Labor Force and Unemployment in the 1920's and 1930's: A Re-Examination Based on Postwar Experience Author(s): Robert M. Coen Source: The Review of Economics and Statistics, Vol. 55, No. 1 (Feb., 1973), pp. 46-55 Published by: MIT Press Stable URL: http://www.jstor.org/stable/1927993 Accessed: 29-02-2016 02:24 UTC Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/ info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. MIT Press is collaborating with JSTOR to digitize, preserve and extend access to The Review of Economics and Statistics. http://www.jstor.org This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions LABOR FORCE AND UNEMPLOYMENT IN THE 1920'S AND 1930'S: A RE-EXAMINATION BASED ON POSTWAR EXPERIENCE Robert M. Coen * I Introduction virtually no cyclical variation, since annual changes in labor market conditions are not A NNUAL data on the labor force were not taken into account. Thus, Lebergott's annual collected by the United States govern- estimates are not of the same character as the ment before 1940. The decennial censuses prior annual (or monthly) estimates prepared by to 1940 did include labor force enumerations, the Census Bureau after 1940,2 and unemploy- although the definitional basis of these figures ment rates derived from them are not truly was the "gainful worker" concept rather than comparable with the official series for the post- the labor force concept which underlies census 1940 years. data beginning in 1940.' Stanley Lebergott Empirical studies of post-World War II data (1964) has undertaken the most careful and indicate that the size of the labor force in the exhaustive adjustments of the gainful worker short run depends in part on the degree of figures to make them consistent with the labor tightness in the job market.3 The search for force concept and with the sampling procedures used by the Census Bureau in its Current Popu- 2 This point was raised some time ago by Martin Gains- brugh. In commenting on an earlier but similar set of lation Surveys which began in 1940. He has Lebergott's estimates, he noted that, "The basic informa- also constructed an annual labor force series for tion (employed by Lebergott) is static in character as the 1900-1939 period which is consistent with compared with the dynamic concepts of labor force in- corporated in census or sample enumerations after the 1940 his adjusted decennial census figures. One census" (1957, pp. 239-240). shortcoming of these annual labor force estiLebergott's unemployment rate estimates for the 1920's mates is that they do not fully reflect the in- and 1930's have been used in the notable econometric study of the labor market by Lucas and Rapping (1969) covering fluence of economic variables on labor force the period 1929-1965, despite the fundamental difference participation. His procedure was basically to between this series and the official annual estimates be- (1) obtain detailed labor force participation ginning in 1940. The controversy which has arisen over whether the Lucas and Rapping model explains labor market rates by age, sex, and in some cases, nativity, behavior in the 1929-1939 period (see Rees (1970) and for decennial census years, (2 ) interpolate these Lucas and Rapping (1972)) cannot be fully resolved, in my rates linearly between census years, and (3) opinion, until more satisfactory estimates of the unemploy- ment rates in these years (such as the ones presented in apply the interpolated rates to population data this paper) are used in statistical tests of the mnodel. for the intercensoral years. As one might ex- 3 See, for example, the studies of Tella (1964, 1965), pect, the resulting labor force series displays Dernburg and Strand (1966), Strand and Dernburg (1964), Kuh (1966), Black and Russell (1970), and Bowen Received for publication April 25, 1972. Accepted for and Finegan (1969) who present a critical review of earlier publication September 1, 1972. * This research was supported by the National Science works. In recent years the Department of Labor has undertaken Foundation, Grant 1686. I am indebted to Dana N. surveys of nonparticipants in the labor force in an effort to Stevens and Michael J. Varn for computational assistance determine directly reasons for nonparticipation (see Flaim and to Bert G. Hickman for helpful comments. 1971). Data for 1967-1971 confirm to some extent the 1 The gainful worker concept placed an individual in the labor force if (1) he was working or (2) he was not work- ing but was "willing and able to work." Measurements beginning in 1940 place an individual in the labor force if (1) he is working, (2) he is not working but is actively seeking work, (3) he is not working but is waiting to be recalled to a job from which he has been laid off or is waiting to report to a new job, or (4) he is not working and not actively seeking work because he is temporarily ill or because he believes there is no work in his line available findings of the econometric studies noted above. Nonparti- cipants classified as discouraged because they could find no jobs or thought that no jobs were available decreased in number as the unemployment rate fell from 3.8 per cent in 1967 to 3.5 per cent in 1969, then increased in 1970 and 1971 as the unemployment rate climbed to 5.9 per cent. The magnitude of these changes, however, is much smaller than those suggested by estimates of participation relations. Also, the Labor Department's survey data indicate that the absolute number of discouraged workers of the type noted in his community. above averaged about 355 thousand during the period [46] This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions LABOR FORCE AND UNEMPLOYMENT 47 work is, after all, a costly undertaking, and the willingness of individuals to engage in job outline the determinants of the potential labor force, i.e., the labor force that would be avail- search ought to depend on the probability of able at full employment. We then specify the success of the venture. When unemployment relation between the potential labor force and increases and job opportunities decline, some the actual labor force. The potential labor force relative to the work- individuals who were seeking work may become discouraged and leave the labor force, and ing-age population ought to depend on a variety conversely when employment increases. On the of economic, institutional, and demographic other hand, unemployment means loss of in- variables. The decisions of households regard- come, so that an increase in unemployment may bring some individuals into the labor force as families attempt to maintain their incomes. The evidence suggests that the discouraged- worker effect prevails over the added-worker effect. ing labor supply involve choices between income and leisure, given the level of real wages.4 If real wages, which may be determined largely by labor productivity, are increasing more rapidly than desired income, members of house- holds who are employed, or seeking employ- This study presents new annual estimates of ment, may choose to work fewer hours or some labor force and unemployment in the 1920's may withdraw entirely from the labor force. and 1930's which allow for the dependence of Desired average hours per worker and labor labor force participation on the state of the job force participation might then both depend on market. The approach to be taken makes use real wages. In practice, the link between labor of a labor force participation equation fitted to force participation and real wages is likely to post-World War II data as an interpolator be- be rather weak, since determination of the tween Lebergott's adjusted decennial census household members actually assigned the task figures. The estimates appear to stand up well when judged in the light of several reasonable criteria, and they lead to a view of this histori- cal period which differs in some important re- spects from that suggested by Lebergott's series. Also, the initial analysis of the pos;twar data provides some new insights into the de- terminants of labor force participation. of earning income is greatly influenced by in- stitutional factors and social customs which establish the nature of family life. Nonethe- less, the real wage deserves to be tested and is included in our empirical analysis. Institutional factors are, of course, difficult to quantify, and they will necessarily be treated indirectly by including a constant term and a trend in the potential labor force equation. Section II of the paper presents the empiri- Demographic variables measuring the age or cal analysis of participation behavior in the postwar years. The new labor force and un- employment estimates for the 1920's and 1930's are discussed in section III where they are compared with Lebergott's series. Section IV summarizes the findings. sex composition of the working-age population or the ratio of the working-age population to the total population are likely candidates for in- clusion in the potential labor force equation; but none of these was found to be an important causal variable in the relations fitted to post- war data. Significant changes have taken place II Determinants of Labor Force Participation in the age structure of the population and in The labor force participation relation utilized labor force participation rates of different age- in this study is derived in two steps. First, we sex groups, but the net influence of these changes on aggregate participation appears to 1967-1969, even though the unemployment rate remained be small and is accounted for by other variables below 4 per cent. Estimates of the number of discouraged included in this study. workers derived from the econometric studies of participa- tion vary considerably, but they are all based on the as- sumption that there are no discouraged workers (at least on a net basis) when the unemployment rate equals 4 per cent. Thus, the new survey data raise some doubts about both the cyclical variability and absolute number of discouraged workers found in previous econometric studies. 'These remarks, like the empirical work which follows, ignore the possible dependence of the work-leisure choice on the desired rate of saving of the household. In the more general case, the supply of labor might be influenced by the level of household wealth and the rate of interest, as well as by the real wage. This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions 48 THE REVIEW OF ECONOMICS AND STATISTICS One exception to the above remarks is the potential average hours per worker, that is, inclusion of military emplovment as a determi- average hours at full employment. In a cyclical nant of the potential labor force, defined to in- downturn employers generally reduce both clude only civilians. Military employment is average hours and the number of employees, so not a perfect substitute for civilian employ- that deviations of either of these variables from ment and is not entirely voluntary. The effect their full employment levels should serve as of an increase in military employment on the indicators to workers of the state of the job civilian labor force cannot be established from market. Furthermore, reductions in either em- a priori considerations. If the increase is ac- ployment or average hours will result in declines complished through the draft and the draftees in workers' incomes. Stating this relationship were not initially in the civilian labor force, the algebraically, we have size of the civilian labor force may be little L LP E EP\ N -N =a5N -N Ja6(H -HP), (2) affected. However, the draftees may be taken in part from the civilian labor force. In this where case, the decrease in the civilian labor force L = actual labor force may be partially offset by the entrance of sec- E = actual employment ondary family workers attempting to maintain EP = potential employment family income, given that rates of pay in the H = actual average hours military are below civilian levels. If the inHP = potential average hours. crease in military employment is accomplished If full employment existed, both expressions without resorting to the draft, the effect is again on the right-hand side of (2) would be zero, quite uncertain. In view of these consider- and the actual participation rate would be equal ations, it was decided to attempt to explain the to the potential participation rate. The param- size of the civilian labor force, in which par- eters a5 and a6 measure the strength of the ticipation is voluntary, and to estimate the discouraged-worker or added-worker effects. impact of military employment on the civilian Positive values indicate that the discouraged- labor force. worker effect is dominant, and negative values Bringing together these hypotheses regarding that the added-worker effect is dominant. determination of the potential labor force, we To complete the model we must express EP can specify the relation algebraically as: and HP in terms of observable variables. To define potential employment, let us assume that NL= a, + a2t + a3 - + a4W, N 1N(1 e is the proportion of the labor force that would be employed at full employment; that is, e is where the employment rate that workers consider to LP =-potential labor force be the full-employment employment rate; then N -noninstitutional population of working age EP = eLP. (3) t = time trend A -armed forces Note that e is not the full-employment employ- W -real wage rate. ment rate as viewed by policy makers, but by Our second relationship states that the differ- ence between the actual labor force participa- workers. The views of these two groups might well differ, and it would be interesting if we tion rate and the potential participation rate could obtain an estimate of e as it is defined depends on cyclical variations in the job market here. Unfortunately, this is not possible, as we as they effect the probability of obtaining work shall see. and workers' incomes. In particular, the differ- ence between the actual and potential partici- The potential average hours variable is some- what more difficult to specify. Let us suppose pation rates depends on (1) the difference that average hours, H, can be partitioned into between the ratio of actual employment to a trend component, H1, and a cyclical compo- population and the ratio of potential employ- nent, H2, and that the trend component is ment to population and (2) the difference be- tween actual average hours per worker and simply the sum of a constant and a linear trend. Then This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions LABOR FORCE AND UNEMPLOYMENT 49 H1 = bo + b1t, (4) age and older.5 Three variants of equation (8) are presented in table L.0 The first equation and (1.A) is similar in form to relations commonly H2 = H - H = H -bo- bit. (5) found in the econometric literature on labor We shall assume that H1 approximates poten- force participation. The effect of the real wage tial average hours, so that the quantity H -H on potential labor force is constrained to zero, is approximated by H - bo- blt. An alter- as is the cyclical effect of average hours on native approach is to replace the trend by the actual labor force. Employment conditions and real wage, in accordance with the above argu- armed forces are both found to be significant ment which links desired average hours to the determinants of the participation rate; the signs real wage. This alternative will also be investi- of the coefficients confirm the discouraged- gated. worker effect observed by other researchers and Substituting from (3) and (5) into (2), we suggest that an increase in armed forces pro- have duces a net reduction in the civilian labor L E LP force.7 The equation shows rather high first- N = a5 - + ( -a5e) N NN 'The sample is limited to data through 1966 because the + a6H - a6bo -a6bit, (6) official series on labor force and employment were changed in that year from a 14-year-old-and-over basis to a 16-yearand substituting from (1) for LP/N, we obtain old-and-over basis. LE The labor force, employment, population, and armed - =a5-+ (1- -a5e)a, forces data are taken from the Economic Report of the NN President, 1967. Average hours per worker are a weighted A average of average hours of private employees and gover+ (1-a5e)a2t + (1 - a5e)a3 - ment civilian employees, the weights being proportional to N the number of employees in each sector. Government civil+ (1-ae)a4W + a6H ian employment figures are taken from the National In- - acbo- a6b1t, (7) come and Product Accounts. Average hours of govern- or ment employees were assumed to be 2,000 per year throughout the period studied. Average hours of private LEA N = co + clt + C2 N + c3N- + C4W + c5H, employees are computed from Bureau of Labor Statistics data on private man-hours and private employment. The (8) final measure of average hours has the dimension of thousands per year. where The money wage rate is the ratio of an estimate of total Co = (1 - a5e)al - a6bo labor income (derived by Bert G. Hickman and the cl= (1 - a5e)a2-a6b, author) to total man-hours of employed persons. Total labor income is compensation of government civilian emC2= a5 ployees plus private labor income, the latter being defined C3= - a5e)a3 as the sum of private employee compensation (exclusive C4= - a5e)a4 of that paid by proprietorships and partnerships) and an C5= a6- imputed labor share originating in proprietorships and partnerships. The imputation is based on the assumption Equation (8) is the basic participation re- that the labor and nonlabor shares in income originating in lation to be investigated here. It should be the noncorporate sector is the same as that in the corporate noted that several of the "sitructural param- sector. The after-tax money wage is the money wage multi- plied by one minus the average tax rate (federal, state, eters" cannot be identified from estimates of and local) on personal income. The tax rate has been the c's. Two parameters, a5 and a6, which measmoothed to remove cyclical variations. The after-tax real sure the cyclical response of the labor force to wage is the after-tax money wage divided by the implicit GNP deflator for consumer expenditures and has the employment and average hours, are identifiable. dimension of dollars per hour. However, the full-employment employment 6 Other variants of the participation equation were ex- rate, e, and the parameters of the potential amined but are not reported here. In particular, a lagged adjustment formulation of equation (lA) was found to be labor force and average hours relations cannot superior to (LA) itself, but there was no evidence of lagged be identified. adjustment in equations (iB) and (1.C). Furthermore, Estimates of equation (8) were obtained equation (lA) with lagged adjustment still yielded a con- siderably poorer fit than equations (1B) and (1C). from aggregate annual data for the period 'Two sources of variation in employment may well 1949-1966. The labor force, population, and produce an upward bias in its estimated coefficient. First, employment data refer to persons 14 years of sampling variation in measured employment will also be This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions 50 THE REVIEW OF ECONOMICS AND STATISTICS TABLE 1. - ESTIMATES OF LABOR FORCE PARTICIPATION uals. The negative and highly significant co- EQUATIONS, ANNUAL DATA, 1949-1966 efficients on the hours variables indicate that Regression Coefficients and t-Values (in parentheses) an added-worker response occurs as hours fluctuate cyclically. Given the highly signifiVariable or Equation Statistic 1.A 1.B 1.C cant discouraged-worker effect evidenced by 'the coefficient of the employment variable, the fol- Constant term .3717 .6348 .6032 (5.817) (11.30) (10.26) lowing picture of labor force behavior over the t -.0003 -.0015 cycle suggests itself. In the early stages of a (2.017) (6.939) downturn in economic activity it is common for E/N .3901 .5701 .6084 employers to reduce average hours of work (3.254) (7.982) (7.585) before they begin to make sharp reductions in A/N -.7217 --.6394 -.6168 the number of employees. Our participation (4.359) (7.077) (6.171) equation indicates that workers respond to the H - -.1701 -.1502 (5.928) (5.108) decline in their incomes resulting from shorter hours by sending other family members out to W -.0234 (6.125) seek employment. However, if the downturn R?2 .6657 .9028 .8823 becomes more severe and leads employers to S.E.E. .0031 .0017 .0018 Durbin-Watson .7767 2.1471 2.0353 reduce their work forces, it appears that some job seekers view this as a very bearish market signal, become discouraged, and withdraw from the labor force. order autocorrelation in the residuals, as evi- We can note further that the trend in equa- denced by the low Durbin-Watson statistic, but tion (1.B) is highly significant. Recalling no effort was made to correct for this since this equations (7) and (8), we know that the co- equation was judged inferior in any event to efficient of the trend reflects both the trend, if the remaining relations. any, in the potential participation rate and the Equations (1.B) and (1.C) in table 1 intro- trend in average hours. If there is no trend in duce the cyclical effect of average hours and the potential participation rate, then the co- can be discussed together. Note, first of all, efficient of the trend in (1.B) should be equal that the inclusion of average hours greatly im- to the negative of the product of the coefficient proves the fit of the participation function and of average hours and the trend rate of growth also eliminates the autocorrelation in the resid- in average hours. A simple linear trend fitted to present in measured labor force, a point stressed by Mincer postwar average hours indicates a secular de- (1966). This bias should not be very large, however, in cline of 0.0079 (thousands of hours) per year, estimates based on annual data, since the sampling variation so that the product in question is - (-.1701) is not as great in annual as in quarterly or monthly data. Second, we have implicitly treated cyclical variations in (-.0079) =-.0013, which is indeed very close employment as exogenous, that is, as determined by the to the coefficient of the trend in (1.B). Thus, demand for labor, whereas the actual behavior of both we can conclude that there is virtually no trend employment and the labor force may be influenced by demographic influences on the supply side of the market, in the potential participation rate and that the a point stressed by Bowen and Finegan (1969). Bowen trend in the participation equation is simply and Finegan recommend the use of some other variable- picking up the trend in average hours. the unemployment rate, for example-to capture fluctu- ations in labor demand. Replacement of the employment/ Utilizing this interpretation of (1.B), we can population ratio in (1.A) by the unemployment rate was turn to an examination of (1.C), in which the found in this study to result in a coefficient on the un- trend is replaced by the real wage rate. The employment rate that was not significantly different from zero. The ultimate thrust of the criticism of Bowen and real wage carries a highly significant, negative Finegan is that investigators who use the employment/ coefficient, and, in spite of the slightly lower R2, population ratio obtain unbelievably large estimates of equation (1.C) may be preferable to (1.B). hidden unemployment. Our estimates of hidden unemploy- ment (see footnote 16), however, are actually in line with For the small loss in explanatory power over their own, so that the problem seems to center on the use the sample period, we obtain a relation that has of the employment/population ratio alone without taking greater economic content. The long-term moveaverage hours into account. This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions LABOR FORCE AND UNEMPLOYMENT 51 included, 1930 and 1940. Second, we should ment of average hours and the real wage over expect the estimated labor force in 1922 to come the postwar period are closely related.' To the relatively close to Lebergott's estimate for that extent that average hours follow the trend in year, since 1922 is not far in time from 1920, a the real wage, the hours and wage variables in census year on which Lebergott's interpolation (1.C) offset each other. Thus, the influence of is based, and labor market conditions in 1922 the real wage on potential labor supply is lim- were not very different from those prevailing ited largely to its effect on average hours per in 1920. And, a final criterion should be the worker. By allowing for the relation between reasonableness of the behavior of our labor average hours and the real wage, we might very force estimates in the remaining intercensoral well commit smaller errors when using the par- years, considering other information available. ticipation equation to make predictions beyond Labor force estimates for 1922-1940 derived the sample period. For this reason, equation from equation (1.C) were found to be far too (1.C) has been chosen as "the" participation low throughout the period. The 1930 labor equation to be used in the subsequent analysis. force was under-predicted by 2.709 million persons, and that for 1940 by 3.290 million. III Labor Force and Unemployment in the Prewar Period When the prediction errors are stated in terms of participation rates, however, we find an en- Our aggregate labor force participation func- couraging result. The error for 1930 is 0.0305, tion fitted to postwar data can be used to obtain while that for 1940 is 0.0328. These errors are labor force estimates for the prewar years that certainly very similar and suggest that if we incorporate the effects of changing labor mar- merely adjust the constant term in equation ket conditions. Data on all of the indepen- (1.C), we should obtain much better results. dent variables are available for these years, so This was done by taking a simple average of all that is required is to insert the values of the the participation rate errors in 1930 and 1940 independent variables into the equation, calcu- and adding the average error to the constant late the labor force participation rates, and term. The resulting labor force estimates are multiply these rates by the working-age popu- presented in table 2 and are plotted in figure 1, lation for each year.9 Before presenting the along with Lebergott's series. Our estimates results of such calculations for 1922-1940, we for 1930 and 1940 now, of course, lie very close should set forth 'the criteria by which the suc- to the adjusted census figures (Lebergott's cess or failure of the procedure ought to be figures for these years), and our estimate for judged. Perhaps the most stringent test of the 1922 does not differ greatly from Lebergott's. method is its ability to predict accurately the The seemingly arbitrary adjustment of the size of the labor force in the two censuses years constant term deserves some attention. It has already been noted that changes in demo- 8A linear regression of average hours on the real wage for the 1949-1966 period gives the following relation: graphic factors ought to influence aggregate H = 1.7693 - .1354 W R2 = .8271. labor force participation, although the vari- (9.069) ation in such variables during the postwar years If the real wage in equation (1 .C) is capturing only the was not sufficient to obtain meaningful estitrend in average hours, then the coefficient of the real wage (-.0234) should equal the negative of the product of the mates of their impact. Two significant changes coefficient on average hours and the coefficient of the wage in the structure of the population occurred be- in the above equation, -(-.1502) (-.1354) = -.0203. tween the prewar and postwar years. First, the This condition holds approximately. 'The employment and population data are from Leber- proportion of the population 65 years of age and gott's study (1964). The average hours and wage series are over rose substantially, from 5.4 per cent in arrived at by methods described in footnote 5. Govern- 1930 to 9.2 per cent in 1960.10 Since particiment civilian employment data for 1929-1940 are from the National Income and Product Accounts; the 1922-1928 pation rates for individuals 65 and over have figures are estimates prepared by Bert G. Hickman and the generally been well below those for individuals author. The GNP figures used in estimating the wage rate for 1922-1928 are estimates derived by Hickman and me. Since these estimates extend back only to 1922, labor force predictions could not be obtained for earlier years. 10 See U.S. Dept. of Commerce, Bureau of the Census (1960, 1965), series A-1 and A-33. This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions 52 THE REVIEW OF ECONOMICS AND STATISTICS TABLE 2.-ESTIMATES OF THE CIVILIAN LABOR FORCE 0.304 in 1930 and that for the younger age AND UNEMPLOYMENT RATE, 1922-1940 group was 0.548,11 so that the labor force in 1930 would have been smaller by about 1.1 Civilian Labor Force Unemployment Rate (millions of persons) (per cent) million persons. Hence, the shift in the age Equation (1.C) Equation (I.C) Year Lebergott Adjusted Lebergott Adjusted structure of the population between the prewar and postwar periods could partially account for 1922 42496 42.765 6.7 7.3 1923 43.444 44.373 2.4 4.5 the underestimates of the prewar labor force 1924 44.235 44.725 5.0 6.0 obtained from our postwar relation. 1925 a 44.934 45.962 2.7 4.9 The second major demographic change was 1926 45.629 46.722 1.8 4.1 1927 46.375 47.220 3.3 5.0 the decline in the proportion of the population 1928 47.105 47.737 4.2 5.5 living on farms, from 24.5 per cent in 1930 to 1929 47.757 48.873 3.2 5.5 7.5 per cent in 1960.12 If labor force parti1930 48.523 48.622 8.9 9.1 cipation rates are higher for the farm popula1931 49.325 47.500 16.3 13.0 1932 50.098 46.842 24.1 18.8 tion than for the nonfarm population, which I 1933 50.882 47.440 25.2 19.8 suspect is the case although I cannot find data 1934 51.650 51.227 22.0 21.3 to verify this, then again we would expect our 1935 52.283 51.754 20.3 19.5 postwar relation to underpredict the prewar 1936 53.019 52.759 17.0 16.6 1937 53.768 53.599 14.3 14.1 labor force. Even a small excess of the farm 1938 54.532 53.676 19.1 17.8 over the nonfarm participation rate would lead 1939 55.218 54.475 17.2 16.0 to sizable underpredictions, since the popula- 1940 55.640 55.526 14.6 14.4 tion shift being considered is so large. Thus, the a The civilian labor force figure for 1925 in table A-3 of Leber- gott's study (1964) appears to be incorrect. It is not equal to the combined effects of these two demographic total labor force figure minus armed forces. I have assumed that the total labor force figure is correct and recalculated the civilian labor changes could alone account for a large part of force and unemployment rate estimates. the error in our predictions of the level of the prewar labor force. There are, undoubtedly, age 14 to 64, a sizable increase in the propor- other structural changes in the population which tion of the population 65 and over will reduce tend to produce errors in our predictions, some the aggregate participation rate. To give some perhaps working in the opposite direction from idea of the magnitude of the change involved, the two discussed here; but I have been unable suppose that the proportion of the population to discover other changes that might have had 65 and over in 1930 had been 9.2 per cent a major influence. rather than 5.4 per cent. There would then have There are, of course, other factors that could been an additional elderly population of about explain, or help to explain, our systematic un- 4.7 million persons and, let us assume, 4.7 mil- derpredictions of the prewar labor force. Re- lion fewer persons age 14 to 64. The participa- calling equation (7), we see that the constant tion rate for the elderly was approximately term in the participation equation is equal to (1 - a5e)al - acbo, where a5 and a6 measure FIGURE 1. CIVILIAN LABOR FORCE, 192 2-1940 the responses of participation to cyclical vari- MiI hohs of ations in employment and average hours, e is pers ons S56 the full-employment employment rate as viewed S5 ~ by workers, and a, and bo are the constant terms '4 / '3 in the potential participation equation and the 52 Lebergots tortes potential average hours equation. The above arguments regarding structural shifts in the population indicate that a1 was probably larger in the prewar than in the postwar period. Since 49 / / \~~~~~~~ ~Adjusted estinsues 46 / from equotlon (I.C) predictions from our postwar relation appear 44 // "See U.S. Dept. of Commerce, Bureau of the Census 43 - / (1960, 1965), series D-14 to D-25 and A-27 to A-33. 42_ 12 See U.S. Dept. of Commerce, Bureau of the Census 22 23 24 *25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 (1960, 1965), Series A-1 and A-38. Year This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions LABOR FORCE AND UNEMPLOYMENT 53 to be off only in terms of their level, there is reason to believe that the cyclical response co- efficients a, and a6 in the prewar period were not very different from those estimated from post- war data. Thus, the increase in the prewar con- stant term in the actual participation relation that cannot be attributed to a rise in a1 must be due to either a lower full-employment employ- ment rate or a higher bo. There is little basis for speculating about changes in bo, although a smaller proportion of part-time workers in the prewar labor force might account for a larger bo in those years. What about e? The possibil- ity that workers in the prewar years had a less optimistic view of what constitutes full employ- ment is intriguing and perhaps even plausible. We might expect the unemployment rate at full employment to be higher if workers are less mobile or if production patterns by industry and region are changing rapidly because of shifts in demand or differences among indus- tries in rates of technical change. Such con- siderations suggest that if the prewar and post- war e's differed at all, the former was probably smaller than the latter.13 This issue certainly changed.15 Since unemployment rates as low as 2-4 per cent have typically been accompanied by considerable inflationary pressure in the United States economy, our higher unemploy- ment rates appear to be more consistent with observed wage-price behavior than are Leber- gott's. It must be admitted, however, that this decade in our history remains somewhat puz- zling. If the full-employment unemployment rate was higher in the prewar than in the post- war period, as we argued above, then these higher unemployment rate estimates might still indicate that a relatively tight labor market ex- isted during most of the 1920's, thus under- mining what might appear to be an explanation of wage-price stability during the period. The degree of inflationary pressure in the economy should be more closely related to the difference between the actual unemployment rate and the "normal" or full-employment unemployment rate than to the level of the actual unemploy- ment rate. Our labor force estimates fall below Leber- gott's throughout the 1930's and therefore yield lower unemployment rate estimates for those deserves careful study. years. This is to be expected, since Lebergott's Returning to our labor force estimates in table 2, we note first that they exceed Leber- gott's figures throughout the 1920's and there- interpolation method utilizes participation rates from 1930 and 1940, both years of high employ- ment relative to the deeply depressed years of fore yield considerably higher estimates of un- the 1930's. The peak unemployment rate in our employment rates in these years (see columns 3 and 4 of table 2). For the 1922-1929 period Lebergott's unemployment rates reach a peak series occurs in 1934 rather than 1933. This seems odd since the trough in real GNP is placed by the Commerce Department in 1933 of 5.0 per cent in 1924 and hit a low of 1.8 per and Knowles' work (1960) indicates that the cent in 1926; our series has its peak and trough trough in the ratio of actual to potential GNP in the same years, but the levels are 6.0 per also occurred in 1933. The rise in output in cent and 4.1 per cent respectively.14 This hap- 1934 was, of course, accompanied by a substan- pened to be a period of rather remarkable wage tial increase in employment which leads to an and price stability; the money wage increased increase in the size of the labor force but by a at a rate of only 1.4 per cent per year, the de- smaller amount than the increase in employ- flator for consumer expenditures decreased ment. Thus, one would think that the unem- slightly at a rate of 0.1 per cent per year, and ployment rate should have declined. However, the deflator for GNP remained practically un- there was a dramatic decline in average hours in 1934, no doubt due in large part to federal 13 Calculation of the possible magnitude by which the wage-and-hours legislation that went into effect postwar e exceeds the prewar e is complicated and risky, since at a minimum it requires an estimate of a, the con- in that year. According to our participation re- stant term in the potential participation relation. lation, a reduction in average hours also leads to 14 Our estimates of unemployment in the 1920's come an enlargement in the labor force, and it is this closer to those of Daniel Carson and Meredith Givens than to Lebergott's. See Lebergott's study (1957) in which he compares various estimates of unemployment for this 15 These figures are derived from the Hickman-Coen esti- period. mates of the underlying series for the period 1922-1928. This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions 54 THE REVIEW OF ECONOMICS AND STATISTICS factor which accounts for the unemployment to employment and hours evidenced in the post- rate peak in 1934. war data give sensible predictions of cyclical changes in the prewar era. Our systematic IV Conclusions underestimation of the prewar labor force can be partially explained by differences in the Our examination of the determinants of structure of the population between the prewar labor force participation in the postwar pe- and postwar periods, but underestimation might riod has turned up several new findings. Both also have resulted from workers' expectations armed forces and cyclical variations in average of a higher normal unemployment rate in the hours per worker, variables which have gener- prewar years than in the postwar years. Esti- ally been ignored in previous studies, appear mates of prewar unemployment rates derived to have significant effects on participation. The from our adjusted labor force predictions sug- added-worker response to average hours tends gest that contrary to Lebergott's findings and to moderate cyclical changes in participation the views of many economists, the din of the that would be predicted from participation re- "Roaring Twenties" may have included a few lations which include only an employment whimpers from workers. In five of the eight variable. Hence, estimates of the number of dis- years from 1922 to 1929, the unemployment couraged workers and of so-called hidden rate exceeded 5 per cent, and in one remaining unemployment based on our participation re- year it was just a shade under 5 per cent. Our lation are considerably smaller than those results present few fresh insights into the derived by most other investigators."6 The in- 1930's, although they locate the peak unem- fluence of the real wage on long-run labor sup- ployment rate in 1934 rather than 1933. These ply decisions appears to be confined to its role new estimates of labor force and unemploy- in determining desired hours of work; potential ment should prove superior to those currently participation is virtually unaffected by the real available in studies of wage-price behavior and wage. Finally, our careful derivation of the potential output in the prewar years. participation equation indicates that many of the underlying parameters of interest cannot be REFERENCES identified from estimates of the equation. How- ever, knowledge of the relations between the Black, S. W., and R. R. Russell, "Participation Func- estimated coefficients and the underlying parameters is seen to be useful in interpreting tions and Potential Labor Force," Industrial and Labor Relations Review, 24, no. 1 (Oct. 1970), 84-94. the estimates and in speculating about changes Bowen, W. G., and T. A. Finegan, The Economics of in the underlying parameters. Labor Force Participation (Princeton: Princeton "Backcasting" from our postwar participaUniversity Press), 1969. tion equation to obtain estimates of the prewar Coen, R. M., "Aggregate Labor Supply in the United labor force, we find that the cyclical responses States Economy," Memorandum no. 117, Research Center in Economic Growth, Stanford University 16 Preliminary estimates of hidden unemployment for the (Aug. 1971). prewar and postwar years derived from our participation Dernburg, T., and K. Strand, "Hidden Unemployment relation are presented in Coen (1971), table 6. Calculation 1953-62: A Quantitative Analysis by Age and Sex," of hidden unemployment in our model requires estimation American Economic Review, LVI, no. 1 (Mar. of full employment average hours per worker. The proce- 1966), 71-95. dure followed is described in the paper just referred to. Our estimate of hidden unemployment in 1960, for ex- ample, is 538 thousand. By comparison, Dernburg and Strand's estimate (1966) is 1.2 million; and for the census Flaim, P. O., Employment in Perspective: Discouraged Workers and Recent Changes in Labor Force Growth, Report 396, U. S. Department of Labor, week of 1960, Tella's estimate (1965) is 780 thousand and Bureau of Labor Statistics (Washington, D.C.: Bowen and Finegan's estimate (1969) is 386 thousand. The Government Printing Office, 1971). Tella and Dernburg-Strand estimates do not appear in the Gainsbrugh, M., "Annual Estimates of Unemployment references just cited but were calculated by Bowen and in the United States, 1900-1950: Comment," in Finegan (see p. 515, footnote 14). Since the unemployment The Measurement and Behavior of Unemployment, rate in the census week was about 0.4 of a percentage point below that for the year, our estimate clearly comes very close to the Bowen and Finegan figure, that is, to the low end of the range of these estimates. National Bureau of Economic Research, Special Conference Series no. 8 (Princeton: Princeton University Press, 1957), 239-241. This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions LABOR FORCE AND UNEMPLOYMENT 55 Knowles, J. W., "The Potential Economic Growth in Mincer, J., "Labor-Force Participation and Unemploy- the United States," Study Paper no. 20, Joint Economic Committee, 86th Congress, 2nd Session, ment: A Review of Recent Evidence," in R. A. Gordon and M. S. Gordon (eds.), Prosperity and Unemployment (New York: John Wiley and Sons, January 30, 1960. Kuh, E., "Measurement of Potential Output," American Economic Review, LVI, no. 4 (Sept. 1966), 758- Inc., 1966) 73-112. Rees, A.. "On Equilibrium in Labor Markets," Journal of Political Economy, 78 (Mar./Apr. 1970), 306- 776. 310. Lebergott, S., "Annual Estimates of Unemployment in the United States, 1900-1950," in The Mleasure- ment and Behavior of Unemployment, National Bureau of Economic Research, Special Conference Series no. 8, (Princeton: Princeton University Press, 1957), 213-239. , Manpower in Economic Growth: The American Strand, K., and T. Dernburg, "Cyclical Variation in Labor Force Participation," this REVIEW, XLVI (Nov. 1964), 378-391. Tella, A., "Labor Force Sensitivity to Employment by Age and Sex," Industrial Relations, 4 (Feb. 1965), 69-83. , "The Relation of Labor Force to Employment," Record Since 1800 (New York, McGraw-Hill, 1964). Indutstrial and Labor Relations Review, XVII Lucas, R. E., Jr., and L. A. Rapping, "Unemploy- ment in the Great Depression: Is There a Full Explanation?" Journal of Political Economy, 80, no. 1 (Jan./Feb. 1972), 186-191. , "Real Wages, Employment and Inflation," (Apr. 1964), 454-469. U.S. Department of Commerce, Bureau of the Census, Historical Statistics of the United States: Colonial Times to 1957 and Continuation to 1962 and Journal of Political Economy, 77, no. 5 (Sept./ Revisions (Washington, D.C.: Government Print- Oct. 1969), 721-754. ing Office), 1960 and 1965. This content downloaded from 134.71.247.199 on Mon, 29 Feb 2016 02:24:01 UTC All use subject to JSTOR Terms and Conditions
© Copyright 2025 Paperzz