Behavior, Robustness, and Sufficient Statistics in Welfare Measurement

The joint implications for welfare measurement of three recent literatures are considered: the behavioral welfare economics literature, the structural versus reduced form debate in econometrics, and the use of sufficient statistics for characterizing behavior. The first permits the revealed preference paradigm to include a wide variety of so-called anomalous behavioral criteria. Sufficient statistics permit aggregation under heterogeneous behavioral criteria, not only heterogeneous characteristics, to facilitate meaningful economic welfare analysis. Atheoretic reduced form econometrics and traditional treatment-effect econometrics do not support welfare analysis, but marginal treatment-effect econometrics is useful for a certain type of policy problems. More generally, estimation of sufficient statistics requires theory-based reduced form econometrics. Sufficient statistics are further suggested for general equilibrium welfare measurement. Under certain assumptions about endogenous government behavior, robust concepts of economic welfare measurement emerge that rationalize the terminology and concepts of traditional welfare economics, although with a much broader conceptual basis.