Working Paper Series no. 285: The Growth-Volatility Relationship: New Evidence Based on Stochastic Volatility in Mean Models

Abstract

This paper models the relationship between growth and volatility for G7 economies in the time period 1960-2009. It delivers for the first time estimates of this relationship based on a logarithm variant of stochastic volatility in mean (SV-M) models. The relationship appears significantly positive in Germany and Italy, but insignificant in other countries. We also show that output volatility has increased in all countries since the beginning of the financial crisis, which illustrates the end of the great moderation. For comparison, the paper also delivers estimates based on a logarithm variant of GARCH in mean (log-GARCH-M) models, the class of time series models previously used in the literature to estimate the growth-volatility relationship. We show that SV-M models deliver results preferable to those of log-GARCH-M models, despite the high computational cost of their estimation. SV-M models fit generally better data than log-GARCH-M ones. As their residuals do not violate distribution assumptions, they do not deliver dubious conclusions concerning the significance of the relationship, which is the case of the log-GARCH-model for France, the UK and the US. Finally, SV-M models suggest a positive impact of unexpected volatility on output growth, which is not taken into account by log-GARCH-M models.

Matthieu Lemoine and Christophe Mougin
July 2010

Classification JEL : O40, E32, C15

Keywords : growth, volatility, sequential Monte-Carlo methods.