Nested Models and Model Uncertainty

Abstract

This paper studies how model uncertainty over nested structural models affects policy analysis and model choice. We consider a situation where a simple benchmark model is nested within a richer alternative and show that, when the true data-generating process is unknown, standard likelihood-based model selection criteria can favour misspecified models whose policy prescriptions differ from those of the benchmark. We derive decision rules that account explicitly for model uncertainty and characterize conditions under which ignoring this uncertainty leads to substantial welfare losses. Applying our framework to monetary policy analysis, we illustrate how different assumptions about the underlying structural model map into different optimal policy rules and how robust policy recommendations can be obtained by integrating out model uncertainty.

Citation

Kriwoluzky, Alexander, and Christian A. Stoltenberg (2016). “Nested Models and Model Uncertainty.” The Scandinavian Journal of Economics 118(2): 324–353. [web:355][web:364]

@article{KS2016,
author = {Kriwoluzky, Alexander and Stoltenberg, Christian A.},
title = {Nested Models and Model Uncertainty},
journal = {The Scandinavian Journal of Economics},
volume = {118},
number = {2},
pages = {324-353},
keywords = {Bayesian model estimation, model uncertainty, optimal monetary policy, C51, E32, E52},
doi = {https://doi.org/10.1111/sjoe.12134},
year = {2016}
}