The Limits of Inference Without Theory

In this rigorous and well-crafted work, Kenneth Wolpin examines the role of theory ininferential empirical work in economics and the social sciences in general -- that is, any researchthat uses raw data to go beyond the mere statement of fact or the tabulation of statistics. Heconsiders in particular the limits that eschewing the use of theory places on inference. Wolpinfinds that the absence of theory in inferential work that addresses microeconomic issues ispervasive. That theory is unnecessary for inference is exemplified by the expression "let thedata speak for themselves." This approach is often called "reduced form." A morenuanced view is based on the use of experiments or quasi-experiments to draw inferences.Atheoretical approaches stand in contrast to what is known as the structuralist approach, whichrequires that a researcher specify an explicit model of economic behavior -- that is, a theory.Wolpin offers a rigorous examination of both structuralist and nonstructuralist approaches. He firstconsiders ex ante policy evaluation, highlighting the role of theory in theimplementation of parametric and nonparametric estimation strategies. He illustrates thesestrategies with two examples, a wage tax and a school attendance subsidy, and summarizes the resultsfrom applications. He then presents a number of examples that illustrate the limits of inferencewithout theory: the effect of unemployment benefits on unemployment duration; the effect of publicwelfare on women's labor market and demographic outcomes; the effect of school attainment onearnings; and a famous field experiment in education dealing with class size. Placing each examplewithin the context of the broader literature, he contrasts them to recent work that relies on theoryfor inference.