Abstract
During the 1990s and early 2000s an international body of literature explored property portfolio diversification strategies, motivated by the view that reliance on administrative regions to represent asset classes is suboptimum because they are based on historic and governmental factors rather than market fundamentals. Thus, individual ‘asset classes’ may contain highly heterogeneous areas, violating the basic tenets of investment theory. Asset classes based on alternative market groupings were proposed and are re-evaluated here. A sample of 73 local markets is analysed covering 1998–2007, with temporal stability additionally explored. Comparative portfolio performance opportunities are assessed using efficient frontiers, with the alternative market groupings most often found to offer superior performance to traditional strategies. Further than this, the optimal weightings suggested by the classifications are compared to observed aggregate institutional investment weightings. The differences in allocation are found to be considerable. Subsequent hypothetical benchmark portfolios, constructed using observed allocations, are found to statistically significantly violate the mean variance criterion in volatile market phases. Explanations are proposed, drawing on the investment characteristics of property and consequences for portfolio management, as well as an exploration of behavioural factors to include benchmarking. A re-visiting of investment theory and practical strategies is urged.
Acknowledgements
The author wishes to thank Professor Craig Watkins, Department of Town and Regional Planning, University of Sheffield, during the initial inception stage of the study and also Dr Ed Trevillion, Head of Real Estate Research for Scottish Widows Investment Partnership (SWIP) for his helpful and encouraging comments at the 2011 ERES conference. A debt of gratitude goes to Chris Adcock, Professor of Financial Econometrics, University of Sheffield, for his time and help with the bootstrapping analysis.
Notes
1. Jackson and White’s (2005a, 2005b) data series end in 2000 but, for consistency with the other studies, and to enable a full 10-year study period, the current data begins in 1998, thus representing a three-year overlap.
2. In the tables, presentation of the details used to determine the curve of the efficient frontiers has been reduced. This aids interpretation and allows greater clarity in the patterns evident, without the reader becoming lost in the minutiae of the repetitive calculations.