Detecting contagion with correlation: Volatility and timing matter

Abstract

The detection of contagion effects is sensitive to controlling for volatility changes between periods of tranquility and periods of crisis. An additional consideration is the use of synchronised data for geographically separated markets. We demonstrate how these effects can combine in a practical application to detecting contagion in European equity markets in the period of 2007-2009. Without controlling for volatility clustering synchronization does not apparently matter. Once volatility clustering is accounted for synchronized data dramatically changes results. Our preferred results indicate relatively little evidence for contagion effects flowing directly from US equity markets to those of Europe during the crisis itself, and more evidence of continued transmission during the post crisis period-potentially reflecting unsettled conditions associated with the burgeoning Greek debt crisis.