Model-Free Impulse Responses

This paper introduces methods for computing impulse response functions that do not require specificationand estimation of the unknown dynamic multivariate system itself. The central idea behind these methodsis to estimate flexible local projections at each period of interest rather than extrapolating into increasinglydistant horizons from a given model, as it is usually done in vector autoregressions (VAR). The advantagesof local projections are numerous: (1) they can be estimated by simple regression techniques with standardregression packages; (2) they are more robust to misspecification; (3) standard error calculation is direct;and (4) they easily accommodate experimentation with highly non-linear and flexible specifications thatmay be impractical in a multivariate context. Therefore, these methods are a natural alternative toestimating impulse responses from VARs. An application to a simple, closed-economy monetary modelsuggests that the output loss and inflation effects of an interest rate shock depend on the stage of thebusiness cycle.