| Abstract: | In this study, we use a vector autoregressive modeling framework to test the zero restrictions implied by alternative criteria for ranking regions into hierarchies in the wage-transmission debate. This approach allows formal statistical tests to be carried out for competing criteria suggested for determining leading and following regions. The ability of the modeling technique proposed here to produce a set of nested hypothesis tests of alternative criteria is in stark contrast to the historical literature in this debate. Researchers have traditionally proposed a criterion for ranking regions into a hierarchy, and then argued for the merits of their particular criterion by demonstrating that an econometric model of wage formation produces statistically significant coefficient estimates when their criterion is used to rank the regions of their data sample. We apply the methods proposed here to a sample of eight midwestern cities in the U.S. in order to test the following criteria: Beaumont (1983), Hart and MacKay (1977), Reed and Hutchinson (1976), unemployment rate, and earnings level. The test is for consistency with the Granger-causal structure of wage interactions inherent in the wage diffusion idea. We argue that the technique set forth here is a real step forward that should allow a resolution of this particular debate. The proposed procedures might also be applied to empirically test other regional science hypotheses concerning for example, intercity and interindustry causal structures. |
