Gains from Diversifying into Real Estate: Three Decades of Portfolio Returns Based on the Dynamic Investment Model

Article excerpt

Nils H. Hakansson [**]

Robert R. Grauer [*]

This paper compares the investment policies and returns for portfolios of stocks and bonds with and without up to three categories of real estate. Both domestic and global settings are examined, with and without the possibility of leverage. The portfolios were generated via the dynamic investment model based on the empirical probability assessment approach applied to past (joint) realizations of returns, both with and without correction for "smoothing" in the real estate data series. Our principal findings are: (1) the gains from adding real estate, on a semi-passive (equal-weighted) basis, to portfolios of either U.S. or global financial assets were relatively modest; in contrast, (2) the gains from adding real estate to the universe of U.S. financial assets under an active strategy were rather large (in some cases highly statistically significant), especially for the very risk-averse strategies; (3) the gains from adding U.S. real estate to a universe of global financial assets under an active strategy were mixed, although generally favorable for the highly risk-averse strategies; (4) correcting for second-moment smoothing in the real estate returns series had a relatively small impact for the more risk-tolerant strategies; and (5) there was some evidence that desmoothing resulted in improved probability estimates.

In several previous studies, discrete-time dynamic portfolio theory [1] was applied to the asset allocation problem, in conjunction with the empirical probability assessment approach (EPAA), to implement a set of active investment strategies. In the domestic setting (Grauer and Hakansson 1982, 1985 and 1986) the model was employed to construct and rebalance portfolios composed of U.S. stocks, corporate bonds, government bonds, and a riskfree asset. Borrowing was ruled out in the first article, while margin purchases were permitted in the other two. The third article also included small stocks as a separate investment vehicle. The probability distributions were naively estimated from past realized returns in the Ibbotson and Sinquefield 1926-1984 data base, and both annual and quarterly holding periods were employed from the mid-thirties forward. The results revealed that the gains from active diversification among the major asset categories were substantial, especially for the highly risk-averse strategies. I n some cases, the returns from the active strategies were significantly higher than for fixed-weight rebalancing policies of similar riskiness.

In Grauer and Hakansson (1987) the model was applied to a global environment by including in the universe the four principal U.S. asset categories and up to fourteen non-U.S. equity and bond categories. The results showed that: (1) the gains from including non-U.S. asset classes in the universe were surprisingly large (in some cases statistically significant), especially for the highly risk-averse strategies; (2) the gains from removing the no leverage constraint were more substantial than they were in the absence of non-U.S. securities; and (3) there was strong evidence of market segmentation in that the optimal levels of investment in U.S. securities were mostly zero in the presence of the non-U.S. asset categories.

Finally, Grauer, Hakansson and Shen (1990) examined the asset allocation problem when the universe of risky assets was composed of twelve equal- and value-weighted industry components of the U.S. stock market. The results indicated that the active strategies of the dynamic investment model performed moderately well when applied to the value-weighted industry indices. Furthermore, the majority of the power policies generated statistically significant positive abnormal returns when managing the equal-weighted industry indices over both the full 1934-1986 period and the 1966-1986 subperiod. In the latter time frame, with one exception, the abnormal returns averaged two-thirds of total excess returns. …