Jumps Versus Bursts: Dissection and Origins via a New Endogenous Thresholding Approach

Abstract

We study the different origins of two closely related extreme financial risk factors: volatility bursts and price jumps. We propose a new method to separate these quantities from ultra-high-frequency data via a novel endogenous thresholding approach in the presence of market microstructure noise and staleness. Our daily jump statistic proxies volatility bursts when intraday jumps are accurately controlled by our local jump test (which proves to be highly powerful with extremely low misclassification rates due to its timely detections). We find that news is more related to volatility bursts; while high-frequency trading variables, especially volume and bid/ask spread, are prominent signals for price jumps.