Evaluating Liquidity Indicators in Predicting Trade Durations and Market Stability: A Case Study of the 2010 Flash Crash

Abstract

This master’s thesis investigates the potential of liquidity indicators to help forecast the trade duration and their impact on the probability of flash crashes in financial markets. Our study uses autoregressive conditional duration (ACD) models for modelling the trade duration and compares the standard ACD with the Log-ACD model. Our analysis is based on high-frequency trading data over the period of the May 2010 flash crash. Our research incorporates the percentage effective spread (PES), the volume-synchronized probability of informed trading (VPIN) and the average depth liquidity indicators into the ACD models. The results indicate that, although these indicators offer some understanding of trade duration, incorporating liquidity indicators into the models did not significantly improve model performances. Our study also reveals that longer trading durations are not systematically correlated with significant price variations, making trading duration alone an impractical predictor of flash crashes. The results suggest that the dynamics leading to extreme market events are probably influenced by a broader set of factors than liquidity and trading duration. This emphasizes the need for more comprehensive models that incorporate additional market variables.