Do Proxies for Informed Trading Measure Informed Trading? Evidence from Illegal Insider Trades

This paper exploits hand-collected data on illegal insider trades to test whether standard illiquidity measures can detect informed trading. Controlling for unobserved cross-sectional and time-series variation, sampling bias, and strategic timing of insider trades, I find that only absolute order imbalance and the negative autocorrelation of order flows are statistically and economically robust predictors of insider trading. However, this result only holds for short-lived information. When information is long-lived, none of the measures of illiquidity I consider detect informed trading, including bid-ask spreads, Kyle's lambda, and Amihud illiquidity. These results suggest that standard measures of illiquidity have limited applications.