Assessments of the distributional effects of public spending reforms have g
enerally been based on average rates of program participation by income or
expenditure group. This practice can be deceptive because the socioeconomic
composition of participants can change as a social program expands or cont
racts. The geographic variation found in 1993-94 household survey data for
rural India is used to estimate the marginal odds of participating in schoo
ling and antipoverty programs. The results suggest early capture of these p
rograms by the nonpoor. It is shown that conventional methods for assessing
benefit incidence underestimate the gains to the poor from higher public o
utlays and underestimate the losses from cuts.