On the premise that investment first takes place in June 1977, annual data from the previous ten years was used to compute the variance-covariance matrix and expected returns, with the position rebalanced annually.

Presentation of the Data

The data used to estimate the probabilities of the next period's forecast returns, and to calculate the assets actual return in each period, came from several sources. Constrained by both the frequency and length of real estate data - as detailed in section 2.4.2 - the series of returns runs from June 1967 to June 1994. The UK Long-Term Government bond index was obtained from International Monetary Fund tables.16

The series of returns on the other UK financial asset categories employed were obtained from Datastream International. The risk-free asset was assumed to be a 360-day Local Authority deposit rate maturing at the end of each period. The Financial Times All - Share (“FTA”) index was used as a proxy of the equity market, while the Morgan Stanley Far East Capital Index (“MSCI FE”) was used as a proxy for the Far East equity market. As the purpose was to investigate how real estate influences the asset mix and performance of the optimal portfolio, the inability to invest directly in these indices was ignored.17 As the central factor driving the results will be that of the relative size of the common movement between real estate and the financial markets,the proxy of the market should not a priori affect the general result. This is discussed further in section 5.3.2 and section 4.5.6.

The series of returns for UK commercial real estate for the period 1967-1994 was obtained from Jones Lang Wootton (“JLW”). All returns are expressed in pounds sterling. Total returns represent capital appreciation or depreciation plus net income. In the UK there are several returns series for commercial real estate, although the choice of series seems unimportant as they are highly correlated (Morrel [1991]). In this case, however, the JLW annual index is of particular value due to its historical depth.

The Adjustment of Commercial Real Estate Returns

The majority of studies based on commercial real estate returns use appraisal-based returns.18 As many have discussed, data on real estate return is less accurate than that for financial asset returns; this is discussed further in section 2.4.2.

Allowance for this bias was made by employing the simple exponential unsmoothing procedure in Barkham & Geltner [1994]; the reader is therefore referred to their paper for details. The annual capital returns component of the JLW series was unsmoothed using the following methodology:

Reliance was placed on the annual smoothing parameters, argued for in Barkham & Geltner [1994]. Specifically,α factors of 0.75, 0.625 and 0.5 were used.19 Adjustment was then made to the rental yield series in order to obtain an unsmoothed total returns series.20 Table 3.2 contains the geometric mean and standard deviation of each series.


17The FTA All-Share and MSCI FE are not traded, therefore investment would have to be made in their constituent securities. However, the FT-SE 100 Stock Future Index is traded on the LIFFE, as is the FT-SE Mid 250. The FT-SE 100 represents over 75% of the total market capitalisation; the Mid 250 over 90%.

18Although appraisal-based returns are used, in the majority of recent studies an attempt is made to ‘unsmooth’ the returns series. Examples, include Brueggerman, Chen & Thibodeau [1984], Ibbotson & Siegel [1984], Hartzell, Hekman & Miles [1987] and Liu, Hartzell, Grisson & Greig [1990].

19Barkham & Geltner [1994] regard a=0.625 as the ‘most likely’ case, and the other two parameters as upper and lower bounds of a plausible sensitivity analysis. However, they do not provide any formal statistical confidence intervals on these estimates.

20Note that this is an appraisal-based series which has been adjusted. However, it is clear that much work remains to be done in the area of `unsmoothing' real estate returns. This issue is discussed in section 2.4.2 . Furthermore, as a practical matter the heterogeneous nature of real estate (particularly lot size) restricts the vast majority of investors from holding a diversified real estate portfolio. This may make it harder to match the returns on real estate indices than on indices of equities or bonds; see section 2.4.3.