Z:\grogger3\ddrive\Duncan\Results\Work Conditioned Welfare\APSA Version\Work Conditioned Welfare 7.doc
Printed On: 08/14/03
29
salient features of state welfare programs. It means that we can have more confidence not just in
the maximum benefit results but in the results for work incentives as well.
C
ONCLUSIONS
Recent reforms to U.S. redistributive programs have radically altered the way in which the
government delivers social assistance to the poor. One of the most dramatic changes is the
increased degree to which benefits are conditioned on employment. Examples of these reforms
include: expansions to the Earned Income Tax Credit, which only provides benefits to people with
earnings; the requirement that most AFDC/TANF work while on aid; and changes the structure of
AFDC/TANF benefits that allow welfare recipients to keep more of their earnings.
To determine what caused these changes, I pursued two different empirical approaches. The
first was a difference model that compares changes in AFDC/TANF maximum benefits to changes
in inequality, unemployment, government ideology, and a number of other variables between 1980
and 1999. Since we can expect that the labor force participation of welfare recipients will decline as
benefits increase, changes to maximum benefits provide an indicator of the degree to which states
are worried about work incentives. The second, more targeted, approach directly examined
differences in the effective tax rate on earnings of AFDC/TANF recipients in the late 1990s.
The evidence from these studies suggests that the recent trend towards work incentives in
welfare benefits is at least partly a function of ideology, the unemployment rate, and increasing
earnings inequality. The evidence on ideology is the weakest, but there is some reason to believe
that states with Republican governors implemented stronger work incentives as a part of welfare
reform. States with high levels of unemployment also significantly reduced the effective tax rates on
welfare recipients’ earnings during the 1990s, but did not significantly alter benefit levels over the