Optimal REIT Portfolio Selection Using Machine Learning

Abstract

Listen

Translate article

This study examines whether machine learning approaches can effectively solve the portfolio selection and optimization problems related to Real Estate Investment Trusts (REITs)-mixed portfolios. It also investigates the impact of different proportions of Equity-REIT (EREIT) and Mortgage-REIT (MREIT) on portfolio returns. This research collects daily data from 157 EREITs and MREITs, totaling 1,283,800 observations, and construct 126 portfolios in the U.S. market, covering the period from January 1, 2000, to July 31, 2022. Using a genetic algorithm (GA), we optimize REITs combined with stocks, bonds, and gold, and compare GA’s performance with the classical mean-variance (MV) method. To assess robustness, authors also perform sub-period testing to examine portfolio performance under adverse economic conditions. The results show that the GA outperforms the MV method in optimizing REIT-mixed portfolios, especially during periods such as the pre-Global Financial Crisis (GFC) and COVID-19. Portfolios with higher allocations to REITs exhibit superior returns, with EREIT outperforming MREIT in enhancing returns and controlling risk. However, MV shows better results in some low-volatility periods, such as during the GFC. This study advances existing literature by applying machine learning techniques—specifically genetic algorithms—to the optimization of REIT-mixed portfolios. The contributions include demonstrating the effectiveness of GA in achieving better risk-adjusted returns compared to traditional methods, providing practical insights for investors on optimizing their portfolios in volatile markets. Additionally, it highlights opportunities for researchers to further explore the integration of heuristic methods in portfolio management, expanding the understanding of portfolio optimization in the context of REITs. Our findings provide practical implications for investors and portfolio managers, demonstrating the benefits of GA-based optimization in the REIT sector, particularly in turbulent market conditions.