Testing for Strong Exogeneity in Proxy-Vars

Abstract

Proxy variables have gained widespread prominence as indispensabletools for identifying structural VAR models. Analogous to instrumental variables,proxies need to be exogenous, i.e. uncorrelated with all non-target shocks. Assessingthe exogeneity of proxies has traditionally relied on economic argumentsrather than statistical tests. We argue that the economic rational underlying theconstruction of commonly used proxy variables aligns with a stronger form of exogeneity.Specifically, proxies are typically constructed as variables not containingany information on the expected value of non-target shocks. We show conditionsunder which this enhanced concept of proxy exogeneity is testable without additional identifying assumptions.

Keywords: Structural vector autoregression, proxy VAR, exogeneity test