Men stand in a bread line in a black and white image from the Great Depression.

About the Author

Gabriel Mathy is an associate professor in the Department of Economics at American University, Washington, District of Columbia.

Article Citations

Crossref 0

top Back to Top

09/15/2026

Private-sector estimates of unemployment in the 1930s and the development of official labor force statistics

Various private-sector organizations developed their own estimates of unemployment and the labor force during the Great Depression because official government statistics for those measures were not yet available in most years. The modern concepts of unemployment and the labor force arose from discussions between the private sector, academia, and the government. The manner of construction for these unemployment and labor force figures is discussed, as well as why organizations came up with different figures.

It may be a surprise that modern unemployment statistics in the United States only begin in 1948. After all, gross domestic product (GDP) and the national income accounts produced by the U.S. Bureau of Economic Analysis (BEA) begin much earlier—in 1929. How could the U.S. Bureau of Labor Statistics (BLS) have only collected labor force statistics almost two decades after the start of the Great Depression? This article discusses the history of the unemployment and labor force concepts, highlighting the importance of private-sector organizations developing and defining their own estimates of unemployment and the labor force. Government statistics in other areas were essential to developing these estimates, and this article discusses the interactions between the private sector (including universities) and government statisticians.

The development of the unemployment and labor force concepts began before the 20th century. In the decennial censuses from 1880 to 1910, census questions asked about those who were out of work.^¹ Although the 1920 census did not include any questions on unemployment, in the wake of the 1920–21 recession then-Secretary of Commerce Herbert Hoover established a special commission on unemployment, which set the stage for the 1930 unemployment census during Hoover’s presidency. The results of this survey were released in June 1930 and were immediately controversial for showing an unemployment figure that seemed too low. As a result, Congress immediately authorized another census of unemployment for the next year. The 1931 census was less controversial, and there were no unemployment censuses between 1931 and 1937.^²

Given the lack of official unemployment data throughout most of the 1930s, researchers and organizations at the time, like government-turned-private economist Robert Nathan,^³ the National Industrial Conference Board (NICB, known today as the Conference Board),^⁴ the Alexander Hamilton Institute (AHI), and the American Federation of Labor (AFL),^⁵ among others, all created their own estimates of unemployment. These were not based on modern methods to calculate unemployment but were instead based on a statistical definition: Unemployment is the whole labor force minus employment. However, “the labor force” did not have a universal definition at that time.

1929 to 1940, timeline of estimates and censuses

As the Depression dragged on into the 1930s, estimates of unemployment were increasingly made in both the private and public sectors. Given the need to piece these data together, estimates varied widely across different organizations.

BLS had been collecting broad nonagricultural employment figures since at least 1929.^⁶ Preliminary figures were released in real time, and could be found in various government publications, like Employment and Earnings and the Survey of Current Business.^⁷ Discussion among economists and statisticians about these figures allowed BLS to compile a revised and updated estimate of nonagricultural employment for each month from 1929 through 1940.^⁸ The series was for employees in nonagricultural establishments by sector and so did not cover all employment. The series excluded agricultural workers, the self-employed, self-proprietors, domestic workers and servants, and gig workers.^⁹ These estimates for 1929 were only available much later, but for the purposes of this article starting in 1929 is useful, because this was the first year of the Great Depression and corresponds to the start of U.S. GDP figures and the National Income and Product Accounts data.^¹⁰ Unlike the 1929 estimates, estimates for the 1930s were produced in the 1930s, and the decennial census was the basis for those estimates.

The 1930 census had several features that made it useful for calculating labor force statistics. One feature was a question about gainful workers (someone who was generally in the workforce, whether working or not). Although this would include certain categories of workers who would be excluded from the later labor force concept (like discouraged workers), it was a step closer to a labor force concept, rather than simply focusing on current employment information. Most importantly, the 1930 census included a survey on unemployment.^¹¹ This survey allowed researchers to compare their methods for estimating unemployment against a direct measure of unemployment. This direct measurement also meant that, theoretically, all the unemployment estimates would have a fixed point from which to carry forward changes in unemployment.

Once researchers dug into the census figures, however, it became clear that the concept of unemployment was underdeveloped and that, between poorly formulated questions and poorly trained enumerators, the figures developed in the 1930 census were not useful for giving a complete picture of American unemployment.^¹²

The 1931 census of unemployment provided a more accurate estimate and found a much larger number of unemployed compared with the 1930 census.^¹³ BLS further refined its methods for unemployment surveys in some local trials in 1934.^¹⁴ The reaction to and the changes between the 1930 and 1931 censuses and the local trials are an example of the discussion that occurred among researchers and analysts inside and outside of government that led to a more refined labor force concept. There was a follow-up census in 1937 to gauge how pervasive unemployment was once recovery was underway after 1933.^¹⁵

The 1940 census provided another comprehensive estimate of labor market conditions, with a formal labor force concept and even questions about unemployment duration as well as unemployment itself. Because the 1930 census had happened about a decade before, and the unemployment surveys did not provide explicit estimates of the labor force, the 1940 census provided a wealth of data, which led to significant revisions to private-sector unemployment estimates.^¹⁶

Constructing unemployment and the labor force

Though the 1930, 1931, and 1940 censuses were extensive, they only provided a snapshot in time. In total, they represented only 3 months in a 10-year span. Using whatever data they had available, Nathan, the NICB, the AHI, and the AFL, among others, attempted to interpolate and extrapolate estimates of unemployment for the noncensus months and years and quantify unemployment. (See table 1.)

Table 1. Comparison of 1930s unemployment estimate construction methods
Category	Robert Nathan	National Industrial Conference Board	American Federation of Labor	Alexander Hamilton Institute
Labor force	Assumed the ratio of working age population to overall labor force is unchanged from 1930 census. Adds 325,000 per year to labor force as a result, later raised to 425,000 as growth in working age population.	Adjusted for age, race, and sex to obtain growth in labor force from 1930.	Added 60,000 per month to labor force.	Assumed constant ratio of labor force to population, like Robert Nathan.
1930 census employment baseline	Used 1930 census of unemployment, with a broad definition of unemployed, then subtracted this measure of unemployment from labor force to obtain the employment baseline.	Similar to Robert Nathan, but with a less extensive view of unemployment.	Same as Robert Nathan.	Just reports 5,860,000.
Ad hoc adjustment to 1930 census estimate of unemployment	Added a million initially, then later in the 1930s revised it downward.	None.	600,000.	^¹
Nonagricultural employment source	United States Bureau of Labor Statistics nonagricultural employees from establishment survey/Consumer Expenditure Surveys.	Own employment adjustment.	Own employment adjustment.	^²
Agricultural employment	Adjusted for changes in farm population from net births and movements between city and farms and used the United States Department of Agriculture measure of hired labor on farms.	Allowed for inclusion new entrants from new farms and changes in farm population.	Added new farms but considered unpaid family workers constant since 1930.	Set the number of farmers constant and all other changes proportional to changes in the Department of Agriculture's index of demand for farm labor.
^¹ Not applicable. ^² The Alexander Hamilton Institute used its own method of getting variable employment in sectors it had data for. Employment of farmers, trade employers, and professionals was assumed to be unchanging throughout the decade. Then changes from the 1930 census were calculated assuming all these sectors were proportional to the variable employment sector. Source: U.S. Bureau of Labor Statistics.

Labor force

First, the organizations had to estimate the labor force. Nathan, the NICB, and the AFL used the estimates from the U.S. Census Bureau (Census) for the labor force in April 1930 as their starting point. The Census had estimates for the share of the population that were gainful workers (those generally in the workforce whether working or not), and most organizations assumed this same proportion applied to the labor force. Because the Depression had already begun, there may have been a higher or lower number of workers in the labor force—depending on whether the added worker effect or the discouraged worker effect dominated.^¹⁷

In subsequent years, most organizations assumed that labor force growth matched population growth. Nathan pinned his estimate to linear population growth, and his estimates of labor force growth increased by 100,000 each year from 1930 to 1940, eventually reaching 425,000 per year. The AHI also followed a similar method of using the working-age population to labor force ratio from the 1930 census and assumed that ratio was unchanging for the rest of the decade. The NICB and AFL both estimated larger increases in the labor force, ranging from roughly 530,000 to 620,000 per year from 1930 to 1940. The NICB figures adjust for changing propensities to participate in the labor force by age, sex, race, and other demographic features, while the AFL just considers a few age-related factors.^¹⁸ The AFL only considered two age bins for its working age population estimates: under 15 years old and over 15 years old, with the latter being the relevant group for the labor force. Given the lack of widespread retirements outside of the wealthy at the time (especially during the Depression), this binning was not unreasonable. The AFL also made some other adjustments in the 1920s for immigration restrictions, changes in school attendance patterns, reductions in child labor, and other social changes, though these did not change the series much.^¹⁹ However, once the 1940 census results were available, it became clear that the growth in the working-age population (ages 20 to 64) was higher than that of the overall population.

Employment

Once the researchers had established their estimates for the labor force, they then needed to estimate employment. The first step in this process began in 1930. The 1930 census included a survey of the number of unemployed and the number in the labor force. Although no question was asked regarding employment, it was estimated by taking the difference between the number in the labor force and the number of unemployed. However, the 1930 census was widely viewed as underestimating unemployment, because the follow-up 1931 census found a much higher level of unemployment. Still, most contemporaneous estimates are based on the 1930 census.^²⁰

Most organizations added some upward adjustment to the unemployment level from the 1930 unemployment census because of concerns over underenumeration.^²¹ Robert Nathan originally added 1 million to the unemployed, but he then adjusted his estimate downward to be in line with other estimators in the mid-1930s. The NICB, however, decided that the 1930 figures produced by Census did not require adjustment. There are some reasons to be skeptical of this approach, however, as the NICB estimated negative unemployment in fall 1929.^²² The AFL increased its unemployment estimate by 600,000 because the 1931 census had a much higher level of unemployment compared with the 1930 census. When there was a 1937 unemployment census, the AFL had the unemployment estimate that was closest to the Census figures (though that is not necessarily the ground truth, given issues with enumeration and lack of responses to the surveys at the time).

The employment estimates were carried forward to construct unemployment estimates instead of carrying forward the unemployment estimates themselves because employment was directly measured and unemployment was an indirect residual.

Unemployment

Finally, to construct unemployment estimates, a procedure to interpolate employment between census years was also required, and employment estimates had to be combined from separate agricultural and nonagricultural estimates.^²³

Nonagricultural data

The BLS Current Employment Statistics program began publishing employment estimates for some manufacturing industries in 1915.^²⁴ This collection was later extended to production workers in 54 manufacturing industries in 1927 and was extended to nonmanufacturing industries in the 1929.^²⁵ BLS continued extending its data collection (year by year) to the entire manufacturing and service-providing sector as the need to provide estimates of employment became more pressing.^²⁶

For nonagricultural employment, two approaches were taken. Nathan used the BLS monthly figures for nonagricultural employment.^²⁷ The NICB, AFL, and AHI constructed their own employment figures based on aggregation of various sectoral employment indexes or indexes of business activity based on both private-sector and government sources. In general, the 1930 census and other censuses of economic activity, like the Census of Manufactures, were used as a benchmark, and then changes in employment were carried forward (until the next census in 1940). This reliance on the 1930 census resulted in errors that grew until the 1940 census was released, which resulted in sharp revisions after these crucial data were available.^²⁸

Agricultural data

Agricultural estimates were collected by the Bureau of Agricultural Economics, U.S. Department of Agriculture (USDA), which was established in 1922.^²⁹ However, estimating agricultural employment also posed its own challenges. Many workers in agriculture were family workers in which enumeration of their employment status was difficult, as there was no formal employment contract. Even among agricultural workers with formal employment contracts, complete unemployment was rare because some agricultural work was available, even if for only a few days a week. Moreover, the farm population did not continue to fall during the Depression, as employment was more readily available in agriculture than in other sectors. Employment on farms was essentially unchanged between 1929 and 1933 (during the Great Contraction), while the overall unemployment rate reached 25 percent.^³⁰

The family farm worker issue with mismeasurement of unemployment is less severe than it might appear at first, however. The difference between counting or not counting these unpaid workers as both employed (and thus in the labor force) or not employed (and not in the labor force) does not affect the number of unemployed. Either classification would result in the same number of unemployed due to the subtraction method used. An issue would arise, though, if these workers were being counted in the labor force even if they were not employed or vice versa. Fluctuations in family agricultural labor in this period arose from a few factors related to the Depression, including reduced migration from rural to urban areas, increased subsistence farming due to low agricultural prices, and even migration from cities to farms as food was more available on farms even if incomes were low.

For agricultural workers, Nathan adjusted his estimate of agricultural employment for movements in population from urban to rural areas. He assumed that all increases in agricultural population were changes in agricultural employment and incorporated a hired-labor measure from the USDA. The NICB adjusted agricultural employment upwards because of additional farms and changes in the agricultural population, but it also estimated that there was no increase in employment in agriculture after 1930.^³¹ The AHI held the number of farmers constant and assumed that all other farm labor varied according to USDA’s index of demand for farm labor. For agricultural workers, the AFL adjusted for additional farms but froze the number of unpaid family workers at 1930 levels.

Overall, estimates of agricultural employment did not vary much between the different organizations, and agricultural employment was roughly 20 percent of the labor force at the time, so roughly 80 percent of the labor force was in the easier to measure nonagricultural sector.^³²

Emergency workers

The 1930s also saw the emergence of another debate about unemployment, this time over the status of those working on New Deal federal relief projects like the Works Progress Administration (WPA) and the Civilian Conservation Corps (CCC). Should these workers be counted as employed or unemployed? On the one hand, they were employed by the federal government and received a paycheck in exchange for their labor, just like other government workers who were considered employed. On the other hand, these workers were not regularly employed, but instead were part of emergency relief employment that was never meant to be permanent. Indeed, this was a form of relief for the able-bodied, mostly men. Those that could not work, like the elderly, orphans, and the disabled, got relief without a work requirement and were not considered employed. At its peak, the difference between the two methods would have accounted for over 3.5 million workers being classified as employed or unemployed. (See chart 1.)

Chart 1. Annual unemployment, with and without including emergency relief workers, 1929–43
Year	Unemployment (relief workers considered unemployed)	Unemployment (relief workers considered employed)
1929	1,550,000	1,550,000
1930	4,340,000	4,320,000
1931	8,020,000	7,721,000
1932	12,060,000	11,468,000
1933	12,830,000	10,635,000
1934	11,340,000	8,366,000
1935	10,610,000	7,523,000
1936	9,030,000	5,286,000
1937	7,700,000	4,937,000
1938	10,390,000	6,799,000
1939	9,480,000	6,225,000
1940	8,120,000	5,290,000
1941	5,560,000	3,351,000
1942	2,660,000	1,746,000
1943	1,070,000	985,000
Source: Michael Darby, "Three-and-a-half million U.S. employees have been mislaid: or, an explanation of unemployment, 1934–1941" (National Bureau of Economic Research, 1975), https://doi.org/10.3386/w0088.

In practice, most unemployment estimates counted those working on relief projects as unemployed. This includes government sources (like the U.S. Census Bureau), labor unions (like the CIO), researchers like (Robert Nathan), and many other private-sector organizations (like the NICB and AHI). However, others chose to classify relief workers as employed, as these emergency workers were employed by the same government that employed regular workers in the U.S. Weather Bureau, the U.S. Post Office, and so on. This disagreement was not settled at the time, though most classified relief workers as unemployed.

Comparing nongovernment estimates of unemployment

Overall, Nathan’s unemployment estimates were generally on the lower end of the distribution. The NICB tended to find the highest estimates of employment and lowest levels of unemployment.The AHI found the highest estimates of unemployment.^³³ And the AFL was generally in the middle, bounded on both sides by the other estimators. (See chart 2.)

Chart 2. Monthly unemployment, contemporaneous estimates, January 1929–October 1939
Period	Alexander Hamilton Institute	American Federation of Labor	National Industrial Conference Board	Robert Nathan
Jan 1929	4,204,000	3,060,000	^[1]	3,140,000
Feb 1929	4,105,000	3,119,000	^[1]	3,224,000
Mar 1929	3,801,000	2,560,000	^[1]	2,685,000
Apr 1929	3,622,000	2,043,000	^[1]	1,860,000
May 1929	3,418,000	1,754,000	^[1]	1,504,000
Jun 1929	3,286,000	1,447,000	^[1]	1,253,000
Jul 1929	3,197,000	1,214,000	^[1]	960,000
Aug 1929	2,851,000	1,064,000	^[1]	904,000
Sep 1929	2,768,000	614,000	^[1]	898,000
Oct 1929	2,851,000	910,000	^[1]	217,000
Nov 1929	3,549,000	1,949,000	^[1]	1,647,000
Dec 1929	3,923,000	2,629,000	^[1]	2,738,000
Jan 1930	5,262,000	3,919,000	3,144,000	4,035,000
Feb 1930	5,688,000	4,286,000	3,426,000	4,481,000
Mar 1930	5,813,000	4,323,000	3,338,000	4,450,000
Apr 1930	5,860,000	4,049,000	2,932,000	4,014,000
May 1930	6,072,000	3,756,000	2,433,000	3,886,000
Jun 1930	6,526,000	3,905,000	2,331,000	3,896,000
Jul 1930	7,248,000	4,441,000	3,253,000	4,152,000
Aug 1930	7,734,000	4,919,000	3,764,000	4,663,000
Sep 1930	7,659,000	4,983,000	3,671,000	4,829,000
Oct 1930	7,880,000	5,525,000	4,529,000	5,054,000
Nov 1930	8,529,000	6,293,000	6,080,000	5,900,000
Dec 1930	8,877,000	6,841,000	6,802,000	6,397,000
Jan 1931	10,082,000	8,169,000	7,923,000	7,648,000
Feb 1931	10,349,000	8,274,000	8,179,000	8,120,000
Mar 1931	10,340,000	8,133,000	7,838,000	7,872,000
Apr 1931	10,158,000	7,815,000	7,270,000	7,466,000
May 1931	10,245,000	7,811,000	6,890,000	7,393,000
Jun 1931	10,512,000	7,894,000	6,850,000	7,403,000
Jul 1931	10,984,000	8,367,000	7,483,000	7,606,000
Aug 1931	11,266,000	8,760,000	7,948,000	8,002,000
Sep 1931	11,292,000	8,846,000	7,838,000	8,135,000
Oct 1931	11,558,000	9,484,000	8,682,000	8,423,000
Nov 1931	12,220,000	10,410,000	9,931,000	9,275,000
Dec 1931	12,265,000	10,889,000	10,522,000	10,067,000
Jan 1932	13,425,000	11,926,000	11,397,000	10,905,000
Feb 1932	13,574,000	12,169,000	11,670,000	11,103,000
Mar 1932	13,792,000	12,387,000	11,859,000	11,229,000
Apr 1932	14,254,000	12,519,000	11,982,000	11,299,000
May 1932	14,741,000	13,004,000	12,069,000	11,483,000
Jun 1932	15,216,000	13,373,000	12,321,000	11,725,000
Jul 1932	15,757,000	13,793,000	12,906,000	11,990,000
Aug 1932	15,667,000	13,968,000	12,941,000	12,129,000
Sep 1932	15,128,000	13,458,000	12,469,000	11,804,000
Oct 1932	14,855,000	13,415,000	12,961,000	11,481,000
Nov 1932	15,152,000	13,925,000	13,404,000	11,988,000
Dec 1932	15,170,000	14,240,000	13,768,000	12,531,000
Jan 1933	16,103,000	15,166,000	14,262,000	13,390,000
Feb 1933	16,047,000	15,319,000	14,535,000	13,488,000
Mar 1933	16,467,000	15,653,000	14,762,000	13,808,000
Apr 1933	16,018,000	15,125,000	14,410,000	13,345,000
May 1933	15,667,000	14,615,000	13,891,000	12,849,000
Jun 1933	14,858,000	13,843,000	13,025,000	12,084,000
Jul 1933	14,120,000	13,458,000	12,377,000	11,571,000
Aug 1933	13,239,000	12,662,000	11,475,000	10,975,000
Sep 1933	12,388,000	11,854,000	10,647,000	10,394,000
Oct 1933	12,261,000	11,842,000	10,796,000	9,993,000
Nov 1933	12,735,000	12,374,000	11,248,000	10,494,000
Dec 1933	12,824,000	12,760,000	11,500,000	10,913,000
Jan 1934	13,528,000	13,382,000	11,544,000	11,409,000
Feb 1934	12,904,000	12,964,000	11,200,000	10,968,000
Mar 1934	12,255,000	12,420,000	10,213,000	10,427,000
Apr 1934	11,987,000	12,004,000	9,965,000	9,920,000
May 1934	11,858,000	11,711,000	9,484,000	9,528,000
Jun 1934	12,021,000	11,714,000	9,277,000	9,247,000
Jul 1934	12,418,000	12,222,000	9,898,000	9,437,000
Aug 1934	12,373,000	12,362,000	10,086,000	9,709,000
Sep 1934	12,673,000	12,429,000	10,232,000	9,909,000
Oct 1934	12,353,000	12,213,000	10,597,000	9,441,000
Nov 1934	12,556,000	12,581,000	11,211,000	9,878,000
Dec 1934	12,102,000	12,359,000	11,086,000	10,089,000
Jan 1935	12,659,000	11,695,000	11,234,000	10,485,000
Feb 1935	12,283,000	11,419,000	10,777,000	10,291,000
Mar 1935	12,018,000	11,233,000	10,395,000	9,973,000
Apr 1935	11,804,000	10,876,000	9,942,000	9,451,000
May 1935	11,902,000	10,819,000	9,572,000	9,208,000
Jun 1935	11,924,000	10,758,000	9,168,000	8,971,000
Jul 1935	11,983,000	10,759,000	9,018,000	8,960,000
Aug 1935	11,640,000	10,492,000	8,947,000	8,661,000
Sep 1935	11,127,000	10,031,000	8,640,000	8,437,000
Oct 1935	10,792,000	9,775,000	8,729,000	8,088,000
Nov 1935	10,909,000	9,961,000	8,737,000	8,315,000
Dec 1935	10,499,000	10,007,000	9,099,000	8,389,000
Jan 1936	11,464,000	10,952,000	9,434,000	9,688,000
Feb 1936	11,374,000	10,862,000	9,479,000	9,509,000
Mar 1936	11,019,000	10,482,000	8,883,000	9,047,000
Apr 1936	10,534,000	9,819,000	8,346,000	8,125,000
May 1936	10,388,000	9,555,000	7,705,000	7,792,000
Jun 1936	10,318,000	9,419,000	7,296,000	7,449,000
Jul 1936	10,179,000	9,381,000	7,034,000	7,292,000
Aug 1936	9,771,000	9,027,000	6,393,000	7,147,000
Sep 1936	9,151,000	8,494,000	6,294,000	6,720,000
Oct 1936	8,870,000	8,176,000	6,521,000	6,849,000
Nov 1936	8,846,000	8,287,000	6,676,000	6,828,000
Dec 1936	8,188,000	8,274,000	7,120,000	6,727,000
Jan 1937	9,250,000	9,241,000	7,574,000	8,076,000
Feb 1937	8,764,000	8,960,000	7,507,000	7,655,000
Mar 1937	8,838,000	8,604,000	6,865,000	7,210,000
Apr 1937	8,130,000	8,313,000	6,431,000	6,768,000
May 1937	7,959,000	7,909,000	5,585,000	6,430,000
Jun 1937	7,964,000	7,824,000	5,441,000	6,277,000
Jul 1937	7,974,000	7,782,000	5,155,000	6,133,000
Aug 1937	7,802,000	7,746,000	5,134,000	6,171,000
Sep 1937	7,618,000	7,513,000	5,066,000	6,145,000
Oct 1937	7,883,000	7,706,000	5,691,000	6,145,000
Nov 1937	8,977,000	8,479,000	7,175,000	7,177,000
Dec 1937	9,671,000	9,307,000	8,841,000	8,068,000
Jan 1938	11,830,000	10,926,000	10,328,000	10,265,000
Feb 1938	12,046,000	11,123,000	10,687,000	10,343,000
Mar 1938	12,122,000	11,226,000	10,721,000	10,354,000
Apr 1938	12,159,000	11,065,000	10,680,000	10,068,000
May 1938	12,719,000	11,404,000	10,754,000	10,191,000
Jun 1938	12,903,000	11,445,000	10,352,000	10,255,000
Jul 1938	12,932,000	11,274,000	10,347,000	9,859,000
Aug 1938	12,414,000	11,087,000	10,023,000	9,725,000
Sep 1938	11,704,000	10,465,000	9,314,000	9,331,000
Oct 1938	11,487,000	10,371,000	9,214,000	9,046,000
Nov 1938	11,317,000	10,515,000	9,429,000	9,324,000
Dec 1938	10,601,000	10,335,000	9,304,000	9,615,000
Jan 1939	11,754,000	11,590,000	10,012,000	11,029,000
Feb 1939	11,689,000	11,444,000	10,105,000	10,856,000
Mar 1939	11,420,000	11,145,000	9,800,000	10,615,000
Apr 1939	11,724,000	11,091,000	9,595,000	10,398,000
May 1939	11,740,000	10,852,000	9,382,000	9,860,000
Jun 1939	11,412,000	10,363,000	8,933,000	9,334,000
Jul 1939	11,493,000	10,305,000	9,384,000	9,649,000
Aug 1939	11,136,000	10,132,000	8,838,000	9,274,000
Sep 1939	10,388,000	9,451,000	8,196,000	8,805,000
Oct 1939	9,827,000	9,221,000	8,149,000	8,837,000
^[1] Not applicable. Source: U.S. Bureau of Labor Statistics.

No matter which estimator, however, by 1940 all unemployment estimates had drifted and were higher than the amount surveyed in the 1940 census. Not only were the unemployment estimates higher than the census figures, the private-sector estimates of the total labor supply differed from each other increasingly throughout the 1930s. From the 1930 census to the 1940 census, the variation between the largest and smallest estimates of total labor supply went from under 0.25 million to over 1.5 million.^³⁴

Transition to modern unemployment and labor force statistics

In preparation for the 1940 census, the Works Progress Administration (WPA) began a survey in December 1939, attempting to better understand the scope of the unemployment problem their agency was tasked with addressing.^³⁵ After a testing period of 3 months, the “Monthly Report of Unemployment” was published in March 1940. The 1940 census asked detailed questions about labor force status, including employment, unemployment, membership in the labor force, and emergency relief employment. After all the discussion in the 1930s about labor force, based on the old gainful worker concept, the Census Bureau now based labor force status on activity, asking workers about their behaviors and preferences regarding their desire to work and whether they were actively working in the sample period. The WPA continued its survey after the census was finished, publishing their estimates in the Monthly Labor Review periodically. (See chart 3.)

Chart 3. U.S. unemployment rate, U.S. Census Bureau estimate, March 1940–February 1947
Period	Unemployment rate (in percent)
Mar 1940	15.0
Apr 1940	14.6
May 1940	13.9
Jun 1940	13.9
Jul 1940	14.9
Aug 1940	14.2
Sep 1940	11.4
Oct 1940	12.1
Nov 1940	12.4
Dec 1940	12.0
Jan 1941	13.0
Feb 1941	12.2
Mar 1941	11.5
Apr 1941	10.9
May 1941	9.5
Jun 1941	9.8
Jul 1941	9.3
Aug 1941	8.8
Sep 1941	7.6
Oct 1941	6.4
Nov 1941	6.4
Dec 1941	6.2
Jan 1942	7.3
Feb 1942	6.9
Mar 1942	6.0
Apr 1942	5.1
May 1942	4.3
Jun 1942	4.5
Jul 1942	4.3
Aug 1942	3.5
Sep 1942	2.7
Oct 1942	2.7
Nov 1942	2.7
Dec 1942	2.6
Jan 1943	2.6
Feb 1943	2.5
Mar 1943	2.0
Apr 1943	1.8
May 1943	1.7
Jun 1943	2.2
Jul 1943	2.3
Aug 1943	1.9
Sep 1943	1.8
Oct 1943	1.7
Nov 1943	1.7
Dec 1943	1.7
Jan 1944	2.1
Feb 1944	1.7
Mar 1944	1.7
Apr 1944	1.5
May 1944	1.7
Jun 1944	1.8
Jul 1944	1.8
Aug 1944	1.6
Sep 1944	1.5
Oct 1944	1.2
Nov 1944	1.3
Dec 1944	1.3
Jan 1945	1.7
Feb 1945	1.9
Mar 1945	1.6
Apr 1945	1.5
May 1945	1.4
Jun 1945	2.0
Jul 1945	1.7
Aug 1945	1.5
Sep 1945	3.1
Oct 1945	2.9
Nov 1945	3.3
Dec 1945	3.7
Jan 1946	4.3
Feb 1946	4.9
Mar 1946	4.9
Apr 1946	4.1
May 1946	4.0
Jun 1946	4.4
Jul 1946	3.8
Aug 1946	3.5
Sep 1946	3.5
Oct 1946	3.3
Nov 1946	3.3
Dec 1946	3.6
Jan 1947	4.2
Feb 1947	4.3
Source: U.S. Census Bureau (1945–47)

In the run up to World War II, the emergency unemployment problem was waning, and it became clear that the WPA would be dissolved. The Social Security Administration, the Census Bureau, and BLS were all candidates to take over the administration of the survey, with the Census Bureau ultimately being chosen. The survey was expanded based on the Census Bureau’s statistical expertise, and the name was changed to “Monthly Report on the Labor Force” before becoming the Current Population Survey (CPS) in 1948, the name it still holds today.

As the official labor force statistics became formalized, Stanley Lebergott compared the private-sector estimates with the then-new official BLS series in 1948.^³⁶ While doing so, Lebergott acknowledged the value of those nongovernmental sources directly: "A wide variety of estimates of labor force and unemployment had previously been prepared by private agencies and individuals. These statistics were of considerable value in informing the public about the gravity of the unemployment situation in the 1930s.” BLS estimates for the 1930s were done with the benefit of hindsight and were estimated to conform with the Census Monthly Report of the Labor Force. (See table 2.)

Table 2. Selected estimates of unemployment in the United States, 1929–39
Year	U.S. Bureau of Labor Statistics	Alexander Hamilton Institute	American Federation of Labor	Congress of Industrial Organizations	National Industrial Conference Board	Robert Nathan
1929	1,550,000	3,456,000	1,864,000	1,831,000	429,000	1,752,000
1930	4,340,000	6,929,000	4,735,000	4,710,000	2,896,000	4,646,000
1931	8,020,000	10,939,000	8,568,000	8,322,000	7,037,000	8,118,000
1932	12,060,000	14,728,000	12,870,000	12,120,000	11,386,000	11,639,000
1933	12,830,000	14,394,000	13,271,000	12,643,000	11,842,000	11,942,000
1934	11,340,000	12,419,000	11,424,000	10,845,000	9,761,000	9,998,000
1935	10,610,000	11,629,000	10,652,000	10,050,000	9,092,000	9,102,000
1936	9,030,000	10,008,000	9,395,000	8,756,000	7,386,000	7,723,000
1937	7,700,000	8,366,000	8,282,000	8,109,000	6,403,000	6,856,000
1938	10,390,000	11,934,000	10,836,000	11,030,000	9,796,000	9,865,000
1939	9,480,000	10,696,000	9,979,000	10,813,000	8,786,000	9,835,000
Source: U.S. Bureau of Labor Statistics.

Like other estimators, BLS interpolated its estimates for the 1930s and extrapolated them back to 1929. And like other estimators, BLS benchmarked its estimates to Census data and used a wide variety of sources: BLS used figures for nonagricultural employees that were benchmarked to nonagricultural employees in the 1930 and 1940 censuses, used agricultural employment from the U.S. Bureau of Agricultural Economics, and calculated a ratio of self-employed Americans per employee to go from employees to total employment.^³⁷ In 1959, as part of a larger reorganization of responsibilities for collecting economic statistics, BLS took over the primary responsibility for the CPS.^³⁸

Conclusion

Many private-sector organizations developed estimates of unemployment and the labor force throughout the 1930s. This ultimately prompted the government to define the concepts of unemployment and the labor force, and to create modern labor force statistics that have been published since the 1940s. No one true estimate was created in the years before there were consistently produced official government statistics because economic actors made different choices in creating their estimates of unemployment and the labor force. A multitude of private-sector organizations, like labor unions, business groups, political organizations, and corporate economists, all participated in the discussions regarding these choices. These discussions, especially related to the treatment of emergency workers in the WPA and CCC, shaped the construction of the official labor force statistics we use today.

Many thanks to the staff of the Library of Congress and Tamiment Library for their assistance in locating sources, to Peter Meyer, the Society of Government Economists and to many members of the Bureau of Labor Statistics for helpful comments, and to Sarah al-Masri and Yves-Myriam Millien for their excellent research assistance. The author gratefully acknowledges funding by the Institute of New Economic Thinking and the American University College of Arts and Sciences Mellon Fund.

Appendix: Other estimators and estimates of the 1930s

The Congress of Industrial Organizations (CIO) relied on USDA figures for hired and unpaid workers and for the number of new farms. As with Robert Nathan, it relied on BLS for nonagricultural employment to supplement its agricultural employment estimates.^³⁹

Estimates of unemployment from the Labor Research Association (LRA) largely matched other estimates. LRA adjusted the labor force upward by 1 million, consistent with Robert Nathan’s original estimates and the National Industrial Conference Board’s estimates. For nonagricultural employment, LRA used BLS data and other measures of employment by sector. Its estimate for professional unemployment was somewhat higher than those of other estimators, some of which found negligible unemployment among professionals. For agricultural workers, the LRA included them with the possibility to be unemployed, which increased unemployment relative to other estimators. In line with other estimators, the LRA considered emergency workers as unemployed.^⁴⁰

Similar to LRA, the National Research League (NRL) adjusted the 1930 unemployment census figure upward by 1 million. It also found that the labor force was 1.5 million larger than standard Census Bureau estimates. Although the NICB found that there was no increase in unemployment in agriculture, forestry, fisheries, government, and professionals after 1930, the NRL found an increase in unemployment of 1.15 million from 1930 to 1935. Relative to the NICB, the NRL estimated unemployment in manufacturing and construction to be 1 million lower and the economy-wide level to be roughly 4.5 million lower. Otherwise, the NRL method largely followed the procedures of other estimates, using the 1930 and 1931 unemployment censuses and sectoral employment indexes to create estimates of employment.^⁴¹

The Cleveland Trust Company (CTC), its estimates constructed by economic statistician Leonard Porter Ayres, found a relatively high level of unemployment.^⁴² The U.S. Chamber of Commerce (a nongovernmental business association) produced a very low estimate of unemployment but did not publicly disclose its methodology.^⁴³ In September 1936, The New York Sun found that unemployment was effectively no longer an issue and were widely criticized. The newspaper looked at four sectors, which made up about 57 percent of employed workers in 1930. It sent out employment surveys in 1929 and in 1935 and found that employment declined by 2,350,000 and the labor force increased by 1,234,000 between 1929 and 1935. Thus, it estimated a total unemployment increase of just 3.6 million. This estimate did not account for establishments that went out of business and were not available to be surveyed between 1929 and 1935, which likely led to a large undercount of the jobs lost.^⁴⁴

Suggested citation:

Gabriel Mathy, "Private-sector estimates of unemployment in the 1930s and the development of official labor force statistics," Monthly Labor Review, U.S. Bureau of Labor Statistics, September 2026, https://doi.org/10.21916/mlr.2026.27

Notes

¹ Some early studies of unemployment include Hornell Norris Hart, Fluctuations in Unemployment in Cities of the United States, 1902 to 1917, Studies from the Helen S. Trounstine Foundation, vol. 1, no. 2, (Ohio, 1918), https://openlibrary.org/books/OL24176249M/Fluctuations_in_unemployment_in_cities_of_the_United_States_1902_to_1917; and William A. Berridge, Cycles of Unemployment in the United States, 1903–1922, Publications of the Pollak Foundation for Economic Research; No. 4, (Boston and New York: Houghton Mifflin Co., 1923), https://archive.org/details/cyclesofunemploy00berr.

² See David Card, “Origins of the unemployment rate: the lasting legacy of measurement without theory,” American Economic Review 101, no. 3, May 1, 2011, p. 552–57, https://doi.org/10.1257/aer.101.3.552.

³ Robert Nathan emerged as a central figure in the world of unemployment estimates. In 1929, Nathan was brought on to perform an unemployment survey for the Industrial Research Department at the Wharton School as a college junior. This was a survey using enumeration techniques, which were largely pioneered by researchers in Philadelphia based on previous experience with unemployment statistics. Enumeration would later form the basis for modern U.S. government unemployment statistics.

Nathan would lead the Philadelphia survey in 1930 and 1931, which found that the 1930 census of unemployment was flawed. In 1934, Nathan was hired to work on unemployment statistics with the President’s Committee on Economic Security to bolster the case for social programs like Social Security. His unemployment estimates, both as a government economist and as a private-sector researcher after World War II, were not based on enumeration but were based on constructing an estimate of unemployment as the difference between estimated labor force and employment. See J. Frederic Dewhurst and Ernest A. Tupper, “Social and economic character of unemployment in Philadelphia April, 1929,” June 1930, https://fraser.stlouisfed.org/title/social-economic-character-unemployment-philadelphia-april-1929-4078; and J. Frederic Dewhurst and Robert R. Nathan, “Social and economic character of unemployment in Philadelphia, April, 1930,” March 1, 1932, https://fraser.stlouisfed.org/title/social-economic-character-unemployment-philadelphia-april-1930-4094.

⁴ The Conference Board’s estimates are outlined in National Industrial Conference Board, “Employment and unemployment in the labor force 1900–1940,” The Conference Board Economic Record 2, no. 8, March 20, 1940, p. 89–92, https://catalog.hathitrust.org/Record/009663264.

⁵ The American Federation of Labor (AFL) was a craft union based in the skilled trades. The Congress of Industrial Organizations (CIO) was a labor union that espoused industrial unionism, in which entire factories or industries would be unionized, including both high-skill and low-skill workers together. The AFL and CIO would merge into the AFL-CIO in 1955, but were known to be rivals in the 1930s and both produced estimates of unemployment. A detailed description of the AFL’s unemployment estimates can be found in American Federation of Labor, “The Federation’s revised unemployment estimate,” American Federationist 43, January 1936,https://hdl.handle.net/2027/coo.31924054366954.

⁶ Aryness Joy, “Recent Progress in Employment Statistics,” Journal of the American Statistical Association 29, no. 188 (1934): 355–71, https://doi.org/10.2307/2278606.

⁷ See Employment and Payrolls, U.S. Bureau of Labor Statistics, https://fraser.stlouisfed.org/title/employment-payrolls-152?browse=1930s; and see Survey of Current Business, U.S. Department of Commerce, https://fraser.stlouisfed.org/title/survey-current-business-46?browse=1920s.

⁸ See table 11 in “March 1941,” Survey of Current Business 21, no. 3, U.S. Department of Commerce, March 1941, https://fraser.stlouisfed.org/files/docs/publications/SCB/1940-41/SCB_031941.pdf?utm_source=direct_download.

⁹ BLS also published a series for military employment, but those workers were excluded from nonagricultural employment, a classification that continued when the civilian labor force was developed.

¹⁰ To use this BLS series in unemployment estimates, researchers first had to scale nonagricultural employees to an overall employment figure and scale up the figure to incorporate nonemployee employment and agricultural employment. They then had to bring labor force estimates from the 1930 census backwards to 1929 and construct unemployment with a difference method or by subtracting employment from the labor force.

¹¹ For a list of questions from the 1930s census, see “FAQs about the 1930 Census,” National Archives, August 15, 2016, https://www.archives.gov/research/census/1930/general-faqs.html. The unemployment schedules have been lost.

¹² For a contemporaneous critique of the census, see Mary van Kleeck, “The federal unemployment census of 1930,” Journal of the American Statistical Association 26, no. 173, 1931, https://doi.org/10.2307/2277617. For a later, fuller analysis, see Card, “Origins of the unemployment rate.”

George Arner argued that it was not a difference in measurement accuracy so much as a deepening of the recession that led to the change between the 1930 and 1931 censuses. See George B. L. Arner, “The census of unemployment,” Journal of the American Statistical Association 28, no. 181A (1933), p. 48–53, https://doi.org/10.1080/01621459.1933.10502649. W.S. Woytinsky claimed that the discrepancy between the 1930 and 1931 census estimates of unemployment was due to the added workers that were induced into the labor force as the Depression worsened. See Woytisnksy, Additional workers and the volume of unemployment in the depression, Social Science Research Council (U.S.); No. 1, (Washington: Committee on Social Security, 1940. This supposition was not universally supported, even at the time. See Russell A. Nixon and Paul A. Samuelson, “Estimates of unemployment in the United States,” The Review of Economics and Statistics 22, no. 3, August 1940, p. 101, https://doi.org/10.2307/1926605; and Don D. Humphrey, “Alleged ‘additional workers’ in the measurement of unemployment,” Journal of Political Economy 48, no. 3 (1940), p. 412–19, https://doi.org/10.1086/255563. Clarence D. Long argued that the depression would have had an effect opposite of what Woytinsky claimed. See Clarence D. Long, The labor force under changing income and employment, NBER Books (National Bureau of Economic Research, 1958), https://EconPapers.repec.org/RePEc:nbr:nberbk:long58-1.

¹³ See Nixon and Samuelson, “Estimates of unemployment in the United States.”

¹⁴ See George Arner, “The census of unemployment”; and see Florence M. Clark, “Employment conditions and unemployment relief: unemployment survey of Bridgeport, Conn., 1934,” Monthly Labor Review 40, no. 3, March 1935, p. 625–45.

¹⁵ See Nixon and Samuelson “Estimates of unemployment in the United States” for an extensive discussion around the censuses of the 1930s and the views of nongovernmental economists on these issues.

¹⁶ Security, Work, and Relief Policies, 1942: Report of the Committee on Long-Range Work and Relief Policies to the National Resources Planning Board, prepared under the direction of Eveline M. Burns (Washington, DC: National Resources Planning Board, 1942).

¹⁷ See W.S. Woytinsky, Additional Workers. For context from a more recent work, see T. Aldrich Finegan and Robert Margo, “Added and discouraged workers in the late 1930s: a re-examination,” (Cambridge, MA: National Bureau of Economic Research, April 1993), https://doi.org/10.3386/h0045.

¹⁸ See Nixon and Samuelson, “Estimates of unemployment in the United States.”

¹⁹ Most estimators calculated population growth as linear. This will cause some degree of error, because population growth is exponential (i.e., the logarithm of population growth is linear). See Nixon and Samuelson, “Estimates of unemployment in the United States.”

²⁰ The exception to using the 1930 census as the basis for calculation was the Alexander Hamilton Institute, which instead measured unemployment as the difference between the estimated labor force and estimated employment, even in a year when unemployment was directly measured.

²¹ The Alexander Hamilton Institute assumed that unemployment in 1930 was 5,860,000, which was 2–3 million lower than what the census of 1930 implied for employment.

²² If one follows the modern definition of the labor force, any civilian employed person should be a part of the civilian labor force, and unemployment can never be less than zero. Given the presence of some frictional unemployment in practice, this is suggestive evidence that the upward adjustments to unemployment by other organizations were reasonable and that underenumeration was an issue. See Nixon and Samuelson, “Estimates of unemployment in the United States.”

²³ Though the values are interpolated between the values anchored by the 1930 census and the 1940 census, at the time the estimates were extrapolated from the 1930 census and other data.

²⁴ See Lyda Ghanbari and Mike McCall, “Current Employment Statistics survey: 100 years of employment, hours, and earnings,” Monthly Labor Review, August 2016, https://doi.org/10.21916/mlr.2016.38.

²⁵ See “Current Employment Statistics—National” in Handbook of Methods (U.S. Bureau of Labor Statistics), https://www.bls.gov/opub/hom/ces/history.htm; Ghanbari and McCall, "100 years of employment, hours, and earnings"; and Joy, “Recent Progress in Employment Statistics.

²⁶ See Ghanbari and McCall, “100 years of employment, hours, and earnings.”

²⁷ Security, Work, and Relief Policies, 1942: Report of the Committee on Long-Range Work and Relief Policies to the National Resources Planning Board, prepared under the direction of Eveline M. Burns (Washington, DC: National Resources Planning Board, 1942), p. 554.

²⁸ See Nixon and Samuelson, “Estimates of unemployment in the United States.”

²⁹ The BAE (Bureau of Agricultural Economics) would continue as the USDA Economic Research Service (ERS) in 1961. Employment estimates stretch back to 1909, based on the Censuses of Agriculture. Monthly changes are then derived from the monthly reports of USDA Crop Reporting Board correspondents starting in 1925. See Eldon E. Shaw and John Abel Hopkins, Trends in employment in agriculture, 1909–36, Works Progress Administration, National Research Project, 1938, p. 86–92 and the Farm Labor publication of the USDA, various issues.

³⁰ See John W. Kendrick, Productivity Trends in the United States, General Series (National Bureau of Economic Research); No. 71, (Princeton N.J.: Princeton University Press, 1961), p. 363, https://archive.org/details/productivitytren0000kend. For the 25 percent unemployment rate figure, see the BLS page “Where can I find the unemployment rate for previous years?,” particularly the subsection “How do I get unemployment data for the years of the Great Depression?,” https://www.bls.gov/cps/prev_yrs.htm.

³¹ As well as no employment increases in forestry, fisheries, government, and professionals.

³² See Nixon and Samuelson, “Estimates of unemployment in the United States.”

³³ This difference was largely driven by the AHI’s assumptions about the 1930 census.

³⁴ Security, Work, and Relief Policies, 1942: Report of the Committee on Long-Range Work and Relief Policies to the National Resources Planning Board, prepared under the direction of Eveline M. Burns (Washington, DC: National Resources Planning Board, 1942), p. 553

³⁵ “WPA unemployment estimates,” Monthly Labor Review 53, no. 4, October 1941, p. 893–899, https://fraser.stlouisfed.org/title/monthly-labor-review-6130/october-1941-608492?page=79.

³⁶ Stanley Lebergott, "Labor force, employment, and unemployment 1929–39: Estimating methods," Monthly Labor Review 67, no. 1 (1948), p. 50-53, https://fraser.stlouisfed.org/title/monthly-labor-review-6130/july-1948-610272?page=56.

³⁷ Lebergott, "Labor force, employment, and unemployment 1929–39, p. 52–3.

³⁸ John E. Bregger, “The Current Population Survey: a historical perspective and BLS’ role,” Monthly Labor Review 107, no. 6, June 1984, p. 8–15, https://www.bls.gov/opub/mlr/1984/06/art2full.pdf.

³⁹ Security, Work, and Relief Policies, 1942: Report of the Committee on Long-Range Work and Relief Policies to the National Resources Planning Board, prepared under the direction of Eveline M. Burns (Washington, DC: National Resources Planning Board, 1942) p. 554.

⁴⁰ Labor Research Association, “16,000,000 looking for work,” Labor Notes 6, no. 1, January 1938, https://hdl.handle.net/2027/umn.31951d00314612d.

⁴¹ See Daniel Carson, Unemployment in the United States, (New York: National Research League, 1935), p. 4–7, 15–16, https://catalog.hathitrust.org/Record/102925401.

⁴² See W. Randolph Burgess, “Leonard P. Ayres: an appreciation,” Journal of the American Statistical Association 42, no. 237, March 1947, p. 128–33, https://www.jstor.org/stable/2280193.

⁴³ See Carson, Unemployment in the United States.

⁴⁴ Labor Research Association, “Unemployment ‘estimates’ of New York Sun,” Labor Notes 4, no. 10, October 1936, https://hdl.handle.net/2027/umn.31951d00314612d.