A0180
Title: Detecting event-driven systemic cojumps and market crashes
Authors: Xiye Yang - Rutgers University (United States)
Deniz Erdemlioglu - IESEG School of Management (France) [presenting]
Christopher Neely - Federal Reserve Bank of St Louis (United States)
Abstract: A new testing framework is developed to detect the presence of systemic cojumps in a large panel of financial assets. The essential feature of the approach is that the test statistics are conditional on the release times of information events, which in turn allows us to pinpoint when precisely individual stocks or portfolio indices jump at the same time, or in a downward direction, when assets crash together systemically at high frequency. For inference, we introduce a computationally feasible StepM-type recursive bootstrapping procedure in high (asset and event) dimension, control for the multiple testing problem and eliminate spurious detection. We establish the bootstrap consistency of the tests and show in simulations that the tests have good size and considerably high power. Based on the high-frequency data on Dow Jones constituents and sector-specific ETFs, our empirical analysis provides strong evidence of systemic cojumps driven by the FOMC news announcements and monetary policy shocks. We find that a large fraction of the detected systemic cojumps exhibit downward pattern, indicating that monetary policy actions can lead to sudden crashes and generate systemic downside risk. We discuss the practical implications of our results for portfolio diversification, (news-driven) systemic risk monitoring and realized stress testing.
