Given the illiquidity of private real estate, investors are increasingly recognising the importance of REITs. However, for a meaningful REIT allocation investors must understand the composition of their multi-asset portfolios. Stephen Lee reports

Many studies worldwide indicate that the allocation to private real estate in the mixed-asset portfolio, suggested by mean-variance portfolio analysis, is far in excess of that currently seen in most investor portfolios.

However, while portfolio analysis has always tended to favour private real estate, as an asset class it suffers from several major deficiencies relative to public markets. Chief among these are illiquidity, poor transparency and higher transaction costs, hence the interest of investors in real estate investment trusts (REITs) as a way of introducing real estate into the mixed-asset portfolio.

REITs offer investors a tax-efficient vehicle with a high-income yield similar to that of private real estate, with the advantage of greater liquidity. The continued growth of REIT structures across the world has created an opportunity to re-examine how and to what magnitude public and private real estate fits into institutional mixed-asset portfolios.

The starting point for most consultants and plan sponsors in setting their asset allocation is mean-variance optimisation by which the expected risk, returns and correlations of the asset classes under consideration are used to determine the optimum allocation that offers the greatest level of return for each level of risk.

Such studies typically indicate a significant allocation to private real estate due to its competitive returns, low volatility and low correlation with stocks and bonds. In fact, asset allocation studies using mean-variance optimisation techniques tend to produce unrealistically high allocations to private real estate. As a result, most investors discount or reject the results from such studies on the basis of weaknesses in the private real estate data.

Indeed, it has long been recognised that the low volatility displayed by private real estate data is an illusion because the private real estate indices are primarily appraisal-based rather than transaction-based (ie, asset values are based on relatively infrequent appraisals rather than actual transactions). The effect of this is that the appraisal-based private real estate indices understate the true volatility of real estate investments. This ‘appraisal-bias' inherent in the real estate indices also undermines the reliability of the correlation estimates of private real estate with the alternative asset classes, since the indices fails to capture the true movements of asset values.

Any appraisal-bias or ‘appraisal-smoothing' is detected by examining the correlation of the current returns of an asset class with its lagged returns, ie, its serial correlation. For example, the serial correlation of transaction-based stock market returns is essentially zero, whereas the serial correlation of private real estate is significantly positive.

Hence, to overcome the shortcomings identified in private real estate data it is now usual to de-smooth the real estate data by reducing the serial correlation in returns, which increases the volatility of private real estate and its correlation with the other asset classes and so reduces its attractiveness in mean-variance optimisation.

Public real estate index data, in contrast, do not suffer from the same appraisal-bias issues that plague private real estate data. However, despite years of debate, there is still no industry consensus as to whether REITs are real estate or stocks. In truth, they are probably a mixture of the two. REIT shares clearly are stocks since they trade in the public markets, and REITs themselves are listed on most major stock market indices.

But REITs are also real estate companies deriving most of their revenues from the ownership and management of property. Consequently, REITs share prices are determined by both property market fundamentals and stock market effects. Table 1 brings this point home by examining the returns from 1972-2007 for a number of US asset classes: large-capital stocks (LCS), long-term government bonds (LTGB), private real estate, both appraisal-based (PRE) and de-smoothed (DPRE); and REITs.

As shown, the historical correlations between private real estate, as measured by the NCREIF Property Index (NPI) and the other assets, stocks, bonds, and REITs are very low. The volatility of private real estate returns is also extremely low. Indeed, since 1972, the average annual volatility of the private real estate appears to have been only half that of long-term US government bonds.

However, the serial correlation of the appraisal-based data is significantly positive at 0.80, which indicates that the private real estate data suffer from severe appraisal-bias. In contrast, the serial correlation of stocks and REITs is essentially zero.

Once the appraisal-bias is addressed, by reducing the serial correlation in the private real estate data, the volatility of the de-smoothed real estate data is substantially increased, without affecting average returns. That is, since 1972 the de-smoothing process indicates that the ‘true' volatility of the NPI is probably more that three times that suggested by the appraisal-based series.

Nonetheless, the correlation of de-smoothed private real estate returns with stocks and bonds shows little change. This implies that private real estate, even after adjustment for the appraisal-bias in the original data, is still an attractive diversifier in the mixed-asset portfolio.

The volatility of REITs is considerably greater than that for private real estate. Nonetheless, REITs still offer relatively low correlations with stocks and bonds (0.50 and 0.27, respectively). In addition, the correlation of REITs with the private real estate data is zero, which adds weight to the argument that REITs behave more like stocks than real estate, at least in the short term. All of which implies that public and private real estate (appraisal-based or de-smoothed) may both have a place in the mixed-asset portfolio. Indeed, the results of the mean-variance optimisations shown in table 2 confirm these impressions.

Table 2 shows the optimum mixed-asset portfolio allocations over the historical period for the private real estate and public real estate data, individually and together, by maximising the mixed-asset portfolios' Sharpe ratio, ie, (portfolio return-T-bill rate)/portfolio standard deviation. Two things are evident in the analysis.

First, on an individual basis the optimal portfolio in all instances includes a substantial allocation to real estate, the allocation highest for the appraisal-based private real estate data and lowest for REITs 73.2% and 17.5%, respectively. Second, and more important, the historical analysis shows that the optimal mixed-asset portfolio could have contained both public and private real estate. In other words, it appears that any form of real estate has a place in the mixed-asset portfolio.

The results in table 2 are typical of most studies of the case for public and private real estate in the mixed-asset portfolios, suggesting an allocation of real estate far in excess of that seen in most investor portfolios. However, the results suffer from at least two major shortcomings:

A limited set of alternative asset classes; Short holding periods.
Most portfolio studies, including the results in table 2, have usually examined the case for real estate against few alternative asset classes; typically large cap stocks and long-term government bonds, whereas investors hold a much larger number of asset classes, especially different types of domestic stocks, such as small-cap stocks and/or value and growth stocks.

This has particular relevance for REITs as studies indicate that REITs are relatively small companies with a high income component, which suggests that REITs can be classified as a small-cap value stock. Indeed, the correlation of REITs with small-cap stocks over the historical period from 1972 to 2007 is 0.78. In addition, over the same period small-cap stocks showed an average return higher than that of REITs (16.5%) with much the same volatility (22.5%). Hence, if investors are already holding small-cap stocks in their mixed-asset portfolio the addition of REITs may be superfluous to the optimal allocation. Indeed the results in tables 3 provide confirmation of this supposition.

Table 3 shows the optimum allocations over the historical period and indicates that the simple addition of small-cap stocks to the alternative asset classes leads to the elimination of REITs from the optimum portfolio. Further analysis (not reported) using large-cap and small-cap value and growth stocks simply reinforces the view that REITs are small-cap value stocks and so are redundant in a portfolio already containing such asset classes.

Table 3, however, still shows an extremely large allocation to private real estate. Nonetheless, if further asset classes are considered, such as international stocks, the allocation to private real estate shows a decline but only to a level still far in excess of that seen in institutional portfolios. This implies that while institutional investors may recognise that private real estate is an attractive addition to the mixed-asset portfolio, such investors still avoid private real estate because of its perceived deficiencies, especially illiquidity.

The second problem with the simple case for real estate in the mixed-asset portfolio is the use of relatively short holding periods (annual in this case); whereas the typical holding period for real estate is upwards of 10 years. Again this is of importance for the case for REITs in the mixed-asset portfolio as previous work indicates that over the long term the risk-return characteristics of REITs equate to those of private real estate, suggesting that REITs are ultimately driven by property market fundamentals, once the short-term impacts of the stock market are removed. This implies that a mixed-asset portfolio, which includes REITs and is held for 10 years or more, could benefit from the diversification advantages of private real estate without the disadvantage of

However, this assumes that while the correlation of REITs with private real estate increases with the holding period, its correlation with small-cap stocks declines. Unfortunately this proves not to be the case - figure 1 shows the correlation coefficients of REITs with the other asset classes over holding periods from one to 10 years.
The results in figure 1 show that the correlation of REITs with private real estate increases with the holding period, which supports the view that in the long term REITs can be considered as a real estate investment.

Additionally, the correlation coefficients of REITs with large-cap stocks and long-term government bonds decline significantly with the increase in the holding period, reinforcing the case for REITs in the mixed-asset portfolio in the long term.
But more importantly, the results in figure 1 show that the correlation of REITs with small-cap stocks does not change with increases in the holding period and stays at about 0.80.

Unfortunatly, this implies that REIT returns still behave like small-cap stocks, even in the long term. Thus, in any mixed-asset portfolio that already as an allocation to small-cap stocks, the inclusion of REITs will add nothing to portfolio performance.

When the case for real estate in the mixed-asset portfolio is examined by modern portfolio theory, private real estate has risk and return characteristics that make it extremely attractive as a portfolio diversifier.

Yet many investors have avoided private real estate as an asset class. Initially the case against private real estate was based on the poor quality of the index data. But the use of de-smoothed data, and more recently the development of transaction-based indices in the US, suggests that private real estate should have a prominent place in the optimum portfolio. This has led to investors questioning the case for private real estate due to its lack of liquidity.

Investors have therefore looked to public real estate securities as a panacea, with REITs seemingly offering all of the portfolio diversification advantages of private real estate without the disadvantages of illiquidity. Indeed, a simple portfolio optimisation would suggest that both public and private real estate have a place in the mixed-asset portfolio.

However, simple adjustments to the number of alternative asset classes considered and changes in the length of the holding period show that although private real estate is always part of the optimum mixed-asset portfolio, REITs are always eliminated. These results are unequivocal in showing that private real estate is simply too beneficial to the mixed-asset portfolio for investors to ignore, so long as they are large enough to purchase sufficient assets to hold a well-diversified property portfolio.

In contrast, for institutional investors with long-term holding periods, but who cannot afford to purchase private real estate, REITs may provide access to real estate, as long as the investment is undertaken carefully regarding which assets in their existing portfolio REITs will either reduce or replace.

Stephen Lee is a professor of real estate finance at the Cass Business School, City University, London