Another Look at the Relationship between Cross-market Correlation and Volatility

While increases in cross-market correlations during periods of market crises are well documented, Forbes and Rigobon (2002) show that correlation coefficients are biased measures of dependence when markets become more volatile, and that there is no evidence of contagion in recent financial crises once corrected for these effects. This paper explores the impact of volatility on market dependence more broadly, both analytically and empirically using simulated time-series of financial asset returns that follow alternative stochastic processes commonly used in financial research. The results show that market dependence is not generally conditional on volatility regimes and that a bias in dependence measures occurs only for particular assumptions about the time-series dynamics. Since real world data may often not be characterized by homoskedasticity, a correction of estimated unconditional correlations during market crises may not always be needed. Consequently, the results provide evidence that contagion indeed exists as a real phenomenon during financial crises, reducing the benefits of portfolio diversification when needed most. In contrast, if the return data generating process is invariant but displays conditional heteroskedasticity, a conditioning bias exists that cannot be distinguished from a fundamental change in market dependence.