A distinctive feature of OAA was that states set benefit levels. In contrast, Social Security and SSI, the other social insurance programs for the elderly, are federally administered, which makes identification of behavioral responses difficult. The response to changes over time in program rules may be confounded with other factors that influenced labor supply. That may explain Moffitt’s (1987) findings that much of the time series correlation between rising Social Security benefits and rising retirement rates appears spurious. Furthermore, everyone faces the same rules at a point in time, which makes the variation in actual benefits across individuals problematic. As Bound (1989) noted, Social Security benefits are a positive but declining function of past earnings. This makes it difficult to identify whether a high replacement rate or low current earnings, which are likely to be correlated with low past earnings, are driving observed retirement behavior.

Studying OAA makes it easier to understand how benefit levels – which varied across states at a point in time and also within states over time – are related to changes in labor supply. The cross-state variation in OAA benefits also makes it possible to distinguish its impact from that of the early Social Security program. This strategy is quite different from the structural modeling approach that many papers use to analyze retirement. Structural estimates can be used to make detailed predictions about a wide range of policy changes; however, the results depend crucially on identifying the underlying behavioral responses. The alternative approach of focusing on reduced form estimates of the response to policy changes has yielded insights into a range of contemporary public programs.