This paper is a case study of the use of vector autoregression (VAR) models to test economic theories. It focuses on the work of Christopher A. Sims, who in 1980 found that relationships in economic data generated by a small VAR model were inconsistent with those implied by a simple form of monetarist theory. The paper describes the work of researchers who criticized Sims' results as not robust and Sims' response to these critics. The paper reexamines all of this work by estimating hundreds of variations of Sims' model. The paper concludes that both Sims and his critics are right: Sims' conclusion about monetarism is robust, but some of his other statistical results are not. In general, the paper concludes that VAR models can be used to test theories, but that any relationships they uncover in the data must be carefully checked for robustness.