Queen Mary University of London, School of Economics and Finance, London
This paper evaluates the performance of structural VAR models in estimating the impact of credit supply shocks. In a simple Monte-Carlo experiment, we generate data from a DSGE model that features bank lending and credit supply shocks and use SVARs to try and recover the impulse responses to these shocks. The experiment suggests that a proxy VAR that uses an instrumental variable procedure to estimate the impact of the credit shock performs well and is relatively robust to measurement error in the instrument. A structural VAR with sign restrictions also performs well under some circumstances. In contrast, VARs of the narrative variety, i.e. VAR models that include measures of the credit shock as endogenous variables are highly sensitive to ordering and measurement error. An application of the proxy VAR model and the VAR with sign restrictions to US data suggests, however, that the credit supply shock is hard to identify in practice.