Diversification in Commercial Real Estate: Realizing Continuous Spatial Correlation

In a mean-variance state investment managers must thoroughly specify portfolio risk to obtain full diversification and select investment options on the efficient frontier. However, diversification within real estate is not as costless as other financial assets such as stocks or bonds. The local nature of real estate implies that investment managers may need to specialize for simplicity or by portfolio objective. To address this trade-off between diversification and specialization this study quantifies the relation between commercial property returns and distance. The returns of four types of commercial properties are modeled as a function of the distance of separation by measuring the covariation between each commercial return. Results are obtained for the effects of diversifying based upon distance and also upon property type, size and distance. The results indicate that correlation for the same type of commercial properties located within the same or adjacent zip codes can be as high as 0.93. After controlling for property type and size, the correlation reduces significantly. However, when residential data are matched by location to the commercial returns and added to the controls for property type and size, the spatial autocorrelation between different property types in the same or adjacent zip codes is 0.89. In all instances, the distance at which the spatial autocorrelation decays to zero is approximately 60 miles. JEL classification: G11, R12