a full
probability sample of retail stores and other outlets
to be used in collecting data for the monthly index
will be developed. Optimization principles are being
used to assure proper balance between the number
of outlets and the number of price quotations col­
lected in each. Here again the Bureau of the Census
serves as collection agent, under contract with BLS.
Rent survey. Still another survey is underway, to
provide more accurate and current data for the rent
index. Under the present system, change in rents is
measured from large samples of rental units which
include the same units at successive periods. No




191

sumer Price Index some time before completion of
the entire revision program.

study projects. These data are expected to be avail­
able beginning in 1976.

Output of the surveys. In addition to its application
to the expenditure weights and the market basket,
information from the Consumer Expenditure Survey
will be analyzed and published in a number of other
formats. Comparisons of the changing expenditure,
savings, and income patterns which have occurred
since the last such survey (in 1960-61) will provide
a wealth of material for sociologists, urban planners,
and other economic and manpower policymakers.
These will include analyses of the differences in levels
of living among various demographic groups using
characteristics such as family income, family size, age
of family head, occupation of head, and so forth.
The quarterly Consumer Expenditure Survey will
provide data similar to that obtained in previous
such surveys, though in some cases in greater detail
and of greater applicability. Information on clothing,
for example, will be collected with great specificity—
Coats: heavy-weight coats, light-weight coats, snowski suits, all-weather coats, plastic raincoats, and
other coats— and will carry an age-sex code for pur­
chases for members of the household as well as for
gifts purchased for persons outside the household.
Major household equipment items will be identi­
fied as to whether they were purchased new or old,
whether they were purchased for own use, received
as a gift, included with dwelling, rented, or pur­
chased as a gift to others. Purchases of this type will
also carry a code to indicate whether the item was
bought for cash, on 30-day credit, installment credit,
or other credit.
The diary survey uses a complex system of more
than 1,700 commodity codes. Examples of the level
of detail provided by this coding structure are:

Sample of cities

Improvements are being planned for the sample
of Standard Metropolitan Statistical Areas (desig­
nated by the names of their central cities). The
present sample of “cities,” which numbers 56, was
selected on the basis of the 1960 Census of Popula­
tion using probability methods. It was designed to
represent the entire urban portion of the country.16
For the revised index, prices will be collected in
85 areas, with the area selection based on the 1970
Census of Population. The 85-area design lends
itself to further expansion to at least 156 areas,
if needed. Of the 85 areas, 28 are self-representing
and 57 are representative of the balance of the
SMSA’s and the remainder of the urban population.
The increase in the number of areas to be sam­
pled will make it possible not only to improve the
reliability of the national Consumer Price Index and
the indexes recently introduced for different regions
of the country and for urban areas classified by size
of population, but also to provide, for the first time,
regional indexes for cities of different populationsize classes. Monthly or quarterly indexes will be
published for 28 cities in 1977, compared with 23
at present.
Population coverage

One of the major problems related to the current
revision program has been to determine just who
should make up the index population. Historically,
the index has been oriented toward the urban
worker. As the characteristics of the urban wageearner family have changed over the years, this fact
has been reflected both in the titles of the index and
in the index structure.
In earlier periods, wage earners and clerical work­
ers could be characterized realistically as being of
“low income.” Clerical and salesworkers were
identified as “lower salaried” employees, and the
index was referred to as one for “low and moderate
income” families. These were renters primarily,
living in the more densely populated city centers, and
including relatively more of the older established
households and larger families.
The large increase since World War II in the
size of the middle-income group and population

M ilk —buttermilk, chocolate, condensed, evaporated,

imitation, malted, powdered, or skim-whole;
B eefsteak —chuck, rib round, sirloin, T-bone, or other

steak.
These food products will also include a net weight
or volume per unit identification. (The quantity
code may appear only on the tapes for public use
and not in published data.)
In addition to BLS publication of these data,
computerized data of extraordinary detail and spec­
ificity will be available on public use tapes to
econometricians and other researchers from outside
the Bureau of Labor Statistics for their individual




192

movement to the suburbs reflected to a large degree
the improving economic status of the “worker” in­
cluded in the index population. Expenditures are
by and large based on income, and the large increase
in the number of two-earner families has raised many
wage-earner and clerical-worker families into the
middle-income group. Also, the shift toward a serv­
ice economy and the increasing unionization of
salaried white-collar employees has caused the oc­
cupational classification of “wage earner and clerical
worker” to lose much of its significance, because of
the similarity in the manner of living of this group
and that of the total urban population in the middleincome range. As a result of these demographic and
economic changes, questions were raised about the
coverage of the index.
More than a decade ago, the Stigler Committee
recommended that an index with broad population
coverage be developed.17
A more comprehensive index for the entire popu­
lation, not only the wage and salary earners, should
be made. . . .
From the viewpoint of general public policy and
scientific study, our basic need is for a comprehensive
Consumer Price Index . . . that is appropriate to the
measurement of the changes in welfare of the Nation
and to the measurement of inflation (and hence the
guidance of monetary and fiscal policy).18
Along with this recommendation, the committee
stated that “a price index comparable to the present
CPI, suitable to the current wage escalation clauses,
should be maintained for several years even if an
extensive revision of the scope of the index is under­
taken by the Bureau of Labor Statistics.”
During 1961 appropriations hearings, Ewan
Clague, then Commissioner of Labor Statistics, made
a point of the Bureau’s plans for extending the scope
of the price indexes to cover single-person families
and “further extensions which may eventually expand
the index to represent all nonfarm families.” In May
1966, then Commissioner Arthur M. Ross, testifying
before the Senate Subcommittee on Economic Statis­
tics, indicated an urgent need to expand the Con­
sumer Price Index to represent purchases by all con­
sumers and all retail sales,19 as well as the need for
separate indexes to be compiled within this frame­
work.
The idea of broadening the population coverage
of the index was introduced early in the current
revision program. In May 1970 the question was
discussed with the Research Advisory Councils
(from business and from labor) that regularly meet




with the Bureau of Labor Statistics. In a Review
article in March 1972, discussing Bureau programs,
then Commissioner Geoffrey H. Moore wrote:
In the past the index has reflected expenditures only
for urban wage earners and clerical workers, but con­
sideration is being given to broadening this base by
expanding the coverage to include other types of
workers or retired persons.20
Views on the coverage issue

In April 1974 the Bureau of Labor Statistics an­
nounced its intention to broaden the coverage of the
Consumer Price Index to include all urban house­
holds beginning in April 1977. The index limited to
urban wage earners and clerical workers would have
to be discontinued because of time and cost con­
straints.21 This announcement stirred up a lively de­
bate and led to the surfacing of the many different
points of view on this issue. To a large extent, the
interest in the issue was prompted by the recent high
rate of inflation and the great increase in the use of
the index as an escalator for many different types of
income payments.
In discussion of background papers prepared by
the Bureau of Labor Statistics for its Research Ad­
visory Councils, it became clear that there was gen­
eral agreement among council members that broad­
ening the population coverage of the Consumer Price
Index would be acceptable. However, union spokes­
men on the Labor Research Advisory Council urged
that the Bureau reconsider its plans to discontinue
the urban wage-earner index.
The controversy was brought to public attention
by AFL-CIO President George Meany and United
Auto Workers President Leonard Woodcock. Mr.
Meany stated his opposition to the dropping of the
wage-earner and clerical-worker index in a letter to
Secretary of Labor Peter Brennan, which was widely
reported in the press. Mr. Woodcock, in testimony
before a subcommittee of the Joint Economic Com­
mittee, argued forcefully for continuation of the
wage-earner index:
Trade unions have a vital interest in the CPI as it
currently stands. It is absolutely essential for effective,
responsible, and rational collective bargaining that
there be available a consistent and reliable index re­
flecting changes in the cost of living actually experi­
enced by working people. . . .
. . . We have had nearly 30 years of experience
dealing with [the Consumer Price Index]; we under­
stand its strengths and weaknesses; we are familiar
with its behavior and we know how to incorporate it

193

responsibly into our contracts. . . . In principle we
are totally opposed to the abolition of a CPI geared
to workers in favor of one geared to nobody. . . .
In practice neither we nor anyone else have any con­
crete experience as to how this new index would
behave. However, there is a presumption that it would
record lower rates of inflation than the current OPI,
at least if prices continue to behave as they have done
in the last decade. This is because items whose prices
have generally been rising fas*: l are precisely those
which figure most prominently in the budgets of lower
income families.
The UAW, and the labor movement in general,
clearly recognize that there are purposes for which
the existing CPI is not suited. Certain macroeconomic
analyses require more general indicators along the
lines of the proposed All-Consumer Index. Other
purposes (for example, setting the appropriate level
for social security payments) require more specific
measures covering groups currently excluded from
the CPI. Such functions are legitimate and we would
support the creation of indexes appropriate to them.22

Cost-of-living adjustments are not generally con­
sidered by labor leaders to represent a real gain for
the worker in terms of the labor share of expendi­
tures. Escalator clauses— which in effect freeze
real wages— are merely a defensive factor to prevent
there being any loss in real income. Improvement in
the workers’ share must be negotiated in other forms
(in some contracts the escalator clause is tied in with
an annual improvement factor).24 Thus, labor lead­
ers felt it essential that this defensive tool not be
weakened by changing its effective coverage.
Many users, however, spoke out in favor of the
expansion of population coverage. For example, the
Interagency Subcommittee on Economic Statistics
(headed by Gary L. Seevers) of the Council on
Economic Policy (then headed by George P. Shultz)
expressed “general agreement that the family defini­
tion in the CPI should be as broad as possible. The
Subcommittee encouraged the Bureau of Labor Sta­
tistics to take steps to enlarge the definition in the
revised index.” 25
Senator William Proxmire, at appropriations hear­
ings, noted the value of both the current wage-earner
index and the comprehensive index with broader
coverage:

In hearings before another Senate subcommittee,
later in the month, union spokesmen stressed the
need for continuity of statistical data, especially in
an area such as labor-management negotiations
where faith in the reliability of that data is basic to
its acceptance as a tool in bargaining. Lazare Teper,
Director of Research for the International Ladies’
Garment Workers’ Union, pointed out that ‘‘Neither
workers nor management are likely to accept the
new set of figures just because it covers other groups
whose marketplace experience is different from
that of wage and salaried workers.” 28

You know how very concerned some people are be­
cause their escalation clauses are tied to it [the present
index for urban wage earners and clerical workers].
Fifty million people in this country, including 5 mil­
lion in organized labor, and a very large number of
government workers, and many others, have their

Population coverage in other countries
Although the coverage is not complete, it is very broad
scale in Austria, the United Kingdom, and Japan. Austria
covers all urban households. The United Kingdom con­
ducts a continuous expenditure survey o f all households
and publishes two indexes. One index excludes pensioners
and upper level income earners; this refers to about ninetenths o f the population. The United Kingdom also pub­
lishes quarterly indexes for pensioner households. Japan
excludes only agricultural and single-person households.
Canada’s index includes urban, middle-income families
ranging in size from two to six persons, living in metro­
politan areas with over 30,000 inhabitants. Middle income
is considered to be $4,000 to $12,000 as o f the base year
1967.
France publishes a monthly index representing house­
holds headed by urban wage earners or non-supervisory
salaried employees. Their index excludes upper-income
salaried workers.

In other countries that publish consumer price indexes,
the most common variables used to determine whether
or not a family should be included in the weights o f an
index are (1 ) location— urban or rural; (2 ) source of
income; (3 ) income level; and (4 ) family structure or
size. The definitions used range from total population to
quite narrow definitions, but in most industrial countries
coverage is broader than it is in the United States at
present.
Among the countries with complete coverage o f all
consumers are the Federal Republic of^Germany, Bel­
gium, Denmark, Norway, Sweden, Italy, and the Nether­
lands. The Federal Republic of Germany regularly pub­
lishes indexes for three subclasses in addition to its
overall index.
Italy and the Netherlands publish two
indexes, one for the total population and one for non­
farm wage earners and salaried workers. Both countries
utilize a maximum income cutoff.




194

com pensation directly tied to this index. T hey know
it, they have faith in it, and they feel it relates to
their actual incom e. A nd I think you are very w ise,
a n d the governm ent is w ise, in deciding that they
s h o u ld have this new index to m ake it m ore com pre­
hensive, and cover 80 percent o f the people— and
have a m uch more representative index o f the cost
o f living.26

At its April 24, 1974, meeting the BLS Business
Research Advisory Council passed a resolution in
“support [of] the efforts of the Commissioner to ex­
pand the population coverage of the revised Con­
sumer Price Index,” and laid the ground for further
exploration of the desirability of maintaining more
than one index.27

geneous group; it is made up of many individuals,
each with an individual way of life. If a price index
were calculated for each of the individuals in this
group, some of these indexes would rise more rapidly
than others. So even under the current wage-earner
and clerical-worker index, there are some covered
individuals who gain when wages are escalated by
the Consumer Price Index, and some who lose, in
relation to their actual expenditures.28
New BLS plan— two indexes in 1977

On May 24, 1974, the Bureau of Labor Statistics
announced a decision to issue two indexes starting
in April 1977— an updated version of the current
Consumer Price Index for Urban Wage Earners and
Clerical Workers, and a broader Consumer Price
Index for All Urban Households.
Both indexes will incorporate improvements be­
ing developed as part of the revision program— for
instance, it is anticipated that both will be produced
with smaller measurement errors than the present
index, and that the full array of city and other detail
will be included in both indexes.29 In addition, an
evaluation plan will be built into the program.
Both indexes will be calculated and published for
at least 3 years, 1977-80. During that period, the
comparative movements of these two indexes and
their components will be studied. Results of these
studies will be discussed periodically with the Re­
search Advisory Councils, as well as with Adminis­
tration officials, the Congress, and professional eco­
nomic and statistical groups. Finally, a determination
will be made as to whether one index is adequate, or
whether both and perhaps an index representing the
difference between them is needed, or whether a
whole family of indexes best meets the demands
placed upon the CPI index program.
The comprehensive index will cover all urban
households in Standard Metropolitan Statistical
Areas. Some of these include rural areas as well as
cities and suburbs. Nonfarm families living in these
rural areas within SMSA’s will be included, but the
index will exclude other rural families and the mili­
tary and institutional population. The result will be
to increase the population coverage to about 80
percent of the total noninstitutional population (from
the past coverage of under 45 percent). Other
rural families make up about 18 percent of the total
and military personnel about 2 percent.
As in 1964, the change in population coverage
will change the average annual income of the index

A family of indexes

The determination of the target population for
the Consumer Price Index must recognize the major
uses of the index— the traditional as well as the
recent ones. In theory, one way of satisfying this
need is through a family of indexes. In this approach,
indexes would be developed that represent not only
the price experience of all consumer units, but also
the separate experience of particular subgroups of
the population, such as wage earners and clerical
workers, the aged, the poor, and the rural popula­
tion. In practice, production of such a family of
indexes does not appear feasible as part of the 1977
revision program, given current constraints of time
and funds. BLS will be studying the prospects for
such indexes as time and resources become available.
A family of CPI indexes would be roughly analo­
gous to the complex of unemployment data published
by the Bureau of Labor Statistics. There, in addition
to the total unemployment rate, numerous compo­
nents are shown— for example, the unemployment
rates for household heads, for adult men, for women,
for Negroes and other races, for veterans. In order to
show this large variety of data, the total size of the
sample must be large enough that the figures for each
component are reliable. (The Current Population
Survey sample used to obtain the unemployment
data now includes 47,000 households monthly, com­
pared with 15,000 retail outlets in the current CPI
sample.) Technically and operationally, a similar
program could be developed for the Consumer Price
Index; it is a matter of time and money, on the one
hand, and the usefulness of the additional output, on
the other.
Even the urban wage-earner and clerical-worker
segment of the population is not a completely homo­




195

than an index for wage earners and clerical workers
alone. Some students of the index speculate that
movement of the comprehensive index would closely
parallel that of the urban wage-earner index. But no
one can speak authoritatively on this until there is
empirical evidence.

population. Rough estimates based on 1971 data
from the Current Population Survey indicate that
broadening the population coverage to all urban
households would lower the mean income of the
index population from about $10,500 to $10,100.
Although the other workers (professional and selfemployed, for instance) added to the covered popula­
tion had 1971 average annual incomes higher than
the wage-earner and clerical-worker group, incomes
of the unemployed and those not in the labor force,
who will also be included in the index, were markedly
lower. The following tabulation shows total money
income in 1971 for consumer units (families and
unrelated individuals):
In
C onsum er units headed by

All of above groups...........

In

1960-61

1971

dollars

dollars

Wage earners and clerical workers. $ 7,745
Other salaried and self-employed
workers ...................................... 11,803
5,544
Unemployed persons.....................
Persons not in the labor force . . . .
3,760
7,396

The issue depends on more than just the weights
assigned to various items— it depends also on the
items priced and the kinds of outlets sampled. Some
people have argued, for example, that the prices of
lobster and imported caviar have risen much more
rapidly than have the prices of other, more prosaic
food items; others have noted that prices of some
very low-cost items, not now priced, have also risen
more than the average. This implies that prices of
goods purchased by groups not covered by the
present index— professional workers, the unem­
ployed, retired persons— have risen more than aver­
age. But in fact we know very little about differences
in the movements of price indexes which might be
constructed for different population groups.

$10,539
16,062
7,544
5,116
10,064

Other conceptual problems

The difference in 1971 average annual income be­
tween the urban wage-earner and clerical-worker
group and the all-urban-households group was $475
in 1971 dollars and $349 in 1960-61 dollars. As
noted above, a similar effect (of lowering the aver­
age family income of the population group covered
by the index) occurred in the previous revision, when
inclusion of single urban wage earners and clerical
workers lowered the income of the index population
group during 1960 and 1961 by $267 (in 1960-61
dollars).
To produce separate indexes for wage earners
and clerical workers and for all urban households
will increase the costs in terms of both the revision
program and the ongoing program after 1977. How­
ever, the increase should be relatively modest. The
two indexes will be of high quality and both are
planned to be as good as or better than the present
index.
No one today can tell which components of the
index— foods, fuels, services— are likely to be rising
most rapidly in the future. Thus in the 1960’s, food
price increases averaged 2.7 percent a year, fuels
and utilities 1.0 percent, and services 3.5 percent a
year. From 1972 to 1973, foods rose 20.1 percent,
fuels and utilities 11.5 percent, and services 6.2
percent. Nor can anyone say whether an all-urbanhouseholds index would rise more or less rapidly




A number of other basic conceptual issues remain,
on which there is also considerable controversy. Two
of the most vexatious are briefly described below.
Housing. The treatment of owner-occupied housing
presents a two-tiered problem. At the first level, a
decision must be made as to the concept under
which housing is to be priced. After that decision
is made, a second is required on the most accurate
and most efficient way of measuring prices and price
changes under that concept.
Up to the present time, the price of the house
itself has been used. For other index items, a loaf of
bread, for instance, purchase of the bread implies
consumption of the bread within that month.
The problem with housing— and with all durable
goods— is that the purchase of a house is not the
same as consumption. In effect, the index treatment
of housing has said that those individuals who pur­
chase a house this year consume the total pur­
chase price, as well as total financing costs, this year.
And those individuals living in previously pur­
chased houses spend nothing on housing in this
year. In other words, the entire “consumption” of
the purchase price plus financing costs is attributed
to the year of purchase.
196

Another way of looking at it is that what is really
being “consumed” by the owner living in a house
are housing services— that is. shelter, and accommo­
dation for food preparation and consumption,

recreation, entertaining, laundry, and so forth. Ob­
viously, the owner does not consume all these serv­
ices in the year of purchase, but continues to con­
sume them over the years of living in that house.

Some questions for BLS raised by the “indexing” proposal
more rapidly than the broadly based index. Then tying
wage escalation agreements for this subgroup to the all­
consumers CPI would result in income payments for this
group that are smaller than would be the case if their
incomes were moved by their own index. This loss would
be offset, in the aggregate, by the fact that the other
groups would receive larger income increases than those
which would be triggered by their own CPI, though this
would provide little solace for the groups that lose. The
loss by one group would be offset by the gains o f the
other, so use of a broadly based measure would result in
the appropriate aggregate income adjustment— though it
would also involve a shift in income shares. These hypo­
thetical examples do not, of course, take into account the
dynamic effects of the indexing which add to the com­
plexities.
Thus far we have assumed that all income payments
are escalated by the CPI. But what if some are and
some are not, as is, in fact, the situation in the U.S.
today. And, especially, how would those whose income
payments are not pegged to the CPI fare in an economy
where most income payments are pegged. The unesca­
lated groups may very well be starting off with a handi­
cap in obtaining their income share— unlike the others,
they would not have any automatic increases nor a floor
to their income payment increases.
Events may lead us to a statistical program in which
indexes are developed that represent not only the price
experience of all consumer units, but also the separate
experience of many subgroups of the population. The
existence of multiple indexes would create uncertainties
in the minds of many groups regarding the particular
index to which it would be most appropriate to tie their
own income payments. Expansion in the number of CPI
indexes would, however, only complicate problems that
already exist because of the availability o f city and com­
modity indexes. In a recent contract, the wages of New
York transit workers were tied to the CPI for New
York-Northeastern New Jersey area, rather than the
national index. The Food Stamp Allotment program is
escalated on the basis of average price data from the
food-at-home component series o f the CPI and the chil­
dren’s lunch program by the food-away-from-home com­
ponent. A degree of familiarity with the statistical
methods used in compiling the CPI far above what exists
today will be required for effective use of a multiple index
approach.
A single all-consumer-units index would probably be
easiest to administer, but it will be hard to convince
groups who think their cost of living will rise more
rapidly than the average that this is the best route to take.

The possibility of “indexing” the U.S. economy was
recently brought to public attention by Milton Friedman
of the University of Chicago. Under such a system— the
most notable current example is the one in use in Brazil
— when the CPI rises, so do not only salaries and wages,
but also the tax structure, rents, interest rates, and other
items. Thus the objective is to keep all or at least most
of the economy in step, by reference to the Consumer
Price Index.
The basic question is, of course, what effect would
indexing have on inflation and, in turn, on unemployment,
real economic growth, distribution of real income, and
other economic conditions. The debate on this basic
question is just getting under way.
Answers are lacking to some troublesome questions
that arise from the increasing use of the Consumer Price
Index as a basis for escalating income for an increasingly
large proportion of the population. What is the effect if
an inaccurate index is used, or if an index is used which
represents a portion of the population whose costs are
rising either slower or faster than the average for the
country as a whole? What is the impact of escalation by
a single index upon groups which experience changes in
living costs different from the average? Upon those groups
in the population whose incomes are not escalated? What
additional requirements would indexing put upon the
accuracy of price indexes?
Beyond the effect on income payments, what is the
impact of an incorrect or inappropriate index on statis­
tics on real economic growth— especially real personal
consumption expenditures and real retail sales, which are
deflated in part or in whole by the CPI? If an inappro­
priate or inaccurate index is used as a deflator, measures
of real economic growth will be correspondingly off the
mark. What distortions appear in income distribution
data when the same CPI is used to deflate the incomes of
all classes?
Let us consider some of the possible economic impli­
cations of a situation in which the economy is indexed
and there is only one CPI, with coverage limited to a
subgroup of the population. Let us assume that this price
index rises more rapidly than an index which covered the
other segments of the population. Under this assumption,
these additional groups would be getting a CPI adjust­
ment exceeding that which they would receive if their
incomes were escalated by their own index. This would
give them a greater than warranted increase in money
income and, in this way, the measure of price change
could become a source of inflation in itself.
On the other hand, suppose that the single index
covered all consumer units, and suppose that prices for
an important subgroup of the population were rising




197

(In the same way, a renter consumes housing serv­
ices during the time of residence in a rented house
or apartment.)
If a decision is made to price the flow of housing
services, the problem will be to develop a technique
for estimating the price of owner-occupied housing.
There are two methods which can be used. The first
is to use a rental equivalence technique— in effect,
measure what you would charge if you rented the
house to yourself in an assumed arms-length transac­
tion. The second is to establish a user-cost function
for the provision of housing services— that is, to
measure the major cost components that an owner
incurs in providing himself housing. These would in­
clude mortgage and equity financing costs, mainte­
nance costs, taxes, and the variety of other expenses
that go into providing housing services.
Both approaches present considerable data prob­
lems. The rental equivalence approach requires the
development of a sample of rental units which can
provide an adequate measure of the changes in
owner-occupied housing costs. Another aspect that
must be considered is the increasing share of owneroccupied apartments and townhouses in condomin­
ium developments. It is difficult to construct a good
sample for this purpose since housing units which are
typically rented differ in various ways from those
which are normally owner-occupied. For example,
owner-occupied houses are often located in areas
where there is very little rental housing such as,
for instance, suburban developments.
Implementation of this pricing technique does not
require that the average owner-occupied house be
equal to the average rental house, but only that there
is enough overlap to pick from the sample of rental
housing the houses which are similar in their most
important aspects to those that are owner-occupied.
We must determine whether there is sufficient over­
lap between the distribution of rental single-family
housing and the distribution of owner-occupied
housing.
There are measurement problems associated with
the user cost approach also. First of all, this ap­
proach requires a source of house prices. The cur­
rent CPI obtains price data from the Federal Hous­
ing Administration on FHA-insured houses. But,
these houses represent a small and unrepresentative
segment of the market. Similarly, the user cost ap­
proach must take into account in some way the
capital gains which arise from appreciating home
values. In addition, since the same houses are not
sold in successive periods, it would be necessary to




develop methods for factoring quality change out
of the house price data collected.
Some of these same problems of data collection
also exist with the current method of pricing housing
costs. The problem of quality change is a particularly
difficult one. Also, as pointed out above, data from
the Federal Housing Authority on prices for new and
existing housing purchased under FHA commitment
have serious limitations for use in the Consumer Price
Index. These FHA-guaranteed purchases represented
only about 6 percent of the home purchase market
in 1973. In addition, there are considerable differ­
ences between the typical house financed under the
FHA program and those financed under conven­
tional mortgages and thus the FHA sample may not
be representative of all houses sold.
In the current method, prices of houses, classified
by age and size, are converted to price per square
foot. This is reflected in the index by a 3-month mov­
ing average, to eliminate erratic fluctuations in each
month’s data.
Investigations have been made into the availability
of data from other sources, such as lending institu­
tions and real estate associations, as well as the cen­
sus series on new housing prices. They have not,
as yet, uncovered data that would be useful for the
CPI.
Quality change. The Bureau of Labor Statistics is
also investigating new methods to improve the hanComparison of costs and income effect
T o put the costs o f the revision program into per­
spective as they relate to the amounts affected by
changes in the Consumer Price Index, let us assume
that all escalators we know o f today had been in ef­
fect at the beginning o f 1974, when the index had
risen by some 10 percent over the early part o f 1973.
Under that assumption, income payments would have
been increased by at least $10 billion.
If that figure is used, the cost o f preparing the
monthly Consumer Price Index, including the author­
ized cost of updating and revising the index decen­
nially, works out to something like 70 cents for each
$1,000 increase in payments as a result o f escalation
alone. If two indexes (the present one plus the more
comprehensive one covering all urban households)
had been calculated during this period, the cost per
$1,000 increase in escalated income payments would
have been between 85 and 90 cents. Of course, if in­
flation rates are lower, as has usually been the case,
these cost figures would be higher. However, they do
not take into acount the important uses o f the CPI
as a measure of inflation.

198

dling of quality change. Quality change is one of the
most difficult problems faced in compiling a price
index, since both products and consumption patterns
are constantly changing. An example familiar to
many consumers is that of passenger automobiles,
where— with each model change— the Bureau faces
the problem of separating out the actual price rise
from the changes in quality, some of the latter
necessitated by statute (such as emission control
and safety belt legislation)
Frequently a model currently being priced must
be replaced, either because it is discontinued or
because consumption patterns have changed to such
an extent that the model no longer accurately reflects
consumer spending patterns. The value of the quality
change in the new model should not be reflected as
a price change, since the goal of the index is to
measure the cost to consumers of purchasing a con­
stant market basket of goods and services of con­
stant quality through time. Ideally, estimates would
be obtained for the value of each change in quality
that occurred as a result of a change in the model or
item priced; and this estimate would be based on
the consumers’ valuation of a change in quality,
rather than that of the producer.
At present, most changes in quality are handled
in one of two ways:1
1. The quality change is deem ed to be minor, and
any price change that occurred sim ultaneously is
reflected in the index just as if there w ere no quality
change— the prices are com pared directly; or
2. The quality change is judged to be significant,
and the sim ultaneous price change is assumed to be
an accurate measure o f the value o f the quality
change— no price change is reflected in the index.

Since in most cases a large number of different
models are priced for each item (such as a console
color television set), and since sales of these models
are not discontinued at the same time in various
stores, the problem created by the above procedure
can easily be overemphasized. However, it is true
that many times price changes do occur simul­
taneously with model changes and that in order to
have an accurate measure of price change (holding
quality constant), an estimate of the consumers’
valuation of the quality change is needed to separate
out the price and quality components of the price
change. The problem is even more difficult when
quality changes and there is no price change. For
some important items in the index, producers’ cost
estimates of quality changes are being used for this
purpose. Although this eliminates the “all or noth­




ing” nature of the usual procedure, there is still no
reason to assume that the producers’ estimates will
reflect accurately the consumers’ valuation of the
changes.
The Stigler Committee report recommended that
the Bureau investigate another approach to quality
measurement, one which does measure the consum­
ers’ valuation. For certain items such as houses,
cars, and major household appliances, there are at
any time a wide variety of models available, models
which possess a large number of different character­
istics. From cross-sections of data on retail prices
and characteristics of models, taken over a period
of several years, it is possible to estimate the con­
sumers’ marginal valuation of quality change— that
is, what the consumer is willing to pay for the addi­
tion of a particular characteristic, such as a meatkeeper in a refrigerator. This is done using standard
statistical techniques. These marginal prices for the
characteristics can then be used as estimates in the
current index of the value of changes in quality.
Research on using this approach to measuring
quality change is currently underway in the Bureau.
The technique has been applied to data for rental
housing, automobiles, and refrigerator-freezers. How­
ever, results are preliminary and, as yet, new price
indexes have not been computed using the implicit
prices yielded from the research.
There is great interest in whether quality change
results in any bias in the Consumer Price Index.
A recent article in the Monthly Labor Review
pointed out:
M any econom ists believe that quality changes in
goods and services are not adequately taken into
account in the preparation o f the C onsum er Price
Index ( c p i ) . As a result, they believe, the cpi has a
system atic and persistent upward drift w hich makes
the index a questionable indicator o f the course o f
inflationary price m ovem ents.
T o what extent is the belief that price indexes are
biased upward borne out by existing evidence? N o
assessment o f the quality error in the c p i as a w hole
has yet been made, but a num ber o f investigations
have produced estim ates o f quality error in individual
index com ponents. The present article is a survey o f
existing studies, w hich present contradictory evidence.
Some investigators found upward bias, but others
reported that quality error might be negative— that is,
when the b l s failed to correct adequately for quality
changes, it resulted in a price index that rose too
slowly, rather than too rapidly.

After reviewing key studies in the field, the con­
clusion reached was that there is no conclusive
evidence which indicates a particular bias in the

199

Consumer Price Index due to quality change:
. .
we have not proved that price indexes are biased
either upward or downward; rather, they establish
only that the proposition that indexes are systematic­
ally upward-biased is not conclusively confirmed by
the available evidence.” 30

cennial programs to smaller decennial or quinquen­
nial programs supplemented by annual sample sur­
veys. The President’s budget for fiscal year 1975
includes funds to plan such a shift in the decennial
revision of the Consumer Price Index. An ongoing
quarterly Consumer Expenditure Survey would pro­
vide more timeliness and greater flexibility, at
roughly the same cost over a 10-year period. The
ongoing Consumer Expenditure Survey could also
have the advantage that numerous analytical studies
could be made on a current basis, including prompt
information of the effect of the rise in food and fuel
prices upon spending patterns. Further, a continuing
consumer expenditure survey could, after a break-in
period, be tabulated rapidly, so that shifts in spend­
ing patterns, market baskets, and retail stores sam­
ples could be analyzed and information could be
provided on the need for more frequent updating
of the Consumer Price Index. New market baskets
and new retail store samples could be phased in
more often, say once in 5 years rather than once in
10, but this would not be a necessary part of the new
approach.
□

Future updating and revision programs
Perhaps even more significant than the immediate
problems of this current index revision is the ques­
tion of long-run improvement in the revision process.
BLS records show that the 1950-52 revision took 3
years and cost $4 million; the 1960-64 revision took
5 years and cost $6.5 million; and present estimates
are that the current revision will take 8 years and
cost more than $40 million (after adding the cost
of the second index). The endless delays in issuing
the results and the ever-rising costs suggest that a
better method of updating the CPI and making revi­
sions must be found.
Over the past decade, statistical agencies over the
world have been shifting away from large-scale de-

* Changes in C ost o f L ivin g in Large C ities in the U nited
States, 19 1 3 -4 1 , B u lletin 6 9 9 (B u reau o f L abor Statistics,

1 F o r a m o re co m p reh en siv e h istory o f the C o n su m er
P rice In d ex , a lo n g w ith a d escrip tion o f the 1964 revisio n
and d eta iled d escrip tio n o f tech n iq u es, see The C onsum er
Price Index: H istory and Techniques, B u lletin 1517 (B u reau
o f L ab or S tatistics, 1 9 6 6 ). S ee a lso Prices, Escalation, and
E conom ic Stability (B u rea u o f L abor Statistics, 1 9 7 1 ).
2 S ee, fo r ex a m p le, stu d ies o f fa m ily exp en d itu res co v er ­
ing the y ears 1 8 8 8 -9 0 in the Annual R eport o f the C o m ­
m issioner o f L a b o r in 1890 and 1891, and the im p o sin g c o l­
lec tio n o f w h o le sa le p rice d ata in clu d ed in the “A ld rich
R ep o rts” by the S en ate C o m m ittee on F in an ce in 1892 and
1893.
* T h e term “co st o f liv in g ” w as used to describ e th e B u ­
rea u ’s in d ex u n til its n am e w as ch an ged fo llo w in g c o n tro ­
versy in th e W o rld W ar II p eriod over the in d ex ’s v a lid ity
as a m easu re o f co st o f livin g. It has alw ays b een m erely a
m ea su re o f ch a n g es in p rices fo r go o d s and services pu r­
ch a sed fo r fa m ily livin g.
‘ “L abor and the W ar: A d ju stm en t o f S h ip b u ild in g D is ­
pu tes o n the P acific C o a st,” M on th ly R eview o f the U.S. Bu­
reau o f L abor Statistics, M arch 1918, pp. 6 7 - 7 6 .
0 C ost o f living in the U n ited States, B u lletin 357 (B u rea u

1 9 4 1 ).
10 “B L S C o st o f L ivin g In d ex in W artim e,” M on th ly L abor
R eview , Ju ly 1943, pp. 8 2 - 9 5 , and C onsum ers’ P rices in
the U n ited States, 1942—48, B u lletin 9 6 6 (B u reau o f L abor
Statistics, 1 9 4 9 ).
11 “R ev isio n

12 Interim A dju stm en t o f C onsum ers’ Price Index, B u lletin
1039 (B u reau o f L abor S tatistics, 1 9 5 2 ).
“ C onsum er P rices in the U nited States, Price Trends and
Indexes, 1 9 53-58, B u lletin 1256 (B u reau o f L ab or Statistics,
1 9 5 9 ).
14 A t the b eh est o f certain labor groups, the old in d ex w as
co n tin u ed fo r an ad d itio n a l 6 m on th s. T h is a ctio n prom p ted
so m e parts o f the p ro fe ssio n a l c o m m u n ity to charge that
p o litica l ju d gm en t w as b ein g substitu ted fo r scien tific d e­
cisio n m a k in g in the statistical field. See, fo r e x a m p le, the
rep ort o f the T ech n ica l C o m m ittee o f the A m erica n Sta­
tistical A sso c ia tio n , ap p oin ted b y A S A P resid en t S im on
K u zn ets in the su m m er o f 1949 to ad vise th e B ureau o f
L ab or S tatistics o n P rice In d ex N u m b er R evision s. T h e
co m m ittee, ch aired by B ruce D . M u d gett, h eld its last m e et­
ing June 3 0 , 1953. O th er m em b ers w ere D u d le y C ow d en ,
R eavis C o x , and S o lo m o n F abricant.
T h e rep ort stated:

o f L a b o r S tatistics, 1 9 2 4 ).
• C h a n g e s in C o st o f L iv in g F rom Sep tem b er 15 to N o ­
v em b er 15, 1 9 4 0 ,” M on th ly L a b o r R eview , January 1941,
p . 146.
7 “R ev isio n o f In d ex o f C o st o f G o o d s P urch ased b y W age
E arn ers and L o w er S alaried W ork ers,” M on th ly L a b o r R e ­
view , S ep tem b er 1 9 3 5 , pp. 8 1 9 - 3 7 .
8 R etail P rices o f F ood, 192 3 -3 6 , B u lletin 635 (B u reau o f

A g o v ern m en tal b u reau can attract scien tific p erso n n el
o f the h igh est co m p e te n c e o n ly if it creates w ork in g c o n ­
d itio n s th at assure th e u n fettered pu rsu it o f their w ork .

L ab or S tatistics, 1 9 3 8 ).




o f th e C o n su m ers’ P rice In d e x ,” M on th ly

L abor R eview , July 1950, pp. 1 2 9 -3 2 .

200

B y the sam e tok en an y restriction u p on their freed o m
arising fro m the pressu res o f sp ecial interests w ill destroy
the v ery co n d itio n s that attract m en o f co m p eten ce, and
any bureau y ield in g to su ch pressu res m ay lo se n ot o n ly
its qualified w ork ers bu t also its rep u tation fo r ob jectivity
and fo r the m a in ten a n ce o f h igh standards o f scien tific
w orkm an ship . S h ou ld n ot the a ssociation alw ays stand
read y to su pp ort an y bu reau resistin g th ese pressures?
T h e se th o u g h ts w ere aroused by an in cid en t w h ich to o k
p la ce w h en it w a s an n ou n ced in January th at the revised
C o n su m ers’ P rice In d ex w o u ld d isp lace the old in d ex, and
that the o ld in d ex w o u ld be d iscon tin u ed . A t the urging
o f a nu m ber o f g rou p s w h ich have co lle ctiv e b argaining
agreem en ts w ith w age escalator cla u ses b ased o n th e old
in d ex , the P resid en t ask ed the D ep artm en t o f L ab or to
resum e c o m p ila tio n and p u b lica tio n o f the o ld in d ex
th rou gh June 3 0 o f 1953, and the D ep artm en t acceded
to this request. . . . T h a t th is m atter cou ld con stitu te a
case o f d an gerou s pressure w as recogn ized b y Secretary
[o f L abor] D u rk in , w h o ca u tio n ed that n o n e w con tracts
sh ou ld be based u p on the old in d ex, and b y C o m m issio n er
[o f L ab or Statistics] C lagu e, w h o urged th at all u sers
co n sid er the revival o f the old in d ex as p u rely tem p orary.
It w ill be reco g n ized a lso b y tech n ical w ork ers in this
field, and it b eco m es their d u ty to support th ese w arnings
and to ca ll a tten tion to the p o ssib ility th at this kind o f
s te p m a y b e th e first a lo n g th e d a n g e r o u s r o a d to w a r d
partisan co n tro l o f e co n o m ic m easu rem en t.
15 A sid e fro m gen era lly h igh er in co m e levels fo r o ccu p a ­
tions w ith in the sco p e o f the ind ex, an in com e lim ita tio n on
fa m ilies in clu d ed (a t a level o f $ 1 0 ,0 0 0 after taxes in 1 9 5 0 )
w as discarded b eca u se o f the high er in co m e p er fa m ily u n it,
resu ltin g fro m the in creased nu m ber o f fam ilies w ith m ore
th an o n e w orker, and greater p recision in the o ccu p ation al
cla ssifica tio n o f the su rvey.
,a B oth the 1 9 6 1 -6 2 C on su m er E xpend iture Survey and
su b seq u en t data c o lle ctio n are based on a 56-area p rob ­
ab ility sa m p le, o f w h ich 18 are self-rep resen tin g areas. T h e
balan ce are areas selected b y a stratified co n tro lled -selection
pro b a b ility proced ure to rep resent the balan ce o f urban
areas, classified b y region and size.
17 A co m m ittee ap p oin ted in 1959 by the N a tio n a l B ureau
o f E c o n o m ic R esearch , u n der con tract w ith the O ffice o f
S tatistical Stan dards o f th e B ureau o f the B ud get, and
ch aired by P ro fe sso r G eorge Stigler o f the U n iv ersity o f
C h ica g o . See The Price Statistics of the Federal Govern­
ment: Review, Appraisal, and Recommendations, H earin g
B efo re the S u b co m m ittee o n E co n o m ic S tatistics o f the
Joint E co n o m ic C o m m ittee, 87th C o n g ., 1st sess., 1961.
18 Hearings on Government Price Statistics, Pt. I. See a lso
Zvi G rilich es, “H ed o n ic P rice In d exes for A u to m o b ile s: A n
E co n o m etric A n a ly sis o f Q u ality C h a n g e,” in th at vo lu m e.
18 H arold S. T a y lo r,
N o te d ,” The New York

“W hat

P rice

D ata?

D isto rtio n s

Times, A u g . 21, 1966.

80 G eo ffrey H . M o o re and M axin e Stew art, “N e w d e v e lo p ­
m en ts in lab or sta tistics,” Monthly Labor Review, M arch
1972, pp. 3 - 1 3 .




21 Statem en t by Julius Sh isk in , C o m m issio n er o f L abor
Statistics, b efore the Senate S u b com m ittee o n P rod u ction
and S tab ilization , C o m m ittee on B ank ing, H o u sin g and
U rb an A ffairs, A p r. 2 3 , 1974.
22 S tatem en t o f L eonard W o o d co ck , P resident, U n ited
A u to m o b ile , A ero sp a ce and A gricu ltu ral Im p lem en t W ork ­
ers o f A m erica ( U A W ) , b efo re the S u b co m m ittee o n P ri­
orities and E c o n o m y in G o v ern m en t o f the Join t E co n o m ic
C o m m ittee, A p r. 5, 1974.
23 “T h e N e ed to P reserve the C urrent C overage o f the
C on su m er P rice In d ex for U rb an W age E arners and C ler­
ical W ork ers,” statem en t b y L azare T ep er, D irecto r o f R e­
search, In tern ation al L a d ies’ G arm en t W ork ers’ U n io n ,
A F L -C I O , to the S en ate S u b com m ittee o n P rod u ction and
S ta b ilization o f the C o m m ittee on B ank ing, H o u sin g and
U rb an A ffairs, A p r. 23, 1974.
28 F or oth er d iscu ssion s o f the use o f the escalator clau se in
c o lle ctiv e bargaining, see H en ry L ow en stern , “A d justin g
w ages to livin g costs: a h istorical n o te ,” and Jerom e M .
Staller and L oren M . S oln ick , “E ffect o f escalators on
w ages in con tracts exp irin g in 1 9 7 4 ,” pp. 21 and 27, in this
issu e.
25 R eport o f m eetin g o f Sept. 10, 1973. T h e su b com m ittee
is chaired by G ary L. S eevers, m em b er o f the C o u n cil o f
E c o n o m ic A d visers, and in clu d es rep resen tatives o f the
T reasury D ep artm en t, the O ffice o f M an agem en t and
B ud get, the F ed eral R eserve B oard , the C ost o f L ivin g C o u n ­
cil, and the D ep artm en ts o f A gricu ltu re, C om m erce, and
L abor.
“ Statem en t by Sen. W illia m P roxm ire ( D ., W is.) at
H earin gs o f Senate A p p rop riation s C om m ittee, A p r. 25,
1974.
27 M in u tes o f the B u sin ess R esearch A d v iso ry C o u n cil,
A p r. 2 4 , 1974.
28 M ost escalator cla u ses are tied to the n a tion al C P I, bu t
so m e attem p t to m ak e the relation sh ip m o re d irectly app li­
cab le to the w ork ers’ o w n exp erien ce b y u sin g the in d ex
fo r the area in w h ich th ey are lo cated or an oth er relevan t
“c ity .” F o r ex a m p le, in the spring o f 1974, a n ew 2-year
con tract b etw een the city o f S alem , O reg., and the Salem
P o lice A sso c ia tio n p rovid ed , in the co n tra ct’s secon d year,
a 5- to 9-p ercen t in crease d ep en d in g on th e in crease in the
January 1974 C P I for the G reater P ortlan d area. (T h e
P ortlan d , O reg.-W ash. S M S A is o n e o f the areas priced
ea ch 3 m on th . S ee table 2 7 , p. 1 1 0 .)
28 S in ce there w ill be tw o in d ex p o p u lation s (o n e o f urban
w age earners and clerical w orkers, and o n e o f all urban
h o u se h o ld s ), item s w ill be selected to be rep resen tative o f
ea ch o f these p o p u la tio n s. S elected item s m ay vary fro m
region to region and b etw een in d ex p o p u lation s, bu t p roba­
b ility sam p lin g p roced u res w ill be used to m axim ize the
overlap for efficien cy in c o lle ctio n . W ith in selected item s,
in general, the goal is to use an ob jectiv e prob ab ility p rocess
for the selectio n o f g o o d s “sp ecified in d e ta il,” in clu d in g
proper rep resen tation o f b o th b ig -v o lu m e and oth er good s.
“ Jack E. T rip lett, “D eterm in in g the effects o f q u ality
ch an ge on the C P I,” Monthly Labor Review. M ay 1971,
pp. 2 7 - 3 8 .

201

Measuring
changes
in industrial
prices

New study adds wealth of data
on transaction prices
for use in continuing BLS effort
to improve wholesale indexes
JOSEPH A. CLORETY, JR.

“ T he reliability of an index number obviously
depends upon the judgm ent and the accuracy
with which the original price quotations were
collected. This . . . work is not only fundamental,
it is also laborious, expensive, and perplexing
beyond any other part of the whole investiga­
tion. . . . To ju d ge from the literature about index
numbers, one would think that the difficult and
important problems concern methods of weighting
and averaging. But those who are practically
concerned with the whole process of making an
index number from start to finish rate this . . .
work lightly in comparison with the . . . work of
getting the original data.”

a u th o r s , G e o rg e J . S tig le r a n d J a m e s K . K in d a h l,
as p r a c tic a l in d e x m a k e r s . T h e m in o r c a v e a t rise s
f ro m th e im p o r t a n t d is tin c tio n b e tw e e n p ro b le m s
in v o lv e d in a o n e -tim e s t u d y a n d th o s e w h ic h
p la g u e th e m a k e r s o f a m o n th ly , c o n tin u in g in d e x .
M o r e im p o r ta n t, S tig le r a n d K in d a h l n o d o u b t
w ill s ti m u la te g o v e r n m e n ta l, p ro fe ssio n a l, a n d
p u b lic i n te r e s t in th e m a jo r p r o b le m o f tr a n s a c tio n
p ric e s. M o s t im p o r ta n t, th e w e a lth o f d a t a
p r e s e n te d p r o v id e s a m in e o f n o m e a n v a lu e fo r
r e s e a r c h e rs in a c a d e m ic , p r iv a te , a n d p u b lic
circ les. I n d e e d , th e B u r e a u ’s c o m m o d ity a n a ly s ts
re s p o n s ib le f o r c o m m o d itie s c o v e re d b y S tig le r
a n d K in d a h l a re s tu d y in g th e d a t a in d e ta il
p r im a r ily to in it ia t e c o r r e c tiv e a c tio n w h e re
in d ic a te d , fe a sib le , a n d n e c e s s a ry .
I n so m e in s ta n c e s , s u b s t a n tia l p r o g re s s in o b ­
ta in in g a c tu a l tr a n s a c t io n p ric e s h a s b e e n m a d e
sin c e 1966— th e la s t y e a r c o v e re d b y S tig le r a n d
K in d a h l. T h is re fle c ts th e v ir tu a lly d a ily a t t e n t i o n
to th e p ro b le m s of p ric e c o lle c tio n o u tlin e d b y
M itc h e ll, w h ic h la rg e ly c h a r a c te r iz e s n o t o n ly th e
c u r r e n t bls s ta ff b u t th e i r p re d e c e s s o rs sin c e
M itc h e ll’s d a y . A lth o u g h p e rs o n a l q u a litie s c o n ­
tr ib u te , th e B u r e a u ’s su c ce sses a n d fa ilu re s a re
la rg e ly a f u n c tio n of th e re s o u rc e s c o m m itte d to
a g iv e n in d e x . F o r e x a m p le , p r io r to W o r ld W a r I I
th e e n tir e s ta ff p r o d u c in g th e w p i w e re h o u s e d
r e a s o n a b ly c o m fo r ta b ly in o n e m e d iu m la rg e office,
w ith a v e r y sm a ll c u b ic le fo r th e ir ch ief.

Thus W esley C . Mitchell in his classic The M aking
and Using of Index Numbers,1 originally published
in 1915 and with minor modifications republished
in 1921, prefaced his discussion of price collection
problems.
C o n c ise ly a n d in c is iv e ly M itc h e ll lim n e d all of
th e m a jo r a n d m a n y of th e m in o r p r o b le m s : th e
m u lt ip lic ity o f p ric e s f o r a n y im p o r t a n t c o m m o d ­
i t y ; th e d iffic u ltie s in s e le c tin g a r e p r e s e n ta tiv e
s a m p le ; th e n e e d f o r q u a l ity a d j u s tm e n ts ; th e
n e e d to “ g u a r d a g a in s t th e p itf a lls of c a s h d is ­
c o u n ts , p r e m iu m s , r e b a te s , d e fe rre d p a y m e n ts , a n d
a llo w a n c e s o f all s o r ts ” ;2 a n d m a in te n a n c e of
c o m p a r a b ility . M itc h e ll w ro te a s a p r a c tic a l in d e x
m a k e r ; h e w a s th e p r in c ip a l a r c h ite c t o f th e
W h o le s a le P ric e I n d e x (w p i ) in th e g e n e ra l f o rm
i t h a s sin c e d e v e lo p e d .

The first major challenge

Publication of The Behavior of Industrial Prices 3
(reviewed in the October 1970 M onthly Labor
Review) in substantial measure qualifies its
J o s e p h A . C lo r e ty , J r ., is C h ief of t h e D iv is io n - of
I n d u s tr ia l P r ic e s a n d P r ic e I n d e x e s , B u r e a u o f L a b o r
S t a t is t ic s .

From the Review of November 1970




202

C o n s id e rin g t h e re s o u rc e s a v a ila b le , th e q u a l i t y
of th e ir w o rk w a s r e m a r k a b le . T h is w a s f o r t u n a t e
sin c e th e f ir s t m a jo r c h a lle n g e to t h e r e li a b ili ty o f
th e w p i as a m e a s u r e of p r ic e c h a n g e s in p r im a r y
m a r k e ts c a m e in t h e la te 1930’s. T h a t c o n tr o v e r s y
ro s e fro m a se rie s o f s tu d ie s b y G a r d in e r C . M e a n s

d e v e lo p in g h is th e o ry of a d m in is te r e d p ric e s, w h ic h
re lie d h e a v ily o n u se of in d iv id u a l c o m m o d ity d a t a
f ro m th e w p i . N o t o n ly t h e th e o r y b u t th e v a lid ­
i t y a n d r e lia b ility of th e d a t a w ere c h a lle n g e d
s h a rp ly . T h e g ist of th e c h a rg e w a s t h a t th e in d e x
d a t a fa ile d d is m a lly to r e fle c t e ith e r th e f r e q u e n c y
o r th e m a g n itu d e of a c tu a l c h a n g e s in t r a n s a c ­
tio n p ric e s. B o th th e m a jo r c o n tr o v e r s y a n d d e ta ils
of th e a c c o m p a n y in g c h a rg e s a re b e y o n d th e sc o p e
of th is a rtic le , b u t r e p r e s e n ta t iv e a rtic le s a re c ite d
in th e a c c o m p a n y in g b o x fo r th o s e in te r e s te d .
B e c a u s e th e M e a n s a p p r o a c h to u se of w p i
d a t a w a s c e n tr a l to th e c h a p te r o n p ric e s tr u c t u r e
in th e N a tio n a l R e s o u rc e s C o m m itte e ’s The
Structure of the Am erican Economy ,4 S a u l N e ls o n
w a s c o m m issio n e d to in v e s tig a te th e v a lid ity of
th e w p i fo r th e p a r ti c u la r u se m a d e of it. S tig le r
a n d K in d a h l n o te h is c o n c lu sio n , w h ic h s u b s t a n t i ­
a te d th e o v e ra ll v a lid ity of th e w p i d a t a fo r “ th e
s ta t e m e n t a n d i n te r p r e ta t io n of s u c h d if fe r e n t
[rigid a n d flexible] ty p e s of p ric e b e h a v io r .” 5 I n
h is p r e s e n ta tio n of th e s u p p o r tin g e v id e n c e (w h ic h
p r o b a b ly w a s a b s tr a c t e d f ro m a c o n s id e r a b ly
g r e a te r b o d y of d a t a ) , N e ls o n c o m m e n te d o n tw o
p o in ts of so m e re le v a n c e to c u r r e n t p ro b le m s . H e
n o te d t h a t th e R o b in s o n - P a tm a n A c t h a d b e e n in
e ffe c t o n ly p a r t of th e p e r io d c o v e re d a n d t h a t
p r e s u m a b ly i t w o u ld in h ib it c e r ta in fo rm s of s e c r e t
c o n c e s sio n s. ( I n te r e s tin g ly , m u c h c u r r e n t d is c u s ­
sio n a s s u m e s t h a t t h a t s t a t u t e in h ib its r e p o r tin g
to th e B u r e a u a c tu a l c o n c e ssio n s m a d e . T h is
im p lie s a g e n e ra l w illin g n e ss to v io la te th e law ,
w h ic h se e m s u n d u ly c y n ic a l. A c tu a lly , d a t a r e ­
p o r te d to th e B u r e a u b y in d iv id u a l b u sin e sse s— o r
in d iv id u a l h o u s e h o ld s fo r th e m a t t e r — a re tr e a te d
as a b s o lu te ly c o n fid e n tia l b y th e B u r e a u .)
N e ls o n w a s also c o n c e rn e d b y th e s u b s ta n tia l
n u m b e r of c o m m o d itie s fo r w h ic h th e r e p o r tin g
s o u rc e w a s a tr a d e p u b lic a tio n . T h e n as n o w , of
c o u rse , a d is tin c tio n is n e c e s s a ry b e tw e e n th o s e
w h ic h r e p o r t p ric e s o n a n o rg a n iz e d e x c h a n g e
(w h ic h o b v io u s ly a re c le a rly a c tu a l tr a n s a c tio n
p ric e s) a n d c e r ta in o th e r s w ith v a r io u s im p e r ­
fe c tio n s . T a b le 1 s u m m a r iz e s d iffe re n c e s b e tw e e n
r e p o r tin g so u rc e s fo r th e w p i in F e b r u a r y 1937,
w h ic h N e ls o n u se d , a n d J a n u a r y 1970, fro m a
c o m p ila tio n m a d e fo r p u rp o s e s o f th is c o m p a ris o n .
D a t a f o r J a n u a r y 1970 a re also a v a ila b le b y
n u m b e r of p ric e se rie s, w h ic h th ro w f u r th e r lig h t
o n c o v e ra g e b y r e p o r tin g so u rc e .




203

Selected readings
Gardiner C. M eans, “ N otes on Inflexible Prices,”
American Economic Review, March 1936, pp. 28-35.
Price Behavior and Business Policy (W ashington,
Temporary N ational Economic Com m ittee, 1940),
Monograph 1, appendix 1, pp. 165-168.
The Structure of Industry (W ashington, Temporary
N ational Economic Com m ittee, 1941), Monograph
27.
Jules Backman, “ Price Inflexibility and Changes
in Production,” American Economic Review, Sep­
tember 1939, pp. 480-486.
John K. Galbraith, “ M onopoly Power and Price
R igidities,” Quarterly Journal of Economics, M ay
1936, pp. 456-475.
Frederick C. Mills, “ Price D ata and Problems of
Price Research,” Econometrica, October 1936, pp.
289-309.
D on D . Humphrey, “ The Nature and M eaning of
Rigid Prices, 1890—1933,” Journal of Political Econ­
omy, February-D ecem ber 1937, pp. 651-661; of
particular interest for use of an earlier study by
Frederick C. Mills covering the period 1890-1925.
Willard L. Thorp, “ Price Theories and M arket
R ealities,” Papers and Proceedings of the American
Economic Association, March 1936, pp. 15-22.

Reopening the controversy
The old controversy flared again in the late
1950’s, when Means presented relatively current
w p i data to the Senate Antitrust Subcommittee of
the Judiciary Committee to sustain his contention
that the industries characterized by administered
prices caused the bulk of the increase in the w p i
during the last half of the 1950’s. To the con­
siderable extent to which the validity and relia­
bility of the w p i for this purpose were the center
of controversy, John M. Blair— then chief econ­
omist for the subcommittee— and Stigler led the
debate.6
Prior to the exchange of views cited above,
the Price Statistics Review Committee (popularly
referred to as the Stigler Committee, reflecting
his chairmanship) functioned during 1959 and
1960. Both their summary recommendation and
a more fully stated version relative to use of
buyers’ prices deserve full quotation in that order:
The individual product prices should, where feasible,
be collected from buyers (not from sellers, as at

present) to get more accurate information on actual
transaction prices.7

of a g o v e r n m e n t a n d o f th e b u y e r s ’ p r ic e d a t a
b e c a u s e o f a n o b v io u s b ia s a s w ell a s o th e r m e a s u r e ­
m e n t p ro b le m s s h a r e d b y s e lle rs ’ a n d b u y e r s ’
p r ic e s , i t is r e g r e tt a b le t h a t a la w y e r w a s n o t a
m e m b e r o f th e c o m m itte e . E c o n o m is ts a n d s t a t i s ­
tic ia n s p e r h a p s s h o u ld b e f o rg iv e n fo r f o r g e ttin g
u p o n w h o m th e b u r d e n of p ro o f r e s ts .

We recommend that a major shift be made to the
collection of buyers’ prices. Large and continuous
buyers of manufactures should be able to supply prices
which truly represent the effective terms on which
transactions are made. We do not believe that this
shift to buyers’ prices will be simple or free of new
difficulties, but it is the m ost promising source of
comprehensive,
continuous, and reliable price
quotations.
Where buyers’ prices are not available, we recom­
mend extensive use of unit values, at least as bench­
marks to which the m onthly prices are adjusted.
U n it values are inferior to specification transaction
prices, but when unit values are calculated for fairly
homogeneous commodities, they are more realistic
than quoted prices in a large number of industrial
m arkets.8

The BLS response
B e t h a t as i t m a y , t h e B u r e a u ’s o fficial re s p o n s e
to th e r e c o m m e n d a tio n s , as g iv e n b y C o m m is ­
s io n e r E w a n C la g u e , w a s to “ p la c e e m p h a s is f ir s t
o n m o re in te n s iv e e ffo rts to o b ta in a c tu a l t r a n s a c ­
tio n s p ric e s f ro m se lle rs, o b ta in in g p ric e s f ro m
b u y e r s o n ly w h e re a b s o lu te ly n e c e s s a r y , b e c a u s e
of th e g r e a t d iffic u lty a n d e x p e n se in v o lv e d in th e
l a t t e r m e th o d ” ; 11 a n d to q u e s tio n th e s o u n d n e s s of
th e r e c o m m e n d e d u s e o f u n i t v a lu e d a t a in th e
w p i “ b e c a u s e te s ts w h ic h w e h a v e m a d e s h o w t h a t
r e a l p ric e c h a n g e s c a n n o t b e s e p a r a te d f ro m
c h a n g e s in p r o d u c t m ix . . .” 12 H is p r e p a r e d s t a t e ­
m e n t d o c u m e n ts b o th p o s itio n s in c o n s id e ra b le
d e ta il.13
L a z a r e T e p e r, a n o u ts t a n d in g e x p e r t o n p ric e
in d e x e s, te s tif ie d a t th e s a m e h e a r in g s t h a t “ th e
s u g g e s tio n t h a t p ric e q u o ta tio n s b e c o lle c te d f ro m
b u y e r s d o e s n o t se e m r e a lis tic . I t c e r ta in ly w o u ld

T h e b rie f t e x t of th e r e p o r t p re c e d in g th e m o re
d e ta ile d r e c o m m e n d a tio n m a k e s c le a r t h a t th e s e
r e c o m m e n d a tio n s r e s t b a s ic a lly o n tw o s ta ff
p a p e r s a t t a c h e d to th e r e p o r t b u t sp e c ific a lly
p r e s e n te d a s th e r e s p o n s ib ility of th e ir in d iv id u a l
a u th o r s . H a r r y E . M c A llis te r (w h o also s e rv e d as
s e c r e ta r y of th e S tig le r C o m m itte e ) e v a lu a te d th e
w p i in te r m s of i n te r n a l d a t a , f re q u e n c y of p ric e
c h a n g e in r e la tio n to n u m b e r of r e p o r te r s , c o m ­
p a r is o n of w p i d a t a b a s e d o n s e lle rs ’ p ric e s w ith
th o s e w h ic h h e c o lle c te d fro m a s a m p le of la rg e
b u y e r s , a n d a c o m p a r is o n w ith C e n s u s u n i t v a lu e s .9
J o h n A . F lu e c k ’s e v a lu a tio n r e s te d o n e n u m e r a tin g
so m e o f th e m a in re a s o n s t h a t a c tu a l tr a n s a c t io n
p ric e s m a y d iffe r f ro m lis t p ric e s ( b u ttr e s s e d b y
q u o ta tio n s f ro m v a r io u s p u b lic a tio n s ) , a n d a
c o m p a r is o n o f w p i p ric e s w ith g o v e r n m e n t b id
p ric e s .10

r e p r e s e n t a v e r y c o s tly p r o c e d u re w h ic h m a y n o t
n e c e s s a rily y ie ld w h a t is e x p e c te d of it. . .

w ro n g p a t h . ” 14 H is fu ll s t a t e m e n t p r e s e n ts h is
re a s o n s . A f o o tn o te , in c id e n ta lly , illu m in a te s p itfa lls in v o lv e d in u se of g o v e r n m e n t b id p ric e s to
e v a lu a te th e w p i . 15
T h e S u b c o m m itte e ’s r e p o r t p ith i ly s u m m a r iz e s

C o n s id e r in g th e d u b io u s q u a lity (fo r e v a lu a tin g
a p ric e in d e x ) of u n i t v a lu e d a t a d u e to th e p r o d u c t
m ix p r o b le m a n d of g o v e r n m e n t b id p ric e s b e ­
c a u s e of m a r k e d d iffe re n c e s in th e m a r k e t le v e r a g e
Table 1.

[as to

u se of u n i t v alu es] T h e C o m m itte e is c le a rly o n a

th e g e n e r a l r e a c tio n to th e r e c o m m e n d a tio n t h a t
th e B u r e a u m o v e a s r a p id l y as p o ssib le to c o lle c -

Reporting sources for the Wholesale Price Index and number of price series used
Sources used
Reporting source

January 1970

February 1937

Number

Price series used

Percent

January 1970

Percent

Number

Number used
per item

Percent

Number

Total____________________________

784

100.0

2,445

100.0

7,726

100.0

Company reports________________________
Trade publications__________________ ____
Trade associations_______________________
Government agencies.. .....................................

383
367
31
3

48.8
46.8
4.0
0.4

1,906
394

78.0
16.1
.4
5.5

7,107
415

92.0
5.4

193

2.5




11

134

204

11

0.1

3.2
3.7

1.1

1.0
1.4

tio n of p ric e s f ro m b u y e r s : “ T h e e n th u s ia s m of th e
Review C o m m itte e fo r th is r e c o m m e n d a tio n w a s
n o t s h a re d g e n e r a lly b y th e w itn e s s e s .” 16
A p p a r e n tly th e A p p r o p r ia tio n s C o m m itte e s
of th e C o n g re s s s h a r e d th is la c k of e n th u s ia s m ,
fo r n o a d d itio n a l r e s o u rc e s w e re m a d e a v a ila b le .
A m o d e s t in c r e m e n t w a s a p p r o v e d to p e r m it th e
B u r e a u to in it ia t e d e v e lo p m e n t of th e I n d u s t r y
S e c to r I n d e x e s (i s p i ). D e v e lo p m e n t of th e s e
in d e x e s, in w h ic h p ric e s a re c lassified b y th e S t a n d ­
a r d I n d u s t r i a l C la s s ific a tio n s y s te m , w a s r e c o m ­
m e n d e d b y th e S tig le r C o m m itte e .

User views
I n 1 9 6 5 -6 6 , th e J o i n t E c o n o m ic C o m m itte e ’s
S u b c o m m itte e o n E c o n o m ic S ta ti s tic s c o n d u c te d
a m a jo r e x p lo r a tio n of th e n e e d s fo r im p r o v e d
s ta tis tic s . A s a f irs t s te p , th e s u b c o m m itte e so ­
lic ite d th e v ie w s o f a la rg e g r o u p of u s e rs o f
g o v e r n m e n t s ta tis tic s . A m o n g o v e r 70 r e s p o n d ­
e n ts , tw o c a lle d a t t e n t i o n to th e p r o b le m o f
tr a n s a c tio n s p ric e s. A r t h u r F . B u r n s c o m m e n te d ,
“ T o o f r e q u e n tly , s ta tis t ic s o n w h o le sa le p ric e s
r e p r e s e n t l i s t p ric e s r a t h e r th a n a c tu a l p ric e s
c h a r g e d .” 17 H e r b e r t S te in s u b m i tte d a r e c e n t
s p e e c h b y A lfre d C . N e a l in w h ic h N e a l a s s e r te d
t h a t “ I n th e a r e a of w h o le s a le p ric e s, I a m in ­
c lin e d to p la c e a t th e to p of th e p r io ritie s a n a t ­
t e m p t to o b ta in d a t a o n a c tu a l p ric e s p a id b y th e
b u y e r , n o t th e p ric e s u p p lie d b y se lle rs . . . ” 18
L ik e o th e r g o v e r n m e n t a g e n cies, th e B u r e a u
w a s in v ite d to c o m m e n t o n th e c o m p e n d iu m of
r e c o m m e n d a tio n s . A s to N e a l’s s u g g e s tio n , th e
B u r e a u r e s p o n s e w a s t h a t “ T h e p r o je c t h a s m e r it,
a t le a s t o n a s e le c tiv e b a s is , b e c a u s e i t w ill e n a b le
a b e t t e r e v a lu a tio n to b e m a d e of p ric e tr e n d s fo r
in d u s tr ie s c h a r a c te r iz e d b y c o m p lic a te d r e b a te
a n d d is c o u n t s tr u c t u r e s . . . . T h is w o u ld b e a
c o s tly p r o je c t b u t o n e w h ic h bls h a s r e c o m ­
m e n d e d f o r s e le c te d p r o je c ts fo r a n u m b e r of
y e a r s .” 19
A s o n e p h a s e of it s p r o je c t, th e s u b c o m m itte e
h e ld h e a r in g s in M a y 1966 o n p ric e s ta tis tic s ,
fo c u s in g o n th e e x t e n t to w h ic h th e S tig le r
C o m m itte e r e c o m m e n d a tio n s h a d b e e n im p le ­
m e n te d d u r in g th e in te r v e n in g 5 y e a r s . W ith
r e s p e c t to th e r e c o m m e n d a tio n to m o v e to w a r d
u se o f b u y e r s ’ p ric e s, R a y m o n d B o w m a n ( th e n
A s s is ta n t D ir e c to r of th e B u d g e t B u r e a u fo r
S ta ti s tic a l S ta n d a r d s ) s t a t e d :




205

There is general agreement with the objective of
this recommendation, i.c., that the w p i should re­
flect realistic, actual transaction prices, not quoted
prices. Price respondents (sellers) arc requested to
report all discounts applicable to quoted prices . . .
It is recognized th at such discounts are not universally
reported.
We agree with h l s that the first step toward im­
plem entation of this recommendation should be
through a limited and experim ental program to identify
com m odity areas in which the differences are im­
portant. . . . Because of the heavy costs and re­
spondent burden involved, collection of price data
from buyers should be undertaken only if other
methods are not successful.20

B o w m a n ’s s t a t e m e n t c o n c ise ly s u m m a r iz e d th e
m o re d e ta ile d s t a t e m e n t b y C o m m is s io n e r A r t h u r
M . R o s s .21 L a z a r e T e p e r, a g re e in g t h a t r e p o r te r s
d o n o t a lw a y s r e p o r t all d is c o u n ts , w e n t o n :
These short-term distortions on the index may
make it at tim es a bit less sensitive to current price
changes. Collection of actual transaction prices, of
course, is a massive and costly operation. On the other
hand, it is conceivable that ways may be found to se­
cure better cooperation from respondents and to get
more accurate responses from them . Experimental
research is called for in this area.22

T h e s u b c o m m itte e ’s r e p o r t, c itin g th e d iffe re n c e s
b e tw e e n lis t a n d a c tu a l tr a n s a c t io n p ric e s as a
d e fic ie n c y in th e w p i , 23 re c o m m e n d e d t h a t “ C o l­
le c tio n of d a t a o n p ric e s p a id b y b u y e r s fo r se­
le c te d p r o d u c ts , s u c h as m e ta ls a n d m a c h in e ry ,
s h o u ld b e i n it ia t e d in o r d e r to in s u r e o b ta in in g th e
te rm s of a c tu a l tr a n s a c tio n s w h ic h o fte n d iffe r
s ig n ific a n tly fro m lis t p ric e s . . . ” 24

Expansion of industry sector indexes
W ith re s o u rc e s u n a u g m e n te d , th e B u r e a u d e ­
v o te d s u c h of its in d u s t r ia l p ric e p r o g ra m re s o u rc e s
as c o u ld b e m u s te r e d to e x p a n d in g th e I n d u s t r y
S e c to r I n d e x e s — a p p r o x im a te ly d o u b lin g th e n u m ­
b e r of s ic 4 - d ig it i n d u s t r y in d e x e s p u b lis h e d ,
w ith a r o u g h ly p r o p o r tio n a te in c re a s e in 5 -d ig it
p r o d u c t c la ss in d e x e s. I m p le m e n ta tio n o f th is
S tig le r C o m m itte e r e c o m m e n d a tio n w a s d e e m e d
th e w ise r u se of sc a rc e re s o u rc e s . T h is r e c o m m e n ­
d a t io n h a s b e e n s u p p o r te d n o t o n ly u n a n im o u s ly
b u t e n th u s ia s tic a lly — e x c e p t fo r p r o v id in g th e
a d d itio n a l r e s o u rc e s w h ic h a re a p r e r e q u is ite fo r
p ro v id in g th e c o m p le te b a t t e r y of o u t p u t p ric e
in d e x e s a n d i n p u t p ric e in d e x e s e s s e n tia l to a c h ie v e
th e ir m a jo r p u rp o s e s . T h u s f a r, th e p r o g r a m h a s
b e e n lim ite d to o u t p u t p ric e in d e x e s, f o r w h ic h

th e re are few er p ro b lem s in u sin g d a ta collected
p rim a rily fo r th e w p i .
W i th o u t f a n f a r e b u t p e r s is te n tly th e B u r e a u
th r o u g h its c o m m o d ity a n a ly s ts h a s p re s s e d to w a r d
o b ta in in g a c tu a l tr a n s a c t io n p ric e s f ro m se lle rs.
A s T e p e r s u g g e s te d , th e r e a re w a y s to im p r o v e
c o o p e r a tio n a n d th u s o b ta in m o re a c c u r a te r e ­
sp o n se s. S u b s ta n t ia l p ro g re s s h a s b e e n m a d e in a
n u m b e r of in d u s tr ie s , th a n k s la rg e ly to c o m m o d ity
a n a ly s t s a b le to c o n v in c e r e p o r te r s t h a t a c c u r a te
r e p o r ts of a c tu a l p ric e s s e rv e n o t o n ly th e p u b lic
i n t e r e s t in a c c u r a te official p ric e in d e x e s, b u t also
th e ir o w n lo n g - te rm b e s t in te r e s ts . S u b s ta n t ia l
c r e d it is d u e to th o s e in th e b u s in e s s c o m m u n ity
Who b y a n d la rg e h a v e d e e p e n e d a n d e x te n d e d
th e ir v o l u n ta r y c o o p e r a tio n . A ll of th e s e r e p o r te r s
a r e v o lu n ta r y , a s h a s b e e n th e c a se t h r o u g h o u t
th e h is t o r y o f th e w p i , b u t th e r e h a s b e e n a
m a r k e d im p r o v e m e n t in th e q u a l ity of c o o p e r a tio n .
T o ill u s tr a te : T h e w p i fo r m o to r v e h ic le s re fle c ts
n o t o n ly a c tu a l tr a n s a c t io n p ric e s b u t also a
s o p h is tic a te d a d j u s t m e n t fo r q u a l ity c h a n g e s , o n ly
b e c a u s e o f th e w e a lth o f d e ta ile d d a t a s u p p lie d
b y th e m a n u f a c tu r e r s , a t n o s m a ll e x p e n se to th e m .
O n a less s t r u c t u r e d a n d m o re in f o rm a l b a s is ,
th e B u r e a u r e c e iv e s th e n e c e s s a r y d a t a fro m m a n y
if n o t m o s t of th e c o m p a n ie s w h ic h r e p o r t m a ­
c h in e r y p ric e s. R e c e n tly , th e r e p r e s e n ta t iv e of a
c o m p a n y w h ic h a t o n e tim e r e p o r te d p r e t t y m u c h
o n a “ ta k e i t o r le a v e i t ” b a s is s p e n t m o s t of a
d a y w ith bls s ta ff a s s is tin g in th e c o r r e c t c a lc u la ­
tio n o f a p a r ti c u la r l y c o m p le x p ric e c h a n g e . T h e
o v e r a ll p ic tu r e is n o t a s r o s y as th e s e e x a m p le s
m a y im p ly ; th e p o in t is t h a t a p p r e c ia b le p ro g re s s
is b e in g m a d e .
A lth o u g h p e r s o n a l v is its to r e p o r tin g c o m p a n ie s
a re lim ite d ( u s u a lly in c o n n e c tio n w ith e s ta b lis h in g
a n e w r e p o r te r ) , e x te n s iv e u s e of th e te le p h o n e is
e ffic ie n t, r e la tiv e ly in e x p e n s iv e (a r o u g h r u le of
th u m b fo r e s tim a tin g c o s ts of field tr ip s is $ 1 0 0
p e r m a n - d a y ) , a n d a c h a n n e l fo r b u ild in g r e la tio n ­
s h ip s c o n d u c iv e to r e p o r tin g a c c u r a te ly . I t is
m o re ty p ic a l t h a n n o t, w h e n th e tr a d e p re s s o r
o th e r p u b lic a tio n s r e p o r t d is c o u n tin g , f o r th e
a p p r o p r ia t e c o m m o d ity a n a ly s t to c h e c k th e
r e p o r ts im m e d ia te ly w ith r e p o r tin g c o m p a n ie s
a n d o th e r so u rc e s of in f o rm a tio n . L ik e q u a l ity
a d j u s tm e n t, th e p u r s u it of a c tu a l tr a n s a c t io n
p ric e s is v i r tu a lly a d a ily p r o b le m in c a lc u la tin g
the wpi.




206

B y the m id-1960’s it was apparent th at b l s
was unable to pursue the recommendation for
use of buyers’ prices. The N ational Bureau of
Economic Research evidently considered such a
study a worthwhile project. Stigler’s own strong
convictions on the relative merits of sellers’ versus
buyers’ prices evidently made him willing to invest
much of his time in the tedious if challenging
task of data collection and index calculation.
Although field work covered the period from the
fall of 1965 to m id-1967, the data and their indexes
are for 1957 through 1966. A few sellers were
included in their sample, but it consisted pre­
dom inately of large manufacturers, governm ent
agencies, and a few hospitals. The study was
lim ited to relatively few industries, constituting
less than one-fifth of the weight of the w p i . For
the com m odity groups to which the indexes for
individual series were aggregated and which are
given the titles corresponding to the appropriate
w p i major groups, the coverage varies sharply.
The bls indexes shown in Stigler and Kindahl
were constructed by them from b ls series cor­
responding to theirs, and at the group level, of
course, are not those published by b l s . Its appear­
ance and the attendant publicity raise a variety
of questions to which relatively brief answers m ay
be given on my personal responsibility. Bureau
review and assessment are not complete.

Assessing the data
Does the study invalidate the w p i ? Absolutely
not. As Stigler and Kindahl very properly point
out, the study was not a test of the w p i .25 Stigler
personally m ay think so, or that Stigler and
Kindahl tends to indicate so.26 For the other pole
of opinion, see testim ony before the Joint Com­
m ittee on July 14, 1970, by Means and Blair.27
D o e s th e s t u d y d e m o n s tr a te t h a t b u y e r s ’ p ric e s
m o re n e a r ly a p p r o x im a te a c tu a l tr a n s a c t io n p ric e s
t h a n d o s e lle rs ’ p ric e s? N o t u n le s s o n e a c c e p ts
e ith e r ty p e o f p ric e as a n a c tu a l tr a n s a c t io n p r ic e
as u s e d in c a lc u la tio n o f a n in d e x of p ric e c h a n g e .
O b v io u s ly , S tig le r a n d K i n d a h l’s in d e x e s d iffe r
f ro m b ls in d e x e s in v a r io u s w a y s a n d b y v e r y
rfa rro w to v e r y w id e a m p litu d e s . G iv e n th e
d iffe re n c e s in s a m p le s , sp e c ific a tio n s , c o lle c tio n
a n d c a lc u la tio n p ro c e d u re s , d iffe re n c e s a r e to b e
e x p e c te d .

T h is d is c u s s io n a s s u m e s t h a t S tig le r a n d K in d a h l a g re e s u b s t a n tia lly w ith th e b l s d e fin itio n of
a n a c tu a l tr a n s a c t io n p ric e , w h ic h is lis t o r b o o k
“ p ric e s less all d is c o u n ts , a llo w a n c e s, r e b a te s , fre e
d e a ls, e tc ., so t h a t th e r e s u ltin g n e t p ric e is th e
a c tu a l se llin g p r ic e o f th e c o m m o d ity fo r th e
sp e cifie d b a s is of q u o ta tio n .28 T o th is m ig h t b e
a d d e d “ p lu s a n y p r e m iu m , e t c . ” T h is d ig re s s io n
m a y se e m u n n e c e s s a r y , b u t th e lit e r a tu r e a b o u n d s
w ith u se of th e te r m in v e r y d iffe re n t a n d u s u a lly
m u c h b r o a d e r c o n n o ta tio n s . A c tu a lly m u c h c r iti­
c ism of th e w p i ( a n d c p i ) s te m s f ro m f r u s t r a tio n
in v o lv e d in a t t e m p t i n g to u se i t fo r la c k of a n
a p p r o p r ia t e in d e x .

The Bureau usually obtains prices f.o.b. pro­
duction or central marketing point, to avoid re­
flecting changes in transportation costs. Stigler
and Kindahl’s data from buyers normally would
include these charges which introduces another
source of differences.
Can b l s use Stigler and Kindahl as a pilot
study? No, but both data in the book and data
which its authors may have in their possession
could be very helpful. A b l s pilot study of buyers’
prices necessarily would be constructed to test the
feasibility of collecting such prices on a continuing
basis (rather than an essentially one time), using
the mail or average b l s agents (rather than two
distinguished professors supported by the presti­
gious n b e r ) , and including small buyers as well as
large.
S tig le r a n d K in d a h l p u b lis h e d th e n u m b e r o f
p r ic e s e rie s u s e d e a c h y e a r fo r e a c h c o m m o d ity .
T h e s e d a t a sh o w a* c o n s is te n t p a t te r n . T h e y rise
f ro m a s m a ll n u m b e r of r e p o r te r s in th e e a r ly y e a r s
to a p e a k in 1964 a n d 1965, fo llo w ed b y a p r o ­
n o u n c e d d e c lin e in 1966. T h e f o rm e r p h e n o m e n o n ,
w h ic h in d ic a te s th e i n a b ilit y o r u n w illin g n e ss o f
th e r e p o r te r to p r o d u c e re c o rd s in e a r lie r y e a r s o r
S tig le r a n d K i n d a h l’s in a b ility to d e te r m in e c o m ­
p a r a b le p ric e s f ro m s u c h r e c o rd s , w o u ld n o t b e
r e le v a n t f o r b l s if d a t a w e re c o lle c te d c u r r e n tly .
T h e d r o p in 1966, h o w e v e r, is d e c id e d ly p e r ti n e n t
in a p p r a is in g th e p o te n tia l su c c e ss in o b ta in in g
r e p o r te r s o n a c o n tin u in g b a s is.

If Stigler and Kindahl recorded man-days spent
in data collection, such information would be most
useful, b l s must consider any major change in
terms of its costs. For the industrial price program,
these must be regarded as almost fixed. Much more




lig h t o n th e p r o b le m s of m a in ta in in g c o m p a r a b il­
ity , e v e n w ith in th e b r o a d e r sp e c ific a tio n s S tig le r
a n d K in d a h l e m p lo y e d , a n d of p ro b le m s in id e n ­
tif y in g th e sp e c ifie d ite m a n d its p r ic e w o u ld b e
m o s t h e lp fu l. I n v o ic e s w e re n o t d e s ig n e d w ith a
v ie w to a id in g th e p ric e c o lle c to r. D u r in g W o r ld
W a r I I , I w a s in v o lv e d in c o lle c tin g b u y e r s ’ p ric e s
a n d c a n te s tif y t h a t th e p ro b le m s p r o g re s s e d a l­
m o s t g e o m e tr ic a lly a s I m o v e d f ro m fo o d to te x tile s
a n d a p p a r e l to m a c h in e ry .
W ill th e B u r e a u m o v e to a w p i b a s e d o n b u y e r s ’
p ric e s? I t is a s u n lik e ly as i t is u n d e s ira b le . T o
p r o d u c e e n o u g h p ric e q u o ta tio n s fo r r e lia b le
m e a s u r e s of p ric e c h a n g e f ro m m o n th to m o n th
w o u ld r e q u ir e a m u c h la rg e r s a m p le of r e p o r te r s .
T o ta k e S tig le r a n d K in d a h l’s a v e ra g e of 17 p e r
c o m m o d ity (a fig u re w ith v e r y w id e v a ria n c e s )
v e r s u s th e b l s a v e r a g e of s lig h tly m o re th a n th r e e
as a v e r y c o n s e r v a tiv e e s tim a te o f th e r e q u ir e d
s a m p le size, b l s w o u ld b e fo rc e d to c u r ta il s h a r p ly
th e n u m b e r o f c o m m o d itie s in c lu d e d in th e w p i —
u n le ss i ts re s o u rc e s w e re e x p a n d e d tr e m e n d o u s ly .
D is c u s s in g th e q u e s tio n of w h e th e r a n in d e x sh o u ld
in c lu d e a sm a ll o r la rg e n u m b e r o f c o m m o d itie s ,
M itc h e ll o b s e rv e d t h a t “ E v e r y r e s tr ic tio n in th e
sc o p e of th e d a t a im p lie s a lim ita tio n in th e
sig n ific a n c e of th e r e s u lt s .” 29
T h e B u r e a u h a s lo n g b e e n in te r e s te d in tw o u se s
o f b u y e r s ’ p ric e s. F o r th e w p i , w h e re i t is e s ta b ­
lis h e d c o n c lu s iv e ly t h a t a c tu a l tr a n s a c t io n p ric e s
c a n n o t b e o b ta in e d in a n y o th e r w a y , b u y e r s ’
p ric e s s h o u ld b e u s e d g iv e n th e n e c e s s a ry re so u rc e s.
L o o k in g f o rw a r d to c o n s tr u c tin g in p u t p ric e s fo r
th e I n d u s t r y S e c to r In d e x e s , i t is q u ite p r o b a b le
t h a t c o lle c tio n of p ric e s f ro m b u y e r s m ig h t b e
u n a v o id a b le in so m e in s ta n c e s . I n b o th cases,
b u y e r s ’ p ric e s e v e n c o lle c te d a t lo n g e r in te r v a ls
m ig h t b e u s e d a d v a n ta g e o u s ly to s p o t c o m m o d ity
a re a s in w h ic h c o r r e c tiv e a c tio n is in d ic a te d . A s
n o te d e a rlie r, th e B u r e a u ’s c o m m o d ity a n a ly s ts
a re u s in g S tig le r a n d K in d a h l d a t a fo r p re c ise ly
th is p u rp o s e .
I n te r m s of a lo n g e r r u n a n d m u c h m o re a m b L
tio u s p r o je c t— c o n s tr u c tin g a G e n e ra l P ric e I n d e x ,
c o v e rin g all s e c to rs of th e e c o n o m y — p ric e d a t a
f ro m th e v e r y im p o r t a n t g o v e r n m e n t s e c to r
p r o b a b ly c a n b e m o s t re lia b ly , effic ie n tly , a n d
e c o n o m ic a lly c o lle c te d f ro m th e p u r c h a s in g g o v e r n ­
m e n t a g e n cies.

The Bureau is no more prejudiced toward sellers’

207

12 I b id ., p . 5 6 0 .

a n d a g a in s t b u y e r s ’ p ric e s th a n i t is w e d d e d to th e
th e o r y of a u n iq u e p ric e . I t is c r ib b e d , c a b in e d , a n d
c o n fin e d b y th e b a s ic e c o n o m ic p r o b le m of
a llo c a tin g sc a rc e re s o u r c e s to m e e t m a n y n e e d s. □

13 I b id ., p p . 6 0 2 - 6 0 3 .
“ I b id ., p . 6 73.
13 I b id ., p p . 6 7 2 - 6 7 3 .
18
Government Price Statistics, R e p o r t o f t h e S u b c o m ­
m it t e e o n E c o n o m ic S t a t is t ic s o f t h e J o in t E c o n o m ic
C o m m itte e (U .S . S e n a te , 1 9 6 1 ), p . 8.

------- —FOO TNO T E S ---------1 The M aking and Use of Index Numbers ( b l s Bulletin
656, 1938), p. 25. (M itchell, of course, gained his greatest
fame in later years for his work on business cycles and on
m any other economic and statistical areas, but was a
principal architect of the Wholesale Price Index in roughly
its present form.)

17 Improved Statistics for Economic Growth, A C o m ­
p e n d iu m o f V ie w s a n d S u g g e s tio n s F ro m I n d iv id u a ls ,
O r g a n iz a tio n s, a n d S t a t is t ic s U s e r s (U .S . S e n a te , J o in t
E c o n o m ic
C o m m itte e ,
S u b c o m m itte e
on
E c o n o m ic
S t a t is t ic s , 1 9 6 5 ), p . 15.

2 Ibid., p . 26.

18 I b id ., p . 130.

3 George J. Stigler and James K. Kindahl, The Behavior
of Industrial Prices (N ew York, N ational Bureau of
Economic Research, 1970), General Series 90.

19 Improved Statistics for Economic Growth, C o m m e n ts b y
G o v e r n m e n t A g e n c ie s o n V ie w s a n d S u g g e s tio n s F r o m
I n d iv id u a ls , O r g a n iz a tio n s, a n d S t a t is t ic s U s e r s ( U .S .
S e n a te , J o in t E c o n o m ic C o m m itte e , S u b c o m m itte e o n
E c o n o m ic S t a t is t ic s , 1 9 6 6 ), p . 4 8 .

4 The Structure of the American Economy (W ashington,
N ational Resources Committee, 1939), pp. 122-152.

20 Government Price Statistics, H e a r in g s b e fo r e t h e S u b ­
c o m m itte e o n E c o n o m ic S t a t is t ic s o f t h e J o in t E c o n o m ic
C o m m itte e (U .S . S e n a te , 1 9 6 6 ), p . 11.

5 Ibid., p. 185.
8 George J. Stigler, “Administered Prices and Oli­
gopolistic Inflation,” Journal of Business, January 1962,
pp. 1-13: John M. Blair, “Administered Prices and
Oligopolistic Inflation: A R eply,” Journal of Business,
January 1964, pp. 68-81 (see also Stigler’s comment, pp.
82-83, and M cAllister’s comment, pp. 84-86, of the same
issue); W alter Adams and Robert F. Lanzillotti, “The
R eality of Administered Prices,” Administered Prices:
A Compendium on Public Policy (U.S. Senate, Committee
on the Judiciary, Subcomm ittee on A ntitrust and
M onopoly, 1963), pp. 5-21.

21 I b id ., p p . 59 a n d 63.
22 I b id ., p . 154.
23 Government Price Statistics, R e p o r t o f t h e S u b c o m m it­
t e e o n E c o n o m ic S t a t is t ic s o f t h e J o in t E c o n o m ic
C o m m itte e ( U .S . S e n a te , 1 9 6 6 ), p . 8.
24 I b id ., p . 17.
25 S tig le r a n d K in d a h l, o p . c it., p . 4.
28 A s q u o te d in p ress r e le a s e d a te d J u n e 26, 1970, is s u e d
b y th e N a t io n a l B u r e a u o f E c o n o m ic R e se a r c h , a n n o u n c in g
p u b lic a tio n o f The Behavior of Industrial Prices.

7 Government Price Statistics, Hearings before the Sub­
com m ittee on Economic Statistics of the Joint Economic
C om m ittee (U .S. Senate, 1961), Pt. I, p. 21.

27 G a rd in er C . M e a n s a n d J o h n M . B la ir , in th e ir
in d iv id u a l t e s t im o n y , M idyear Economic Review, H e a r in g s
b e fo r e t h e J o in t E c o n o m ic C o m m itte e ( U .S . S e n a te , 1970)
in p ress.

8 Ibid., p. 71.
8 Ibid., Staff Paper 8, pp. 373-418.
10 Ibid., Staff Paper 9, pp. 419-458.

28 B L S Handbook of Methods for Surveys and Studies

11 Government Price Statistics, Hearings before the Sub­

(b l s

comm ittee on Economic Statistics of the Joint Economic
Com m ittee (U .S. Senate, 1961), Pt. II, p. 559.




B u lle tin 1458, 1 9 6 6 ), p . 9 2 .

29 M itc h e ll, o p . c it ., p . 5 3 .

208

Recent studies cast
doubt on view that Consumer
Price Index shows upward bias
because of inadequate correction
for product improvement
JACK E. TRIPLETT

M any ec o n o m ists believe that quality changes
in goods and services are not adequately taken
into account in the preparation of the Consumer
Price Index ( c p i ). A s a result, they believe, the
c p i has a system atic and persistent upward drift
which makes the index a questionable indicator
of the course of inflationary price movements.
Products and services probably do tend to
improve in quality as the years go by. It is easy,
therefore, to suppose that the “market basket”
priced by the Bureau of Labor Statistics must
experience a similar change in quality. If not
allowed for in some way, such improvements in
quality would cause the computed price index to
rise too rapidly, which is in contrast to the con­
cept the c pi is supposed to measure: the cost of
acquiring a fixed collection of goods and services.
Thus, the argument goes, price indexes will drift
upward even when no inflation is actually taking
place, and they will give an exaggerated notion of
the speed of inflation when prices are in fact
increasing.
To what extent is the belief that price indexes
are biased upward borne out by existing evidence?
No assessment of the quality error in the c p i as a
whole has yet been made, but a number of in­
vestigations have produced estimates of quality
error in individual index components. The present
article is a survey of existing studies, which pre­
sent contradictory evidence. Some investigators
found upward bias, but others reported that
quality error might be negative— that is, when
the bls failed to correct adequately for quality
changes, it resulted in a price index that rose too
slowly, rather than too rapidly. The key studies are
reviewed in the following section. For convenience

Jack E. T riplett is assistant professor of economics,
Washington U niversity, St. Louis, Mo.

the Review of May 1971
DigitizedFrom
for FRASER


209

the effects of
quality change
on the CPI
of presentation they are grouped according to
dates and products covered.
Autom obiles

B y far the best known empirical work on the sub­
ject of price indexes and quality change is Zvi
Griliches’ study of the c p i new automobile compo­
nent.1 Griliches employed what has come to be
known as the “hedonic technique” for measuring
quality change, an approach also used in a number
of other studies. Briefly, this technique involves
searching for variables or attributes which m ay
account for quality differences among varieties
of a product selling at the same point in time.
For example, if we look at the automobile market
as a whole, it appears (from analysis of the statisti­
cal data, as well as from knowledge of the nature of
automobiles and the behavior of consumers in the
aggregate) that a more powerful engine, other
things equal, is generally preferred to a less power­
ful one. People differ in how much they are willing
to pay for more power, but there are clear statis­
tical regularities between power ratings and the
price paid for an automobile, after allowing for
differences in size, comfort, economy, and so on.
Regression analysis is used to isolate an “implicit
price” for power (as well as for the other th ings);
then if the power of a 1970 car exceeds that of the
same car in 1969, the 1969 implicit price for
power can be used to adjust for the value of the
change in power between the 2 years.2 Other
attributes can be allowed for in the same way.
Griliches presented several different estim ates
of quality-adjusted automobile price indexes, but
all his quality-adjusted indexes showed that the
new automobile component of the c p i rose too
rapidly. Indeed, for the 1954-60 period, during
which the c p i index of new automobile prices rose
11.3 percent, Griliches’ quality-adjusted index

numbers actually declined. The decreases he
recorded in his several indexes ranged from 1.0
percent to 26.6 percent. These results were widely
interpreted as strong evidence that the
rose
far too rapidly during the inflation of the late
1950’s, and several economists have suggested
that there m ight not have been any inflation at all.

Table 1. Comparisons of percent changes in price indexes
for automobiles, 1953-60

However, a second study of quality bias in the
new car component reached just the opposite
conclusion.
Philip C agan3 used an entirely
different technique to allow for quality change in
autos: the “vintage price” method.4 When Cagan
compared his quality-adjusted index, for the
years 1954-60, with the
component, he found
his own index had risen more than the
— 16.7
percent compared with 11.3 percent in the
—
suggesting “. . . that the
is not biased upward
and may even overcorrect for quality improve­
ments in automobiles. How that happened is not
clear.” 6

1 9 5 3 - 5 4 ______
1 9 5 4 - 5 5 . .........
1 9 5 5 - 5 6 ............
1 9 5 6 - 5 7 ............
1 9 5 7 - 5 8 ______
1 9 5 8 - 5 9 ............
1 9 5 9 - 6 0 ............
1 9 5 3 - 6 0 ______
1 9 5 4 - 6 0 . ..........

N e w s e r ie s o f t r a n s ­
a c tio n p r ic e s

c p i

c p i

c p i

c p i

c p i

P e r io d

P u b lis h e d
CPI

2 .6
-1 .7
-.9
5 .1
4 .2
4 .2
. 1
1 4 .2
1 1 .3

c p i

c p i

c p i

c p i

Which is the appropriate basis for adjustment—
list prices or the actual
? There are arguments
either way. Because the
always included some
quality adjustments, applying a quality index to
the published
new auto component probably
introduces double-counting of quality changes.®
On the other hand, an index based on changes in
list prices may not correspond to the movements
of actual transactions prices, although the longer
the time span, the better the approximation.
In view of the divergence in movement between
the index of list prices and the
auto component,
close examination of the
during the 1954-60
period seems imperative. The required information
can be extracted from an unpublished
mem­
orandum by Thomas W. Gavett, on which the
c p i

c p i

c p i




c p i

c p i

b l s

210

0 .9
2 .9
.9
1 2 .9
1 .3
6 .2
2 .4
3 0 .3
2 9 .1

A d ju s t e d
f o r q u a lit y
ch an ge

G r ilic h e s

-3 .2
-4 .5
-.4
9 .9
-.8
3 .4
3 .5
7 .4
1 1 .0

-0 .5
-2 .3
2 .2
4 .7
3 .2
4 .5
-.2
1 1 .2
1 2 .5

Cagan

( i)
-2 .5
6 .3
6 .1
5 .3
.4
. 4
( i)
1 6 .7

> N o t a v a ila b le .
N O T E S : C u m u la t iv e c h a n g e s c o m p u t e d b y c h a in in g .
SOURCES:

C o lu m n 1:
C o lu m n 2 :
C o lu m n 3 :

c p i

Since Cagan and Griliches produced nearly
identical estim ates of the value of quality change,
m ost of the difference in their results stems from
their adjusting different price series. Griliches
applied his quality indexes directly to the pub­
lished
auto component. Cagan, on the other
hand, adjusted an index of list prices for the cars
priced for the
. If we take Griliches’ quality
figures and apply them to an index of lis t prices,
we get an estimated 12.7-percent increase in the
price of cars, corrected for quality changes, be­
tween 1954 and 1960—not far from the actual
estim ate of an 11.3-percent increase, and (like
Cagan’s result) a somewhat greater price increase
than was recorded by the published
.

U n a d ju s te d
f o r q u a lit y
change

L is t p r ic in d e x e s
( a d ju s t e d f o r q u a lit y
ch a r ges)

C o lu m n 4 :

C o lu m n 5 :

C o m p u te d f r o m a n n u a l C P I n e w a u t o m o b ile c o m p o n e n t.
B a s e d o n a n u n p u b lis h e d B L S m e m o r a n d u m b y T h o m a s W .
G a v e tt.
C o lu m n 2 d iv id e d b y an in d e x o f q u a lit y , c o m p u t e d b y
G a v e tt u s in g m a t e r ia l f r o m G r ilic h e s , o p . c it . (1 9 6 1 , 1964),
a n d s p e c ific a t io n s o f c a r s in th e N e w P r ic e S e r ie s .
F r o m a n in d e x o f l i s t p r ic e s c o m p u t e d f r o m G r ilic h e s , o p .
c it . (1 9 6 1 ), p . 185, t a b le 6, d iv id e d b y a q u a lit y in d e x ta k e n
f r o m G r ilic h e s , o p . c it . (1 9 6 4 ), p . 186, t a b le 8, w it h c o r r e c ­
tio n o f a c le r ic a l e r r o r n o te d b y T h o m a s W . G a v e tt. T h e
c o lu m n r e c o r d s c h a n g e s in th e r e s u lt in g in d e x .
C a g a n , o p . c it., p. 2 30 , t a b le 5.

following paragraphs are based.
In mid-1954,
began to gather information
on the typical or average price concession (dis­
count, or over allowance on trade-ins) allowed by
dealers on the car selected by
for pricing.
When price concession information first became
available, it was not “linked” out of the index.
Instead, prices including concessions were com­
pared directly with previous prices (which were,
in effect, list prices). The full amount of the
reported price concession was treated as a price
decrease for the m onth when data on price con­
cessions were first collected. Because the overall
impact on the index was spread over a period of
time, the error introduced into the index affects it
from'the end of 1953 through 1955.7
In order to have an index for 1953-56 free of the
b l s

b l s

concession error, we made use of information on
price concessions compiled by Gavett to construct
a new price index for the whole 1953-60 period.8
This is an index of transactions prices, without
quality adjustment, and free (or as nearly free as
it can be made) from the price-concession error
affecting the
. The results are shown in the
second column of table 1 . While the published
for cars rose by 14 percent, from 1953-60, the
new price series shows an increase about double
that figure. The next step was to deflate the new
price series by use of a quality index appropriate
c p i

c p i

to the cars of the price series. The results, detailed
in column 3 of table 1, are in m y opinion the best
estimate that can be made of the true course of
price movement for automobiles for the 1953—60
period.
This index indicates that there was substantial
quality error in the automobile price index during
this period. But, contrary to some economists’
opinion, it also indicates that inflation was real
and not just a product of faulty engineering in
the price indexes. Cagan’s result (column 5 of
table 1), and what I have labeled the “ Griliches’
Quality-Ad justed L ist Price Index,” are both not
far from the mark, though Griliches’ original
comparisons were thrown off by the priceconcession error, just as was the
itself.
Around 1960, major revisions were accomplished
in the way quality changes were made in the new
automobile com ponent,9 so one would not expect
the Griliches-Cagan conclusions on quality bias in
the automobile component necessarily to hold
beyond the period they studied. Several more
recent studies 10 have examined automotive com­
ponents of the price indexes for the period after
1960. All of them agree in finding that automobile
price indexes seem biased downward since 1960,
with the major part of the discrepancy occurring
soon after cost-based quality adjustments were
incorporated into the indexes. (See table 2.)
Although the
component declined, the indexes
reported in the studies increased. This group of
studies should be interpreted with caution,
especially since there is the possibility of bias in
both quality measurement techniques employed by
the investigators.11 However, they do strongly
suggest that quality adjustments in the
(and
in th e w p i , sin ce a d ju stm e n ts in both indexes are
based on the same data) m ay have been too large.
If the studies are correct, one should be wary of
concluding that price indexes are always rising too
rapidly.
c p i

c p i

c p i

c p i

Table 2. Percent changes in price indexes for vehicles,
selected periods, 1960-67
A u t h o r o r s o u rc e

A c t u a l C P I ( a u t o c o m p o n e n t ) ______
T r ip le t t ( a u t o m o b ile s ) ......... .. ..........
D h r y m e s ( a u t o m o b ile s ) ............. . . .
H a ll ( p ic k u p t r u c k s ) __________ . . .

P e r io d

1 9 6 0 -6 6
1 9 6 0 -6 6
1 96 1-64
1 9 6 1 -6 7

P e rce n t change

-7 .2
+ 9 .6
1 + 1 3 .2

Appliances offer a fertile field for investigations
into quality change. Two studies have evaluated
quality bias in the
refrigerator com ponent.12
The refrigerator index has been falling continu­
ously through m ost of the postwar period;
interestingly, both studies of refrigerator prices
suggest that the
component m ay have declined
too fast.
Burstein computed his indexes from refrigerator
prices in mail-order catalogs. He expressed sur­
prise at finding that his quality-adjusted index fell
(in a period of falling prices) more than an index
which lacked quality adjustments: “W hy should
procedures ignoring quality changes impart a
downward bias to a price index? A plausible
explanation is that the prices of the . . . models
chosen for pricing by the Bureau fell relative to
the prices of refrigerators and freezers as a
group.” 13
Burstein’s paper opens up some intriguing ques­
tions, but the specific result (that the
m ay be
biased downward) relies exclusively on the be­
havior of indexes of mail-order prices and these
indexes could be given a different interpretation.
However, Burstein’s interpretation is consistent
with a later study by D hrym es.14
Using a variant of the hedonic technique
Dhrym es produced a quality-adjusted price index
for refrigerators covering most of the postwar
period. B etw een 1950 and 1960, Dhrym es’ re­
frigerator index fluctuates, but shows no clear
trend at all. Over the same period, the
re­
frigerator component declined by over one-third.
After 1960, D hrym es’ index moves downward at
about the same rate as the
,
except for a
precipitous and unexplained drop in the final year
of his study. Overall, Dhyrm es’ data are consistent
with downward quality bias in the
refrigerator
component.
Although I have reservations about each of the
studies on refrigerators, both point to downward,
not upward, bias in the
. In the face of such
evidence, one has less faith that price indexes
always drift upward because of quality change.
c p i

c p i

c p i

c p i

c p i

c p i

c

p i

O ther studies

+ 8 .9

i T h is f ig u r e w a s c o m p u t e d f r o m th e la t e r v e r s io n o f D h r y m e s ' p a p e r. In th e e a r lie r
v e r s io n , a 7 .0 - p e r c e n t d e c r e a s e w a s r e p o r t e d .
S O U R C E : S e e th e s t u d ie s c it e d in fo o t n o te 10.




Studies of refrigerator prices

211

Although there has been much recent research
on the problem of quality measurement,15 some of
the studies did not present results that can be

Table 4. Summary of conclusions of several studies of
price indexes and quality change, various periods 1947-66

compared with the
. A noteworthy group
of
studies (unpublished) were carried out within
by Thomas W. G avett.16 Covering automatic
washing machines, men’s suits, and carpets, these
studies are the only ones conducted in conjunction
with close examination of actual pricing and
processing procedures within the indexes for the
products studied. The latter two employed actual
price quotations from the
. Condensed portions
of the results are presented in table 3.
The analyses of washing machines and suits
gave quality-adjusted indexes slightly below the
corresponding
and
components, but their
author judged the differences not statistically
significant, especially since the indexes computed
were, for various reasons, not precisely comparable
with the published indexes with which they are
compared. For the carpet study, the discrepancy
between the quality-adjusted index and the
and
com ponents was considerably larger.
All three of the recomputed indexes are con­
sistent w ith upward quality bias in the respective
components.
c

p i

b l s

w

c p i

w

A u th o r

G a v e t t . ___________
_.
G a v e t t __________
G a v e t t ________
. _____
M a r t in _____
S c it o v s k y _______
B a r z e l. .
.
L a m s o n ___ .

N O T E : “ U p w a rd

M e d ic a l S e r v ic e s ____
T h e a t e r a d m is s io n s . .

196 3-66
1958-6 6
1 95 9- 66
1 9 5 4 -6 1
1 95 1-65
1 9 4 5 -6 4
1 94 7-64

S lig h t u p w a r d b ia s .
S lig h t u p w a r d b ia s .
U p w a r d b ia s .
U p w a r d b ia s .
D o w n w a r d b ia s .
U p w a r d b ia s
U p w a r d b ia s .

c p i

c p i

However, a second study found just the opposite
result: “In the 14 years from 1951-52 to 1964-65,
the costs of treatm ent of all five illnesses covered
by the study (with one minor exception) increased
more— some of them substantially more— than
the
medical care, price index.” 18 A number of
objections have been raised concerning the
m ethodology of the Scitovsky study.19 B u t the
debate actually indicates that the appropriate
pricing concept in the medical services area is not
patently obvious. Nor is the concept of quality in
medical care w ithout serious conceptual dif­
ficulties, which are intertwined with the problem
of defining the appropriate measurement units for
transactions and output.

Table 3. Comparison of percent changes in CPI and
WPI with percent changes in indexes adjusted for quality
change, various periods, 1958-66
C h a n g e In
W PI

p r ic e in d e x c o m p u t e d in th e

c p i

c p i

Q u a lity
a d ju s t e d "

b i a s " m e a n s th e q u a lit y - a d ju s t e d

It has often been argued that one of the defects
in the
medical care components is that they
price units such as “hospital room” and “physi­
cian’s fee,” and that the correct transaction unit—
and therefore the appropriate unit for pricing—is
the cost of a cure. It has been alleged, furthermore,
that a move toward pricing the cost of recovery
from an illness would produce a lesser increase
than the present
procedures.
Such allegations seemed to be borne out in a
somewhat tentative study which related hospital
costs to the length of stay for particular illnesses.17
Though daily charges, in a sample of hospitals,
rose about the same as the
hospital room
component, adjusting for changes in the length of
stay cut the increase in the price index almost in
half. (The years studied were 1954 to 1961.)

p i

Next, we examine a few investigations into
price changes in services. The services components
of the
have consistently risen more rapidly
than price indexes for commodities, taken as a
whole. Of major importance are the medical care
components.

N o q u a lit y
a d ju s t m e n t "

C a rp e ts

C o n c lu s io n

A n ? s e e ss ™ a n
r e ^e v a n t C P I c o m p o n e n t i f p r ic e s w e r e r is in g , o r f e ll m o re th a n
th e C P I c o m p o n e n t , i f p r ic e s w e r e f a llin g . “ D o w n w a r d b i a s " in d ic a t e s th e o p p o s ite
f in d in g .

p i

P e r io d

W a s h in g m a c h in e s . . .

P e r io d

p i

c p i

w

P r o d u c t o r s e r v ic e

C h a n g e in
CPI

b l s

A u t o m a t ic w a s h in g m a c h in e s

1 9 6 3 - 6 4 ____________
1 9 6 4 - 6 5 ____________
1 9 6 5 - 6 6 _____
1 9 6 3 - 6 6 .......................

-1 .7 4
-1 .4 8
-1 .4 6
-4 .6 1

0. 11
- 2 . 23
- 0 .3 3
-2 .4 5

-1 .3 5
-1 .3 6
-.5 8
-3 .2 5

M e n ’ s s u it s

1 9 5 8 - 6 6 _____

2 .0 1

19. 27

22. 07

2 4 .2 5

3 - 4 . 38

* 1 .2 0

C a rp e ts

1 9 5 9 -6 6 ...

2 .9 1

w i l f i t q L n|ityS a d j Us?mef n
0 tr sC a rP e tS ’

-1 1 .7 9

C° m P U te d f r ° m W P ' P riC e q u o t a t io n s ’

?
a u t o m a t ic w a s h in g m a c h in e s , h e d o n ic q u a lit y a d ju s t m e n t s w e r e a p p lie d to a
p r ic e in d e x c o m p u t e d f r o m p r ic e s ta k e n f r o m Consumer's Digest Price Buying
Directory. F o r m e n s s u it s a n d f o r c a r p e ts , h e d o n ic q u a lit y a d ju s t m e n t s w e r e a p p lie d
to th e in d e x d e s c r ib e d in th e f ir s t fo o tn o te .
3 “ S o f t s u r f a c e f lo o r c o v e r i n g s " c o m p o n e n t o f th e W P I
4

R u g s , s o f t s u r f a c e " c o m p o n e n t o f th e C P I ( p e r c e n t c h a n g e f r o m 1 9 5 7 - 5 9 a v e r a g e .)

S O U R C E : U n p u b lis h e d B L S m e m o r a n d u m b y T h o m a s W . G a v e tt.




In order to get around some of these problems,
Yoram B arzel20 constructed a quality-adjusted
medical price index based on insurance rates (for
“Blue Shield” plans). Between 1945-64, the
“Physician’s Fees” index rose by 85 percent, while
Barzel’s index rose only 66 percent.
A final study on services was concerned with
c p i

212

motion picture theaters.21 The results are espe­
cially questionable, since they are derived
almost exclusively from a sample of theaters in
Seattle which were “unchanged” over the period
1947-64. For w hat the findings are worth, the
adjusted (for some quality changes) index for
these theaters rose less than either an index of
their actual ticket prices or the
theater
admissions index.
c p i

Conclusions of the studies noted in this section
are summarized in table 4. W ithout attempting
to evaluate these studies, m ost point to upward
bias in price index components.

Conclusions

The empirical studies surveyed in this paper
do not exhaust the investigations that have a
bearing on the problem of quality error in price
indexes. B ut they show that the index may have
negative as well as positive errors due to quality
changes. The studies of the quality problem are
themselves of uneven quality, so some should
carry more weight than others. Also, a simple
count shows more conclusions of “upward bias”
than of “downward bias.” Nevertheless, those
that show downward quality error indicate that
the widespread view that price indexes always
overstate the degree of inflation may be incorrect.
Notice that these studies do not point to a
positive conclusion: we have not proved that price
indexes are biased either upward or downward;
rather, they establish only that the proposition
that indexes are system atically upward-biased is
not conclusively confirmed by the available evi­
dence. If individual components show both upward

and downward errors, the overall error m ay go
either way.
The reader m ay wonder, however, how it is
possible that quality error can cause price indexes
to be biased downward. Clearly downward bias
can result when deterioration in products and
services is not fully allowed for in the indexes.
There are frequent allegations of this, particularly
in services.
B u t it is also quite possible for quality errors to
cause the index to understate price increases even
when the quality of products is improving. The
reason is that b l s does not simply price whatever
products m ay appear in stores and ignore any
change in quality that m ay occur. Instead, for
m ost products there is an attem pt to control for
quality differences. This means that the prices
that are compared for the index are not neces­
sarily prices of product varieties which show the
average rate of quality improvement.
In order to establish the direction or the size
of quality error in the indexes, we need to examine
the actual quality errors that creep into individual
components. These errors will be determined, not
solely by the extent and rapidity of quality
change in the marketplace, but also by the
particular marketing arrangements for different
products and by the interaction of these factors
with the mechanisms set up by
to try to
control the size of quality errors permitted in
index comparisons. Any extended discussion of
these m atters is beyond the scope of the present
paper. Elsewhere it has been shown that even
when the quality of a product is improving rapidly,
the quality errors that get into the index m ay give
it a downward error instead of the upward bias
so frequently assumed.22
□
b l s

FOOTNOTES
nique. For a com plete discussion of concepts and problems,
see Griliches, op. cit. (1961); Jack E. Triplett, “The
Theory of H edonic Quality M easurem ent and Its U se in
Price Indexes,” BLS Staff Paper Num ber 6; Richard
Stone, Quantity and Price Indexes in N ational Accounts
(Paris, Organization for Economic Cooperation and D e­
velopm ent, 1965); and Thomas W. G avett, “ Quality and a
Pure Price Index,” M onthly Labor Review, March 1967,
pp. 1&-20.

1 “Hedonic Price Indexes for Automobiles: An Econo­
metric Analysis of Quality Change,” Staff Paper 3 in
Price Statistics R eview Com m ittee, The Price Statistics
of the Federal Government, General Series Number 73
(N ew York, N ational Bureau of Economic Research,
1961). Additional results from the study were published
in Zvi Griliches, “ N otes on the M easurement of Price
and Quality C hanges,” in N ational Bureau of Economic
Research, Conference on R esearch in Incom e and W ealth,
Models of Income Determination, Studies in Incom e and
W ealth, Volum e 28 (Princeton, N .J ., Princeton U niver­
sity Press, 1964).

3
“ Measuring Quality Changes and the Purchasing
Power of M oney: An Exploratory Study of A utom obiles,”
N ational B anking Review, D ecem ber 1965, pp. 217-236;
reprinted in Zvi Griliches, ed., Price Indexes and Quality

2 This is an intuitive explanation of the hedonic tech­




213

Change: Studies in New Methods of Measurement (Cam­
bridge, Harvard U niversity Press, 1971).

pp. 408-417; Phoebus J. Dhrymes, “ On the M easurem ent
of Price and Quality Changes in Some Consumer C apital
4
This technique rests on the hypothesis th at prices Goods,” American Economic Review, M ay 1967, pp. 50 1 518, and “Price and Quality Changes in Consumer Goods:
for used cars of different ages differ partly because the
An Empirical S tu d y,” in Griliches, ed., op. cit. (1971);
older cars have less useful life remaining and partly be­
Richard J. Olsen, “ Some Aspects of Q uality Change as an
cause newer cars differ in quality from older ones. If a
Economic Variable,” (R utgers U niversity, unpublished
pure rate of depreciation can be established to allow for
Ph. D . dissertation, 1968); and (on pickup trucks, a closely
the aging effect, the remainder of the price difference can
related product entering the w p i but not the c p i ) , R obert E.
be taken as an estim ate of the value of quality change.
Hall, “ The M easurem ent of Quality Change from V intage
* Cagan, op. cit., p. 231.
Price D a ta ,” in Griliches, ed., op. cit. (1971).
6 An estim ate of the value of quality changes made in
11 See Triplett, op. cit. (1969). An elaboration of this
the auto com ponent over an extended period of tim e is
point appears in the full report from which the present
contained in Olga A. Larsgaard and Louise J. M ack,
article is condensed.
“ Compact Cars in the Consumer Price Index,” Monthly
12 M eyer L. Burstein, “ M easurem ent of Quality Changes
Labor Review, M ay 1961, pp. 519-523.
in Consumer D urables,” The Manchester School, Septem ber
7 I do not w ish to convey the impression th at concessions
1961, pp. 267-279, and Dhrym es, op. cit. (1971). A number
data were introduced into the index in this fashion simply
of other studies were carried out on various w p i com ponents
because the b l s was unaware of its im pact. When the
and are discussed in the full report.
concession information becam e available, it was argued
13 Burstein, p. 279.
within the Bureau th at the price after any concession
had always been the appropriate concept for the c p i ,
14 D hrym es, op. cit. (1971).
and that linking concessions would ignore real price
15 See Zvi Griliches, ed., op. cit. (1971).
change that occurred before the first reports on dealer
price concessions. T hat is, it was believed th at at some
18
These are contained in the unpublished G avett m em ­
orandum referred to in the text.
previous period (perhaps around 1950) cars had actually
sold at list prices. (Indeed, popular reports indicate trans­
17 Leonard W. M artin, “Pure Price Indexes, Quality
actions prices were well above list prices in the im m ediate
Change, and H ospital C osts,” Proceedings of the American
postwar years.) Since these price decreases had been
Statistical Association, Business and Economic Statistics
missed by the c p i , bringing concessions in by direct com­
Section, 1966, pp. 479-487.
parison, it was argued, preserved the valid ity of the index
in making comparisons, such as for the period 1950-57.
18 Anne A. Scitovskv, “ Changes in the Costs of Treat­
ment of Selected Illnesses, 1951-65,” American Economic
8 It should be emphasized th at the com putations reported
Review, D ecem ber 1967, pp. 1182-1195.
here are based on G avett’s data, but th ey are n ot the con­
structions produced by G avett (except where indicated),
nor are th ey used for the same purposes. Therefore, it
should not be inferred th at G avett is necessarily in agree­
m ent w ith anything expressed in this section.
8 E stim ates of the cost of quality changes were obtained
from manufacturers and used as adjustm ents in both the
c p i and the w p i . See M argaret S. Stotz, “Introductory
Prices of 1966 Automobile M odels,” M onthly Labor Review,
February 1966, pp. 178-181; and E thel D . H oover, “The
C PI and Problems of Q uality Change,” M onthly Labor
Review, N ovem ber 1961, pp. 1175-1185.

19 See, for example, Yoram Barzel, “ Costs of M edical
Treatm ent: C om m ent,” American Economic Review, Sep­
tem ber 1968, pp. 936-938.
20 “P roductivity and Price of M edical Services,” Journal
of P olitical Economy, N ovem ber-D ecem ber 1969, pp.
1014-1027.
21 R obert D . Lamson, “ M easured P roductivity and Price
Change: Some Empirical E vidence on Service Industry
Bias, M otion Picture T heaters,” Journal of P olitical Econ­
omy, M arch-A pril 1970, pp. 291-305.

22 Jack E. Triplett, “Quality Bias in Price Indexes and
10
These include (in order of com pletion): Jack E.
N ew M ethods of Quality M easurem ent,” in Zvi Griliches,
Triplett, “Automobiles and Hedonic Quality M easure­
ed., op. cit. (1971).
m ent,” Journal of Political Economy, M ay-Ju n e 1969,




214

Technical Note

The U se of Price Indexes

contracting parties to decide. But, for those who
are interested in escalation, the article highlights
some of the essential qualities of the data they
m ay specify in the contract, shows how these
data should be described in the agreement, and
suggests techniques for adjustments.

in Escalator Contracts
I n lo n g - t e r m contracts governing wages, rents,
continuous or future delivery of a product,
alimony payments, administration of legacies,
and delivery of a new product for which the seller
has no satisfactory cost estimate, to name a few
examples, changes in the purchasing power of the
dollar pose a problem because they are beyond the
control of the contracting parties. One method
the parties use to protect themselves against un­
foreseen price change, especially in times of inflation,
is the escalator clause. Essentially, this attem pts
to have the transaction price represent constantvalue units as measured by the quantity of goods
and services which a given amount of money will
buy. It usually employs a price index as an
objective means of adjusting the actual price.
One advantage of escalation is that the techniques
and measures used to convert monetary units
into constant-value units are normally mechanical
and, once established, cannot be manipulated by
either party. Another is that it is relatively in­
expensive to administer since, after the mechanics
have been agreed upon, very little computing is
required.
This article discusses the techniques of escala­
tion using the two major price indexes published
by the Bureau of Labor Statistics— the Consumer
Price Index (CPI) and the Wholesale Price
Index (W PI). It does not discuss the pros and
cons of using one type of index or data as against
another, or the desirability of escalation in pref­
erence to other means of protecting against price
change.1 Both of these are matters for the

BLS Data for Escalation Purposes
The Consumer Price Index. The CPI measures
changes in the cost of a list of goods and services
which represents the item s important in the
expenditures of urban wage earners and clerical
workers and their families. It does not measure
their actual expenditures or their total cost of
living, both of which include outlays for such
purposes as income taxes, contributions to charity,
and personal insurance— things which the workers
and their families do not “consume.” Nor does
it measure the cost of changes in the manner or
level of living which are typically associated with
changes in income, size of family, the age of
fam ily members, etc. It does, however, measure
changes in the prices of things which the '‘average”
fam ily normally buys for current consumption
and, conversely, the purchasing power of the
dollar spent by workers and their families' as a
group.
The same items are priced month after month,
using specifications to insure that identical quali­
ties are priced, in about 50 cities These cities
represent all urban areas from metropolitan New
York C ity to communities with as few as 2,500 i
i O th e r m e th o d s in c lu d e h e d g in g , w h ic h in v o lv e s u se o f a c o u n te r b a la n c in g
tra n sa ctio n ; co st p lu s c o n tr a c ts , w h ic h p la c e s th e risk o n o n e o f th e parties;
ta r g e t or in c e n tiv e c o n tr a c ts , w h ic h s tip u la te th e o rig in a l p rice a n d a fee,
w it h th e fee in cr ea sed if c o s ts a re d ecrea sed ; a n d d e liv e r y p rice c o n tr a c ts,
w h ic h p r o v id e t h a t price w ill b e d e te r m in e d b y m a r k e t o r c o s t c o n d itio n s at
th e tim e o f fu tu r e d e liv e r y .

From the Review of August 1963




215

R ecently, an insurance com pany began issuing
life insurance policies which contain a provision
that benefits will be increased in proportion to the
rises in the CPI.

residents. Price trends in each city affect the
U nited States index according to population.
W ith in e a c h c ity , p r ic e c h a n g e s f o r t h e s a m p le
g o o d s a n d s e rv ic e s a r e c o m b in e d w ith w e ig h ts
b a s e d o n th e im p o r ta n c e in f a m ily e x p e n d itu r e s
o f t h e s a m p le ite m s a n d th e r e la te d ite m s w h ic h
t h e y r e p r e s e n t. T h u s , if fa m ilie s m a k e 1 p e r c e n t
o f th e ir o u tla y s f o r m ilk a n d 20 p e r c e n t f o r r e n t,
a 5 - p e r c e n t ris e in r e n t s w o u ld h a v e 20 tim e s as
m u c h e ffe c t o n t h e in d e x as a 5 - p e r c e n t ris e in
m ilk p ric e s.

Once the item sample has been determined, it
stays fixed until the next major weight revision 2
or until there is clear evidence that an alteration
in the list of goods or services is called for. For
example, as wages and their purchasing power
rise, workers begin to spend proportionately less
for food and other necessities and eventually the
index weights m ust be revised. Also, new items
m ust be added from tim e to time as they become
important— such as television sets and nylon hose.
T h e C P I is p u b lis h e d a b o u t th e 2 5 th o f th e
m o n t h fo llo w in g t h a t to w h ic h th e in d e x a p p lie s
a n d re fle c ts p ric e s c o lle c te d a t v a r y in g d a te s d u r in g
t h e e n t ir e m o n th . S e p a r a te in d e x e s a r e a v a il­
a b le f o r m o s t o f t h e la r g e s t c itie s: f o r N e w Y o rk ,
C h ic a g o , L o s A n g e le s, D e tr o it , a n d P h ila d e lp h ia
o n a m o n t h l y b a s is ; f o r t h e o th e r s o n a q u a r te r l y
b a s is . I n a d d i tio n to th e to ta l, o r th e A ll I te m s
in d e x , s e p a r a te in d e x e s a r e c a lc u la te d f o r m a jo r
c a te g o rie s q f f a m ily s p e n d in g : F o o d , h o u s in g ,
a p p a r e l, t r a n s p o r t a t i o n , e tc .3

The C PI has been used for m any years as a
wage escalator in labor-management contracts.
I t is estim ated that about 2 million workers are
now covered b y such agreements. The index is
also used to a lesser extent to adjust rent pay­
m ents, royalties, pensions, and alimony payments. *

The Wholesale Price Index. The W PI is a general
purpose index designed to provide a continuous
m onthly series showing price changes, singly and
in combination, for all commodities sold in primary
markets of the United States. The index meas­
ures the general rate and direction of price m ove­
ments in primary markets and the specific changes
for individual commodities or groups of commodi­
ties. It is based on a sample of over 2,100 com­
modities chosen to represent a wide variety of
com modity specifications and markets. The
prices used in constructing this index are those
which apply at the first important commercial
transaction for each commodity. M ost are the
selling prices of representative manufacturers
or producers or prices quoted on organized ex­
changes or markets. The basic weights are total
transactions as reported in the latest industrial
censuses.
The index is intended to measure price changes
between two periods of time, excluding the in­
fluence of changes in quality, quantity, terms of
delivery, level of distribution, unit priced, or
source of price. To accomplish this, the index
calculations are based on the relative change from
one period to the next in prices of identical or
nearly identical item s, as defined by precise
specifications.

s T h e s e r e v is io n s , b a se d o n d e ta ile d s u r v e y s o f w o r k e r s ’ in c o m e s a n d e x p e n d ­
itu r e s , are m a d e a t in te r v a ls o f a b o u t 10 y e a r s. T h e n e x t r e v is io n is s c h e d ­
u le d for c o m p le tio n w it h t h e J a n u a r y 1964 in d e x . A s u m m a r y o f th e m a jo r
c h a n g e s in c id e n t to t h e r e v is io n w a s p u b lis h e d in t h e J u ly is s u e o f t h e R e v i e w ,
p p . 794-795.
8 A m o r e d e ta ile d d e s c r ip tio n o f t h e in d e x a s c u r r e n tly c a lc u la te d is a v a il­
a b le o n r e q u e s t.
* O fficia l m o n t h l y in d e x e s are a v a ila b le s e p a r a te ly for s o m e o f t h e m a jo r
g r o u p s o f c o m m o d itie s , a s w e ll a s for t h e t o t a l, c o n t in u o u s ly s in c e 1890. A
fin er c la s s ific a tio n b y s u b g r o u p s o f c o m m o d itie s is a v a ila b le s in c e 1913.
I n 1952, t h e th ir d le v e l o f c la ss ific a tio n — p r o d u c t cla ss— w a s in tr o d u c e d ;
t h e s e h a v e b e e n e x te n d e d b a c k t o 1947.
5 Q u e s tio n n a ir e s w e r e s e n t t o t h e 2,700 n a m e s o n t h e m a ilin g lis t for th e
m o n t h l y p r e ss r e lea se a n d t o th e 4,200 w h o r e c e iv e th e d e ta ile d r e p o r t. T h e
n u m b e r o f u s a b le r e tu r n s to ta le d 3,026.




216

T h e b a s ic A ll C o m m o d itie s in d e x is d iv id e d in to
15 m a jo r g r o u p s a n d a b o u t 80 s u b g r o u p s . I n
a d d itio n , s o m e 3 0 0 “ p r o d u c t c la s s ” in d e x e s ,
w h ic h g r o u p c o m m o d itie s c h a r a c te r iz e d b y s im ila r ­
i t y o f r a w m a te r ia ls , p r o d u c tio n p ro c e sse s, o r
e n d u se , a r e o f p a r ti c u la r i n te r e s t to u s e rs o f
e s c a la to r c la u s e s .4 T h e B u r e a u w ill also , u n d e r
c o n t r a c t, c o n s tr u c t in d e x e s f o r s p e c ia l c o m b in a ­
tio n s o f in d iv id u a l se rie s to m e e t t h e sp e c ific a ­
tio n s of th e p a r ti e s to a n e s c a la to r a g r e e m e n t.
A 1961 s u r v e y o f W P I u s e rs 5 r e v e a le d t h a t
932 c o m p a n ie s o r in d iv id u a ls u s e d t h e W P I f o r
th e e s c a la tio n o f s a le s o r p u r c h a s e c o n t r a c ts
t o ta lin g n e a r ly $14 b illio n . T h e in d e x e s m o s t
f r e q u e n tl y sp e c ifie d in e s c a la to r c o n t r a c ts a r e
sh o w n o n th e fo llo w in g p a g e .

N um ber of
con tracts

Metals and metal products_________________________
All com m odities___________________________________
All commodities other than farm
products and foods_______________________________
Finished steel products_____________________________
Steel mill products_________________________________
Iron and steel______________________________________
Machinery and m otive products____________________
Electrical machinery and equipm ent________________
Machinery and equipm ent—________________________
Structural steel shapes______________________________
Petroleum and products____________________________
Industrial chem icals________________________________
General purpose machinery and equipm ent_________
Chemicals and allied products______________________
Carbon plates______________________________________
Crude petroleum ___________________________________
Lumber and wood products_________________________
Specially constructed index for
metals and metal products________________________

178
124
87
87
73
66
24
23
19
19
19
16
15
14
13
10
10
10

Elements of Escalation
There are three major elements in an escalator
clause contract:
1. Establishm ent of the initial price or rate at
the time of the contract. The escalator clause
protects against radical changes in real costs from
the original estim ate; it cannot correct an errone­
ous or inequitable original price. In fact, if the
original price is incorrect, almost any escalator
clause will exaggerate the error over time.
2. Selection of an appropriate escalating index.
Escalation is usually based on an index which is
assumed to represent the com modity or service
being escalated. The escalator, then, is subject
to any lim itations inherent in the escalator index.
The CPI is generally used for escalating wages
and items sold at retail levels; the W PI is more
often used for adjusting prices of raw materials
or production equipment, industrial rents, etc.
3. Procedures for carrying out the escalation.
Six basic points are usually defined in escalator
mechanisms:

W eigh t
B L S

Carbon steel p late-------------------------Carbon steel sh eet_________________
Electrical sh eet____________________
Steel forgings______________________
Copper w ire_______________________

C ode

10-14-26
1 0-14-46
1 0-14-50
10-15-71
10-26-01

{p e r c e n t)

50
20
15
7

B. Reference Dates of Escalation. The date to which
the index being used as the base of the escalator
refers should always be indicated. The reference
base period is usually not the same as the base
period of the price index or series, for example, the
reference index may be the CPI for March 1962,
stated on the 1957-59 official base period. The
reference date of indexes on which subsequent
changes are to be computed should also be speci­
fied. The parties may prefer— for the escalator
base or the subsequent adjustments— to use a
particular m onth’s index, an annual average, or
an average for 3 months, or 6 months, or 5 years,
or any other period or date, whatever suits their
purpose. In any event, the contract should
specify precisely the reference dates of the indexes
to be used.

A. Identification of the Index To Be Used. The
index used as the escalator should always be
completely identified regardless of whether it is
a widely known index or a special combination of
individual series or categories. Exact title and
the index base period should be indicated. For
example, an adequate identification would be:
The Consumer Price Index, All Item s, U .S., 1957-




5 9 = 1 0 0 , or the Wholesale Price Index, All Com­
modities (Except harm and Food Products),
1957-59= 100, issued by the U.S. Bureau of Labor
Statistics. The indexes are published first in a month­
ly press release, about 2 weeks later in a detailed
report, and, about a month thereafter, in the
Current Labor Statistics section of the Monthly
Labor Review. The publication to be used should
be specifically named in the contract.
If the contract is based on the W PI, it is safer
to specify whether the preliminary or final index 6
is to be used. The CPI is final on first publica­
tion. If an item or group or specially computed
combination of individual W PI series is to be
used, the BLS code or category numbers should
be included in the identification.
In specially computed indexes based on either
the CPI or W PI, the relative weights of items
should be specified. For example, an index for
escalating the price of turbines might assign the
major components of the product the following
relative importances to reflect changes in material
costs:

« I n d e x e s are c o n sid ere d p r e lim in a r y for 1 m o n th , or u n t il t h e Ind ex for th e
m o n th fo llo w in g t h e d a te o f r eferen ce Is p u b lis h e d .

217

C. Frequency of Adjustment. It should also specify
the effective dates of adjustments. The parties
m ay agree that adjustments are to be made
quarterly. For example, if the index goes up
or down by a specified number of points or fraction
of a point by the end of the quarter, the change in
wages, rents, etc., takes place automatically at a
stipulated time. If the index does not change by
at least this amount, then no change is called for
in the wage rate, rent, or product price.
On the other hand, the change in the paym ent
m ay be required whenever the index reaches a
certain point— 118.0 (1947-49= 100), 120.0, etc.—
or when it changes b y a specified amount. Thus,
if an increase of 1 cent in a wage rate is called for
whenever the index moves up 0.5 point, the time
interval is immaterial— it m ay be 1 m onth or 6,
or 12, etc.
Two factors should be considered in deciding
the frequency of adjustments:
1. Too frequent adjustments m ay create some
difficulties because of the seasonal or erratic
m ovem ents of prices, particularly for farm prod­
ucts and foods. As commodities move up the
processing scale away from raw materials into
more highly fabricated goods, seasonal price
changes tend to become progressively less impor­
tant. Use of quarterly, semiannual, or annual
average indexes will minimize such periodic
fluctuations and result in a smoother adjustment
pattern.
Conversely, in a period of continuous price
m ovem ent in one direction, infrequent adjust­
m ents m ay understate the true change somewhat,
since escalator clauses adjust only for what has
already happened. When prices are rising, pay­
ments do not increase as rapidly as the index;
when prices are falling, payments do not
decrease as fast as the index.
2. The time lag between collection of the basic
price data and publication of the indexes does not
permit the contracting parties to time adjustments
to coincide with the occurrence of price changes.
For the CPI and the W PI, 4 to 6 weeks elapse
between the collection and the release of the
index, even in preliminary form. If a final
index is used, then the lag increases by at least a
7 T h e referen ce b a se for b o t h t h e C P I a n d t h e W P I w a s c h a n g e d In 1962
fr o m 1 9 4 7 -49= 100 t o 1 9 57-59= 1 0 0 . A lth o u g h t h e In d e x e s are a lso a v a ila b le
o n t h e 1947-49 b a se , u se rs s h o u ld c o n s id e r s h iftin g t o t h e n e w b a se a s s o o n as
p r a c tic a b le .




month. Unless provision is made for this re­
porting lag, the understatement of the true price
change 'will be intensified when prices are changing
rapidly. In m any instances, particularly for
rents, retroactive payments are called for in
order to correct the time lag.
D. The Mechanics of Adjustment. The heart of the
escalator clause is the method of adjustment,
which can be varied in m any ways, depending on
the purpose for which the index is to be used and
the wishes of the contracting parties. There are
two basic methods of adjusting paym ents in
accordance with a price index— one is to apply to
the price (or the wage rate) some multiple of the
percentage change in the index; the other is to
provide that for each specified absolute change in
the index, the price will change by some specified
amount. Either method is satisfactory. H ow­
ever, unless the base index is exactly 100.0, a
change of one index point is not the same as a
change of 1 percent. When the index is greater
than 100, a change of an index point is less than
a change of 1 percent and when the index is less
than 100, a change of a point is more than a
change of 1 percent. Therefore, a clause might
read that prices will change 1 percent for each
1-percent change in the index or that prices will
change a given dollar-and-cents amount for each
1-point change in the index.
M any wage agreements contain a cents-to-point
relationship, requiring that wage rates be upped
by a cent for every change of one-half (or 0.5)
point from the base index. Others condition the
change on 0.6 point, a whole point, etc. Such a
relationship frequently is predicated on the wage
rate-price index relationship at the time of the
agreement. For example, if the base index for
escalation is 120.0 and the average wage is $2.40
per hour, then one index point is equivalent to 2
cents and one-half point equals 1 cent. To main­
tain this relationship, the parties m ay agree on a
1-cent wage increase if the index moves up onehalf point. Under this type of agreement, a new
cents-to-point relationship m ust be calculated
when the BLS changes its index base period,
because rebasing changes the value of an index
point.7 The index-wage relationship can also be
of the percentage type. This is less frequently
used in wage agreements than the cents-to-point
relationship— perhaps because the resulting wage

218

The
escalator clause should specify whether adjust­
ments will be made for index changes in either
direction or only one. If an agreement mentions
increases only, presumably decreases are not
contemplated. Some clauses, on .the other hand,
specify that wages, rents, etc., are to move down
as well as up but are not to drop below a specified
minimum; for example, a wage escalation clause
may call for a 1-cent decrease for every 0.4 index
point down to an index level of 97.8.

viously published as final, because of late reports
or errors. The contract should specify whether or
not account is to be taken of such corrections.
For statistical accuracy, the Bureau is com­
mitted to keep the composition of the indexes in
line with prevailing conditions. In both the CPI
and the WPI, the BLS revises commodity specifi­
cations, adds new products, discontinues obsolete
items, and, from time to time, revises the weighting
structure and reference base.
Escalator mechanisms cannot be controlled by
either pai;ty, so agreements often stipulate a
procedure to follow if the escalator mechanism
changes or disappears. In most cases, this pro­
cedure simply states that the original issuing
agency will be sole judge of the comparability of
successive indexes, and that if the agency cannot
supply indexes which are comparable, a named inde­
pendent authority (such as the dean of the business
school or the head of the economics department in
the State university) will select a method of con­
tinuing the contract. When the relationship is
one of cents to point as in a wage contract, the
parties may want to renegotiate. For this reason,
the Bureau gives notice of anticipated changes in
the official indexes at least 6 months in advance.

F. P ro v isio n f o r R evisio n o f the In dex. The Bureau
occasionally publishes corrections of indexes pre­

D iv isio n o f In d u strial Prices an d P rice In d exes

rates might result in fractions of a cent or because
the concept is not quite as simple.
For leases and other long-term price agreements,
the percent-of-change technique offers no diffi­
culty. If this method is adopted, the clause
should specify how the change is to be computed;
for example, the parties may decide that an index
change of 12.5 percent is to result in a price ad­
justment of 12 percent, 12.5'percent, or 13 percent,
or they may work out a schedule of changes, as in
the cents-to-point contracts. As both the CPI
and the WPI are published to one decimal place,
it is desirable that contracts refer to the indexes
in these terms.
E . U p p e r a n d L o w er L im its o f A d ju stm e n t.




— F

219

r a n c is

S.

C

u n n in g h a m

Postwar
price cycles:
a new
chronology

Fluctuations in
the rate of change
of consumer prices
generally match changes
in economic activity
GEOFFREY H. MOORE

I n f l a t i o n is characterized by a general and widely
diffused rise in prices and costs. However, all
prices and factors affecting prices do not begin to
rise or fall at the same time. Moreover, prices
do not all move at the same pace. These differ­
ences in price behavior have significant conse­
quence. Real wages— m oney wages adjusted for
price changes— m ay rise or fall, with vital effects
on the wage earner and his family. Profit margins,
dependent on the difference between prices and
costs, m ay rise or fall, thereby encouraging or
discouraging expansion of production, hiring of
workers, developm ent of investm ent plans, or
shifts of resources from one activity to another.
This article sets forth some of the results of a
recent study of the cyclical behavior of prices.1
I t describes a new chronology of fluctuations in
the rates of change in the price level, relates these
fluctuations to those in overall economic activity,
examines the extent to which price changes involve
the entire price system, measures the tendencies
of some prices to lead and others to lag, and
sketches the relationship of price cycles to changes
in costs and profits. Recent developments are
touched on wdth a very broad brush.

meaning of “shaded areas” in charts of m onthly
time series.
In view of the usefulness of such a framework,
it seems sensible to adopt a similar strategy for
studying m ovements in the price system . To do
so, however, a number of questions have to be
faced. Should the chronology represent peaks and
troughs in the level of prices or in their rate of
change? If the latter, how should the rate of
change be measured? monthly? quarterly? W hat
type of data should be used: unadjusted or sea­
sonally adjusted? W hat index or set of indexes
of prices should be used to construct the chro­
nology? W hat criteria should be set up to define
the chronology and identify its turning points?
The business cycle chronology was based on the
working definition of business cycles set forth by
M itchell in his 1927 volume, Business Cycles— The
Problem and Its Setting,2 and later refined by
Arthur F. Burns and M itchell in their 1946 mono­
graph, Measuring Business Cycles.3 In brief, the
definition applied three criteria to the problem:
the magnitude, the duration, and the diffusion of
fluctuations in economic activity. One inquired
how large the decline or rise in aggregate activity
was, how long it lasted, and how widely it was
diffused over different economic sectors. Turning
points were identified not by referring to a single
aggregate, such as gross national product, but by
determining the consensus among a number of
series, each of which had some claim to represent
or reflect total economic activity.
There is much to be said for developing a price
chronology in a similar manner. W hether it is the
level of prices or their rate of change that is
selected as the ultimate variable, attention should
naturally be focused upon swings that are of sub­
stantial size, last more than just a few months, and
are widely diffused throughout the price system .

Reference chronology for prices

A simple y et effective device for studying busi­
ness cycles is the National Bureau of Economic
Research’s reference chronology of peaks and
troughs in economic activity, created by Wesley
Clair Mitchell. It is a widely used device in tracing
fluctuations in the economy and has imprinted
upon the minds of many economic statisticians the

Geoffrey H . Moore is Commissioner of Labor Statistics,
Bureau of Labor Statistics.

From the R ev iew of December 1970



220

To aid us in identifying turning points in the
price cycle, we turned to a N ational Bureau of
Economic Research computer program recently
developed by Charlotte Boschan and Gerhard Bry.
Essentially, this program reproduces, in an ob­
jective and mechanical fashion, m ost of the choices
of “specific cycle” turning points that used to be
entirely dependent upon the judgm ent of National
Bureau staff. Of course, it uses criteria th at are
similar to those used by the staff. It bases its
choices upon whether the fluctuations in the data
are sufficiently large and long enough to be re­
flected in various moving averages, but does not
explicitly use any criterion as to the size of a
swing. D espite this, it is rather uncanny in its
ability to detect and identify turning points
independently selected by experts. We used the
turns selected by the computer program in a
large m ajority of instances. The exceptions were
due to the occasional failure of the program to
mark a large m ovem ent because it is too short,
or (more frequently) to mark very small m ove­
ments simply because they last quite long.
After deciding upon the rate of change in prices
as the variable that the chronology would rep­
resent, several other decisions remained. First, the
rates of change had to be seasonally adjusted or
derived from seasonally adjusted indexes. During
the past year the Bureau of Labor Statistics has
been reporting the seasonally adjusted rate of
change in the c p i . The seasonal pattern has a
relatively small effect upon the level of the index
(currently the largest and the smallest seasonal
factors are, respectively, 100.12 in July and 99.83
in January and February). Nevertheless, it has a
substantial effect upon rates of change over short
periods. For example, the rate of change from
July 1969 to January 1970 is raised from an annual
H§tte of 5.7 percent to 6.3 percent after seasonal
adjustment, which is equivalent to dividing a
seasonal index of 90 into the unadjusted rate.
This seasonal effect has been powerful enough to
cause the unadjusted July to January rates to be
lower than either the preceding or the following
January to July rates in 4 years out of the past 5.4

I t would be convenient to depend upon a single
general price index for this purpose. Unfortunately,
although the idea of an index of the general price
level is an ancient one, there is today no single
widely accepted measure of it. The three leading
candidates would be the Consumer Price Index,
the W holesale Price Index, and the Gross National
Product Deflator. Each of these has its merits and
deficiencies for the purpose.
The Deflator is quarterly, whereas the other two
indexes are monthly, and other things being equal,
a m onthly chronology would be preferred. The
Deflator has the largest economic coverage, but
that also means it includes some dubious elements,
notably in the government sector where the “price”
is really a wage rate. For this reason many consider
the Private g n p Deflator a better price index.
The Deflator is affected not only by changing
prices but also by changing weights, because it is
derived by dividing current dollar g n p by an
estimate of g n p in constant dollars, whereas the
other two indexes use fixed weights and hence
reflect price changes alone.
The Wholesale Price Index, of course, covers
only one part of the price system — commodities,
not services— has some gaps in its industrial
coverage and depends in part upon list prices
rather than actual transaction prices. The Con­
sumer Price Index is the closest approximation of
the three to an actual transaction price index but
is limited to prices paid by urban wage earners’ and
clerical workers’ families. Unlike the other two,
it includes prices for existing goods, such as houses
and used cars, as well as for newly produced goods
and services.
These considerations do not point to a clear-cut
conclusion, except to suggest a real need for a
m onthly general price index. Lacking this, I have
based the chronology in this paper upon the Con­
sumer Price Index, using the g n p Deflator and
the Wholesale Price Index, and some of their
principal components (for example, the Private
Deflator and the wpi for industrial commodities)
to provide supplementary evidence. The c p i has
risen almost continuously since 1954, but there
have been sizable fluctuations in its rate of in­
crease, and the chronology identifies these fluctua­
tions. The rate of inflation is, of course, of major
concern. The chronology shows when this rate, as
measured by the c p i , reached high points and low
points since 1947.




Next, it is necessary to determine precisely
how the rate of change is to be measured. The
range of possibilities is wide. The interval over
which change is measured can be as short as 1
month or as long as 12 months or more. Monthly
indexes can be averaged over calendar quarters,

221

Postwar price cycles

or over moving 3-month intervals and rates of
change measured between these averages. More
complicated smoothing formulas can be applied.
Generally, m onth-to-m onth changes are highly
erratic, so some form of smoothing is desirable.
On the other hand, smoothing formulas can twist
and distort cyclical patterns and timing relation­
ships. After some experimentation I concluded
that the rate of change over a 6-month span m et
reasonably well such criteria as smoothness,
sim plicity, and limited distorting effects, for the
c p i and m ost other price and wage series. For
series that are available only in quarterly form,
quarter-to-quarter changes are used. Occasionally
we use changes over 12-month or 4-quarter spans,
when these are the only data available or when
the 6-month or 1-quarter rates are unduly erratic.

Chart 1.

Taking into account the foregoing considera­
tions, chart 1 shows the reference chronology,
based upon the rate of change in the Consumer
Price Index, together with rates of change in the
other comprehensive indexes mentioned earlier.
Six contractions in the rate of change are identified:
in 1947-48, 1950-52, 1953-54, 1956-58, 1959-61,
and 1966-67. We have marked a tentative peak in
February 1970. If this peak is confirmed b y data
later this year and in 1971, this will m ark the
beginning of the seventh contraction since 1947.
Taking the 23-year period between the 1947 and
1970 peaks, we find that expansions in the Irate of
change lasted 162 months in the aggregate, while
contractions covered 106 months. T hat is, although

Rates of change in comprehensive price indexes

1947 48

49




50

51

52

53

54

55

56

57

58

59 60

222

61

62

63

64

65

66

67

68

69 1970

importance of foods in family budgets has had the
effect of preventing declines in the rate of change of
the cpi from reaching as low a level in recent
years as they did earlier in the postwar period.

the Consumer Price Index has been generally
rising during this period, the rate of increase has
declined over long stretches— aggregating nearly
9 years.
The other indexes show broadly similar fluctua­
tions, but with exceptions, especially in the period
1959-64. In terms of these comprehensive indexes,
therefore, the chronology seems to represent
fluctuations that are widely diffused in the price
system . This m atter will be examined more
directly later.
During the first three contractions in the rate
of change in the cpi, the rate fell below zero;
that is, the index declined. B ut the rate barely
reached zero in the next two contractions (1958
and 1961), and did not do so at all in the last one
(1967). Indeed, the level of the rate at its succes­
sive low points becomes progressively higher
throughout the period. There is a related tendency
for the declines in the rate to become progressively
smaller. In the first two contractions the rate
dropped 18 and 15 percentage points; in the next
two, 3 and 4% percentage points; and in the last
two, 2 and 2% percentage points. (See table 1.)
However, the high points in the rate have not
become progressively higher, nor have the expan­
sions become progressively larger. If there has
been a rising floor under the rate, there has not
been a rising ceiling also. One possible explanation,
which needs further exploration, is that the rising
importance of services, and the diminishing

Price cycles and business cycles

How does the price chronology compare with the
business cycle chronology? Four of the price
contractions correspond with the four business
contractions of 1948-49, 1953-54, 1957-58, and
1960-61. B ut the business expansion of 1949-53
was interrupted by the price contraction of
1950-52, and the long business expansion that
began in 1961 was interrupted by the price
contraction of 1966-67. Each of these inter­
ruptions was also characterized by some hesitancy
in business as well. Hence there is a notable
degree of correspondence between the behavior of
the rate of change in the Consumer Price Index
and general economic activity. Since World War
IT, every economic slowdown or actual recession
has been accompanied by a cyclical contraction in
the rate of change in the price level, and cyclical
contractions in the rate of change in the price level
have not occurred at other times.
This is not to say, however, that a business
recession is a necessary condition for a reduction
in the rate of inflation. As already noted, two
such reductions since 1947 have occurred at times
when the economy merely slowed down. Moreover,
several of the declines in the rate of price rise that

Table 1. Comparison of peaks and troughs in the rate of change of the Consumer Price Index (all items) with those for
selected price indexes, 1947-70
Lead ( —) or lag ( + ) in months, at turns in the Consumer Price Index, all items

Peaks and troughs in the rate of change
in the Consumer Price Index, all items

Consumer Price Indexes for—
Peak or
trough

Date

Rate
(percent)

Food

Peak..
Trough
Peak..
Trough
Peak..
Trough
Peak..
Trough
Peak..
Trough___
Peak...........
Trough___
Peak...........

October 1947.........
November 1948...
November 1950...
November 1952...
July 1953................
August 1954..........
July 1956...............
July 1958...............
July 1959...............
March 1961............
January 1966.........
January 1967.........
February 1970«...

13.8
- 4 .3
14.3
- 0.6

0
0
0
+3

-

2.1
1.2

4.3
-

0.2

2.3

0

4.1
1.6
6.7

Median lead ( - ) or lag (+ ), in months.
1 Comparable data not available prior to 1956.
3 Tentative.
3 No timing comparison.




+2
-3
0
+10
0
1
0

0

Other
commodities1

+5
0
-3
-1 0
(»)
<*)

-1 .5

Wholesale Price Indexes for—

Services *

+7
+2
-1
+2
+5
+3

+ 2 .5

All
commodities
0
+3
0
-1 7
-3
-1
-5
(>)
(*)
0
0
-1

- 0 .5

Industrial
commodities
0
+5
-i
-1 5
3
-1 0
-1 1
-8
-5
-6
+3
-3

-4

Consumer
finished goods
0
0
-2
0
+7
-1
+17
+1
+7
-1
-1
-1

- 0 .5

GNP implicit
price deflator

0

+5
+2

+5

+6

-1
0
<s)
(>)

+4
+3
+3

+3

NOTE: Rates of change in the Consumer and Wholesale Price Indexes are computed
over 6-month spans, centered, seasonally adjusted at annual rate. Rates of change
in the GNP deflator are computed from quarter to quarter, centered, seasonally adjusted
at annual rate.

223

were associated with business cycle contractions
began well before the contraction in business
activity got under way. The 1947 and 1956 peaks
in the rate of change in the Consumer Price Index
both came about a year before the business cycle
peak, and the 1959 price peak came 10 months
before the business peak. In fact, in 1948, all of
the decline in the rate of change in prices— and it
was substantial— occurred before the recession
began. In 1953, the two peaks coincided. More
often than not, then, the c pi has begun to de­
celerate while business activity was still expanding.
On the other hand, low points in the rate of
price change have coincided rather closely with
business cycle troughs, at least on three out of four
occasions. The 1948 upturn in the rate of price
change (from a level of minus 4 percent) came 11
months before the business upturn, but the 1954
price upturn coincided with the business upturn,
while the 1958 and 1961 price upturns followed
the business turn by 3 months and 1 month,
respectively. In short, declines in the rate of price
change have typically started earlier and hence
have continued somewhat longer than business
cycle contractions.
I t is important to note, however, that the rate of
price change has usually persisted at a low level,
even a negative level, beyond the point of upturn.
Perhaps the m ost striking finding is that a year
or a year and a half after the business peak the
rates of price change have all been in the vicinity
of zero, plus, or minus 1 percent.

this question by examining diffusion indexes of
prices, for such indexes report how many out of a
given population of prices are rising at a particular
time and how many are falling. In terms of the
popular conception of whether or not the economy
is experiencing inflation, or whether it is getting
worse or better, variations in the degree of general­
ity of price increases are perhaps of more signifi­
cance than variations in the rate of change in a
price index.
The price diffusion indexes constructed in this
analysis illustrate several propositions. First,
at all times some prices are falling and some are
rising, but the proportions that are in the one
category or the other vary greatly. Second, the
most widespread increases in prices have generally
occurred during the periods marked off as ex­
pansions in our price chronology, while the most
widespread reductions in prices have generally
occurred during the contractions. This reflects
the fact that the Consumer Price Index increases
more rapidly at some times than at others partly
because price increases are more widespread at
those times, not only because the increases are
larger.
Third, there are discernible sequences in the
process whereby price changes spread through
the economy: prices of industrial materials take
an early position, wholesale prices of manufactured
goods move somewhat later, and retail prices of
consumer goods and services come still later.
The sequences among those parts of the price
system are so long drawn out, in fact, that on
several occasions (notably during 1957-58) the
most widespread declines in the early m oving
prices came almost at the same time as the m ost
widespread increases in consumer prices. Unless the
sequences in the price system are taken into
account, therefore, one could be misled into
thinking that the cyclical swings in prices are less
general than they are in fact.

The diffusion of price change
One of the characteristics of business cycles
that W esley M itchell deemed important, and
which he demonstrated empirically time and
again, was their generality. Mitchell and Burns
wrote in their 1946 volume: “A business cycle con­
sists of expansions occurring at about the same
time in m any economic activities, followed by
similarly general recessions, contractions, and
revivals. . . .” Among the many activities are
prices, and we have just seen that the rate of
change in the price level is clearly one of the
participants in the ebb and flow of business cycles.
This observation does not, however, directly
answer the question whether the price chronology
we have constructed reflects widespread, similar
movem ents among different prices. We can get at




Leads and lags
The diffusion indexes depict some of the
sequences in the price system. But we can examine
the matter more thoroughly by referring to the
rates of change in a larger array of price indexes
using the price chronology as a reference frame in
the same way that the business cycle chronology
has been used to study leads and lags in economic

224

activities generally. In this manner we can observe
not only the leads and lags of other prices vis-a-vis
the Consumer Price Index, but eflso their leads and
lags with respect to one another.
Looking first at certain major components of
the Consumer Price Index, we find that the
turns in the com modity component match those
in the total index very closely. On five occasions
since 1956 (when the commodity-service grouping
first became available) the turns in the rate of
change in the commodity index and in the total
index came in exactly the same month, while on
the remaining occasion the commodity turn was
1 month earlier. This correspondence is due more
to food prices, whose volatile movements have a
marked effect on both the commodities com­
ponent and the total, than to commodities other
than food. As for prices of services, their wellknown tendency to lag is apparent. Perhaps less
well known is the fact that the rate of change in
service prices undergoes cyclical movements that
correspond closely, except for the lag, to those in
com modity prices. The lag of service prices behind
commodity prices averages about 3 months.
Turning to wholesale prices, we find that the
total
exhibits a slight tendency to lead the
total
.
T hat is, it leads on five occasions,
exactly coincides 4 times, and lags only once.
The lead appears to derive more from the in­
dustrial commodities in the
than from the
farm products, processed foods and feeds com­
ponent. The latter component, however, matches
the
quite closely, and of course compares most
directly with the food price component of the
, which, as we have seen, itself has a dominant
effect on the
.
The behavior of consumer
prices depends, to an extent most city dwellers
are probably unaware of, on the behavior of farm
prices.
The industrial commodities component of the
has turned before the
9 times since
1948, coincided once, and lagged twice. The
tendency to lead is imparted primarily by the
prices for crude and intermediate materials other
than foods, rather than for finished goods. Prices
for crude materials other than food have led 9
out of 10 turns in the
since 1947; the average
lead is about 4 months. This index is similar in
its movements and timing to the weekly spot
market index of industrial materials prices. On
w

most occasions the turns in the rates of change in
these two materials price indexes have occurred
within a month or two of each other. Prices for
producer finished goods— that is, machinery,
equipment, trucks, office furniture, and so on—
show about as much tendency to lag behind as to
lead the movements in the
.
The rate of change in the
Deflator is a
lagging indicator in comparison with the rate of
change in the
. This is true also of the Private
Deflator, since its turns usually coincide with
those of the total. The Deflators have lagged
behind the turns in the
far more frequently
than they have led or coincided with it, and the
average lag has been about 3 months. The reason
for the lag m ay lie in the fact that personal con­
sumption expenditures— that is, the type of
expenditure reflected in the
— constitute less
than two-thirds of total
, while the prices for
the two largest elements in the remainder— fixed
investment goods and government services— are
relatively sticky.
Our review of the complex structure of leads
and lags in the price system has merely scratched
the surface of the subject. Very generally, the
discernible sequences in the manner in which
price changes spread through the economy are as
follows: Prices of industrial materials move first.
Wholesale prices of manufactured goods move
somewhat later. Retail prices of foods and other
commodities follow shortly thereafter, and retail
prices of services, such as passenger fares and
medical fees, bring up the rear. In this review, we
have dealt with prices for fairly large groupings of
goods and services and have not dealt at all with
the prices of fixed assets, such as land or buildings,
or the price of labor, or of interest rates. There is
much room for further investigation.
c p i

g

c p i

c p i

g

p i

p i

c p i

c p i

c p i

w

p i




p

c p i

c p i

w

n

n

p

Costs and profits

c p i

During the past few years, economists and
statisticians have developed a system atic body of
data that connects the rate of change in the price
level with rates of change in labor compensation,
output per man-hour, labor costs, profits and
other costs per unit of output. From these data,
a fairly clear picture of the general behavior of
costs and profits in the United States emerges.

c p i

When prices are relatively stable or declining—

225

the bottom of the price cycle— the rate of increase
in output per man-hour is high. It diminishes,
however, as prices rise. Rates of increase in hourly
compensation for workers, on the other hand, are
usually at a moderate level but soon begin to rise,
partly in response to the upward m ovem ent of
prices. The ra te’of change in unit labor costs is
low and oft n declining during the beginning
phase of price expansion but rises sharply in the
later phase as a result of the opposing movements
of the rates of change in labor compensation and
in productivity. Other unit costs follow a similar
path. The effect of all this on unit profits is to
produce a rapid rise at the start of a price
expansion, but a decline at the end.
The situation is reversed when inflation starts
to subside. W hen the rate of price increase first
starts down, output per man-hour continues to
grow at lower rates for a while but presently
starts up, contributing to a reduction in physical
costs. N o t long afterward, the rate of increase in
hourly compensation turns down. The output rise
and the compensation slowdown generate a decline
in the rate of increase in unit labor costs and other
unit costs start showing more moderate rates of
increase. In summary, at the start of a price
contraction, increases in total unit costs exceed
those of prices— with unit profits therefore
declining—but the downswing in costs exceeds
that in prices before the bottom of the price
contraction is reached.
H ow does the current situation in the United
States stack up in terms of the price chronology
we have outlined? As stated earlier in the article,
we have placed a tentative recent peak for the
chronology of prices in February 1970, based upon
the rate of change in th e Consumer Price Index.
This is the month when the seasonally adjusted
rate of change over a 6-month interval reached its
highest level in the current upswing, 6.7 percent
per year. (February is simply the central m onth of
that interval, which runs from November 1969 to
M ay 1970.) Since then, the 6-month rate has begun
to decline, and the m ost recent observation on it
(covering the period March to September 1970)
is 5.0 percent.

We do not consider this peak to be firmly
established as yet, since the decline has not been
very large or very long. But there is evidence to
support it in the behavior of the Wholesale Price




Index, the Gross National Product Deflator, and
indexes of unit labor costs and unit profits.
Moreover, all of the price diffusion indexes for the
current period have receded from their highs,
which were reached during 1968 and 1969. T hat is,
fewer prices have been rising in recent months,
and more have been declining. The general trend
has been one of a slowing in the pace of price and
cost inflation, and that is the reason for recognizing
it, at least tentatively, in our chronology of price
change.
□
----------F O O T N O T E S---------1 T h is a r tic le is a d a p te d fr o m a p a p e r p r e s e n te d a t a
c o llo q u iu m “ T h e B u s in e s s C y c le T o d a y ,” w h ic h w a s
sp o n s o r e d b y t h e N a t io n a l B u r e a u o f E c o n o m ic R e s e a r c h
in S e p te m b e r 197 0 . T h e fu ll p a p e r w ill b e p u b lis h e d a s
The Cyclical Behavior of Prices (b l s - R e p o r t 3 8 4 , 1 9 7 0 ).
2 W e s le y C . M itc h e ll, Business Cycles— The Problem and
Y o r k , N a t io n a l B u r e a u o f E c o n o m ic
R e se a r c h , 1 9 2 7 ).

Its Setting ( N e w

3 A r th u r F . B u r n s a n d W e s le y C . M itc h e ll, M easuring
Business Cycles ( N e w Y o r k , N a t io n a l B u r e a u o f E c o n o m ic
R e se a r c h , 1 9 4 6 ).

* T h e s u b s t a n t ia l s e a so n a l e ffe c t o n t h e r a te o f c h a n g e
ca n b e illu s tr a te d a s fo llo w s. T h e in c r e a se in t h e se a so n a l
fa c to r fr o m 9 9 .8 3 in J a n u a r y t o 1 0 0 .1 2 in J u ly is 0 .6 p e r c e n t
a t a n a n n u a l r a te . I f t h e in c r e a se in t h e u n a d ju s te d in d e x
is a t a 6 -p e r c e n t a n n u a l r a te , th e se a s o n a l fa c to r a c c o u n ts
fo r a b o u t 10 p e r c e n t o f t h e rise. O f co u rse, it h a s a n e q u a l
a n d o p p o s ite e ffe c t o n t h e in c r e a se fr o m J u ly to J a n u a r y .
T h e u p s a n d d o w n s in t h e r a te o f in c r e a se t h a t are a t t r ib u t ­
a b le t o s e a s o n a l fa c to r s c a n b e q u ite m is le a d in g in j u d g in g
tr e n d s in t h e r a te o f in fla tio n . A s t h e fig u res g iv e n b e lo w
in d ic a te , t h e s e a s o n a lly a d ju s te d r a te s sh o w fa r m o r e
c le a r ly t h e o n s e t o f in fla tio n in 1965, its in te r r u p tio n in
1967, a n d it s c o n tin u a tio n th e r e a fte r , th a n d o t h e u n ­
a d ju s te d r a te s.

Percent change at annual
rate, c p i , all items

1964— J an uary-J u ly _____
Ju ly-Jan u ary____
1965— J a n u a r y -J u ly ___
July-January_____
1966— J a n u a r y -J u ly ___
Ju ly-Jan u ary____
1967— J a n u a r y -J u ly ___
July-January__ _
1968— Jan u ary-Ju ly____
July-January _ . _.
1969— Jan u ary-Ju ly____
Ju ly-Jan u ary__ _
1970— J a n u a r y -J u ly ___

226

Unadjusted

Seasonally
adjusted

1.1
1.1
2 .4
1 .5
4 .2
2 .5
3 .2
3 .6
5 .0
4 .3
6 .7
5 .7
6 .0

0. 6
1. 6
1. 9
2. 0
3. 7
3. 1
2. 6
4. 2
4. 4
5. 0
6. 1
6. 3
5 .4

Prices in 1972:
An analysis
of changes
during Phase 2

Prices of finished goods rose less
at retail than at manufacturers’ level;
spread narrowed between
price increases for services
and those for nonfood commodities
JOEL POPKIN

Price behavior in 1972 was marked by the exist­
ence of Phase 2 of the Economic Stabilization Pro­
gram put into effect by the President on August 15,
1971. Phase 1 of that program consisted of a freeze
of virtually all wages and prices that lasted until
November 13, 1971. Phase 2, which ensued imme­
diately, consisted of a varied program of regulation
ranging from the exemption of prices of certain raw
commodities, particularly farm products, to the im­
position of absolute control of the rate of price in­
crease in areas such as medical care. In between
there were regulations governing the extent to which
cost increases could be passed through as price in­
creases and the extent to which profit margins could
rise.
Prices for most major groups of commodities and
services in the Consumer (CPI) and Wholesale
(WPI) Price Indexes rose at a slower rate in 1972
than in the first 8 months of 1971 up to the freeze.
The major exceptions were food prices at retail and
wholesale and the price index for crude nonfood ma­
terials, a component of the Industrial Commodities
Index of the WPI.
During 1972 the Consumer Price Index rose 3.4
percent. The annual rate of increase for the 13
months after Phase 2 began in November 1971 was
also 3.4 percent. The analysis of price behavior
throughout this article will be based largely on
movements during the 13 months from November
1971 when Phase 2 began through December 1972.
How the rate of advance during this time period
compares with periods before and during the Phase
1 freeze is shown in table 1 for the CPI and WPI
and their major components. Changes are expressed
at annual rates (all seasonally adjusted except
services).

From December 1969 to December 1970 the rise
in the CPI slowed by 0.6 percentage points. From
December 1970 to August 1971 (up to the start of
the freeze) the rate dropped further by 1.7 percent­
age points. From the start of the freeze through De­
cember 1972, the first 16 months of the Economic
Stabilization Program, the CPI increased at an an­
nual rate of 3.2 percent, a drop of 0.6 percent­
age points from the pace of the first 8 months of
1971. During the 13 months of Phase 2 through
December of 1972 the rate was down 0.4 percent­
age points from the rate in the 8 months before the
freeze.
In the spring of 1971, there was a sharp decline in
mortgage interest rates which affects the comparisons
made between the first 8 months of 1971 and periods
preceding and following them. If the change in the
CPI in the first 8 months of 1971 is recalculated to
exclude mortgage interest costs and the effect the
elimination of die excise tax on autos had on the
August CPI, the rate of advance is higher, 4.8 rather
than 3.8 percent.1 With these exclusions during Jan­
uary-August of 1971 and the exclusion of mortgage
interest costs from the CPI during Phase 2, the
decline in the rate of advance of the CPI in Phase 2
through December compared to the pre-controls pe­
riod is slightly more than 1 percentage point. A de­
cline in the rate of advance of the CPI from 1970
to the first 8 months of 1971 still shows up, even
when mortgage interest costs are eliminated from
1970 data, but that decline is much smaller.*
Consumer prices

The pattern of movement in the CPI was varied
throughout 1972. In the first 3 months of Phase 2 it
rose at an annual rate of 4.8 percent. In spring as
food prices fell and price rises for services deceler­
ated, the pace of increase slowed to 2.2 percent in
the 3 months ending in June. In the 3 months

Joel Popkin is assistant commissioner for Prices and
Living Conditions, Bureau of Labor Statistics. Toshiko
Nakayama, an economist in the Division of Consumer Prices
and Price Indexes, assisted.

From the Review of February 1973



227

ending in September, the pace quickened to a rate of
4.6 percent as prices of food began to rise again
sharply and those of nonfood commodities advanced
at a faster rate than in the second quarter. Increases
in the food and nonfood components slowed in the
final 3 months of the year with the result that the CPI
rose at a lower rate of 3.2 percent from September
to December.
Implicit price deflator

Another measure of price change, available quar­
terly, is the implicit price deflator (IPD ) for private
gross national product. From the fourth quarter of
1971 to the fourth quarter of 1972, a period roughly
commensurate with Phase 2, the rise in the IPD was
lower than the 3.9-percent rate from the end (fourth
quarter) of 1970 to the third quarter of 1971, the
quarter in which the freeze was imposed (table 2).
A major factor contributing to this slowdown was a
slackening in the advance in unit labor costs. The
rate of advance for compensation per man-hour fell
to 6.6 percent during 1972 from 7.1 percent dur­
ing 1971 up to the quarter in which the freeze was
ordered. Moreover, throughout most of 1972 the
rate of increase in compensation per man-hour fell,
reaching a low of 4.4 percent in the third quarter.
Output per man-hour advanced more rapidly during
Phase 2— 4.9 percent— than in 1971, before the
freeze— 4.3 percent. Because of these two factors,
unit labor costs rose at the lowest rate for any fourquarter period since 1965. Since the rise in the defla­
Table 1.

tor exceeded the increase in unit labor costs during
Phase 2 through December 1972, unit nonlabor
costs advanced in 1972, but less rapidly— based on
indications from preliminary data— than in pre­
freeze 1971.
The deflator for personal consumption expendi­
tures (PCE), a component of the deflator for private
GNP, rose 2.5 percent in 1972. Because both this
deflator and the one for private GNP have moving
weights they are not measures of price change alone,
but both are calculated, alternatively, on a fixed
weight basis. On this basis the deflator for PCE is
more nearly comparable to the fixed weight CPI.
From the fourth quarter of 1971 to the fourth quarter
of 1972, the fixed weight PCE deflator increased less
than the CPI. Much of this difference is attributable
to the treatment of owner-occupied housing in the
two indexes.3 The annual rate of change in the fixed
weight deflator for PCE from the third to the fourth
quarter of 1972 was 2.9 percent as compared
with 3.9 percent between the third and fourth
quarter averages of the CPI.
Wholesale prices

The Wholesale Price Index for all commodities
rose at a faster rate in the 13 months after the freeze
than in the 8 months before it. This occurred as a
result of an acceleration during Phase 2 in price rises
for farm products and processed foods and feeds that
more than offset a deceleration in the rate of
advance for industrial commodity prices. The rate of

Changes in Consumer and Wholesale Price Indexes, selected periods 1968-72

[Seasonally adjusted, except services, compound annual rate]

Item

Second
First
3 months,
13 months, 16 months,
• months
6 months,
Phases 1
7 months,
prior to
Phase 1,
Phase 2,
12 months, 12 months,
Phase_2,
and 2,
Phase 2,
Aug. 1971 to Nov. 1971 to
Dec. 1968 to Dec. 1969 to
Phase 1,
Dec. 1972 Aug. 1971 to Nov. 1971 to June 1972 to
Dec. 1969
Dec. 1970 Dec. 1970 to Nov. 1971
Dec.
1972
June
1972
Dec. 1972
Aug. 1971

CONSUMER PRICE INDEX
All items...........................................................................
Food..........................................................................
Commodities less food.............................................
Services....................................................................

6.1
7.2
4.5
7.4

5.5
2.2
4.8
8.2

3.8
5.0
2.9
4.5

1.9
1.7
0
3.1

3.4
5.0
2.5
3.6

3.2
4.4
2.0
3.5

3.1
4.0
2.5
3.7

3.9
6.1
2.5
3.5

4.8
7.5
3.9

2.2
-1.4
3.6

5.2
6.5
4.7

- .2
1.1
- .5

6.6
14.7
3.5

5.3
12.0
2.7

5.3
7.6
4.4

8.1
23.6
2.6

10.2

4.6
3.1
4.9
4.0
-2.5

3.3
6.5
3.7
2.2
6.8

2.3
- .7
-2.0
- .4
.3

10.3
4.0
2.3
2.4
8.8

8.8
3.1
1.5
1.9
7.1

8.5
4.8
4.1
2.9
5.4

12.6
3.0
.2
1.8
12.9

WHOLESALE PRICE INDEX
All commodities...............................................................
Farm products and processed foods and feeds___
Industrial commodities.............................................
Selected Stage of Processing Indexes:
Crude materials except food............................
Intermediate materials except food.................
Producers’ finished goods................................
Consumers goods except food..........................
Consumer foods................................................




3.9

4.6
2.9
8.2

228

Table 2.

The anatomy of price change
Chang* in quarterly averages
Ken

IV-68 to
IV-C9

IV-69 to
IV-70

IV-70 to
111-71 ‘

Phase 1
111-71 to
IV-71 *

Phase 2
1972
IV-71 to
IV-72 *

Phases 1
and 2
IU-71 to
IV-72»*

Deflator: Private 6NP...........................................................................................
Personal consumption expenditures.........