Bayesian estimation of the Eurozone currency union effect

This paper uses Bayesian methods to estimate the European (Monetary) Union effect on trade. The high dimensionality of the parameter space when estimating gravity equations with many dummy variables results in standard hypothesis tests with a large Type I (false positive) error. Bayesian methods are able to handle this problem; they also provide a principled method of model selection that can be applied to different specifications of the dummy variables. Bayesian model selection tests prefer our most unrestricted dummy specification, which includes asymmetric bilateral effects, as well as time‐varying, country‐specific factors. Our estimate shows a zero Euro effect on trade, but a 14.8% increase in imports for a member of the European Union during 1980–2004.