A Comprehensive Examination of Public and Private Risk Measures for Distinct Commercial Real Estate Asset Classes: Insights from Mixed-Effects Regressions

This paper employs mixed-effects regression models (MRMs) to examine the multifaceted components of public and private commercial real estate (CRE) risk. It does so through two exemplar datasets: public Real Estate Investment Trusts (REITs) and private Commercial Mortgage Backed Security (CMBS) loans. For public REITs, traditional MRMs investigate the effects of (1) dividends, (2) market capitalization, (3) acquisitions, and (4) percentage of unsecured debt on the capitalization rates of varying REIT sectors. For private CMBS loans, logistic MRMs assess the effects of (1) occupancy-rates, (2) debt-rates, and (3) principal on loan default and delinquency rates. For both datasets, these risk metrics are implemented as both fixed and random-effects to investigate their net impacts and degree of property-type heterogeneity; public market analysis also examines within and between-subject effects. Public results show that the effects of two risk metrics, dividends and market capitalization, exhibited disparate impacts on the capitalization rates of different REIT sectors. For example, Empirical Bayes random-effect estimates show that dividends significantly influence hotel capitalization rates, while their significance is not observed in the retail sector. Across sectors, most metrics' within-subject fixed-effect component yielded statistically significant results, except for the percentage of unsecured debt. Finally, the significance of acquisition's between-subject effect offers evidence of a potential correlation between higher mean acquisition rates and lower capitalization rates. Similarly, private market analysis also revealed substantial property-type and region variability in CMBS defaults. However, logistic MRM results indicate that solely the risk metrics' fixed-effect components are significant in predicting CMBS default rates, indicating consistent impacts across sectors. As such, incorporating random-effects for private risk metrics proved unnecessary, yet property-type clustering remained essential. Overall, this paper underscores the nuanced nature of CRE risk assessment, arguing for increased consideration of property-specific dynamics.