The role of hypothesis testing in the molding of econometric models
Keywords:statistical testing, hypothesis tests, models, general-to-specific specification search, optional stopping, severe tests, costs of search, costs of inference, extreme-bounds analysis, LSE econometric methodology
This paper addresses the role of specification tests in the selection of a statistically admissible model used to evaluate economic hypotheses. The issue is formulated in the context of recent philosophical accounts on the nature of models and related to some results in the literature on specification search. In contrast to enumerative induction and a priori theory, powerful search methodologies are often adequate substitutes for experimental methods. They underwrite and support, rather than distort, statistical hypothesis tests. Their success is grounded in a systematic effort to mold appropriate models to, and test them against, constraints of the data.