Heteroskedastic Structural Vector Autoregressions Identified via Long-Run Restrictions

Abstract

A central assumption for identifying structural shocks in vector autoregressive (VAR) models via heteroskedasticity is the time-invariance of the impact effects of the shocks. It is shown how that assumption can be tested when long-run restrictions based on the cointegration structure of the variables are available for identifying structural shocks. The importance of performing such tests is illustrated by investigating the impact of fundamental shocks on stock prices in the U.S. It is found that fundamental shocks post-1986 have become more important than in the pre-1986 period.