THE EFFECT OF OLD AGE ASSISTANCE ON RETIREMENT: EMPIRICAL RESULTS

In this section I present estimates of the effect of OAA on labor force participation and offer evidence against the potential endogeneity of state benefit levels. I also discuss the implications of the estimates for early expansions in Social Security, and I conclude by analyzing how OAA affected people with different personal characteristics.

Estimates of OAA’s Impact

Table 6 reports a series of probit estimates of equation (2). In column (1) with no other covariates, the probit coefficient is significant and negative, demonstrating the strong negative correlation of OAA benefits and labor force participation in the raw data. Its magnitude is easier to interpret by considering the resulting prediction of aggregate participation in 1950 if benefits had stayed at 1940 levels, reported at the bottom of the table.32 The estimate in column (1) implies that labor force participation would have been 50.3% in 1950, instead of falling to 48.7%, if OAA benefits had not been raised 19.7% between 1940 and 1950 – so participation would have risen instead of falling.

Column (2) adds state and year fixed effects, which control flexibly for factors which vary across states and which influence labor force participation. The fixed effects might proxy for unmeasured individual characteristics which vary across states and for local economic conditions. The impact of OAA is now identified from changes in benefit levels within states between 1940 and 1950. The estimated fixed effects are jointly highly significant and raise the estimate of OAA’s impact, although not significantly. They raise the standard error of the estimate, which is not surprising since analysis of variance reveals that the state effects absorb 86.1% of the variation in OAA benefits, and the year effect absorbs 7.4%. Even with only 5.5% of residual variance of OAA benefits left to explain participation, the coefficient estimate remains statistically significant at better than a 99% confidence level.

TABLE 6Probit Results

(1) (2) (3) (4)
Log annual OAA benefit, no interactions* -0.226 -0.264 -0.169 -0.188
State manufacturing value added per capita ($1,000) (0.015) (0.062) (0.074) (0.085)0.687

(0.201)

State average farm value ($1,000) 0.017(0.006)
Log state per capita income 0.311(0.121)
State unemployment rate -0.866(0.720)
Log likelihood (per observation) -37875(-0.691) -37689(-0.687) -33813(-0.616) -33802(-0.616)
Includes state and year effects? No Yes[3031 Yes[3941 Yes[2671
Includes demographic variables and interactions with OAA?* No No Yes[971 Yes[971
Predicted labor force participation in 1950, with benefit levels of 1940* 50.3% 51.1% 49.9% 50.1%
Actual labor force participation in 1940: 49.7%
Actual labor force participation in 1950: 48.7%

The fixed effects will not capture influences on participation which shift within states over time, and which could be endogenous with benefits. Therefore, columns (3) and (4) add additional regressors that are available to control for individual and state level characteristics. Column (3) includes individual level demographic characteristics such as age dummies, education, and household characteristics. The individual level controls are allowed to have different effects in 1940 and 1950.