THE EFFECT OF OLD AGE ASSISTANCE ON RETIREMENT: EMPIRICAL RESULTS 3

Endogenous Legislation

Endogenous benefit levels would make the estimates reported here inconsistent. OAA benefits and participation may be associated not because rising benefits induced people to leave the labor force, but instead because rising retirement rates led states to raise support levels in response. Yet, it is difficult to come up with useful and available instruments to control for potentially endogenous benefits.

An alternative is to test whether OAA is correlated with trends in labor force participation for other people who are ineligible for OAA. I estimated the same model for slightly younger men, aged 55-59. This group should not respond to OAA, but the labor market conditions they face are likely to be linked to those of older men. The hypothesis that changes in benefits drove changes in participation, and not that changes in labor market conditions led policymakers to adjust benefit levels, would be reflected in coefficients close to zero on log benefits and its interaction terms.

Estimates of a probit analogous to column (4), but with 55-59 year olds, yields a coefficient on the OAA main effect of 0.077 (0.175), slightly positive and highly insignificant. The aggregate effect, incorporating the OAA interaction terms, also reflects no relationship between OAA and labor force participation for 55-59 year olds. These results support the hypothesis that OAA benefits were not influenced by labor market conditions.

The results suggest a further check on the OAA estimates. If 55-59 year olds are a reasonable comparison group for the older group eligible for OAA, then the two samples can be combined. Including the OAA ineligibles allows state-year interactions to be identified separately from the effect of benefits, which controls more directly for the possibility that changes in benefits are endogenous with changes in participation. The state-year interactions account for influences on participation that varied over time within states. Changes in participation for 55-59 year olds identify the general trends in participation within states, which might have been correlated with benefit changes. Additional changes that only occur for OAA eligibles remain to identify the direct impact of OAA on participation. The strategy presumes that the labor supply of 55-59 year olds is otherwise similar to the labor supply of OAA eligibles – except for age effects (both main effects and interactions with year and state) and except for the systematic influence of OAA. The “difference-in-difference-in-differences” estimator employing within state control groups has been used to study the impact of other programs as well.

TABLE 7

Additional Specifications

(4) From Table 6 (5)55-59, 66-73 year olds (6)Nonwhites

excluded

(7)Dependent

variable:

Employed

Log annual OAA benefit -0.188 -0.222 -0.179 -0.163
(0.085) (0.162) (0.092) (0.085)
Log likelihood -33802 -40452 -30990 -33612
(per observation) (-0.616) (-0.485) (-0.618) (-0.613)
Includes state*year effects? No Yes No No
[961
Predicted labor force participation in 1950, with benefit levels of 1940 50.1% 53.1% 49.7% 48.2%
Actual labor force participation in 1940: 49.7% 52.6% 49.4% 46.0%
Actual labor force participation in 1950: 48.7% 51.4% 48.8% 47.0%

Column (5) of Table 7 reports the results with state*year fixed effects. Because of limits in computing capacity, the sample is restricted to 66-73 year olds, together with the 55-59 year olds.37 The state*year interactions are jointly highly significant, as are the state*age interactions, while the age*year interactions are not significant. Those results imply that crossstate differences in labor force participation varied over time and also by age, but that age-related patterns were similar over time. The coefficient on the OAA main effect increases to -0.222 (0.162) instead of declining, as would be expected if benefits were set endogenously. The OAA main effect is statistically significant only at a confidence level of 83%, but the OAA interaction terms remain jointly significant at greater than 99% confidence. The boost in the OAA coefficient carries through in the aggregate: the participation rate of 66-73 year olds fell from 52.6% in 1940 to 51.4% in 1950 but would have risen to 53.1% if benefits had not been raised. Therefore, including the 55-59 year olds in the estimation supports the conclusion of a strong effect of OAA on participation and suggests that OAA benefits were not set in response to changes in labor force participation.td