Modeling industrial location decisions in U.S. counties

Given its sound theoretical underpinnings, the RandomUtilityMaximizationbased conditional logit model (CLM) serves as the principal method for applied research on industrial location decisions. Studies that implemented this methodology, however, had to confront the underlying Independence of Irrelevant Alternatives (IIA) assumption and were unable to fully accommodate this problem. This paper shows that by taking advantage of an equivalent relation between the CLM and Poisson regression likelihood functions one can more e.ectively control for the potential IIA violation in complex choice scenarios where the decision-maker confronts a large number of spatial alternatives. The paper also provides an illustration, demonstrating the advantages of this relation in investigation of location determinants of new manufacturing plant births in the U.S. counties.