Modeling behavioral biases of stock investors

Abstract

naïve reinforcement learning, overconfidence and risk aversion. Naïve reinforcement learning is a simple probable principle for learning behavior in decision problems. The investors who follow the naïve reinforcement heuristics, Naïve Learners, pay more attention to their experiences of actions and payoffs than other factors that are considered by rational investors. Naïve learners are pleased to repeat the actions that was successful and avoid to repeat the investment decision which was painful. I also focuse on two psychological phenomena, overconfidence and risk aversion, to examine the emotional process of evaluating gains and losses. Overconfidence is one of the most documented biases (Daniel and Titman 2000). Investors who are overconfident in their investing abilities are more willing to make risky decisions. Conversely, risk aversion is the tendency of investors to avoid risky choices. To address these two conflicting concepts, overconfidence and risk aversion, I use the reference price as the pivot position for psychological recognition by investors.

I propose three proxies

PNLR (Proxy of Naïve Reinforcement Learning), DOC (Degree of Overconfidence) and DRA (Degree of Risk Aversion). These proxies are estimating the behavioral biases of irrational investors. Furthermore, they can predict future stock returns. The empirical results are economically and statistically significant even after controlling various risk factors such as size, value, profitability, investment pattern, turnover ratio, short-term return, and long-term return.

For psychological and emotional reasons, human beings do not always make decisions rationally. The vagarious nature of human behavior has been studied in psychology, economics and even finance. In the stock market, behavioral biases interrupt the price equilibrium process and cause price momentum.

In my thesis, I concentrate on three behavioral biase