However, actual patterns of benefit levels, participation, and average income suggest that this potential source of spurious benefit changes is not a major problem. Average benefits B rose between 1940 and 1950. If trends in income Y influenced B more than statutory changes in Bmax did, it must be that income among the poor elderly fell. However, per capita income rose substantially during the decade, as did other measures of economic activity reported in Appendix Table 4. Moreover, rising average benefits and rising state per capita income were positively correlated, so in states that grew richer overall the poor elderly would have had to grow systematically poorer, raising B , if changes in Y were driving changes in B ; if not, then increases in statutory benefit levels were dominating changes in B . Lastly, as demonstrated in Section IV, increases in per capita income and other measures of the economy were associated with increases in labor force participation among the elderly – which should channel through to lower average benefits. In sum, average benefits rose even while income and participation were rising.

Therefore, it does not appear that the average benefit measure B was seriously contaminated by changes in income distribution correlated with labor force participation, but rather that changes in statutory benefit levels drove benefit changes.

Another potentially important rule governs the treatment of migrants. If people retired and then chose to move to high benefit states, the effect of benefits on retirement would be overestimated. Federal law prevented states from requiring more than one year of in-state residence preceding application, but did allow them to require up to five years’ residence within the last nine. The Social Security Bulletin (1944) noted that the great majority of states imposed the maximum requirements, which limits the extent of benefit-driven migration, though many states did reduce the requirement to one year by 1950. Furthermore, cross-state migration among the elderly was low. In the 1950 Census sample of men aged 66-75 used in this paper, 1.0% had moved to a new state in the past year. Among those, slightly more than half moved to states with lower 1950 benefits than their own. Lastly, even with a more mobile population today, Walker (1994) found that differential AFDC benefits could not explain migration to neighboring states by poor young women. Therefore, it does not appear likely that OAA benefits drove migration.