It is important to remember, however, the estimates in Table 4 do not account
for any behavioral effects induced by these programs. The EITC may reduce
poverty more than it appears because by subsidizing earnings, it provides a greater
incentive to work. On the other hand, cash and noncash means-tested transfers may
reduce poverty rates by less than the already small estimates above because the high
benefit-reduction rates as people earn additional income discourage work.
Can trends in these government programs over time explain trends in poverty
rates? Spending on government programs has varied over time, and (for some
programs) across states. Following our analysis of labor market opportunities
above, we used the March Current Population Survey to construct the same
variables at the state level for 1977–2003, along with several different measures of
the generosity of government programs. We then ran regressions of the poverty
rate on these different measures of government spending, including both state and
year fixed effects. Not surprisingly given the relatively small effects of the programs
themselves on poverty, we also find that changes in government spending over time
explain very little of the trends in poverty rates (Hoynes, Page and Stevens, 2005).