Measuring the Impact of Welfare Benefits on Welfare Durations: State Stratified Partial Likelihood and Fixed Effect Approaches

Previous studies of welfare durations in the U.S. have analyzed the role of welfare benefits on the length of time recipients remain on welfare. Typically these studies use individual data and estimate the benefit effects based on both cross-sectional variation across locations and time variation within locations. Cross sectional results on benefit effects may be biased due to unobserved state specific differences in welfare programs or attitudes. This paper compares two strategies for dealing with the problem when estimating welfare exit hazard models: stratification of baseline hazards by state and the inclusion of state fixed effects in the hazards. Although the results show that high benefit states appear to have unobserved characteristics that lead to longer spells, the results also show that increases in welfare benefits do not directly increase welfare durations once one controls for state specific effects.