Social Security In America

Part I

UNEMPLOYMENT COMPENSATION

Chapter III

ESTIMATES OF UNEMPLOYMENT IN THE UNITED STATES

ADEQUATE statistics of unemployment can be obtained only by complete censuses or a compulsory system of registration in connection with an unemployment compensation system covering the entire industrial population. Present data on the extent of unemployment in this country are so incomplete that it is necessary to resort to estimates of the number out of work. Such figures as are available indicate the existence of a large volume of unemployment in both good years and bad, with a concentration of unemployment in the years of cyclical depression. The estimates of unemployment set forth in this chapter indicate that about l0 percent of the unemployment from 1922 to 1933, inclusive, occurred in the years 1930 to 1933.

TOTAL VOLUME OF UNEMPLOYMENT

Adequate statistics of unemployment are lacking in the United States for a number of reasons. The trade-union movement has not maintained satisfactory records of unemployment among its members; a Nation-wide public employment service is only now in process of development; and unemployment compensation with its corollary of registration of unemployed persons has not been in existence except on a limited basis. Even employment figures, which, until recently, have been collected much more widely than unemployment figures, are chiefly in the field of manufacturing as reported to the United States Bureau of Labor Statistics and to some State labor departments by representative firms. The national census of unemployment in April 1930 and occasional censuses in a few localities constitute practically the only other sources of information.

The lack of accurate data has necessitated recourse to estimates, among which are the studies of William A. Berridge in Cycles of Unemploywnent in the United States, 1903-1922, and of Hornell Hart in Fluctuations in Employment in Cities of the United States, 1902-1917, the estimate of Leo Wolman and Meredith B. Givens of unemployment

56

among non-agricultural labor from 1920 to 1927 in Recent Economic Changes, and Paul H. Douglas' estimate of unemployment in manufacturing, transportation, building trades, and mining from 1897 to 1926 in Real Wages in the United States, 1890-1926. Douglas' estimate, which covers the longest period of years, is given in table 4. Unfortunately, it does not include the most recent years and is limited in industrial scope. The estimate of Wolman and Givens is wider in coverage, including all employees except those attached to agricultural pursuits, but does not go beyond 1927. Comprehensive estimates of the unemployment in the United States from 1929 through 1933 have been published by Robert R. Nathan.{1} His

SOURCE: Douglas, Paul H., and Director, Aaron, The Problem of Unemployment (Macmillan Company, New York, 1931), p. 28.

estimates, together with those of Wolman and Givens, are shown in the following tabulation:

{1} Nathan, Robert R., "Estimates of Unemployment in the United States, 1929-1935", International Labour Review, vol. XXXIII, no. 1, January 1936, p. 49.

{2} Committee of the President's Conference on Unemployment, Recent Economic Changes (McGraw-Hill Book Co., New York, l929), vol. II, p. 478. Excludes employees engaged to agricultural pursuits.

{3} No estimate available.
Nathan, Robert R., op. oft., table 1.

57

In addition to the above estimates of the average annual volume of unemployment, month-to-month estimates are now issued regularly by the American Federation of Labor and the National Industrial Conference Board. The monthly estimates of the former date back to January 1930; those of the latter to January 1933. Both are based on the census of unemployment of 1930, but have deviated considerably, primarily because of differing estimates as to the number of new workers who have entered industry since this census was taken and because of differences in the definition of unemployment. The American Federation of Labor estimates are published regularly in its official monthly magazine, The American Federationist; those of the National Industrial Conference Board have appeared in newspaper releases issued from its offices in New York.

VARIATIONS BY INDUSTRIES

The extent of unemployment varies greatly among different industrial groups. Wolman and Givens in Recent Economic Changes estimated that unemployment in 1921 was 23.7 percent in manufacturing; 26.6 percent in construction; 14.4 percent in transportation and communication; 38.1 percent in mines, quarries, and oil wells; and only 3.7 percent in public service, mercantile, and miscellaneous industries. In July 1934, on the basis of the American Federation of Labor estimates, 64.9 percent of all persons engaged in construction industries were unemployed, 38.1 percent in service industries, 37.4 percent in mining, 36.2 percent in railroads, 27.4 percent in manufacturing, 19.5 percent in trade, 5.1 percent in public service, and 1.1 percent in agriculture.

Within the manufacturing group a wide dissimilarity is also found among the different branches. An analysis of the fluctuations in the employment indexes of the Bureau of Labor Statistics, taking the 1929 and the 1933 averages, indicates decreases in employment since 1929 ranging from 53.7 percent in the lumber industry, 53.2 percent in the machinery industries, and 52 percent in the stone, clay, and glass products industries to 15.4 percent in the leather industries, 16.1 percent in textiles, and 16.3 percent in food manufacturing. Higher unemployment rates are found for industries manufacturing durable goods and lower rates for those manufacturing nondurable goods. For these groups as a whole the decline in employment between 1929 and 1933 was 48.4 percent in the durable goods industries and only 19.4 percent in the nondurable goods industries. Over a period of years, however, the relative divergences between these industries would not be so great.

Individual establishments within a branch, of any industry also vary widely in their employment experience. Some companies in

58

any given year are expanding and consequently increasing their working force, whereas others are losing business or possibly extending mechanization and cutting down employment.

VARIATIONS BY STATES

Virtually no data on unemployment by States are available except those in the April 1930 census of unemployment and some occasional State or city unemployment censuses or surveys. Prior to 1930 the compilation of employment statistics was very limited. Several States published employment indicators of factory employees, representative of manufacturing industries only, but most of these do not go back beyond the 1920's.

During the last 4 or 5 years more complete and reliable employment statistics have been gathered. Not only have States recognized the need for such data, but the United States Bureau of Labor Statistics has been compiling national employment indexes month by month, since 1929, for the following industrial groups: Manufacturing, wholesale trade, retail trade, mining, transportation, telephone and telegraph, light and power, and hotels. In 1933 the scope of the field covered was enlarged by the addition of real estate, banking, insurance, and canning and preserving industries. Since 1932 these indexes have been broken down by States, resulting in the first comprehensive monthly State employment indexes.

Utilizing the Bureau of Labor Statistics indexes, the United States census of occupations, and the census of unemployment of April 1930, the yearly average employment and unemployment by States was estimated for each year, 1930 through 1933, the results of which are shown in table 5. The dependability of these estimates is subject to question, at least for part of the period, because of the inadequacy of the data upon which they were based and also because of the crude statistical methods by which they were necessarily treated in the limited time available. Several checks with other information indicate, however, that these estimates are at.least fair approximations. They, therefore, may be used as a guide in the absence of better data. More accurate indicators could probably be obtained from the statistical treatment of city surveys of unemployment where such are available or from an examination of actual employment records within a State.

It is at least immediately clear from the estimates that there has been a wide variation in the degree of unemployment in individual States during a cyclical decline. To illustrate this, the States were arrayed according to the rate of unemployment in April 1930, the average for 1933, and the average for 1930-33. (See table 6.) Comparing the extremes of the array, the rate in Michigan was more

(IMAGE OF TABLE 5)

59

than twice as high as in South Dakota in 1930, and more than three times as high as in Georgia in 1933.

Over a period of years, of course, the deviation from average unemployment tends to diminish. For example, Michigan, with the highest rate of unemployment in the 4-year period, was 39.2 percent above the country as a whole in April 1930, but its average unemployment for the 4 years (1930-33) was only 30.5 percent greater than that of the country. South Dakota's unemployment, at the other extreme in the array, rose from 38.5 percent below the average in April 1930 to within 30.2 percent of the average for the 4 years. If data for a complete cycle could be included, it is evident that the variation would be reduced still further. However, it is equally apparent that even over a long period, the individual States will experience fairly large differences in unemployment rate, because of the individuality of their industrial structures.

In general, the more highly industrialized States experienced the worst unemployment over the period. Unemployment was most severe in 1933 in Michigan, which also stands at the head of the array of the 1930-33 average with a rate of 35.9 percent. Michigan, Rhode Island, New Jersey, Illinois, Pennsylvania, New York, Massachusetts, and Ohio are among the States with the highest percentage of unemployment, ranging from 28.7 to 35.9. Unemployment in Georgia, South Dakota, and South Carolina, at the lowest end of the array, averaged about 19 percent from 1930 to 1933. Delaware is probably most highly industrialized of the States found at the lower end of the array. Mississippi, Kansas, North Dakota, Tennessee, and Virginia also had low percentages of unemployment. A part of the tendency toward less unemployment in agricultural States is probably explained by the fact that these States presented better opportunities for unemployed city workers to move to farms and out of the industrial labor market. In part, of course, the tendency in some States results from relatively smaller decreases in employment.

It should be remembered, then, that these estimates are most reliable as indicators of the relative intensity of unemployment in the States over the 4-year period. The estimates of the actual volume of unemployment are subject to the limitations outlined above and are presented here as crude approximations to the actual figures, which may serve as a basis for some of the "informed guess work" necessarily attendant on planning for unemployment compensation. The unemployment variable is a function of two other estimated variables, the average number of gainful workers and the average number of employed workers. The estimates of employment, which are essential to unemployment compensation legislation, are probably the most reliable of the three variables.

(IMAGE OF TABLE 6)

62

LONG-TIME TRENDS IN EMPLOYMENT

Distinct trends may be discerned in employment in different industries over a period of years, with employment increasing in some industries and decreasing in others. This results from a variety of causes such as shifts in consumer demand, new products, displacement of labor by technological improvements, and the effects of tariff policy. During the 1920's there was much talk of technological unemployment, since in those relatively prosperous years employment was decreasing in agriculture, manufacturing, railroading, and coal mining. Most of the labor dispensed with because of mechanization was reabsorbed by manufacturing through an enormous in-

crease in the volume of physical output. David Weintraub, when with the National Bureau of Economic Research, {5} estimated that of a total of 2,832,000 workers who were displaced by technological improvements from 1920 to 1929, a total of 2,416,000 were brought back into other manufacturing employment, so that there were only 416,000 fewer workers in manufacturing in 1929 than in 1920. At the same time, however, employment was increasing in transportation other than railroading, in communication, in mining other than coal, in quarrying and oil-well drilling, in construction, in mercantile trades, in banking and government, and in various unclassified industries including the service trades (table 7). There is,

{5} Weintraub, David, "The Displacement of Workers Through Increases in Efficiency and Their Absorption by Industry, 1920-1931", Journal of the American Statistical Association, vol. XXVII, no. 180, December 1932, pp. 383-.400.

63

therefore, no clear evidence that unemployment was increasing on the whole as a result of technological improvements, although there was undoubtedly a lag in reabsorption into the same industry or transfer to other industries on the part of those displaced, with much attendant unemployment.

Opposite trends in employment will also be found in different branches of manufacturing. During the 1920's, while there was a downward trend in employment in manufacturing as a whole and in most industries in that group, a few branches of manufacturing showed an increase in employment. Thus, in contrast to decreased

employment in carriage making, there was an increase in employment in automobile manufacture (table 8). Likewise, employment was declining on steam railroads, in express companies, and in water transportation, and increasing on street railways, in the Pullman Company, in bus and truck transportation, and in the telephone, telegraph, and electric light and power industries (table 9).

It should also be pointed out that geographical shifts take place within an industry so that employment decreases in one area while it is increasing in another. For example, combined employment in the 13 leading branches of the textile industry declined in New England from 322,946 in 1914 to 265,313 in 1929, and increased over the same period in the southern States of North and South Carolina, Georgia, Alabama, and Tennessee from 111,611 to 253,379.

64

PART-TIME EMPLOYMENT

The record of total unemployment does not represent the entire unemployment problem. Account must also be taken of unemployment resulting from working short time. Information on partial or under employment is even more fragmentary than for total unemployment. The few data which have been collected provide little definite information about part-time unemployment, other than a general substantiation of the commonplace observation that there is a considerable volume of part-time work in "normal" times, and that this volume increases materially in depression years.

Ideally, it would be desirable to know (1) the number of part-time workers; (2) the proportion of full time worked by part-time workers; (3) the proportion of total full-time man-hours lost through part-time employment; (4) the incidence of part-time employment by industries. Detailed information of this character is

available for only four cities in the country--Columbus, Buffalo, Syracuse, and Philadelphia. Virtually no information exists about part-time work in the country as a whole.

Until late in 1933, the United States Bureau of Labor Statistics published a monthly report on the average percent of full time reported by all operating manufacturing establishments and reported by establishments operating part time only. This series, however, is only a rough index of plant operating time and has no relation to actual part-time hours worked by employees. The reporting firms indicate to the Bureau simply the proportion of full-time hours during which the doors of the plant are open. Thus, if any one department of an establishment were running full time, that plant would report 100 percent full-time operation, regardless of short hours in any of its other departments. Finally, the figures as pub-

65

lished for the several manufacturing industries are simple averages of the percentages reported by the individual plants; that is, the averages are not weighted by the number of wage earners in each establishment or department.

The series was discontinued by the Bureau in 1933, chiefly because of the impossibility of determining actual full-time hours from one period to another. Although the National Recovery Administration codes of fair competition theoretically established standard hours of operation, at the same time they made numerous allowances for extending the hours during peak ,production periods and for contracting them below the code standard when orders were slack. The confusion in reporting the proportion of full time which

(IMAGE OF TABLE 10)

The United States Bureau of Labor Statistics from time to time has made sample studies of wages and hours of labor in some 24 manufacturing and 5 nonmanufacturing industries. In these studies data have been collected indicating actual and full-time man-hours in the sample plants covered. These data were used to compute the average percentage of full-time man-hours worked by all employees, which is reported by the Bureau in the bulletins of the "Wages and Hours of Labor" series. (See table 10.)

66

These averages have the advantage of representing actual, observed employee working hours. Their chief disadvantages derive from the fact that the surveys have been spot studies and no industry has been covered more frequently than at 2-year intervals. Because of this fact and because not more than two or three observations, at periods of varying comparability, are available for most of the industries, it is difficult to make legitimate chronological comparisons with the data.

One general tendency in the individual industries covered is apparent: The proportion of part-time employment, which was relatively small until 1929, showed a considerable increase during the subsequent depression years.

Volume II of the unemployment section of the Fifteenth United States Census includes a brief section estimating the number presumably working part time in April 1930. The estimates are based on the report of "days worked last week" by persons having a job but on lay-off without pay, excluding those sick or voluntarily idle. {6}

The length of the full-time week, of course, varies, and the census estimates were made on the basis of 5-, 5 1/2-, 6-, or 7-day full weeks. The combined estimates, therefore, indicate only the number of persons presumably working part time in April 1930, and they include no indication of the proportion of full-time or number of working days lost.

The most complete information on part-time employment is found in sample unemployment surveys made in three cities under the direction of Fred C. Croxton. A survey of Columbus, Ohio, covers the years from 1921 to 1925; a Buffalo, N. Y., study has been conducted each year since 1929; and a survey of Syracuse, N. Y., was made in 1931. The surveys all were made about the first of November, and, since they cover carefully selected sections of the respective cities, they probably present fairly accurate results for the particular areas. (See table 11.)

Tabulations were made of the employment status (full time, part time, or wholly unemployed) of all gainful workers enumerated. Part-time workers were tabulated also according to the proportion of full. time worked, the frequency class intervals ranging as follows: Less than one-third time; one-third to one-half time; one-half to two-thirds time; two-thirds but less than full time. Table 12 displays this distribution of part-time workers for the three cities.

{6} U. 8. Department of Commerce, Bureau of the Census, Fifteenth Census of the United States: 1930, "Unemployment", vol. II, General Report (U. S. Government Printing Office, Washington, D. C., 1932), pp. 355-360.

67

The industrial research department of the Wharton School of Finance and Commerce has compiled estimates of part-time unemployment in Philadelphia at four different periods since April 1930. (See table 11.) However, it has released no estimates of the proportion of full time worked on these dates.

From the several reports listed above the statistical and actuarial staff of the Committee on Economic Security constructed estimates of the proportion of full time worked by all workers and by part-time workers in all of these cities except Philadelphia. In general,

(IMAGE OF TABLE 11)

it appeared that part-time work tended to increase in close proportion to the increase in unemployment. In Columbus the proportion of part-time work and unemployed workers tended to move together from the depression of 1921 through the winter recession of 1924. In Buffalo the same similarity was apparent during the depression until November 1932, when unemployment increased at a faster rate than part-time employment. The average loss from total full-time man-hours apparently reached about 15 percent in 1932 compared with 3 or 4 percent from 1921 to 1925.

The proportion of employed workers on part time remained fairly constant from 1921 to 1925 in Columbus and increased in Buffalo from 7 percent in 1929 to 33 percent in 1932. Similar distributions occurred in the other cities.

(IMAGE OF TABLE 12)

While the number of part-time workers tended to increase with unemployment, the variation in the proportion of full time worked by part-time workers was surprisingly small. Averaging about 65 percent from. 1921 to 1925 in Columbus, it fell no lower than 57 percent in Buffalo in 1932 and 54 percent in Syracuse in 1931. The distribution of part-time workers according to proportion of full time worked indicated that roughly 85 percent of them worked half time or more in Columbus. The percentage working half time or more had fallen to 70 percent in Buffalo and Syracuse in 1931.

PERIODIC AND SEASONAL FLUCTUATIONS IN UNEMPLOYMENT

It is common knowledge that unemployment fluctuates widely from year to year. It is evident also that a large proportion of the total volume of unemployment over a considerable period of years appears in depressions. In the 16-year period 1900-1915 the unemployment of 5 of those years-1900, 1904, 1908, 1914, and 1915--constituted approximately 45 percent of the total, and about 70 percent of the unemployment in the years 1922-33, inclusive, was concentrated in the years 1930-33. There is also considerable fluctuation in employment from month to month during any year. During the year 1928, for example, employment in manufacturing for the United States as a whole was 7.5 percent higher in September than in January, and the total number of persons employed in Ohio was 13.8 percent greater in the best month than in the poorest month.{7}

The estimates by Simon Kuznets in Seasonal Variations in Industry and Trade indicate that during the period 1923 to 1931 the range in monthly fluctuations in pay rolls from a yearly average index of 100 was 55 in women's clothing, 35 in automobiles, 25 in cement, 19 in steamfitting, 19 in furniture, and 17 in cigars and cigarettes (table 13).

Fluctuations in seasonal employment are especially marked in the construction industries. The 1930 census of the construction industry showed that the number employed in January 1929 was only 56.2 percent of the maximum number employed in August of that year. Large seasonal fluctuations in employment also occur in mining, railroading, and retailing. All industries have some form of seasonal variation; the difference is one of degree. Some measure must be established to select for special study those industries evidencing seasonal variation in employment of sufficient amplitude to be especially expensive in the operation of a plan of

{7} "Fluctuations in Employment in Ohio, 1914 to 1929", U. S. Bureau of Labor Statistics. Bulletin No. 553.

70

unemployment compensation. Of chief concern here is the industry, which has a definite period of the year when employment is materially curtailed, such as is true for the fertilizer or the canning and preserving industries. On the other hand, the employment record which tends to be level for most of the year but rises rapidly and for a very brief time, such as might be reflected by department store employment at Christmas, is of no significance or importance for present purposes.

(IMAGE OF TABLE 13)

DURATION OF UNEMPLOYMENT

The staff of the Committee on Economic Security studied all available data for the purpose of determining the most probable distribution of the unemployed in the United States by the duration of unemployment from 1922 to 1933, inclusive. Appendix II gives the sources of available information on the duration of unemployment and describes the procedures by which the various data were adjusted to comparable bases for computation purposes.

In an analysis of 92 available surveys, covering about 5,000,000 persons, 5 groups were segregated according to the percentage of gainful workers unemployed at the time the surveys were made. Surveys in which the percentage of employment ranged from 3.0

71

to 6.9 percent of the total gainful workers were included in the first group. For the second, third, fourth, and fifth groups the percentages of gainful workers unemployed fell within the ranges of 7.0-10.9, 11.0-19.9, 20.0-29.9, and 30.0-43.0 percent, respectively. The percentage distribution of the unemployed was tabulated by duration of unemployment for each of these five groups. (See table 18, p. 84, and appendix II, fig. II-3.)

These five duration distributions were then applied to national employment, on the basis of two assumptions: (1) That the duration curve varies with the degree of unemployment and (2) that a distribution curve at a certain percentage of unemployment is representative of the duration experience of the unemployed in the United States when a corresponding percentage of unemployment exists in that group. Since a fair correlation was found between the degree and the duration of unemployment, it was possible to construct estimates of duration for the United States as a whole for each of the. years 1923-33, even though no surveys were available for 1926 and 1927, and for all years except 1929, 1930, and 1931 only a limited area had been covered in the sample studies. The fairly close agreement between the average rate of unemployment calculated for the assumed total compensable labor force of the United States and for the cities in which there were surveys in the corresponding years indicates that employment conditions in these cities were representative of the conditions in the country as a whole in the same period.

The estimated time lost by the unemployed in each of the years 1923-33 was calculated on the assumption that the estimated percentage of the compensable labor force unemployed each year and the percentage distribution of duration of unemployment assigned to each year represented the average unemployment and duration situations throughout the year. This assumption disregards seasonal fluctuations except as they are reflected in the situation at the time of the surveys utilized as basic data. It is also recognized that the individuals who comprise the average number of unemployed shift constantly throughout the year, but for purposes of computing total benefit costs it was not considered essential, even if it were possible, to trace the history of separate unemployed individuals who experience unemployment of specific duration.

If a census of the unemployed were taken every week in a year, a set of consecutive duration distributions would result which would give a picture of the shifting in the size of the duration intervals. This also would afford more accurate means of determining the total weeks lost by the unemployed in each different duration classification for the entire year.