Anomalous waiting times in high-frequency financial data

TL;DR

This paper investigates the non-exponential nature of waiting times between trades in high-frequency financial data, revealing limitations for existing agent-based market models and proposing CTRWs as a suitable framework.

Contribution

It provides empirical evidence that waiting times are non-exponential in high-frequency data, challenging assumptions in traditional market models and suggesting CTRWs for better modeling.

Findings

Waiting times are non-exponential in high-frequency data

Limits are identified for agent-based market models

CTRWs are proposed as a suitable modeling framework

Abstract

In high-frequency financial data not only returns, but also waiting times between consecutive trades are random variables. Therefore, it is possible to apply continuous-time random walks (CTRWs) as phenomenological models of the high-frequency price dynamics. An empirical analysis performed on the 30 DJIA stocks shows that the waiting-time survival probability for high-frequency data is non-exponential. This fact sets limits for agent-based models of financial markets.