Khashanah, Khaldoun, Chen, Jing ![]() ![]() |
Preview |
PDF
- Published Version
Available under License Creative Commons Attribution. Download (1MB) | Preview |
Abstract
Financial jumps have occurred more frequently with the advent of high-frequency trading enabled by technological advancement. Most existing jump detection methods that treat a jump as a singular, random, and isolated shock event were not designed to capture the clustering of jumps related to contagious behaviour, in which the occurrence of jumps increases the probability of further jumps soon after. This paper presents a new method that addresses the challenges of capturing both singular and consecutive jumps. This approach evaluates the size of individual returns with a measure of local volatility based on the median of consecutive absolute returns. We use this method to detect jumps in both S&P 500 and simulated time series, and compare its performance with several classic jump detection methods. Throughout, our consistently outperforms other approaches applied to both real and simulated financial return series. In addition, we demonstrate that the detection results are not biased or compromised by the intraday volatility pattern.
Item Type: | Article |
---|---|
Date Type: | Published Online |
Status: | In Press |
Schools: | Schools > Mathematics |
Publisher: | Royal Statistical Society |
ISSN: | 0035-9254 |
Date of First Compliant Deposit: | 19 March 2025 |
Date of Acceptance: | 17 March 2025 |
Last Modified: | 08 May 2025 13:52 |
URI: | https://orca.cardiff.ac.uk/id/eprint/177005 |
Actions (repository staff only)
![]() |
Edit Item |