Title: Impulse response estimation by smooth local projections
Authors: Christian Brownlees - UPF (Spain) [presenting]
Regis Barnichon - CREI (Spain)
Abstract: Vector autoregressions (VAR) and local projections (LP) are well established methodologies for the estimation of impulse responses (IR). These techniques have complementary features: the VAR approach is more efficient when the model is correctly specified whereas the LP approach is less efficient but is more robust to model misspecification. We propose a semi-parametric impulse response estimator, called smooth local projections (SLP), that attempts to strike a balance between these two extremes. The procedure consists of using local projections under the constraint that the IR is a smooth function of the forecast horizon. Inference is carried out using semi-parametric techniques based on B-splines, and the IR can be estimated by standard (ridge) regression. We also show how SLP may be used in conjunction with common identification schemes such as timing restrictions or instrumental variables to recover structural IRs. We apply our technique to study the effects of monetary shocks and show using out-of-sample validation criteria that smooth local projections provide more precise IR estimates than VAR and LP.