In terms of determinants of the value of a particular property, location (McIntosh & Sykes [1985) is generally considered to be the single most important factor. However,
it should be repeated that an imperfect market in multifarious real estate rights is being dealt with. Changes in law concerning real estate can materially affect the interests of holders of existing real estate rights. Parliament has intervened to alter the balance between parties in the Landlord & Tenant Act 1927 et seq., and in various Housing Acts. A future change of government, or a single piece of legislation, such as the recent amendments to the Land & Property Act 1954 affecting security on assignment, substantially affect real estate.
As the performance of real estate is generally measured using an index of returns, such as those produced by IPD, the results refer to a portfolio of returns of equal composition and weight to those of the index, not individual properties. Due to the high unit cost i.e. indivisibility of individual properties, diversification is difficult for all but the largest funds. Miles & McCue [1984b], MacGregor , inter alios, have argued that to achieve a reasonably diversified portfolio would require a minimum of 20 to 30 properties. Brown [1991b] demonstrated that by holding approximately 30 well balanced properties, it is possible to substantially reduce the unsystemic risk of a property portfolio. Taking the hypothetical portfolio of average properties of equal weighting, the level of risk reduction is 62%-73% dependent upon sector, and approximately 68% at the portfolio level. A similar holding on the ISE would provide a 50% reduction in risk. The high comparative level of real estate portfolio risk reduction is attributed to the low correlation of returns that exist between individual properties. Unfortunately, the hypothetical portfolio is extremely difficult to compile, and the average real estate portfolio tends to be somewhat 'lumpier', with unequal weighting. Furthermore, the majority of real estate portfolios hold less than 30 properties, which infers that they must be poorly diversified. The result is that the performance of the average 'real' portfolio is influenced by real estate-specific factors rather than merely market-wide factors.
Brown [1991a] considered a portfolio to be well diversified if 95% of the variation in returns can be explained by the market, suggesting that it is necessary to hold about 200 properties. He stated:
'... even though levels of diversification of 95% may seem high it should be noted that this means that there could still be about 1% p. a. of specific risk which has yet to be diversified away ... moving from 95% to 99% diversification reduces the specific risk to about 0.5% p. a.'
Table 2.5 suggests that it is easier to produce a highly diversified portfolio with equities than with real estate, even when ignoring the factor of cost.
Brown illustrated that increasing portfolio size increases diversification; by holding some 200 properties specific risk can be contained and tracking the market becomes possible; see table 2.6 below.
Brown's analysis, however, assumed an equal investment in each of the properties of which the portfolio is comprised. This assumption is clearly untenable and, as
highlighted by Morrell , further increases the number of properties required in order to hold a well diversified real estate portfolio.20 Young & Graff [1995, p. 254] took the point one stage further, stating that:
'In real estate portfolios subject to institutional quality standards, the appropriate degree of risk reduction across multiple risk factors (location, economic, etc. ) could only be achieved by purchasing most of the institutional grade properties in the United States -a practical impossibility. ' Brown [1991b] also stated that the market is only responsible for about 10% of the price movement of the average property, compared to 30% by the stock market on an equity. Indeed, a 95% R2 in the stock market can be achieved with a holding of only 44 equities, compared to more than 171 properties. A portfolio of equities will therefore track the market more efficiently.
Brown suggested that diversifying by sector and location within a real estate portfolio is unnecessary to reduce portfolio risk. Geographical diversification within a sector is probably as effective at specific risk reduction as diversification across sectors would be, due to their low correlation. Thus, Brown concluded that investing in different sectors and locations may incur greater expenditure and management costs. However, the process of forecasting diversification to optimise return and reduce risk is full of difficulties. Barber  illustrated this by running a simple mean-variance asset allocation model over different time periods. Highlighting the effect the time period considered, can have on the optimal sector split within a real estate portfolio; see table 2.7.
In the US, Mueller  reconsidered geographical diversification,21 linking real estate performance to local economic conditions by placing cities in economic categories based on their dominant base industry employment type.22 He concluded that diversification across cities with different base industries, is more effective than straightforward geographical diversification.
Also, given the nature of real estate as an investment, it is impractical to reallocate what was an optimal 1979-1989 portfolio to match the optimal 1979-1991 asset allocation. The principal problems are transaction costs - section 2.4.8 - and the unavoidable time lag between the decision to alter a portfolio's weighting and its implementation; see section 2.4.4. This implies that an investor in direct real estate must be prepared to establish a long-term investment position that differs significantly from the accepted normal market portfolio mix. Within this framework, therefore, it could be said that it is safer for fund managers to fail with an asset mix portfolio similar to that of their peers, than to succeed with a strategy which contrasts with them. As discussed in section 2.3.1, the dangers to a fund manager in assuming an abnormal exposure to real estate may outweigh any possible gains.
Therefore, not only may it be harder to replicate real estate indices (and the returns there on) compared to indices of equities or bonds, it could also explain why institutions limit their exposure to real estate. This limitation may not be made on grounds of smoothing, but could be principally due to their inability to hold a diversified portfolio of real estate.23 This point is well illustrated in figure 2.6. However, the difficulties outlined above should not prevent a large institution from investing in direct real estate.
20For example, in order for a fund to allocate 10% of its portfolio to direct real estate investment, it would require a market capitalisation in excess of £600 million; assuming that it holds 200 properties at a market value of exactly. £300,000 each. This greatly limits the number of funds for which diversified direct real estate investment is a viable option; see section 2.3.3.
21An idea first suggested by Hartzell, Shulman & Wurtzebach .
22Mueller defines a base industry as one that exports goods from the local economy, bringing in money which he considers the engine for local economic growth.
23This is in accordance with the 'New Equilibrium Theory' proposed by Ibbotson, Dierrneier &; Siegel , which states that unsystemic risk may have to be priced where it is too difficult to diversify.