A Bayesian approach to calibration

We develop a Bayesian approach to calibration that enables the incorporation of uncertainty regarding the parameters of the theoretical model under investigation. Our procedure involves the specification of prior distributions over parameter values, which in turn induce distributions over the statistical properties of artificial data simulated from the model. These distributions are compared with their empirical counterparts to assess the model's fit. The business-cycle model of King, Plosser, and Rebelo is used to demonstrate our procedure. We find that modest prior uncertainty regarding deep parameters enhances the plausibility of the model's description of the actual data.