Who wins, who loses? Tools for distributional policy evaluation

Most policy changes generate winners and losers. Political economy and optimal policy suggest questions such as: Who wins, who loses? How much? Given a choice of welfare weights, what is the impact of the policy change on social welfare? This paper proposes a framework to empirically answer such questions. The framework is grounded in welfare economics and allows for arbitrary heterogeneity across individuals as well as for endogenous prices and wages (general equilibrium effects). The proposed methods are based on imputation of money-metric welfare impacts for every individual in the data. The key technical contribution of this paper are new identification results for marginal causal effects conditional on a vector of endogenous outcomes. These identification results are required for imputation of individual welfare effects. Based on these identification results, we propose methods for estimation and inference on disaggregated welfare effects, sets of winners and losers, and social welfare effects. We furthermore provide results relating aggregation with social welfare weights to the distributional decomposition literature. %Our framework generalizes approaches used in the empirical optimal tax literature, the distributional decomposition literature, and the literature on determinants of the wage distribution. We apply our methods to analyze the distributional impact of the expansion of the Earned Income Tax Credit (EITC), using variation in state supplements to the federal EITC and the CPS-IPUMS data. We find large negative effects of depressed wages as a consequence of increased labor supply. The estimated effects are largest for earning around 20.000 US$ per year, and for high school dropouts.