The Contagion Box: Measuring Financial Market Co-Movements by Regression Quantiles

Abstract

This paper develops an econometric framework to investigate the structure of co-dependences across markets and to test whether it changes over time or across market conditions. Our approach is based on the computation, over both a test and a benchmark period, of the conditional probability that the returns on one market are lower than a given quantile, when returns on the other market are also lower than their corresponding quantile, for any set of prespecified quantiles. Quantiles are allowed to vary over time using the CAViaR methodology developed by Engle and Manganelli (2004). Graphically, the conditional probabilities can be represented in what we call the contagion box, which is a square of unit side. Since a 45 degree line represents the case of independence, the presence of co-movements is indicated when the conditional probability plots above this line. Differences in the intensity of co-movements can be identified directly from the conditional probability plots for test and benchmark periods. From this insight, rigorous econometric tests of contagion are derived and implemented. In the process, we obtain a new result in the regression quantile literature. We illustrate the methodology by investigating the impact of the tequila (1994/95), Asian (1997) and Russian (1998) crises on the major Latin American equity markets. Our results suggest significant presence of contagion.