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.........