Flash Crashes, Jumps and Running Jumps: A New Method for Jump Detection | Hanlon Financial Systems Center

Flash Crashes, Jumps and Running Jumps: A New Method for Jump Detection

Flash Crashes, Jumps and Running Jumps: A New Method for Jump Detection

seminar date: 
Thursday, October 15, 2015 - 5:30pm
seminar location: 
BC122
Alan Hawkes, Swansea University
Abstract: 

While developing a new jump-diffusion process for asset prices in which identifiable price jumps are likely to show a clustering, or contagious effect, we need to identify the number, location and magnitudes of jumps each trading day in a series of S&P500 prices recorded every two minutes.  We used a jump detection method based on the difference between daily Realized Volatility and Bi-power Variation (call this the RV-BV method) proposed by Barndorff-Nielsen and Shephard (2002, 2004) and Andersen et al (2010). 

Surprisingly, we find no indication of jumps occurred on flash-crash day (06 May, 2010) and other similar days.  The fundamental problem of this method is that it cannot tolerate large variations in consecutive intervals, which can cause the bi-power variation to exceed the daily realized volatility, and consequently leads the jump detection to be null.

Our new jump detection method is based on finding individual 2-minute returns that are large compared to a local volatility measure that uses a median approach to avoid the masking effects that often occur with mean realized volatility based measures. 

We further introduce “running jumps” aimed explicitly at studying the occurrence of sequences of large neighboring returns that effectively form a single jump that evolves over a number of successive intervals. 

We test our methods on S&P500 price behavior between 03 January 2006 and 13 March 2015, both intraday and inter-day. We find that our method, in comparison to RV-BV method, robustly captures significant jumps and jump runs on the Mini Flash Crash day and other days when there are market events triggering sharp volatility variations.

 

Bio: 

After obtaining First Class Honours in Mathematics at King's College London, Alan moved to UCL where he completed a PhD on queuing theory applied to road traffic (supervisor M.S. Bartlett), while also lecturing in the Statistics Department. He was then, for five years, Reader (equivalent to US Associate Professor) in Mathematical Statistics at Durham University before becoming full Professor of Statistics at Swansea in 1974 — still there as Honorary Research Professor.

He has served on the Council of the Royal Statistical Society and the International Committee of the Biometric Society and was formerly a member of the New York Academy of Sciences.

In the early 1970s Alan published a series of five articles on self-exciting and mutually-exciting point processes that later came to be known in the literature as Hawkes Processes. These were quickly taken up by people analysing earthquake sequences, but were otherwise largely ignored

Meanwhile, Alan moved on to other applications of applied stochastic modelling. His major scientific achievement, in conjunction with David Colquhoun FRS, Professor of Pharmacology at UCL has been pioneering the development of stochastic models for the dynamics of ion channels in biological membranes.  These are essential links in the communication system that carries messages around the body, and are important sites of drug action. The ideas and techniques that Alan and David developed are used extensively by Pharmacologists, Physiologists and Neuroscientists.

Since about 2000, Hawkes Processes suddenly became very popular and started to be applied to a wide range of phenomena to describe gang warfare in LA, burglaries, genetics, e-marketing, quantum physics etc. as well as undergoing further theoretical development by mathematicians.

Since about 2005, the most active field of application has been in the finance literature and, in 2012, Dr. Maggie Chen persuaded Alan to rekindle his interest in Hawkes Processes, learn something about finance and make a contribution to the research of the quantitative finance group in Swansea. Maggie and Dr Mike Buckle have recently moved to different universities, but our joint work continues, including collaborative projects with members of Stevens Institute.