By Daniel Waggoner and Tao Zha
We confront model misspecification in macroeconomics by proposing an analytic framework for merging multiple models. This framework allows us to 1) address uncertainty about models and parameters simultaneously and 2) trace out the historical periods in which one model dominates other models. We apply the framework to a richly parameterized DSGE model and a corresponding BVAR model. The merged model, fitting the data better than both individual models, substantially alters economic inferences about the DSGE parameters and about the implied impulse responses.
We all know models are abstractions, and some models thus perform better in some situations. Here, models are combined to let the data tell when a model is more appropriate than another. This improves the fits over using a single model. But why would this be better than using a single model with less abstraction? And how can this be useful for policy experiments, as one is uncertain which model is currently valid?