Estimating the effects of monetary policy shocks: does lag structure matter?

This paper examines the implications of lag structure for estimating the effects of monetary policy shocks in a VAR. A symmetric lag structure in which all variables have the same lag length and an asymmetric lag structure in which the lag length differs across variables but is the same for a particular variable in each equation of the model are examined. This is important in light of the fact that the true lag structure is generally not known. Four commonly used identification schemes are employed to identify monetary policy shocks. Monte Carlo simulations strongly indicate that the lag structure of a VAR model does matter when assessing the quantitative effects of monetary policy shocks. Given the inherent uncertainty about the true lag structure in practice, it is thus important that one compare the impulse response functions from both symmetric lag and asymmetric lag VARs in assessing the effects of monetary policy shocks.