Tail-Aware Portfolio Optimization for Listed Real-Estate Securities Under Downside Risk

This paper has presented a combined empirical framework for measuring the risk-return profile of listed real-estate securities in a non-Gaussian market situation. By leveraging daily data for 30 U.S. and international listed real estate securities from 2021 to 2024, we probe how portfolio outcomes vary according to the optimization criterion and distributional aspects of returns obscured by conventional mean-variance summaries. We build long-only and long-short portfolios using classical Markowitz mean–variance optimization methods and conditional value-at-risk (CVaR) optimization techniques, and compare their realized dynamics cumulative growth and efficient frontiers under alternative risk-free benchmarks. By applying extreme value theory, we quantify extreme-risk exposure via generalized Pareto modeling and Hill tail-index estimation, which is compared to the broader behavior of the equity market. We analyze portfolio stability and reward efficiency using volatility, Sharpe, Sortino, and Rachev ratios, as well as maximum drawdown and information ratio. In addition, robust single-factor regressions are estimated on a sector benchmark, and residual diagnostics are analyzed to define common factor dependence while minimizing the impact of outliers. To introduce a forward-looking dimension, we calibrate the double-subordinated normal inverse Gaussian specification and extract NDIG model-implied option prices and volatility surfaces. We also investigate volatility persistence through ARFIMA-GARCH modeling to assess whether listed real-estate security returns exhibit long-memory features beyond standard volatility clustering. Results indicate that listed real-estate security returns exhibit heavy tails, pronounced downside sensitivity, and persistent volatility, supporting the use of tail-aware optimization, robust estimation, and long-memory-consistent volatility diagnostics beyond standard Gaussian benchmarks.