Recent Unemployment Rate Estimates for the 1920s and 1930s

Abstract

In the 1920s the U.S. government did not attempt to take comprehensive surveys of either the number unemployed or the size of the labor force. In the 1930s the government did make estimates of the number unemployed, but made no estimates of the size of the labor force so as to calculate unemployment rates. Some estimates of the rate of unemployment for the 1920s were constructed-see the unemployment rate estimates for Givens, Douglas, Carson, and Weintraub in Table 1-but these varied in their coverage and accuracy. In the 1950s Stanley Lebergott developed a consistent series of estimates of the labor force and its components, the number unemployed, and the rate of unemployment for the nineteenth and twentieth century. This culminated in his seminal book, Manpower in Economic Growth, and Lebergott's series have become the most widely accepted and used series for the interwar years. I In the 1970s two major revisions of Lebergott's unemployment rate estimates appeared, Robert Coen's 1973 revisions and Michael Darby's 1976 revisions.2 These estimates are presented in columns 6 and 7 in Table 1. It has been suggested that Coen's revised estimates diminish the . . sheen of the prosperity of the twenties . . . Darby's revised estimates are for the 1930s and present a significantly different picture of the level of and changes in unemployment during the recovery from the Great Depression. Because of the strikingly different and contradictory pictures of the interwar period that these revisions present, this note examines each of the revisions. Coen's purpose was to provide estimates for the interwar period that allowed for cyclical variation in the labor force. The size of the labor force in the post-World War II United States tends to vary with business conditions because of the discouraged-worker phenomenon. The census provided the size and characteristics of the labor force at census dates for the interwar years. From this Lebergott was able to construct participation rates by age and sex, interpolate these participation rates between census dates, and with the population data construct estimates of the labor force for each year in the 1920s and 1930s. This allowed no cyclical variation in the labor force since it did not take into account annual variations in labor market conditions.4