The Fiscal Multiplier Morass: A Bayesian Perspective

A Bayesian prior predictive analysis is conducted on a suite of models to assess the probability that a model and corresponding prior distributions bias results toward a specific range of fiscal multipliers. We examine a wide range of DSGE models commonly used to estimate fiscal multipliers, including a real business cycle model, a New Keynesian model with nominal and real rigidities, and open economy models. We decompose changes in multipliers across models into wealth and substitution effects, allowing for a more uniform comparison across models. Through the prior predictive analysis, we show that many of the models and prior distributions impose a very tight range for the multiplier before the models are taken to data. We argue that constraining the multiplier to such a tight range prior to conditioning on data is tantamount to biasing results. A broader message of the paper calls for employing prior predictive analysis when estimating DSGE models.