Bayesian Local Projections

Accepted at Review of Economics and Statistics

Code [GitHub]

MAIN_BayesianLocalProjections

.pdf

Download PDF • 605KB

MAIN_OnlineAppendix

.pdf

Download PDF • 1.25MB

We propose a Bayesian approach to Local Projections that optimally addresses the empirical bias-variance tradeoff inherent in the choice between VARs and LPs. Bayesian Local Projections (BLP) regularise the LP regression models by using informative priors, thus estimating impulse response functions potentially better able to capture the properties of the data as compared to iterative VARs. In doing so, BLP preserve the flexibility of LPs to empirical model misspecifications while retaining a degree of estimation uncertainty comparable to a Bayesian VAR with standard macroeconomic priors. As a regularised direct forecast, this framework is also a valuable alternative to BVARs for multivariate out-of-sample projections.