Detecting Informed Trade by Corporate Insiders

Abstract

Detecting informed trade by corporate insiders is costly and is the subject of significant regulatory and market scrutiny. We introduce a mixture model that leverages the cross-section of insiders' past returns to infer which insiders are more likely to engage in informed trade. The estimation explicitly accounts for the noisiness of insiders' performance histories. Out-of-sample returns are higher for stocks traded by insiders identified as more likely to use information, and prices reflect this information faster over the last decade. The model for insiders implies a person-specific mixture distribution that can be used to classify whether any disclosed trade is informed.

Keywords: Insider trading, Informed trade, Mixture model

JEL Classification: G14, G18, K22