Bayesian Inference of Local Projections with Roughness Penalty Priors

摘要

A local projection is a statistical framework that accounts for therelationship between an exogenous variable and an endogenous variable, measuredat different time points. Local projections are often applied in impulseresponse analyses and direct forecasting. While local projections are becomingincreasingly popular owing to their robustness to misspecification and theirflexibility, they are less statistically efficient than standard methods, suchas vector autoregressions. In this study, we seek to improve the statisticalefficiency of local projections by developing a fully Bayesian approach thatcan be used to estimate local projections using roughness penalty priors. Then,we apply the proposed approach to an analysis of monetary policy in the UnitedStates, showing that the roughness penalty priors successfully estimate theimpulse response functions and improve the predictive accuracy of the localprojections.