Asymptotically Valid Bootstrap Inference for Proxy SVARs

Proxy structural vector autoregressions identify shocks in with external variables that are correlated the of interest but uncorrelated all other shocks. We provide asymptotic theory for this identification approach under mild ?-mixing conditions cover a large class uncorrelated, possibly dependent innovation processes. prove consistency residual-based moving block bootstrap (MBB) inference on statistics such as impulse response functions and forecast error variance decompositions. The MBB serves basis constructing confidence intervals when proxy strongly interest. For case one variable used to shock, we show can be construct sets normalized responses valid regardless strength based inversion Anderson Rubin statistic suggested by Montiel Olea, Stock, Watson.