View original document

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

L 2 J

The Meaning and
Measurem ent
of Productivity

Co.

Bulletin 1714

U. S. D E P A R T M E N T OF LABO R
B u reau of Lab o r S tatistics




DaVt pubttc Libras

D 6
EC

D C N ^ CQUtfW*
OU©




INDIANA UNIVERSITY
INDIANAPOLIS LAW SCHOOL
LIBRARY

The Meaning and
Measurement
of Productivity




Bulletin 1714

Prepared for the
National Commission
on Productivity by
the Bureau of
Labor Statistics
U. S. Department of Labor
September 1971

U. S. DEPARTMENT OF LABOR
J. D. Hodgson, Secretary
BUREAU OF LABOR STATISTICS
Geoffrey H. Moore, Commissioner




For sale by the Superintendent of Documents, U.S. Government Printing Office
Washington, D.C. 20402 - Price 30 cents

Preface
The National Commission on Productivity was established by President Richard
Nixon in June 1970 to develop recommendations for programs and policies to improve
the productivity of the U.S. economy. The Commission is composed of top-level repre­
sentatives of business, labor, government, and the public. In order to aid the members in
their consideration of various topics, staff papers will be prepared by government or
private industry experts in different subject matter fields. These papers serve as back­
ground material for the members but do not necessarily represent their views.
The two papers included here were prepared by Herbert Stein, member of the
President’s Council of Economic Advisers, and Jerome A. Mark, Assistant Commissioner
for Productivity and Technology, Bureau of Labor Statistics, U.S. Department of Labor.




INDIANA UNIVERSITY
INDIANAPOLIS LAW SCHOOL
LIBRARY

Contents
Page
The meaning of productivity b y H erbert Stein .........................................................................................................
The meaning and measurement of p ro d u ctiv ity ....................................................................................................
Why productivity is important n o w ........................................................................................................................

1
1
3

Concepts and measures of productivity b y Jerom e A. M a r k ..................................................................................... 7
Concepts of p ro d u ctiv ity .......................................................................................................................................... 7
Output .................................................................................................................................................................... 8
Labor i n p u t ............................................................................................................................................................ 8
Capital i n p u t .......................................................................................................................................................... 9
Problems of measurement of productivity.............................................................................................................. 9
O u t p u t ...................................................................................................................................................................... 10
Economy and sector level...................................................................................................................................10
Industry le v e l.......................................................................................................................................................11
Labor i n p u t .............................................................................................................................................................. 11
Q u a lity ................................................................................................................................................................. 11
Gaps in coverage..................................................................................................................................................11
Hours paid versus hours w o r k e d ......................................................................................................................12
Total factor measures......................................................•
...................................................................................... 12
Available measures of productivity............................................................................................................................. 12
Labor productivity ..................................................................................................................................................12
National m easures............................................................................................................................................... 12
Industry m easures............................................................................................................................................... 13
Capital productivity..................................................................................................................................................13
Combined factory input productivity................................................................................................................... 14
Recommendations......................................................................................................................................................... 14
Additional price inform ation.................................................................................................................................. 14
Better man-hour and capital d a ta ...........................................................................................................................15
Charts:
1. Output per man-hour and real compensation per man-hour, total private
economy, 1950-70 ........................................................................................................................ , ................
2. Productivity trends, total private economy and selected sectors,
1950-70 and 1965-80 .....................................................................................................................................
3. Relative importance of employment for major sectors, 1950, 1970
and projected 1980 ........................................................................................................................................
4. Annual rates of change in wages, productivity, and unit labor costs,
total private ec o n o m y .....................................................................................................................................




2
3
4
5

The Meaning of Productivity
by Herbert Stein*

sation per hour to productivity per hour except for cy­
clical or other shortterm interruptions. Finally, estimates
of productivity defined in this simple way are available
to permit interesting and analytical comparisons of
different times and countries, whereas more sophisti­
cated measures are not.
The most commonly used measure of productivity
relates the total output of goods and services in the
private economy, that is, private Gross National Product
(GNP) to the man-hours of all persons engaged in the
production of those goods and services. This measure is
expressed in the usual GNP constant dollar terms. Other
measures, at the firm or industry level, may be construc­
ted in a similar conceptual framework but more often
are derived by using some physical concept of output
unique to that firm or industry such as tons of steel
or kilowatt-hours of electricity.
The most common definition and measure of pro­
ductivity is sometimes called inadequate or even irrele­
vant on the ground that it reflects only the “quantity”
and not the “quality” of economic performance. It is
true that like all measurements, the measurement of
productivity relates to a quantity, but it is untrue that
the quantity measured is unrelated to the qualities that
human beings value. The value of gross national product
is a product of the quantities which people buy and the
prices which people are willing to pay for them. The
quantities people buy, and the amounts they are willing
to pay for them, reflect the qualitative values that
people find in the different products. Thus, GNP and
its components reflect a value or “quality” choice
among consumers, subject to limits imposed by income
levels and the available supply of goods and services.
There are, of course, consequences of economic activ­
ity that are not reflected in existing measures of output
per hour and the country has become increasingly con­
scious of some of these in recent years. On the output
side these are generally consequences of activity that
does not pass through a market. The outstanding case
is the deterioration or improvement of the environment.
The deterioration of the environment is not counted as

The rate at which productivity grows is central to
two of the major issues facing the country. One is the
issue of inflation. The other is the speed at which the
society’s demands on the economy are rising— not only
for the traditional purpose of private consumption and
investment but also for improving the environment,
health, domestic security, and general quality of life.
Recognizing the key role of productivity increase in
meeting the Nation’s goals, and the potential contribu­
tion of all sectors of the community, the President on
June 17, 1970, announced his intention to establish a
National Commission on Productivity. The Commission
when established included six members each from busi­
ness, labor, and the public at large, and five members
from the Federal Government.
Subgroups of the Commission have been established to
deal with specific problems. While its work still lies
largely ahead of it, some points raised in early discus­
sions by the Commission, with participation by the
President, deserve public attention at this time.
THE M E A N IN G A N D M EA SU REM EN T
OF P R O D U C T IV IT Y

The most commonly used definition of productivity
is real output per hour of work. Productivity in this
sense is a rough measure of the effectiveness with which
we use our most important productive resource— labor.
It has important social implications because it takes
account not only of the chief source from which individ­
ual and social desires are met— that is, the total output
of the economy— but also of a major source of getting
that output, namely work. We would surely think that
an increase in output achieved by raising the output per
hour of work does us more good than an increase in out­
put achieved by working more hours. The definition of
productivity as real output per hour of work has eco­
nomic significance also. If all terms are consistently
defined, if labor compensation per hour rises at the
same percentage rate as productivity, then unit labor
costs will be stable; and if the shares of compensation
in the national income remain unchanged, then prices
on the average will be stable. In fact, the price level
generally moves very closely with the ratio of compen­



* Member, Council of Economic Advisers.
1

Chart 1.

Output Per Man-Hour and Real Compensation Per Man-Hour, Total Private Economy, 1950-70

Index 1 9 5 0 -1 0 0
(ratio scale)

Nor do we yet know how to assign a correct market
value to products which are not sold such as those of
education and many government activities.
For many analytical and policy purposes the simple
figure of output per hour is inadequate. We would like
to know why total output and output per hour are
larger in one country, industry or time than another and
for this purpose we need additional measures. We need
to measure output per unit of all resources, including at
least capital as well as labor. And we need to recognize
that different kinds of labor, distinguished by skill, in
a sense, are different amounts of labor per hour and
have different productivities. When we measure output
per unit of capital and labor combined and adjusted for
quality werhave another measure of the efficiency of
resource use. When we break down this measure into its
various components we then have a family of measures
which permits better estimation of the contribution of
different factors to the growth of output and provide
insights into the effects of different policies in the
future.

a cost or a deduction from the product, and an improve­
ment of the environment is not counted as an addition
to the product. Thus, as far as this factor alone is con­
cerned, our ordinary figures may have overstated the
growth of productivity in the past and may in the future
understate the growth if more and more resources are
used to improve the environment. However, this is only
one among many omissions in the measurement of out­
put and productivity. Going further, it is obvious that
productivity statistics do not measure justice, security,
happiness, beauty, or the lack of them, and we cannot
be sure in what direction our available measurements
may be biased. But this obvious fact does not belie the
importance of the statistic as an indication of the ability
of society as a whole to achieve its goals.
There are other limitations of the statistical measures
of productivity which are a consequence of our inability,
thus far at least, to perfect our methods of national ac­
counting. In construction, for example, we do not yet
know how to measure properly the output of complex
and diverse structures such as homes, hotels or hospitals.



2

rapid economic growth, which has led to rapidly rising
expectations among workers and consumers. We shall
come closer to meeting these claims if we can increase
the rate of growth of productivity.

Measurements of productivity have been improved
substantially in sophistication, variety, and accuracy in
the past 20 years. They are still inadequate to answer
all the questions that might be asked of them. Further
improvement and dissemination of the statistics will con­
tribute to better understanding of our economic prob­
lems and to better policy. The Commission hopes to
contribute to this.

jk During the past 4 years the rates of growth in pro­
ductivity and GNP slowed down. At the same time the
real income gains which workers has come to expect
with rising wages also deteriorated. A return to a more
normal productivity growth rate can help to restore the
rate of gain in average real income. (See chart 1.)

W HY P R O D U C T IV IT Y IS IM PO R TA N T NOW

By any available measurement the level of produc­
tivity in America has risen persistently over the decades,
except for brief cyclical interruptions, and for many
years has been the highest in the world. One might think
in these circumstances that productivity growth would
be a matter of no great concern. Nevertheless, the dis­
cussions of the Commission have confirmed the premise
which underlay its establishment— namely the great
importance of the productivity question today.
The claims upon the economy, expressed through the
political process or in the market, are very large, even
relative to the great capacity of the American economy.
The size of these claims results at least in part from past
Chart 2.

'^T his striving for higher productivity must not be
viewed as a whip-cracking exhortation to “work harder”
in order to raise some arbitrary abstract measure of
economic performance. Increasing productivity is a way
of increasing the ability of people to do what they want
to do. It can provide the wherewithal for achieving a
higher standard of living for families now living at the
low end of the income scale. It can provide for a choice
of leisure— not idleness— in the form of more holidays
and vacations and entrance to an earlier retirement from
the world of work, and it can provide the resources for
improving the physical quality of the environment.

Productivity Trends, Total Private Economy and Selected Sectors, 1950-70 and 1965-80

7.0

6.0

UJ

C3

5.0

c_>

4.0

3.0
LU

CJ

<

tr
LU

>
<

2.0

1.0

0

TOTAL




FARM

COMMUNICATIONS
AND
PUBLI C UT I L I T I E S

MINING

MANU­
FACTURING

T R A N S ­ WHOLESALE
AND
PORTATION
RETAIL TRADE
U.S. D E P A R T M E N T O F LA B O R
Bureau of Labor Statistics

3

fact that they are behind us in productivity by itself
helps them to grow more rapidly because it means that
they have opportunities to exploit— such as advanced
production techniques— which we are already using.
In some cases where they have embarked on new pro­
duct ventures they have built plants embracing the most
modern technology. But technology and methods may
not be the whole story, and the experience of others
requires us to consider carefully whether there are steps
that could be taken here to speed up the increase of
productivity.
The higher rates of productivity growth in other
countries were accompanied by increases in hourly com­
pensation which, in the past 5 years, tended to exceed
those in the United States. The relationship of trends
between output per man-hour and compensation per
man-hour, however, was closer in those countries than
in the United States so their unit labor costs did not go
up as much in those years.

We may be entering a period in which sustaining the
rate of growth of productivity will be more difficult. For
several decades the shift of workers out of agriculture,
where productivity was below the average, contributed
substantially to the increase of productivity. The number
of workers left in agriculture is so small that this can no
longer be significant. On the other hand, there will now
be an increase in the proportion of the labor force em­
ployed in those industries, loosely called “service in­
dustries,” where productivity and its rate of growth
have been low relative to the national average. (See
charts 2 and 3.)
Productivity, in recent years, has been increasing
more rapidly in Japan and in several Western European
countries than in the United States. The reduction of the
gap between our productivity and their’s is not a matter
of concern; it is to our advantage that their productivity
should be high. Neither does it necessarily indicate any
superior effectiveness of their economic policies. The
Chart 3.

Relative Importance of Employment for Major Sectors, 1950, 1970 and Projected 1980

FARM
C OM M U NICA TIONS AND
PUBLIC U T I L I T I E S
MINING

MANUFACTURING

TRANSPORTATION
FINANCE INSURANCE
A N D RE AL ESTATE
WHOLESALE AND
RETAIL TRADE

CONSTRUCTION

SERVICES




(Projected)
U.S. D E P A R T M E N T OF LA B O R
Bureau of Labor Statistics

4

tion in the rate at which higher real income was being
achieved.
The important point here is not so much that com­
pensation rose rapidly but rather that there was a large
gap between productivity and compensation gains with
a resultant large increase in unit labor costs. (See chart
4.) It is a part of the syndrome of inflation in which
prices and wage rates each rise— and each, at different
points of time, tries to catch up with the other. One way
to break into the syndrome is to increase the rate at
which productivity grows so that wages can rise without
increasing unit costs and the pressure on prices is abated.
Increasing productivity may. thus be regarded as the
keystone to an improved standard of life and environ­
ment for all of society. It is with this broad view in mind
that the National Commission on Productivity has set
its task of finding ways to continue or accelerate the
historical rates of productivity gains in the United States.

/(In comparing the trends of productivity and earnings
among countries, we must draw a distinction between
real income and unit costs. Certainly we must applaud
the rising productivity and standard of living of workers
in the rest of the world. But the gap between changes
in money earnings and changes in productivity which
determines unit labor cost trends, is an important
element in the difference between a stable or improving
or deteriorating position in international trade.
Within the United States, no less than among coun­
tries, to recognize the distinction between costs and in­
come is useful. The last 3 years were witness to the
highest 3-year increase in average compensation per man­
hour (of all persons working in the private economy)
since the early 1950’s. And yet, when these figures are
adjusted for the rise in consumer prices, the resultant
real compensation per man-hour showed the smallest
3-year rise over the same period. There was a deteriora­

Chart 4.

Annual Rates of Change in Wages, Productivity, and Unit Labor Costs, Total Private Economy

Percent Change

Compensation includes wages and salaries and supplemental payments for employees and an
estimate of the salaries and supplements for the self-employed.




U.S. D E P A R T M E N T OF LA B O R
Bureau o f Labor Statistics

5




Concepts and Measures of Productivity
by Jerome A. M a rk *

Despite the wide attention paid to productivity over
the years, confusion prevails as to its meaning and meas­
urement. This is understandable because the concept
does lend itself to ambiguity and a wide range of produc­
tivity measures can and have been developed in response
to different analytical uses.
Productivity is loosely interpreted to be the efficiency
with which output is produced by the resources utilized.
A measure of productivity is generally defined as a ratio
relating output (goods and services) to one or more of the
inputs (labor, capital, energy, etc.) which were associated
with that output. More specifically, it is an expression of
the physical or real volume of goods and services related
to the physical or real quantities of inputs.
A variety of plausible productivity measures can be
developed, the particular form depending on the purpose
} to be served. For example, output per labor input, the
jj£ most familiar measure, is useful in understanding changes
y in employment or labor costs. This measure might be
based on man-hours paid or man-hours worked, with
different results. A more comprehensive measure of in­
put might be more useful in studying how the economy
is using labor and capital combined. Also, there are
various ways of adding up diverse products into a meas­
ure of output. No one measure is the right or best
measure.
Since the interpretation of these statistics depends
on the definitions and data used, an understanding of the
productivity concepts used in relation to the purpose to
be served is always essential.

that factor to production. Rather, they express the joint
effect of a number of interrelated influences on the use
of the factor in the production process— such as changes
in technology, substitution of one factor for another,
utilization of capacity, layout and flow of material, the
skill levels and the efforts of the work force, and mana­
gerial and organizational skills.
Whether for an individual establishment, an industry,
or the entire economy, the most frequently developed
and perhaps most useful productivity measure is an out­
put per unit of labor input measure of what is frequently
termed a labor productivity measure. There are several
reasons for this. Perhaps the most important is that labor
is almost universally required for carrying through all
types of production. There is a labor element of costs in
almost all endeavors; the degree varies but it is always
present. In addition, as a practical matter, it is perhaps
the most measurable input. Other factors, such as capi­
tal, are much more difficult to quantify.
There are, however, various labor productivity meas­
ures, depending on the definition of labor input. A
measure may refer to output per person or it may take
account of changes in hours of work and be based on
total hours. It may cover the hours of the entire labor
force including proprietors, unpaid family workers, and
employers; or it may be limited to selected groups
of workers.
Another set of productivity measures relating output
to a single input is output per unit of capital. These
measures are particularly useful in understanding move­
ments in unit nonlabor costs by relating the measures to
corresponding measures of returns to capital. As in the
case of other single factor productivity measures they
indicate the changes in the use of capital per unit of out­
put not the contribution of capital alone. The measures

C O NC EPTS O F P R O D U C T IV IT Y

( There are two broad classes into which productivity
concepts and in turn measures can be grouped. One in­
cludes those measures which relate output of a produc­
ing enterprise, industry, or economy to one type of in­
put such as labor, capital, energy, etc.; the other in­
cludes those which relate output to a combination of
inputs extending to a weighted aggregate of all associate
V in p u ts.1
Although the former measures relate output to one
input, they do not measure the specific contribution of




* Assistant Commissioner for Productivity and Technology,
Bureau of Labor Statistics, U.S. Department of Labor.
1 Even at the level of an individual craftsman where the
output-input relationships are limited, these two concepts are
present. Productivity can refer to the volume of work the in­
dividual is able to accomplish within a given time span— i.e.,
output per man-hour. It can also refer to the volume of work
completed per unit of his time, his tools, and his materials—
i.e.. an output per total factor input.

7

have been limited to reproduction of what has been
termed “ tangible” capital.
Other single factor measures such as output per
energy input or output per material input are relevant
for plant and industry study where these inputs are of
considerable importance in the production process or
represent relatively scarce resources. For example, in the
aluminum industry where electrical power is an import­
ant element in processing bauxite, output per KWH-is
very useful as an indication of the efficiency with which
electricity is being utilized.

impact on man-hour requirements depending on whether
output is measured by the yard or by the pound.
For the more usual case of a plant or an industry
producing many heterogeneous products, the different
units must be expressed on some common basis. They
can also be combined in terms of their man-hour re­
quirements. The advantage of the latter method for
measures of output per man-hour is that the change in
the productivity of the entire plant or industry is then
a simple arithmetic average of the changes in the pro­
ductivity of the individual components.
When the components are combined with value or
price weights, that is, on the basis of their dollar value,
then the output per man-hour measure for the total
reflects not only changes in the productivity of the
components but shifts in the importance of the com­
ponents.
Physical quantity data are often not readily available,
so deflation of dollar value is used. That is, total value of
production is adjusted for change in price by use of a
price index. This type of index is usually referred to as
constant dollar output or deflated value of output. Such
indexes are conceptually equivalent to indexes which
use physical quantities combined with price weights.
The contribution of a producing unit lies in the value
added, by its own labor and capital, to the materials and
services purchased from other producing units— i.e., its
net addition. Net output, therefore, is the constant
dollar value of production minus the constant dollar
value of purchased goods and services. In measuring
productivity, the net measure would then be related to
the particular input or all associated inputs except the
material inputs. Relating a net output measure to a
single input, when the various commodities produced
and purchased are combined with value or price weights,
will result in a single factor productivity measure that
reflects not only the changes in productivity of the
components and shifts in the importance of these
components but also savings in material consumption.

As mentioned earlier, all single factor productivity
measures reflect the joint effect of a variety of factors
including the substitution of one factor for another. For
some purposes, to develop a measure which eliminates
the effect of that substitution is useful. This type of
measure relates output to a combination of inputs. Thus,
a productivity index of output per labor and capital
combined elininates the effects of changes in amounts
of capital per workers These measures have been termed
multifactor or “total factor” or simply “total” produc­
tivity measures. For both conceptual and statistical
reasons they have generally been limited to labor and
“tangible” capital inputs and have not included as in­
puts activities such as research and education which can
be viewed as intangible capital.2

Output

For all productivity measures, output is measured in
physical or real terms. The concept is one of work done
or the amount of product added in the various enter­
prises, industries, sectors, or economy. It refers not to
activity as such but to the results of activity.
In this sense, at the plant level, production and hence
productivity measurement differs from work measure­
ment. Work measurement generally refers to the analysis
of the stages of activity and the requirements at each of
these stages. Productivity refers to the finished product
(the result of activity) and its relationship to input.
In the case of a producing unit making one homo­
geneous commodity, production in physical terms would
merely be a count of units produced. For a commodity
to be regarded as homogeneous, certain conditions
should be fulfilled. The product should be of a specified
quality (e.g., carbon steel) and it must conform to pre­
cise standards of size, volume, unit, etc. Even though
the measure of production in this case is a single count,
the way of defining the unit of product can have different
implications for productivity measurement. For example,
carpeting can be measured in pounds or square yards.
A change in the density of the carpeting would affect
the weight per yard and, therefore, have a different




/

Labor input

For all productivity measures where labor is rele­
vant, labor input is measured in physical terms. The
measure can refer either to the total number of individ­
uals engaged in production or to only part of the work
force, or it can refer to the man-hours of workers.
It is usually preferable to include the entire employed
work force in the labor input measure— blue-collar and
Denison in his work on the sources of economic growth
has made estimates of the contributions of intangible factor
input such as research, education, organization, etc., to total
output. See Edward Denison, The Sources o f Economic Growth
in the United States and the Alternatives Before Us (New York,
1962) and Why Growth Rates Differ (Washington, D.C., The
Brookings Institution, 1967).
8

white-collar workers, corporate officers, and the selfemployed. The assessment of manpower needs must take
all labor input requirements into account. But of course,
there are times when analysis of labor requirements and
the analysis of cost components suggests the use of mea­
sures which include only a component of the work force.
To analyze the productive capacity of labor and the
effects of changes in working hours, or in use for pro­
jections of manpower needs, an output per man-hour
measure is most relevant. The most suitable unit of
measure is man-hours worked. There are some ambi­
guities or differences of opinion on what to include, for
example, standby time, coffee breaks, etc. In general,
“hours worked” refers to the time spent at the place of
employment, and therefore excludes hours paid for but
used on leave for vacation, holiday, illness, accident, etc.
In some cases, total hours paid are utilized in the pro­
ductivity measures because data on hours worked are
not available.
In developing a labor input measure, in many cases
man-hours are treated as homogeneous and additive.
These measures are particularly relevant to problems
of estimating total man-hour requirements. But merely
adding up the number of hours ignores the qualitative
aspect of an hour worked by different individuals.
Therefore, a productivity measure which is based on
the sum of undifferentiated man-hours will reflect
changes in the composition of the work force with dif­
ferent qualitative characteristics.
For some purposes, it may be desirable to develop a
productivity measure which takes into account the dif­
ferences in the “quality” of an hour of labor. That is,
an hour of high quality labor is counted as propor­
tionately more than an hour of low quality labor. To
do this some methods have to be introduced to differ­
entiate these hours. One way which has been utilized
is to combine the man-hours of various employees in
terms of pay differentials. The man-hours of higher
paid workers are given more weight than lower paid.
This assumes that differences in earnings reflect dif­
ferences in education, experience, skill, and their con­
tribution to o u tp u t.3 Another method is to adjust the
data to take into account changes in vocational training,
length of schooling, or type of education, etc., of the
work force, assuming there is a close relationship be­
tween qualifications and quality. When adjustments are
made for changes in the quality of labor input the re­
sultant productivity measure will not reflect changes
in the composition of the work force as a productivity
change but rather as a change in factor input.

Capital input

Capital stock estimates include the constant dollar
value of structures, plants, and equipment current avail­
able for production. These estimates may also take into
account the value of land, inventories, and working
capital.
Generally capital stock measures are derived by ad­
justing the value of existing plant and equipment for
new investment and the retirement of old assets. There
are different ways of measuring the stock of capital; for
example, they may be gross or net. Net stock estimates are
derived by depreciating assets (and there are various
methods of depreciation). Gross stock estimates are
derived by retaining assets at their full value until they
are retired from use. Since these are physical measures,
the value of capital stock must be adjusted for price
changes.
For productivity analysis, however, the flow of
capital services rather than the stock is the preferred
measure. A capital stock measure does not account for
differences in the intensity of use over time. Equipment,
for example, may be used for several shifts during a
business expansion or may be idle during a contraction.
Then, too, a large part of existing capital capacity may
be standby and employed only during periods when the
economy is operating at very high rates. There is also
a loss of efficiency of assets as they grow older. A
flow measure reflects differences in usage and efficiency
and how they affect varying levels of output, which is
the basis of productivity estimation. Ideally flow meas­
ures should indicate the amount of capital employed to
produce current output.
To derive this capital flow measure, an aggregate of
the capital hours used weighted by the rental value of
each type of structure and piece of equipment is needed.
The data for this measure are often not available in the
detail necessary for a capital flow measure.
A commonly used flow of capital service measure is
depreciation. However, this is based on accounting prin­
ciples which often reflect current income tax regulations
rather than the actual amount of capital used for cur­
rent production. Because of the difficulty of estimating
a capital flow measure, however, most analyses of capital
and production use capital stock estimates.
PROBLEMS OF M EA SU REM EN T
OF P R O D U C T IV IT Y

The measurement of productivity trends involves two
fundamental problems which are applicable to both out­
put and input data. First, because of difficulties in ob­
taining direct quantity measures of output and input,
substitute measures or approximations must be used in
many cases. Second, since most data are collected for

3 Except to the extent that regional or similar wage dif­
ferentials affect average hourly earnings.




9

For the bulk of service activities, however, the de­
flation approach is used and its validity for the resultant
output measure rests on the adequacy of the price in­
dexes. Most of the price indexes used are components of
the Consumer Price Index, which in turn have different
degrees of reliability. The indexes for medical services,
for example, do not adequately take into account
changes in the quality of medical services performed.
As mentioned earlier, the real product measure is
conceptually a net output measure but in many of the
service activities data are not available on the real value
of the material inputs. In such cases estimates of real
product are made on the basis of changes in the total
volume of output. This does not present a serious prob­
lem, however, since in most service industries inter­
mediate purchases constitute a relatively small pro­
portion of total value.
The other major activity in the national accounts
where the output measure has severe limitations for pro­
ductivity measurement is the construction industry. The
constant dollar output measure is obtained by deflation,
but the price index used is really a cost index. For the
most part, these are measures of the change in costs of
materials and labor weighted in terms of their base
period importance. These indexes do not take into ac­
count any savings in the utilization of materials or labor,
and, as a result, there is an overstatement of price in­
creases. Consequently, there is an understatement of
gains in real output and hence productivity.
Productivity indexes based on real product for con­
struction show an average annual decline of 0.2 percent.
This is somewhat inconsistent with studies which the
Bureau of Labor Statistics has conducted of labor re­
quirements for various specific types of construction
during this period. These studies indicate for schools
there was approximately a 2 3 percent per year gain
A
over much -of this period, for highways a 3-percent in­
crease per year, and for hospitals not mcuh change.
With regard to lack of comparability of coverage be­
tween output and labor, perhaps the largest evidence of
this occurs in the real estate activity. For national in­
come accounting purposes, an inputed rent for home
ownership is added to the output of the real estate
industry. There is, however, no corresponding labor in-

purposes other than productivity measurement, defi­
nitions already established and procedures for reporting
information on production and factor inputs must be
used; these may or may not be consistent with concepts
appropriate for productivity measurement.
Output

E conom y and sector level. The problems of using the
gross national product data for productivity measure­
ment involve primarily the inadequacies of the measures
of real output for some components and the lack of
comparability of coverage between output and labor
input measures.
Limitations in measuring output affect the reliability
of productivity statistics in some sectors more than
others. Since the implications for productivity move­
ments can be offsetting among the sectors, the effect on
measures for the overall economy is not as large as it is in
each of the sectors.
The three areas where the real product measures as
derived from the national accounts are particularly weak
for productivity measurement are government, construc­
tion, and services (including business and personal serv­
ices, and finance, insurance, and real estate).
In the absence of market valuation of the services of
general government agencies, the practice in national in­
come accounting is to value government output in terms
of the wages and salaries of government employees. The
deflated, or constant dollar, measure is derived from
changes in employment. Such an output measure, when
related to a labor input measure, results in no statistical
change in productivity. This measure of government out­
put may be increasingly difficult to continue in view of
the reported increases in output per man-hour in certain
government operations which are subject to measure­
ment. 4 Based on these data the trend of output per
man-hour for the national economy would be biased
downward. As a consequence, the available measures of
productivity are limited to the private economy.
Measuring output in the service activities is difficult
because of the absence of a directly quantifiable entity
which describes a unit of service. Consequently, various
substitute indicators are utilized in the national accounts.
These usually involve the use of some “price” index for
deflating the value of the service activities or the use of
an employment index to develop trends in producers of
services.
As in the case of government, the use of the employ­
ment movements as an indicator of the change in real
output implies a constant labor productivity. This ap­
proach is utilized for such activities as security and
commodity brokers, 5 insurance agents, and miscella­
neous business and repair services.




4

Nestor E. Terleckyj “ Recent Trends in Output and Input
of the Federal Government,” in Proceedings of the Business and
Economics Section, American Statistical Association, 1964,
pp. 76-94.
5 In view of the rapid spread in recent years of electronic
data processing in this industry this measure must be very much
understated since the productivity gains undoubtedly were
large, John Kendrick has suggested that data on shares of stocks
and bonds sold appropriately weighted would be a better
measure. See “Production and Productivity in the Service
Industries,” Victor Fuchs, Educational Studies in Income and
Wealth No. 39, National Bureau of Economic Research, 1969.
10

put associated with that output. Rough estimates indi­
cate that the removal of this activity from the output
account would reduce the productivity trend for the
private economy about two-tenths of a percent per year
over the last two decades.

Estimates of the effects of shifts among major
sectors— farming, manufacturing, mining, etc., show
that shifts contributed about 0.3 percent per year of the
output per man-hour growth over the last two decades.
The shift in composition of the work force within man­
ufacturing between production and nonproduction work­
ers contributed 0.1 percent per year to the rate of in­
crease in private output per man-hour over the last two
decades. In recent years this has been reduced consid­
erably to less than 0.05 percentage points.
In view of the limited information on occupational
detail, another approach (followed by Denison) to as­
certain the impact of shifts and changes in the work
force has been to utilize information on changes in age,
sex, and education. He estimates, for example, that the
increase in education of the work force contributed 0.7
percentage points to the trend rate in output per man­
hour from 1950-62. For a longer period, 1929-57, he
estimates the effect to be 0.9 percentage points per
year. Another estimate of the contribution of education
to the growth rate by Schwartzman 6 provides a much
lower figure— three-tenths of a percent per year for a
roughly comparable period, 1929-63. The magnitude of
these differences in this critical area suggests that there
is need for further exploration of the interrelationship
between education skills, training, earnings, and pro­
ductivity.

Industry le v e l The effects of certain measurement
problems are greater at the individual industry level than
at the national level where there is a tendency for errors
and biases to offset each other. On the other hand, at
the industry level, more flexibility is possible because
the output measures are not part of an overall frame­
work (such as the national income and product accounts)
which requires certain definitions and measures not nec­
essarily consistent with the desired productivity measures.
Three major problems are encountered in developing
measures of output from available data for industry
productivity indexes. First, for many industries, the
appropriate detailed product data are not available.
Second, there is the well known quality change problem
which results from the development of new products
and the changing specifications of existing products.
Third, appropriate weights are often not available for
deriving the desired industry measure.
Some of the presently available industry indexes are
based on unit man-hour weights; others are based on unit
value or price. The use of unit value or price weights
is not a serious problem among commodities where labor
consts or inputs are a high proportion of price.

2.
Gaps in coverage. Payroll data on employment and
average weekly hours, which are the primary source of
Labor input
man-hours estimates, do not include the entire economy,
but are limited to nonfarm wage and salary workers.
With regard to labor input measures there are several
These data do not cover farm workers, proprietors, un­
data gaps in presently available measures. They relate
paid family workers, and domestics. Estimates for these
to changes in the composition of labor (the quality),
sectors, for the most part, are taken from the labor force
groups of the work force for which data are lacking or
series (based on household surveys) which is not strictly
incomplete, the relationship of output to the time of
comparable to the payroll series. Employment is a count
research development and other workers whose activities
of persons rather than jobs as in the payroll data, so
are not directed to current production, and finally, the
that appropriate adjustments must be made.
absence of adequate hours worked data on a comprehen­
Average hours for supervisory workers in nonmanu­
sive basis.
facturing industries are not available. The assumption is
1.
Quality. As mentioned earlier, changes in the com­
made that the average workweek for these workers is the
position of labor input are adjusted in some measures by
same as for the nonsupervisory workers in each industry.
weighting industry man-hours with the average hourly
Since 85 percent of all employees in nonmanufacturing
earnings of workers in the industry. Insofar as earnings
industries are nonsupervisory workers, however, the
differentials reflect productivity differences among work­
effect of this imputation may be minimal.
ers, this measure captures changes in the quality of
Sampling procedures also affect the man-hours esti­
workers of different industries. However, this approach
mates. One week of each month is used to represent the
has severe limitations. Pay differentials between indus­
entire month. If anything unusual, such as an unpaid
tries reflect many factors unrelated to productivity dif­
holiday, strike, or bad weather, occurs during this period,
ferences, such as the degree of unionization or regional
the estimates will reflect these aberrations for the entire
and geographical differentials. Moreover, the industry
hourly earnings differential does not take into account
occupational changes which occur within an industry.




6 David Schwartzman, “Education and the Quality of Labor,
1929-63,” A m erican E co n o m ic R eview , June 1968.
11

quality of new capital should be incorporated in the
capital stock measures or treated as a productivity in­
crease. Both interpretations have been used in produc­
tivity analysis.
Total factor measures as currently presented are not
consistent with their treatment of capital and labor.
In general, labor refers to actual man-hours whereas
capital refers to available stock not taking into account
3.
Hours paid versus hours worked. Because of lack varying levels of utilization.
of data, productivity measures for the most part refer
A V A IL A B L E MEASURES OF P R O D U C T IV IT Y
to hours paid rather than the more desirable measure of
hours worked.
Labor productivity
Surveys now are being conducted biannually for the
nonfarm economy where information on leave hours
National measures. Each quarter, the BLS prepares
and hours worked will soon provide a body of data
and publishes indexes of output per man-hour for the
which will fill some gaps in this area. Estimates of the
private economy and for the farm, nonfarm, and man­
effects of the difference between hours paid for but not
ufacturing sectors. 7 For these measures, output per
worked on output per man-hour measures developed by
man-hour refers to the constant dollar value of goods
the Bureau of Labor Statistics indicate that the effect
and services produced in relation to the man-hours of
over the last 15 years has been about 0.1 percent per
all persons employed (including proprietors and unpaid
year for the nonfarm economy.
family workers). Corresponding and comparable indexes
The effects of course can vary substantially by sector
of hourly compensation and unit labor costs are also
and at the industry level. Within manufacturing, the
developed.
annual surveys and censuses of manufactures do provide
The output measure for these productivity indexes
measures of what could be termed plant man-hours,
is real gross national product originating in the private
and these are used in many industry productivity
economy or the individual sectors. It comprises the pur­
measures.
chase of goods and services by consumers, gross private
domestic investment (including the change in business
Total factor measures
inventors), net foreign investment, and government all
deflated separately for changes in prices.
Total factor productivity measures relate output to
Final goods and services are differentiated from inter­
the weighted sum of labor and capital and are therefore
mediate products in that they are usually not purchased
subject to the limitations of each of these data series.
for further fabrication or resale. In addition to purchases
The problems of measuring output and labor input have
in the market, final goods and services also include some
been discussed. Capital measures, however, are probably
items provided but not actually purchased such as food
the most difficult and complex measures to derive. They
furnished to employees, food produced and consumed
contain highly differentiated elements, and to express
on farms, and the rental value of owner occupied homes.
this differentiated stock in physical terms requires ad­
Measures for the farm, nonfarm and manufacturing
justing dollar values of assets for price change.
sectors are derived by subtracting the value o f goods and
For the most part, available data for prices of struc­
services purchased by the sector from the constant
tures are based primarily on cost information. As men­
dollar value of products and services leaving the sector.8
tioned earlier, they do not take into account savings
The labor input measures for these series are based
in utilization of materials and other inputs. Furthermore,
largely on a monthly survey of establishment payroll
the problems of obtaining representative prices for
records. Since this survey does not cover total employ­
equipment which is highly differentiated severely affects
ment in the private economy and because there are gaps
the adequacy of the price measures used in capital
in the hours information, it is necessary to use some sup­
measurement.
plementary data to derive man-hours estimates for all
In addition, technical advances are often built into
persons engaged in producing the output of the private
capital so that a piece of equipment produced in an
earlier period may not be as efficient as one currently
7
produced. In constructing price indexes, some of these
Productivity, Wages and Prices, quarterly release issued by
technical improvements may be incorporated as quality
the Bureau of Labor Statistics, U.S. Department of Labor.
The actual measures are developed according to a variety
changes, but adjusting for quality is often difficult. There
of approaches because of data limitations. However, all are
is some question as to whether improvements in the
attempts to approximate this concept.
month. On the other hand, fluctuations in employment
between survey periods may not be reflected in the
sample estimates. For example, short-term layoffs and
plant shutdowns of 1 to 2 weeks between survey periods
would not be reflected in the man-hour estimates for
the month— leading to an overestimate of man-hours
and an underestimate of productivity.




12

economy. Various sources are utilized and data from
them are adjusted for consistency with the establishment
man-hours.
The establishment man-hours are based on an hours
paid rather than an hours worked concept. That is,
the estimates include paid holidays, vacations, sick
leave and other time off paid for by the employer
in addition to actual hours worked.
Another set of labor productivity indexes is developed
based on man-hours obtained from a monthly survey
of the noninstitutional civilian population. This survey
of households provides information on the labor force,
employment, unemployment as well as man-hours. The
man-hours estimates for the labor force series are based
on an hours worked concept, i.e., hours spent at the
establishment, thus excluding vacation and sick leave
but including such things as rest periods and standby
time.
Since compensation data are derived primarily from
establishment payroll records, when relating labor pro­
ductivity measures to hourly compensation, the appro­
priate series is the one based on establishment man-hours.
On the other hand, when examining the relationship
between productivity changes and displacement of work­
ers, since the employment and unemployment measures
are based on the household survey, the more consistent
output per man-hour measure is the one based on labor
force data.
In addition to the current indexes of output per man­
hour published by .the BLS, John Kendrick of the
National Bureau of Economic Research has published
indexes of output per man-hour for the private economy
and major sectors (as well as total factor productivity)
which include a series that makes adjustments for
changes in the composition of man-hours.9 Using average
hourly earnings for weighting man-hours at the industry
level he derives an index of output per weighted
man-hour. The basic man-hour data for this series are
generally the same as those for the establishment series
and the weights are also derived from BLS average
hourly earnings data. These indexes presently cover
the period 1887 to 1966.
Edward Denison of the Brookings Institution has
published measures of output per labor input in the
form of growth rates for selected periods, the most recent
being 1950-62.10 These measures also take into account
changes in the quality of labor; however, the procedure
differs from Kendrick’s. Adjustment based on age, sex,
education, and other changes in the labor force are
applied to basic employment and man-hours measures
to derive labor input reflecting changes in quality.
These measures are available only at the national level.




13

Industry measures. In addition to the indexes for the
private economy and major sectors, the BLS publishes
annually indexes of output per man-hour for selected
industries.11 At the present time, measures for about 40
manufacturing and nonmanufacturing industries such
as steel, motor vehicles, railroad transportation, coal,
etc., are prepared.
The output measures for these indexes are developed
by combining the data on quantities of commodities
or services within the industry with fixed period weights.
As mentioned earlier, man-hour weights would be pre­
ferred for developing these measures and insofar as pos­
sible these are used. However, where such information
is not available, other weights such as unit labor costs,
or unit value (price) are used. These substitutions are
introduced on the assumption that unit values are good
commodities in an industry.
In addition, for some industries where it is not pos­
sible to obtain any quantity information, indexes of
deflated value of output are developed. For these
industry measures current dollar value estimates are
divided by indexes of price change for the industry to
derive a real output measuer. The adequacy of these
measures is dependent on the quality of the price
measure.
The labor input data for these measures are estab­
lishment man-hours. As in the aggregate measures, they
are derived from payroll records and for the most part
are based on an hours paid concept. For manufacturing
industries, however, additional man-hours information
is available in terms of hours at the plant. These data
which theoretically exclude vacation, holdiays, and such
leave hours are closer in concept to an hours worked
measure. Unfortunately, the information on plant man­
hours is usually not as current as that from other
sources on establishment man-hours.
Capital productivity

Measures of output per unit of capital are not avail­
able on a current basis. Historical measures have been
developed by a number of researchers.
Separate estimates of capital stock and hence meas­
ures of capital productivity are available for the private
economy from sources such as the National Industrial
Conference Board, National Bureau of Economic Re­
search, and many economists doing research in pro­
duction analysis. In addition, the Office of Business
Economics prepares 12 different capital series depending
on alternative options for service lives, depreciation, and
9 John W. Kendrick, Productivity Trends in the United
States, Princeton University Press, 1961.
10 Edward F. Denison, op. cit.
1 Indexes o f Output per Man-Hour, Selected Industries,
1
1939 and 194 7-70 (BLS Bulletin 1692).

ices and for selected industries within these major
groups.
Labor input is measured in man-hours and adjusted
for quality change using industrial hourly earnings.
Capital includes the net stock of structure, plant equip­
ment, inventories, working capital and land. The capital
measures do not include quality inprovements which
can occur because of technical advance. The capital and
labor input are added together with factor prices as
weights. The base period for the weights was changed
periodically to reflect economic conditions of the vari­
ous subperiods under analysis.
The combined factor measure developed by Edward
Denison relates net domestic product (excluding depre­
ciation) to weighted sum of capital and labor. The
weights are the base period share of dollar output of
each of these inputs. He also periodically shifted the base
period for the factor shares to reflect current economic
conditions. Denison’s analysis covers the total economy
for the 1919-62 period, and he is currently updating
his work.
His labor input is employment adjusted for quality
change using relative earnings for selected age-sexeducation groups. He also adjusts the labor input for
intensity of effort as reflected in varying lengths of the
workweek. His assumption is that as the workweek de­
clines productivity improves because the worker is less
fatigued and can work more diligently.

adjustment for price change. These estimates are con­
structed within the framework of the national accounts
and are consistent with output measures used for labor
productivity in the private economy. However, the vari­
ation among these series gives different results for the
growth of capital productivity.
The OBE estimates are prepared annually for the
private economy, farm, manufacturing, nonfarm, and
nonmanufacturing. They include equipment and nonresidential structures, but exclude such items as housing,
motels, and hotels.
These capital stock series are developed using the
perpetual inventory method. That is, each new piece of
equipment or structure is added into the stock estimates
and remains there until retired from use. Retirements of
assets are based on mean useful service lives published
in the Internal Revenue Services Bulletin of Service Lives
of Assets. Because the latter are believed to over­
state asset lives, OBE prepares estimates of capital
series either based on Bulletin F or 85 percent of
Bulletin F service lives.
Net capital measures are derived using either straight
line or double declining balance depreciation.
Most other capital series are developed in a manner
similar to the OBE measures with other variations on
mean service lives and methods of depreciating assets.
Combined factor input productivity

The second group of productivity measures— those
which relate output to several factors— involve the
weighting together of the quantities of the separate
factors. For the most part, these measures have been
limited to output per unit of capital and labor combined.
Just as the separate components of an output index
must be combined with appropriate weights, the separate
components of the input measure also must be appro­
priately weighted together. Capital and labor can be
aggregated using their unit costs (e.g., wages, rate of re­
turns of capital) in a base year as weights. These weights
can also be viewed as the proportion of current dollar
output earned by each input (factor share) in a base
period.
Two sources of combined factor input or total factor
productivity measures exist— the work done by John
Kendrick for the National Bureau of Economic Research
and that by Edward Denison in his work on sources of
growth.
Kendrick provided annual measures for the period
1889 to 1957, and more recent measures, 1957-69, will
be available early in 1972. Estimates cover the private
economy and are based on GNP output measures. Sep­
arate measures were made for farm, manufacturing,
trade, finance, transportation, public utilities, and serv­




REC O M M END ATIO NS

Within the general constraints imposed on all users of
economic data, productivity measures for the total priv­
ate and private nonfarm economics and for selected
major sectors (manufacturing, mining, trade, transporta­
tion, communication, and public utilities) are reliable
and useful for economic analysis. Conversely, produc­
tivity measures for construction and service-type indus­
tries are not reliable measures for identifying either the
magnitude or direction of change in productivity for the
reasons outlined above. To improve these measures, ad­
ditional information must be developed in two areas—
the data base from which output, input, and price
statistics are compiled, and the conceptual base upon
which the output and price data are developed.
Additional price information

More and better price information in the service and
construction industries is of prime urgency. In construc­
tion, work is currently underway by the Census Bureau
to develop price measures in order to improve the meas­
ures of the real volume of residential construction put in
place. Additional research is also necessary for nonresidential construction. In the service sector, more adequate
14

and man-hours. Some research is being carried out using
occupational and wage data to account for some changes
in labor quality, but it is necessary to develop a more
integrated system for collection than in currently in
effect.
Capital information is also needed to make a more com­
plete analysis of factors affecting productivity growth.
Of paramount importance are better data on changes in
the quality of capital equipment and the length of
service lives.
The concepts used to define output need to be changed
to conform to a productivity framework. This is particu­
larly true of government and households and institutions
where actual output definitions are needed rather than
merely relying on an employment measure.
Financial intermediaries also present definitional prob­
lems. For example,banking output is currently measured
as liquidity. If the output reflected changes in the num­
ber of transactions weighted by some value measures, it
would be more compatible with the inputs and provide
a means for making a better productivity measure.
These recommendations will not solve all of the prob­
lems of productivity measurement. However, they will
certainly improve the output and related input measures
and thereby make productivity a more viable tool for
economic analysis.

and extensive price information on personal services is
needed as well as the expansion of wholesale prices to
include business services. The BLS at the present time is
trying to develop an index of the general price level. The
development of this index will materially assist produc­
tivity measurement because it will require the collection
of a wider range o f service prices.
More work is also needed in collecting data on durable
equipment (such as heavy machinery and aircraft) which
is highly differentiated and often custom-made.
Another recommendation is to have timely (quarterly
and annually) estimates on the imputed rental value of
owner occupied housing so that it may be excluded from
the output estimate and not bias productivity meas­
urement.
Better man-hour and capital data

Input data also need to be improved. Most important
to productivity measurement would be better estimates
for certain components of labor input. This would entail
estimating supervisory hours in nonmanufacturing and
expanding employment sampling coverage so that re­
sultant data refer to the entire month rather than
1 week. Adjustment for changes in labor quality calls
for more detailed information on occupational wages




15
* U .S . GOVERNMENT PRINTING O F FIC E : 1971 O - 484-784 (18)




'

.

■

■

.

■

INDIANA UNIVERSITY
INDIANAPOLIS LAW
LIBRARY

Region I
1603-A Federal Building
Government Center
Boston, Mass. 02203
Phone: 223-6762 (Area Code 617

Region V
219 South Dearborn St.
Chicago, III. 60604
Phone: 353-7230 (Area Code 312)

Region II

Region VI
1100 Commerce St., Rm. 6B7
Dallas, Tex. 75202
Phone: 749-3516 (Area Code 214)

341 Ninth Ave., Rm. 1025
New York, N.Y. 10001
Phone: 971-5405 (Area Code 212)

406 Penn Square Building
1317 Filbert St.
Philadelphia, Pa. 19107
Phone: 597-7796 (Area Code 215)

Regions VII and V III
Federal Office Building
911 Walnut St., 10th Floor
Kansas City, Mo. 64106
Phone: 374-2481 (Area Code 816)

Region IV
Suite 540
1371 Peachtree St. NE.
Atlanta, Ga. 30309
Phone: 526-5418 (Area Code 404)

Regions IX and X
450 Golden Gate Ave.
Box 36017
San Francisco, Calif. 94102
Phone: 556-4678 (Area Code 415)

Region III




• Regions VII and V III will be serviced by Kansas City.
* * Regions IX and X will be serviced by San Francisco.

INDIANA UNIVERSITY
INDIANAPOLIS LAW SCHOOL