Riding Wavelets: A New Method to Classify Financial Price Jumps
(Published by JP Bouchaud | May 2024)
Why do stock prices "jump"? And why do they jump sooo often? At the one minute level there is indeed roughly one 4-sigma event per stock every 2 days!
As first observed for daily price jumps in a classic paper by Cutler, Poterba and Larry Summers, most of these jumps seem be endogenous, i.e. are not related to any significant new piece of information.
This is of course in blatant contradiction with the efficient market hypothesis, for which at least large price jumps should reflect new information. But out-of-the-blue events are in line with Shiller's well known observation that prices are much too volatile to be fully explained by “fundamentals”. We have argued many times that the origin of such endogenous jumps is the intrinsic fragility of financial markets, shared with many other socio-technical systems, like firm networks or train networks (on this point, stay tuned for an upcoming paper in Nature with Debabrata Panja José Morán Frank P. Pijpers Utz Weitzel)
In a new paper, Cecilia Aubrun Rudy Morel Michael Benzaquen and myself have introduced an unsupervised classification framework that leverages a multi-scale wavelet representation of time-series and apply it to stock price jumps.
In line with a previous paper with Riccardo Marcaccioli, we recover the fact that the time-asymmetry of volatility is a major feature that separates rare exogenous, news-induced jumps from a majority of endogenously generated jumps. Local mean-reversion and trend are found to be two additional key features, allowing us to identify new classes of jumps.
Using our wavelet-based representation, we investigate the reflexive properties of co-jumps, which occur when multiple stocks experience price jumps within the same minute. Perhaps surprisingly, our results suggest that a significant fraction of co-jumps also result from an endogenous contagion mechanism, and are not induced by major external shocks.
