Liquidity risk measures for high-frequency trading financial markets

Abstract

This research set out to evaluate the effectiveness of existing liquidity risk measures in the context of high-frequency trading (HFT) and to determine their capacity to predict flash crashes. Various predictive models were employed, including Vector Autoregressive (VAR) models, logistic regression, random forests, and artificial neural networks (ANNs). The results indicated that while traditional models like VAR struggle with the high-frequency dynamics of HFT environments, machine learning models, particularly random forests, offer more robust predictive power. The study found that certain liquidity measures, especially Percent Quoted Spread (PQS) and Quote Slope (QS), are significant predictors of flash crashes, reinforcing the need for real-time monitoring and advanced predictive models in managing liquidity risk.