The empirical model in this paper will take the following form:
Equation (2) will be estimated as a probit. The index i refers to the individual, s to his state of residence, and t to the year. LFP takes a value of one if the individual participates in the labor force and zero if he does not. Labor force participation may be influenced by individual characteristics X; time-varying state characteristics Y; state-varying time-constant characteristics; time-varying national characteristics; and the generosity of OAA. The use of state specific intercepts means that the effect of OAA will be identified from the variation in OAA benefits within states across years – the hypothesis being that states with a bigger increase in OAA benefits should have a bigger drop in labor force participation.

Equation (2) will be estimated on individuals from the 1% Census samples of the 1940 and 1950 Integrated Public Use Microdata Samples (IPUMS). Elements of X include dummies for age, education, marital status, household composition, farm residence, urban residence, nativity, and race. Since income and wealth data are lacking for much of the sample, elements of Y include measures of states’ economic conditions: per capita income, manufacturing production, farm values, and the unemployment rate.

The sample includes all male citizens between the ages of 66 and 75 who are not in group or institutional quarters. Sample statistics are reported in Table 4 for 25,426 observations in 1940 and 29,422 observations in 1950. Compared to today, elderly men were somewhat less racially mixed, more foreign born, less educated and lived more rurally. Relative to 1940, the 1950 sample was older and more educated and lived less on farms and more in SMAs.
The dependent variable will be whether someone was in the labor force last week. Labor force participation fell one percentage point, while the employment rate actually rose one percentage point. The proportion reporting they were out of the labor force for reasons besides being unable to work rose ten percentage points, a much larger change than in participation. This trend suggests that the modern concept of retirement – retirement for leisure, or retirement as a Most working respondents worked full-time or more, making it reasonable to focus on retirement instead of the choice of hours.


&le Means (Standard Deviations), Census Data

Number of Observations 1940




Age between:
66-70 0.611 0.597
71-75 0.389 0.403
Married 0.707 0.734
Head of household 0.837 0.850
Lives on a farm 0.283 0.190
Lives in an SMA 0.458 0.508
Foreign born 0.186 0.198
Nonwhite 0.079 0.075
Years of schooling completed 7.01 (3.70) 7.29 (3.88)
0-8 years 0.803 0.745
9-11 years 0.065 0.089
12 years 0.062 0.083
13-15 years 0.033 0.037
16 or more years 0.037 0.046
Years of schooling unknown* 0 0.663
Hours worked last week 19.5 (24.8) 18.9 (23.6)
Hours worked | hours>0 46.2 (14.9) 42.6 (15.9)
Labor force activity:
In the labor force 0.497 0.487
With a job 0.460 0.470
Unemployed 0.037 0.018
Not in the labor force 0.503 0.513
Because unable to work 0.352 0.258
For another reason 0.152 0.254

Table 5 reports labor force participation rates of different groups within the population. Participation declined with age but was higher at all ages than today. Participation rates were higher for those who were married, household heads, farm dwellers, non-SMA dwellers, natives, and nonwhites. Participation rates increased with education, as it does today. Most of the differences in participation across groups narrowed during the decade, especially for nonwhites.


1940 1950
Labor Force Participation 49.7 48.7
Age 66-70 57.6 56.6
Ase 71-75 37.3 37.1
Married 55.1 52.9
Head of household 54.0 51.8
Lives on a farm 68.2 68.4
Lives in an SMA 46.3 47.4
Foreign born 44.7 46.6
Nonwhite 53.0 48.3
Years of schooling completed
0-8 years 47.6 45.2
9-11 years 55.4 50.0
12 years 56.9 52.0
13-15 years 56.4 56.3
16 or more years 66.4 59.2

Program data appear in the Social Security Bulletin. The average OAA benefit divides total dollar obligations, excluding administrative and overhead costs, by the number of recipients. The CPI is used to deflate benefit levels. Additional state level data employed in the estimation appear in the Statistical Abstract and are reported in Appendix Table 4.&