Detecting Contagion Effects During The Subprime Crisis Using Different VAR Size Models: Comparison Between OLS And Bayesian Shrinkage

Abstract

This paper assesses the performance in terms of forecast and structural analysis of different sizes of VAR models while studying contagion effects during the subprime crisis. We compare VAR models of different sizes: a standard VAR estimated by OLS, and a MEDIUM and LARGE VARs estimated by a Bayesian shrinkage procedure.

Using a large database containing national macroeconomic, financial, and trade dynamic variables for 17 OECD countries, we study contagion effects on U.S key variables and OECD housing markets over the period of the subprime crisis.

Our paper unearths that the smallest specification outperforms the largest one, in terms of forecast accuracy and structural analysis. Our results reveal that the SMALL and MEDIUM VARs produce qualitatively similar results; however, the MEDIUM VAR permits us to obtain the responses of a large set of variables and so provides a more comprehensive picture of the effects of the subprime crisis. So, we conclude that the MEDIUM VAR is more successful at extracting pertinent information for a large data set. It appears that dealing with a LARGE VAR provides worst results.

An interesting result that arises is that the standard OLS estimation is interesting; however, the Bayesian shrinkage is a useful tool, in the case of an international contagion study, allowing both a more comprehensive picture of the contagion effects as well as a more complete check of the empirical plausibility of the underlying specification.