Reinforcement learning and risk preference in equity linked notes markets

Abstract Individuals who follow a reinforcement learning heuristic put too much weight on recent failures or successes in placing their next bets. Using a large sample of equity-linked notes (ELNs) investments in South Korea, we find evidence showing a negative effect of reinforcement learning on future investments that lasts longer than one investment period. After losses, investors are less likely to repurchase equity-linked notes and spend less on their repurchases. This behavior also results in reinforcement learners underperforming rational agents. The difference in returns received by reinforcement and non-reinforcement groups is economically large at approximately 10.7%. However, these negative effects of reinforcement learning are mitigated by investors’ higher risk attitudes. We find that more risk-seeking investors are less likely to shun ELNs after undesirable prior returns and that this effect persists for more than one period. The underperformance of reinforcement learners is also reduced with high risk-taking. Overall, our findings highlight how combining different psychological traits can diagnose and improve biases in investor decision-making.