Bootstrapping structural VARs: Avoiding a potential bias in confidence intervals for impulse response functions

► The commonly-used estimator of the VAR error covariance matrix is unbiased in the initial VAR, but biased when used in bootstrapping. ► This can markedly distort reported confidence bands for structural VARs, especially those will small sample size relative to number of regressors. ► A simple degrees of freedom adjustment will eliminate this bias. ► Adjusted bootstrap confidence intervals exhibit improved coverage accuracy.