Detecting Spurious Jumps in High Frequency Data

Mardi | 2010-02-02

P. BAJGROWICZ – Olivier SCAILLET – libre

We propose a technique to avoid spurious detection of jumps in highfrequency data via an explicit thresholding on available test statistics. Weprove that it eliminates asymptotically all spurious jumps. Monte Carloresults show that it performs also well in nite samples. Our empiricalinvestigation of Dow Jones stocks reveals that the spurious detections representup to 50% of the jumps detected initially. After eliminating the spuriousdetections with our method, the average number of jumps amountsto around 40 a year. For the majority of Dow Jones stocks, we do notdetect clustering in time of jumps occurrences. During the three years ofour study, we nd no single cojump a ecting all Dow Jones constituents.However, if we consider industry sectors separately, we observe a number ofcojumps signi cantly larger than if the stocks were independent. Finally,we relate detected jumps to news releases.