Multi-jumps

  • Massimiliano Caporin University of Padova, Department of Economics and Management
  • Aleksey Kolokolov Department of Statistics Lund University School of Economics and Management Box 743, SE-22007 Lund, Sweden
  • Roberto Reno University of Siena, Department of Economics and Statistics

Abstract

We provide clear-cut evidence for economically and statistically significant multivariate jumps (multi-jumps) occurring simultaneously in stock prices by using a novel nonparametric test based on smoothed estimators of integrated variances. Detecting multi-jumps in a panel of liquid stocks is more statistically powerful and economically informative than the detection of univariate jumps in the market index.On the contrary of index jumps, multi-jumps can indeed be associated with sudden and large increases of the variance risk-premium, and possess a statistically significant forecasting power for future volatility and correlations which implies a sizable deterioration in the diversification potential of asset allocation.

Issue
Section
Working Papers in Statistics