Structural Vector Autoregressive Analysis in a Data Rich Environment: A Survey

Abstract

Large panels of variables are used by policy makers in deciding on policy actions. Therefore it is desirable to include large information sets in models for economic analysis. In this survey methods are reviewed for accounting for the information in large sets of variables in vector autoregressive (VAR) models. This can be done by aggregating the variables or by reducing the parameter space to a manageable dimension. Factor models reduce the space of variables whereas large Bayesian VAR models and panel VARs reduce the parameter space. Global VARs use a mixed approach. They aggregate the variables and use a parsimonious parametrisation. All these methods are discussed in this survey although the main emphasize is on factor models.

Keywords: factor models, structural vector autoregressive model, global vector autoregression, panel data, Bayesian vector autoregression

JEL Classification: C32