A Bayesian View of Nominal Money and Real Output Through a New Classical Macroeconomic Window; Comments; Reply

The empirical evidence on the effects of unanticipated changes in nominal money on real output in 47 countries when viewed through a macroeconomic window - the likelihood function - is investigated. Using a Bayesian predictivist approach, a pedagogical Bayesian analysis is presented of generated regressor models in the face of specification uncertainty involving, among other things, multiple unit roots and trend stationary alternatives. All 3 neutrality hypotheses receive nonnegligible support, but the results favor neutrality of unanticipated money. In a comment, Hong argues that Poirier allows too much specification uncertainty for nominal money in light of other studies that suggest insensitivity of output-equation inferences and in light of the auxiliary nature of nominal money equations. Using current available annual postwar data through 1977, Kormendi and Meguire (KM) confirm their 1984 cross-country finding in favor of the Lucas proposition. In reply, Poirier says that KM's analysis would have been more compelling if they addressed the generated regressor problem rather than pursuing the Barro 2-step estimator.