This study examines the impact of location-based risk factors, specifically, local real estate market risk and geographic diversification, on the bond risk premia of US Real Estate Investment Trusts (REITs). It addresses a gap in the literature on how locational factors influence REIT debt pricing.
We quantify local real estate market risk via a REIT-specific local beta, capturing sensitivity of local real estate markets to national real estate shocks. Geographic diversification is measured using the Herfindahl–Hirschman Index (HHI). Using over 30,000 monthly REIT bond yield spreads from 2010 to 2023, we combine machine learning (ML) methods (artificial neural network and accumulated local effect plots) and ordinary least squares regression with polynomial terms.
We find a local real estate market risk premium for REIT bonds, indicating that investors seek compensation for the additional risk. Furthermore, we uncover a non-linear relationship between geographic diversification and bond risk premia. While moderate diversification lowers risk premia, strong diversification increases them. The optimal point of diversification is found at an HHI of 0.25, where REITs face the lowest public debt cost.
REIT managers can reduce borrowing costs by optimising geographic diversification and avoiding overexposure to volatile markets. Bond investors can better price credit risk using REITs' local beta and diversification metrics.
To the best of the authors’ knowledge, this study is the first to show that REIT bondholders price location risk, expanding prior equity-based findings. It uncovers a non-linear relationship between diversification and the cost of public debt. It further highlights how combining ML methods with traditional regression methods can enhance model interpretability and performance.
