Conclusions and Implications for Future Research

The principal motive for this thesis was to seek reason as to the current low weight of direct real estate in life and pension fund portfolios. Real estate as a proportion of total life insurance and pension fund assets has substantially fallen over the past fifteen to twenty years. Although the literature offers many reasons for this decline, these fail to provide sufficient cause to explain the degree by which this has occurred.

Within the framework of conventional investment theory much of the literature persists in the belief in, and promotion of, the long-term diversification benefits offered by direct real estate investment. However, when analysing the benefits of including real estate within a multi-asset portfolio, the majority of studies have failed to consider two important issues, those of the inertia of real estate investment, and an institution's 'natural' tendency to be risk-averse.

Chapter 3 addressed two main issues often cited by fund managers for their low level of real estate investment. One was the extent to which the illiquidity of real estate affects their performance. The second, their questioning of the assumption that transactions occur without cost, another negative effect on performance. Under conventional investment theory placing a constraint on real estate investment reduces portfolio performance. The more risk-averse an investor, the greater this effect. However, on a risk-adjusted basis, a static investment in real estate, of up to 40% of portfolio value, was shown to outperform a portfolio consisting of financial assets alone. When Far East equities were added to the investment universe, the proportion of real estate within a portfolio fell. However, relaxing the constraint on investment to include a more realistic variable constraint provided little further explanation as to the current level of institutional investment in real estate.

A subsidiary point considered in Chapter 3 was the optimal behaviour of treasury managers of companies with a high proportion of assets held in real estate. It concluded that such managers would be better able to diversify company risk by investing cash balances, designated for long-term capital expenditure, in a diversified equity portfolio, or index fund, rather than in more traditional fixed interest investments.

The reduction in institutional investment horizons, coupled with the above mentioned

points, are partly the cause of the current (low) level of life and pension fund investment in real estate. However, when cognisance of their long-term objectives is considered against the background of conventional investment theory, the findings in Chapter 3 agreed with research to date in identifying the substantial long-term benefits offered by direct real estate investment.

The results of this research and the findings in Chapter 3, effectively support the status quo of conventional investment theory as it is applied to real estate investment. This conflicts with the current investment behaviour and practice of life and pension funds, providing limited support for their current low level of investment in real estate. The obvious discrepancy between theory and practice called into question the theoretical foundations upon which these papers are based, and promoted an examination of conventional investment theory.

It was found that the assumptions inherent in the CAPM undermine the practicality of its application in the construction of multi-asset investment portfolios, in which the structure of individual asset classes vary widely. Essentially, having to define and measure an appropriate market proxy for use in the construction of multi-asset portfolios renders the CAPM inapplicable to investment decisions across asset classes as current available benchmarks are unlikely to have any predictive power in the pricing of real estate. The use of the relative β in relation thereto is thus of little, if any, use.

The results obtained from the normality tests, and the existence of regime-switching in value indices of commercial real estate, suggest that they are non-normal. These results further undermine the validity of applying the CAPM to investment decisions that involve real estate. They also call into question the use of unsmoothing real estate value indices with methods that assume a constant relationship between current and past prices.

In addition, traditional investment theory is built upon the assumption that there exists a linear relationship between expected return and variance. The majority of historic work upon this relationship supports the existence of a positive trade-off between risk and return. However, few papers define it as linear. For econometric reasons the majority of tests relied solely on grouped data. Individual asset deviations from linearity may thus cancel out in the formation of portfolios. This disguises the 'true' risk-return relationship, particularly if the underlying process is similar to the linear relationship expected. The risk-return relationship proposed under the DRT model, being linear in part, provides a good example.

As stated, examination of the liability structure of life and pension funds necessarily includes the objective of matching long-term liabilities. However, such an objective does not automatically translate into the traditional definition of risk and the requirement of stability in long-term returns. Diversification as suggested by mean-variance portfolio theory, to reduce the long-term variance of a portfolio - reduces both downside - as well as upside-variance - and may actually hinder the fund's principal objective of matching long-term liabilities.

Alternatively, life and pension funds may benefit from considering the merits of an investment in terms of its effect on downside-variance - limiting the risk of not achieving a target rate-of-return - rather than on total variance.

This led to the proposal of an alternative asset allocation model, to overcome the apparent shortcomings of traditional models, and accommodate the notion of downside risk-aversion in an attempt to provide a more accurate description of investor preferences.

The use of the DRT model as a method of controlling risk may be particularly relevant to life and pension funds. Its approach focuses on returns measured against the principal sources of fund risk. The DRT model offers a framework for integrating alternative assets, such as those with asymmetric payoff distributions, and thus should provide a greater level of risk control. It also promotes the construction of negatively skewed portfolios, weighting the total expected distribution of returns according to an investor's risk preferences, therefore reducing the risk of downside participation. These results are more consistent with life and pension fund objectives. The DRT model also offers a customised approach to risk management by allowing - via the specification of α and κ - an investor to explicitly define risk preferences.

The DRT model offered a statistically significant alternative to traditional asset pricing models, and significantly outperformed the MPI model employed. It also provided an alternative method of measuring performance.

Under the DRT model, and within a national context, the inclusion of real estate within a multi-asset portfolio improved risk-adjusted returns. However, the proportion allocated to real estate was lower than that suggested under traditional asset allocation models. This was mainly due to the DRT model's concern with downside-variance,rather than with total variance. This reduced the importance of an asset's covariance, and consequently the diversification benefits offered by real estate.

However, when the investment horizon was enlarged to include Far East equities, the DRT model severely restricted an investor's allocation to real estate. Foregoing the diversification benefits offered by real estate - reduction in portfolio volatility - in exchange for an increase in the expected return on a portfolio, and hence, probability of achieving the target rate-of-return set.

In addition, the testing of the model was limited to consideration of a portfolio comprised of just four asset classes and a risk-free rate. It did not consider the benefits of real estate investment alongside the plethora of alternative investments currently available to life and pension funds. Theoretically, increasing the spectrum of assets from which investors may form portfolios improves risk-adjusted performance. This was demonstrated by adding Far East equities to an investment universe previously restricted to UK equities, real estate and bonds. By extension, by expanding the investment universe to include not only Far East equities but also e.g. overseas bonds, the proportion allocated to real estate may be even further reduced. Asset classes such as overseas equities also provide diversification benefits and this, when considered against the reduced emphasis placed on variance under the DRT model, increases their relative attraction for life and pension funds.

The underlying factor driving the proposed model - the minimisation of downside variance, the traditional risk factor of volatility being ignored - reduces the attractions of real estate to institutions. The traditional argument in favour of including real estate within life and pension fund portfolios - that of diversification - is thereby weakened.

These results have several implications for theoretical models and empirical research in financial economics. For both multi-period investment models and performance evaluation, the evidence of non-stationarity in real estate value indices implies violations of model specification and estimation procedures. The application of conventional investment theory to asset allocation decisions is also questioned.