Berlin 2024 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
SYEF: Symposium Statistical Physics of Economic and Financial Systems
SYEF 1: Statistical Physics of Economic and Financial Systems
SYEF 1.4: Invited Talk
Thursday, March 21, 2024, 11:15–11:45, H 0105
Statistical-Physics Theory of the Long Memory in Market-Order Flows and its Empirical Validation in the Tokyo Stock Exchange — •Kiyoshi Kanazawa — Kyoto University, Kyoto, Japan
In financial markets, the market-order flow ubiquitously exhibits persistence. In other words, if one observes a buy (sell) market order, it is likely that one observes a buy (sell) market order even in future. This property is called the long-range correlation (LRC) and can be quantitatively characterised by the order-sign auto correlation function (ACF). By writing a buy (sell) at time t as є(t)=+1 (є(t)=−1), the ACF obeys the power-law decay such that C(τ):=E[є(t) є(t+τ)]∝ c0τ−γ with the prefactor c0>0 and the exponet γ∈ (0,1).
In this talk, we study the microscopic origin of the LRC both theoretically and data-analytically. We first report on the theory of a generalised Lillo-Mike-Farmer (LMF) model as the microscopic statistical-physics model for the market-order flow. The LMF model was a theoretical model proposed by Lillo, Mike, and Farmer in 2005 based on the order-splitting hypothesis by assuming the homogeneity of trading strategies. We propose its generalised version by incorporating the heterogeneity of trading strategies and derive the theoretical formula predicting c0 and γ. We then validate the theoretical predictions of the LMF model by studying the microscopic dataset in the Tokyo Stock Exchange.
Keywords: econophysics; data analysis; market-order flow modelling; long-range correlation; long memory