DP13511 Dealing with misspecification in structural macroeconometric models
We consider a set of potentially misspecified structural models, geometrically combine their likelihood functions, and estimate the parameters using composite methods. Composite estimators may be preferable to likelihood-based estimators in the mean squared error.
Composite models may be superior to individual models in the Kullback-Leibler sense. We describe Bayesian quasi-posterior computations and compare the approach to Bayesian model averaging, finite mixture methods, and robustness procedures. We robustify inference using the composite posterior distribution of the parameters and the pool of models. We provide estimates of the marginal propensity to consume and evaluate the role of technology shocks for output fluctuations.