"A man should divide his capital into three parts, and invest one-third in land, employ one-third in merchandise, and reserve one-third in ready money." (Bava Metzia, fol. 42, col. 1.) As the Talmud bears proof, real property has been a natural part of asset allocation since time immemorial. Long present in the books of European institutional investors, real estate began to gain acceptance in the US in the 1970s and investors worldwide are now seeking to increase their allocations to this asset class. Real estate assets are traditionally regarded as real and therefore solid investments that offer attractive risk-adjusted returns, significant portfolio diversification benefits, a high and stable income component and long-term inflation protection.

Our objective is to revisit these ideas and recast them in a modern investment context. One central theme will be the inclusion of real estate within a state-of-the-art asset liability management (ALM) framework. Another key area of interest will be to establish the place of real estate in multi-style multi-class portfolios, design integration methods that optimise the risk/return potential of real estate, and propose innovative asset management approaches to move from the real-value conception to an absolute return framework.

All along, we will be using our expertise in illiquid private markets, index construction techniques and style analysis to look at the quality of existing data, discuss design issues with providers and engage institutional investors on uses of indices and benchmarks.

Lastly, we will be devoting our attention to the emerging property derivatives markets, both to look at structural issues and propose risk management techniques and derivatives-based investment strategies.


Real estate in asset liability management

While real estate has unique and appealing characteristics for asset allocation, some of its specificities make it somewhat challenging to correctly integrate it in an ALM framework.

Any allocation exercise demands that the exact contours of asset classes be established first — while most investors now approach real estate as an asset class, there is yet no consensus on what instruments belong to it: at one end of the spectrum there is the risk of overly focusing on the idiosyncrasies of each property and failing to recognise the potential of real estate as a class; at the other end, there is the danger of using an all-encompassing four-quadrant model that dilutes the unique characteristics of real estate equity assets into a composite index of public and private, debt and equity investments. Between these poles, the now classic question of whether listed property securities should qualify as real estate requires an urgent answer against the backdrop of ballooning capitalisations and the generalisation of the REIT model.

This debate on the contours should not obscure the fact that, unlike hedge funds, which are a disparate collection of strategies, real estate constitutes an ALM class just like bonds and equities. As with the latter, diversity within the real estate class lends itself to analysis based on styles and should be of great relevance at the AM stage.

Once the contours of real estate have been firmed up, we need to define a model of long-term returns for the class and calibrate it. Whereas complex multivariate approaches may be required to optimise AM, stochastic models for ALM demand parsimony and stability in factors. While extracting model parameters is never easy, real estate data poses specific challenges as it is both scarce and noisy. Private real estate markets are decentralised by nature, and the emergence of information providers maintaining property databases and compiling indices is a recent phenomenon in all but a few countries. Due to infrequent trading, leading commercial property indices are based on surveyors' intermittent and subjective valuations which results in late availability, smoothing, lagging and seasonality in data. Private and securitised real estate market data are other possible sources of information. While transaction data has traditionally not been accessible in the opaque private commercial real estate sector, the past year stands out as a watershed with the birth of the MIT Transactions-Based Index (TBI) of US commercial real estate. Although most of the public markets for securitised real estate are too immature to offer much help in the long term modelling of returns, recent research brings new hope about the possibility of recovering information on the direct market from the stock prices of leveraged property companies. Two decades of academic research have documented the issues associated with the structure of the private and public property markets, as well as the biases of the various index construction methodologies. Remedial procedures have been developed and data challenges are no major impediments for the stochastic modelling of long-term real estate returns required for ALM. After the stochastic models for assets and liabilities have been specified and calibrated, the strategic allocation to real estate and other classes is the output of the optimisation that corresponds to the investor's overall objective, eg minimise shortfall risk or optimise the surplus for a pension fund.

EDHEC has developed expertise in dynamic asset allocation strategies for ALM and refined ALM tools to incorporate the impact of parameter and model risk into modelling. These advances are particularly interesting when revisiting the place of real estate in asset allocation given the cyclical nature of real estate markets and the documented lack of normality of property returns. By addressing the contours and data issues in order to build a satisfying model for long-term property returns and by adapting our ALM approach to the specificities of the class, we wish to foster more robust allocations.

We may also want to look at the place of real estate in LDI. Given their (disputed) inflation-hedging characteristics and linkages to demographic factors, real estate may have a natural place in liability-matching portfolios used for immunising defined benefit pension schemes against inflation or longevity risks. At the same time, real estate could enter performance portfolios due to its attractive risk/return characteristics and relatively low correlation to other classes.


Real estate in asset management

Once the ALM exercise has produced a strategic allocation to real estate, the asset management perspective takes the lead with the objective of optimising its contribution to the portfolio. At the heart of AM expertise is the portfolio construction process which aims at optimal asset allocation for maximum diversification benefits. Diversification can be optimised within real estate holdings and at the global portfolio level.

Within the real-estate portfolio, diversification encompasses the elimination of unwanted asset/property specific risk — a difficult task given the heterogeneity, large minimum commitments, non-divisibility and high transaction costs of property investments, but an important one as real estate risk is overwhelmingly idiosyncratic. In theory, listed property securities could be used to establish a diversified and representative exposure to real estate or to diversify the risk from direct holdings.

A scientific approach to diversification presupposes the ability to identify and quantify the miscellaneous sources of risk affecting real estate assets across vehicles, sectors and regions. Risk factors, risk premia and betas must be specified for AM models to become operative.

For investment strategy formulation and implementation, heterogeneous real estate assets have historically been approached by ad hoc groupings based on localisation and property sector. More recently, classifications according to economic rather than geographic regions have appeared and investment style classifications are now emerging.

These classifications must be examined to determine to what extent they correspond to different fundamentals: can they be used as a basis for diversification or are new groupings/models needed? Recent research suggests that property sector distinctions are meaningful but questions the relevance of regional classifications at the country level. Early work on the equity style boxes in real estate concludes that style definitions are too vague to provide much help and stresses the need for a quantitative multi-factor approach to define boundaries. EDHEC has advanced style analysis in the mutual and hedge funds markets and hopes to play a similar role in real estate.

Once investment-relevant classifications have been established based on risk factors, diversification can be optimised within real estate holdings as well as at the global portfolio level.

In a multi-class multi-style framework, two options are available:

■ construct optimal portfolios within each class and then optimally combine them;

■ conduct one-step portfolio-wide optimisation based on the relevant styles identified across all classes. When AM follows ALM, the multi-class portfolio of the first approach combines the optimised class portfolios according to ALM generated optimal weightings, while the alternative one-step multi-style optimisation procedure is carried out under the same constraints of class allocation. While the first approach may be theoretically suboptimal, it facilitates both active management by class specialists and passive approaches using class benchmarks.

In contrast with ALM, the lack of availability of long-time series is less of a problem for AM exercises, but quality and frequency of data become central as models crucially depend on estimates of return, beta moments and co-moments. Since private real estate is illiquid and returns in the property markets have been solidly established as skewed and leptokurtic, standard mean-variance analysis cannot be used and the tools which EDHEC has developed to deal with the illiquidity and non-normality of hedge funds are invaluable in optimising the risk/return characteristics of real estate investment programmes and mixed portfolios.

When managing portfolios that include real estate, one must deal with ‘classic' but exacerbated sample problems, as well as tackle specific data challenges. To improve estimates from small size and dubious quality samples, the respective merits of resampling and Bayesian approaches should be tested. Our tools should be refined to include dimensions of real estate risk whether reflected in prices or not. One aspect of liquidity risk specific to real estate and not reflected in prices is that transaction volumes dwindle, and time on the market lengthens in bear markets. In such conditions, the sample bias present in transaction price indices is exacerbated. Key academic figures have introduced the notion of constant-liquidity indices and the launch of the TBI represents its first major industrial application — exploring and furthering this development are other avenues for research.


Property derivatives and risk management

Derivatives hold great promise for real estate investment and risk management, as they allow investors to take on, reduce and hedge exposure to the overall market or specific property sectors in a time-efficient and cost-effective way. While these are standard benefits of derivatives, the illiquid nature and high transaction costs of real estate investments make property derivatives particularly appealing.

As an investment tool, derivatives not only ease asset allocation, but also allow entry into markets that were previously inaccessible or impractical due to large minimum commitments or lack of expertise. An often overlooked feature of property derivatives is that they can provide portfolio diversification in a class where the large size, lack of divisibility and very high idiosyncratic risk of individual assets make it extremely costly to diversify using the traditional approach.

From a risk management perspective, derivatives allow one to hedge the systematic component of real estate risk, thereby enabling managers to insulate their portfolios from wider market movements. Neutralising market risk allows alpha to be extracted and absolute return products to be structured around real estate.

In Europe, the UK has taken the lead and, with commercial real estate index swaps on IPD indices, boasts a nascent but growing property derivatives market. Market development is hampered by the fact that the underlying IPD indices, as appraisal based measures of private commercial real estate performance, are not tradable. Consequently, these hedging instruments are hard to hedge in a reliable, easy, swift and cost-efficient manner. As a result, intermediaries mostly assume broker roles matching up buyers and sellers.

Issues to be addressed are the merits, feasibility and implementation methods of cash and synthetic hedging. Cash hedging of derivatives written on non-investable indices would rely on direct property and real estate fund investments to manage the derivatives exposure of underwriters, while synthetic hedging would take a risk factor replication approach to create portfolios of assets with high correlation to property. It is perfectly feasible to derive general results on the nature of the optimal trading strategy involving the hedging proxy as well as an estimation of the tracking error introduced by the use of an imperfect substitute of the underlying.

FTSE and MSS may have opened a new route in 2006 by offering indices which they claim are representative of direct property investments and investable through funds with daily NAVs. These players hope to attract investors interested in a (largely) passive exposure to the UK commercial property market, but more importantly, position themselves as the platform of choice for investment banks that wish to offer derivatives and real estate structured products. Work is required to establish to what extent these hybrid products are sufficiently representative and transparent to constitute an attractive alternative to the wider-based and more established IPD indices. Theoretically, investments in the funds can be used for cash hedging of derivatives underwritten on the associated indices. Such hedging is attractive as it minimises tracking variance; however it does not eliminate the need for other hedging techniques unless an active market for short-selling the underlying emerges. In practice, it will be interesting to find out how a property fund established with a base of UK£100m (€150m) will handle massive inflows and outflows of capital.

EDHEC aims to exert its expertise in derivatives, statistical replication techniques and indices to assess existing property derivatives and suggest new derivatives-based investment strategies.