Optimal Portfolio Allocations to Real Estate

Abstract

Executive Summary. The relatively low allocations to real estate in U.S. and U.K. institutional portfolios continues to conflict with those suggested by academic studies. Reasons put forward to explain the discrepancy have included the use of historic data, instead of expectations for the future; the use of smoothed real estate return indices, which underestimate the real estate risk/return relationship and differences in the amount of leverage used to finance real estate. Although these factors may provide some explanation for the allocation problem, they do not appear to fully explain the wide variations between theory and practice.

This study adopts a bootstrapping approach to derive ex-ante estimates of risk and correlations between common stocks and real estate for use in a mean-variance analysis. A dynamic optimisation approach is also undertaken by allowing the inputs to the model to vary. The effects that this has on optimal allocations to real estate in a two-asset portfolio are then explored. The impact of differences in portfolio size is also examined. A simulation model is estimated that shows it is possible to justify a wide range of allocations to real estate. The final choice of allocation depends upon the forecasting skill of real estate investors. If, on average, forecasting skill is poor, it can be shown that the low allocations observed in practice can be justified within a traditional risk/return framework. Introducing the possibility of investing in a riskless asset further justifies the current low allocations to real estate.