Losing from Naive Reinforcement Learning: A Survival Analysis of Individual Repurchase Decisions

Abstract

This paper applies survival analysis to individual trading data from a discount brokerage firm, and documents significant individual-level repurchase bias, investors' tendency to disproportionately repurchase more previously sold winners than losers. Investor sophistication and experience mitigated the bias, but generated asymmetric effects: the most sophisticated/experienced investors' tendency to avoid prior losers were almost completely eliminated, but they were still over twice more likely to repurchase prior winners. Limited attention, chasing past performance and risk-adjusted returns could not justify the asymmetry. This suggests one reason for loss from frequent trading was persistent naive reinforcement learning in repurchasing prior winners.