This paper focuses on explaining the variation across states in labor force participation trends between 1940 and 1950. An alternative strategy would study the labor supply response to OAA’s introduction in the 1930s. However, there are two reasons for emphasizing the 1940s instead. First, only state level aggregates are available from the 1930 Census. In contrast, individual records from the 1940 and 1950 Censuses make it easier to control for characteristics like age structure, education, and household composition that are likely to be correlated with labor supply and perhaps with factors affecting benefit levels. It also makes it possible to analyze the impact of such characteristics on participation and on the sensitivity of participation to OAA.

Another reason to focus on 1940 and later is because labor force participation before 1940 was measured differently. The Census used to ask whether respondents were “gainfully employed”. New questions beginning in 1940 corresponded to the modern economic concept of participation. Various approaches that adjust downward the pre-1940 gainful employment series to match the later participation data are controversial and leave the series with a sharp decline between 1930 and 1940. Such a decline is not surprising in light of the Great Depression, but it is difficult to know how much of the measured drop was driven by economic forces, as opposed to the statistical break in the data series.

On the other hand, during the 1940s there was little change in the national labor force participation rate, which appears to make it a less promising time period to study. World War II drew or kept many older people in the labor force, although it is less clear what happened to retirement patterns after the war. For the sample of 66 to 75 year old men used in this paper, as shown in Table 3, the participation rate was 49.7% in 1940 and 48.7% in 1950 – only a 1.0 percentage point decline.


1940 49.7% 6.0%
1950 48.7% 6.2%
Change from 1940 -1.0 5.6

However, the small change in the national average masks considerable variation across states in each year and especially within states between 1940 and 1950. The standard deviation across states of the changes in participation is 5.6 percentage points. That reflects increases in participation rates of several percentage points in some North Central states, along with large declines in some New England and Southern states. Changes in participation varied a great deal within regions as well. The goal here is to understand how much of the sharp divergence in participation trends across states can be attributed to differences in OAA benefit growth. However, these patterns do raise concerns that benefit determination was endogenous with participation. I will discuss evidence against this hypothesis after presenting the main results.