Why Applied Macroeconomists Should Not Use Bayesian Estimation of DSGE Models

This paper argues that the common practice of Bayesian estimation in applied macroeconomic DSGE modeling can lead to severely biased results when the imposed prior beliefs are misspecified. We demonstrate, through controlled Monte Carlo experiments on two canonical DSGE models (a Real Business Cycle model and a New Keynesian model), that Bayesian estimation may yield misleading parameter estimates, often gravitating toward the researcher’s prior at the expense of empirical accuracy. In contrast, we show that Indirect Inference, a simulation-based classical estimation approach, remains largely unbiased and robust even under substantial model uncertainty. These findings suggest that heavy reliance on Bayesian estimation can perpetuate false conclusions (for example, overstating nominal rigidities) and thereby misguide policy analysis. We advocate greater use of robust estimation and model validation techniques, like Indirect Inference, to ensure that model-based policy guidance rests on credible empirical foundations.