Sufficient Statistics for Welfare Analysis: A Bridge Between Structural and Reduced-Form Methods

   (370 K)

The debate between "structural" and "reduced-form" approaches has generated substantial controversy in applied economics. This article reviews a recent literature in public economics that combines the advantages of reduced-form strategies -- transparent and credible identification -- with an important advantage of structural models -- the ability to make predictions about counterfactual outcomes and welfare. This recent work has developed formulas for the welfare consequences of various policies that are functions of high-level elasticities rather than deep primitives. These formulas provide theoretical guidance for the measurement of treatment effects using program evaluation methods. I present a general framework that shows how many policy questions can be answered by identifying a small set of sufficient statistics. I use this framework to synthesize the modern literature on taxation, social insurance, and behavioral welfare economics. Finally, I discuss topics in labor economics, industrial organization, and macroeconomics that can be tackled using the sufficient statistic approach.

This paper is available as PDF (370 K) or via email.

Acknowledgments

Machine-readable bibliographic record - MARC, RIS, BibTeX