Diversification in Real Estate Portfolios

One of the key research questions in the private commercial real estate market involves the investigation of the amount of specific risk in portfolios of various size. In other words, the number of properties a real estate portfolio needs to diversified, i.e. have no specific risk. Previous studies suggesting that the reduction in specific risk from increasing portfolio size is difficult to achieve and so investors should increase portfolio size almost indefinitely. A conclusion largely based on the linear regression model of Evans and Archer (1968), which infers the amount of specific risk indirectly. This paper therefore re-examine the extent of diversification in private commercial real estate portfolios using two approaches not previous applied in real estate portfolio analysis. First, to overcome the issues with the linear regression approach of Evans and Archer (1968) we use the multivariate curve fitting methodology of Hueng and Yau (2006) to estimate the amount of specific risk in portfolios at each portfolio size directly and find the point at which the specific risk is zero. Second, to avoid timing biases due to changes in the risk/return performance resulting from a fixed holding-periods we follow Chong and Philips (2013) and use randomise start dates with 5-year and 10-year holding periods. Using quarterly returns over the period 2000:1 to 2020:4 for 77 ‘property assets’ and simulation, without replacement, the linear model of Evans and Archer (1968) indicates that even after holding all 77 ‘property assets’ specific risk was still not zero. Confirming the results of previous studies. In contrast, using the multivariate curve fitting approach of Hueng and Yau (2006) the results show that investors can reduce about 95% of the specific risk with only five or six ‘property assets’ and that specific risk is zero at the 11 ‘property asset’ portfolio level. Thus, we reject the idea that investors in private commercial real estate should continue to increase their portfolio almost indefinitely.