A new unique impulse response function in linear vector autoregressive models

Abstract

This article proposes a new unique impulse response function (IRF) measure, or MIRF, based on the popular vector autoregressive model to study interdependency of multivariate time series. Same as the orthogonal IRF, the estimator of MIRF has an analytical form with well-established asymptotics, and is invariant to ordering of series. Compared to alternative unique IRF measures, MIRF does not depend on extreme identifications, and the associated forecast error variance measure is explainable. An illustrative empirical example is also provided.