We develop estimators and asymptotic theory to decompose the quadratic covariation between two assets into its continuous and jump components, in a manner that is robust to the presence of market microstructure noise. Using high frequency data on different assets classes, we find that the recent financial crisis led to an increase in both the quadratic variations of the assets and their correlations. However, we find little evidence to suggest a change between the relative contributions of the Brownian and jump components, as both comove. Co-jumps stem from surprising news announcements that occur primarily before the opening of the US market, and are also accompanied by an increase in Brownian-driven correlations.
All Science Journal Classification (ASJC) codes
- Economics and Econometrics
- Continuous and jump components
- Financial crisis
- News surprises
- Overnight jumps
- Quadratic covariation