Informed Trading Intensity

Abstract

We train a state-of-the-art machine-learning method (ML) on a class of informed trades to develop a new measure of informed trading, the Informed Trading Intensity ("ITI"). Though the measure is trained on a particular class of informed trades, it predicts various informational events, including stock price reactions to earnings surprises, M&A announcements, and unscheduled news releases. The measure also increases on days with opportunistic insider trades and large changes in short interest. Returns on days with high ITI reverse less than returns on other days. In the cross-section, higher ITI is associated with higher returns next month. Our main insight is that learning from data on informed trades can generate an effective measure of informed trading.

Keywords: Informed trading, machine learning, adverse selection, stock returns, intraday data

JEL Classification: G10, G12, G14