Cross-Section of Mini Flash Crashes and Their Detection by a State-Space Approach

Abstract

We define mini flash crashes as significant deviation from a normal price process, and formulate a robust and simple state-space model for efficient detection of mini flash crashes. Compared to existing methodologies, we demonstrate that our method can capture a much broader range of mini flash crashes. As one would expect, mini flash crashes spike when there are major information events in the market. Our research shows that stocks with larger market capitalization, smaller volatility, higher averaged price, and larger liquidity are consistently more robust against these extreme price movements. These market microstructure behaviors are consistent with the literature studying the impact of public information arrivals on price discovery. This suggests that mini flash crashes occur because market assimilates new information at an increasingly faster speed, complementing the view that high-speed trading increases information efficiency.

Keywords: market microstructure, mini flash crashes, state-space model, liquidity, high speed trading, information efficiency