Identifying High Frequency Trading Activity without Proprietary Data

Abstract

Public databases do not identify high frequency traders (HFTs), so researchers use proxies. We assess the reliability of commonly used proxies using data with HFT identifiers. All proxies are highly correlated and correctly identify HFTs’ activities. However, proxies vary by how well each performs at different tasks. Some are better at capturing variations in HFTs’ liquidity demand versus supply although all proxies respond more strongly to shocks in HFTs’ liquidity supply than demand. Some proxies are better at isolating HFT-specific activities from other algorithmic and non-algorithmic activities. Proxies scaled by trading activity are less effective at signaling true HFT activity than unscaled proxies.

Keywords: High-frequency trading; HFT proxies; Quote intensity; Strategic Runs; Monitoring Intensity; Message traffic; Cancellations; Liquidity demand; Liquidity supply

JEL Classification: G10