An Alternative Bootstrap for Proxy Vector Autoregressions

Abstract

We propose a new bootstrap for inference for impulse responses in structural vector autoregressive models identified with an external proxy variable. Simulations show that the new bootstrap provides confidence intervals for impulse responses which often have more precise coverage than and similar length as the competing moving-block bootstrap intervals. An empirical example shows how the new bootstrap can be applied in the context of identifying monetary policy shocks.

Comparison of local projection estimators for proxy vector autoregressions

“…They also show that other bootstraps such as the wild bootstrap, that have been used in structural VAR analysis, do not yield confidence intervals with asymptotically correct coverage. Unfortunately, there is also evidence that the MBB may not be very accurate in small samples (e.g., Bruns and Lütkepohl (2020)). As we are primarily interested in the relative performance of the different estimators, we prefer the asymptotically valid MBB and assume that it imposes a similar small sample handicap on all estimators.…”

Heteroskedastic Proxy Vector Autoregressions: Testing for Time-Varying Impulse Responses in the Presence of Multiple Proxies

“…These authors show the asymptotic validity of the method for inference in structural VAR analysis even under time-varying volatility. We implement the MBB exactly as in Bruns and Lütkepohl (2020). 8 The impulse responses in Figure 8 show that some of the dynamic effects of the two TFP shocks clearly depend on the regime if we allow for time-varying impulse responses.…”