Portfolio optimization with behavioral biases

Abstract

This research aims to understand and quantify the impact of behavioural biases on portfolio performance. Using historical data from U.S. exchange-traded funds (ETFs) representing key sectors, the study employs Monte Carlo simulation to generate simulated returns that reflect underlying assumed market conditions. These simulations feed into two portfolio optimization algorithms. The first algorithm is grounded in Modern Portfolio Theory (MPT) and employs mean-variance optimization to balance risk and return, offering a more traditional view where risk aversion is primarily measured through variance. On the other hand, the second algorithm is built upon Behavioural Portfolio Theory and incorporates key behavioural elements by optimizing the upper or lower percentile returns, thereby capturing the tendencies of real-world investors to prioritize extreme outcomes over average ones. To address the limitation of static risk aversion assumptions in both theories, this study introduces two frameworks: Hidden Markov Model (HMM) and Most Recent Performance (MRP). Findings suggest that higher risk-taking leads to concentrated portfolios which are particularly rewarding during market uptrend. This is especially so if the choice to take on more risk is driven by behavioral bias informed by the performance of the financial market, rather than the broader real economy. This work contributes to behavioral finance literature by integrating shifting risk attitudes into portfolio optimization and advancing the understanding of psychological factors in investment decision-making.