Dynamics of REIT Returns and Volatility: Analyzing Time-Varying Drivers Using an Explainable Machine Learning Approach

In the current context of heighted market tensions driven by rising interest rates, there is vital interest for both researchers and practitioners to understand the dynamics of Real Estate Investment Trust (REIT) returns and their accompanying uncertainties. To address this concern, we examine the drivers of REIT returns and volatility in a time-varying framework, spanning the modern REIT era (1991 to 2022). Our study is the first to simultaneously forecast both REIT returns and their associated volatility using an artificial neural network. We contribute to the literature by opening the black-box character of neural networks, enabling the identification of individual feature impacts on predictions and their evolution over time.The key focus revolves around understanding how the influence of accounting and macroeconomic variables changes during periods of financial crises compared to non-crisis periods. The results showcase superior predictive capabilities of the neural network compared to conventional regression models. We shed light on the intricate interplay of diverse variables influencing the performance of REITs. Our findings hold implications for investors, policymakers and researchers navigating the complex landscape of real estate investments in a dynamically evolving market environment.