Implications for Social Security

Researchers have had surprising difficulty connecting the tremendous growth in the Social Security program to rising retirement rates. Although the estimates in this paper are derived from OAA, they offer a means to extrapolate to Social Security’s impact on early retirement trends.

Both OAA and Social Security were earnings tested income transfers with similar benefit levels until later in the 1950s, when real Social Security benefits rose 50.4%.

Still, the programs differ along other important dimensions. First, OAA was available to all elderly; Social Security covered only about 70% of workers in 1950, and expanded to almost complete coverage over the next decade.39 That factor makes predictions about the resulting change in participation a lower bound. Second, OAA was completely means and wealth tested, while Social Security was only earnings tested.40 The tighter means testing on OAA again implies that its estimated impact can be interpreted as a lower bound on Social Security’s impact. A worker considering retiring and receiving Social Security might not even qualify for OAA, depending on his non-labor income and wealth. However, the ambiguity due to the average benefit measure which must be used here makes it difficult to adjust for this effect.

The estimates from column (3) were used to simulate a 50.4% increase in benefit levels. Such an increase in benefits is predicted to lower labor force participation of 66-75 year olds from 48.7% observed in 1950 to 45.7% predicted in 1960, a 6.2% decline. Rescaling the estimate by 100/70 to account for the expansion in Social Security coverage suggests a 8.8% decline. In fact, the actual participation rate dropped precipitously to 35.3%. Naturally, other factors in the 1950s, such as expanding pension coverage and increasing prosperity – which might have begun to induce wealth effects on labor supply – are likely to have influenced retirement rates. Still, because the simulation is likely to be a lower bound, it suggests that Social Security accounted for at least one-third of the decline.

The evidence about OAA is broadly in line with conclusions from Costa (1995). Exploiting policy variation in the incentive to retire, as in this paper, Costa estimated a sizable income elasticity of retirement for elderly Union Army veterans. Her estimate implies that subsequent increases in per capita income and wealth might explain sixty percent of the decline in elderly labor force participation between 1900 and 1980. Costa further noted that most other estimates of retirement elasticities, based on data from the 1970s on, are much smaller, and she suggested that the income elasticity of retirement might have fallen over time.41 Another possible explanation is simply that the sort of policy variation which she uses for identification yields more conclusive results.