This new series continues as Joseph Hamilton and Pretty Sagoo discuss the importance of real estate in the modern portfolio
It is somewhat ironic that direct real estate is still considered as an ‘alternative asset' by the global pensions industry, as property is arguably one of the oldest asset classes of all.
The evolution of pension scheme asset allocation in the UK has, as is the case globally, tended to focus on fixed-income, given its perceived low risk and matching characteristics with long-term pension scheme liabilities, as well as equities, with their long-term equity risk premium. The major exception to this was a period in the 1970s when disaffection with equities and concerns over fixed-income investing in an era of high inflation led to a sharp rise in property allocations, to 20% or so for many pension fund investors. Since the end of the 1970s, property allocations have tended to be reduced and this is particularly ironic given the strong performance of property over the past 15 years and, more fundamentally, the significant improvements in the transparency and liquidity of the asset class as well as the options for investment.
Any comprehensive evaluation of the relative merits of an asset class should recognise both its asset and liability side characteristics. In this short article we examine the merits of investment in real estate, from both a volatility reduction and return enhancement perspective. Moreover, we overlay a "real world" view of real estate markets in the wake of the recent credit crisis by modifying our assumptions to reflect our outlook for real estate markets.
A second issue that must be considered is the relationship between the two main forms of property investment, these being direct investment in property, and investment in real estate securities. These issues of course, form the basis of a debate that goes far beyond the scope of this short article. However, we endeavour to examine both as far as possible.
Philosophically, we believe the projected liabilities of a pension scheme should form the starting point for its portfolio construction process. Liabilities demonstrate exposure to both interest rates and inflation - does property? We examine this question to ascertain whether property as an asset class might demonstrate valuable interest rate or inflation hedging characteristics that might prove useful in the mitigation of the unrewarded risks to which a pension scheme is exposed.
Property as an inflation hedge
So to what extent can property, in a pensions fund's asset portfolio, be considered as an inflation or interest rate (duration providing) asset? Clearly, direct property investments have a duration, and economic theory would suggest some inflation linkage. However as the global pensions industry moves towards a more precise quantification of these sensitivities - the issue of how closely property investments replicate the inflation measures against which pension liabilities are indexed - gains importance.
We answer this question by employing the statistical tools at our disposal. As the liabilities in question are UK domiciled we examine the relationship between UK property and inflation, where again economic theory should suggest the strongest linkage. At this juncture we draw a necessary distinction between liability management issues and asset allocation. For liability management we restrict our analysis to UK property whereas for asset allocation the full global property investment set is considered.
Perhaps the first most obvious test to apply to the property asset class as a hedge for inflation-linked liabilities is good old fashioned correlation. We investigated both the correlation of real estate investment with realised inflation, and performed a regression analysis on this relationship.
Unfortunately, given the absence of high quality performance data for the direct real
estate market, our enhanced analysis has been confined to investment in property securities. We have tested statistically that such an analysis is sound.
We begin by examining the rolling correlation of realised inflation versus the returns of direct real estate and property security investments. An 18-month period rolling correlation is shown. Other roll periods (12 months and 24 months) were also examined, with all showing the same relative instability. This suggests that the link between inflation and property is not always positive and certainly not stable, hardly a desirable characteristic in a hedging asset. However, as might be expected, direct property is more consistently positively correlated with realised inflation, though again the relationship is unstable, with periods occurring where the two series diverge.
We test the relationship between the variables further through regression.
Regression analysis
Using historical data for real estate, gilts, FTSE and realised inflation indices from November 1989 to July 2007 we attempt, by means of a regression, to replicate UK real estate returns as follows:
UK Real Estate Returns (Direct or Shares) = A x Eq + B x Fi + C x CPIR + ε
Where:
EQ = Equity returns on the FTSE all Share index,
Fi = Gilt Returns and
CPIR = realised year on year inflation. A, B and C are the coefficients that we deduce from the regression and ε is a residual.
The main issue with running the static regression of time series data as described above is nonstationarity of the variables used. If any of the variables are non stationary, then the regression results are invalid. For the times series of EPRA UK Shares total returns there are no such issues. Using the Augmented Dickey Fuller (ADF) test, we can reject nonstationarity to 1% significance level for each variable in the regression. The same is not true, however, for a regression of UK IPD total returns. A potential reason for this is the element of smoothing that occurs in the construction of the index. The index, for this reason, also exhibits very low volatility. As a result of these data issues we focus predominantly on a regression of real estate shares versus equity, bond and inflation returns.
After experimenting with a constant and different lags, we find the regression with the statistics shown in figure 1 to be of best fit.
The validity of the regression is further tested using Durbin-Watson statistic, to test whether residuals are serially correlated. The value of Durbin Watson is 1.967, indicating that serial correlation is not of concern. The ADF and Durbin Watson tests together make us very confident about the validity of our regression results.
As the constant is removed, we cannot use the R2 measure. Instead we have to look at T-statistics, which are calculated by adjusting the coefficient by its standard error. The resulting probabilities show us strength of relationships. The widespread convention is to accept the coefficient as different from 0 only if the probability is smaller than 0.05. Following this, we can only say that FTSE and gilts have some relationship with EPRA UK returns. Even looking at our best regression, we cannot say that there is any relationship between realised RPI and property shares. Thus the clearest conclusion we can draw from regression analysis is that property shares returns are uncorrelated with movements in inflation rate. This implies that while there are periods where property returns have outpaced inflation, the statistical relationship between the two is weak. From the perspective of a UK pension scheme, therefore, liability inflation hedging would be more efficiently executed through the inflation swap market - where inflation exposure and lags can be more precisely replicated. Obviously, in markets where inflation-linked assets are scarce, or hedging is not possible through the derivative markets, a combination of direct property, along with infrastructure and commodities, might well offer the most effective hedge available to a scheme. By disentangling these two issues of liability management and asset allocation, property investment reduces to an asset side optimisation question.
Real estate versus other asset classes
We now turn our attention to the relationship between property and the other most common holdings in a pension fund portfolio - both for the UK and globally.
Figure 2 shows a scatter plot of monthly real estate shares returns (horizontal axis) versus monthly returns from the FTSE, UK all gilts and inflation indices (vertical axis) from data since October 1989. There is no clear relationship between the returns of EPRA shares and FTSE shares or gilt returns and delving further into the statistics demonstrates this further. Property shares should be expected to perform well as a diversifier versus the two main components of a UK pension funds asset portfolio.
We can perform a basic comparison of property performance across broad asset classes through looking at Sharpe ratios (risk premium divided by volatility). In Tables 2 and 3 we show returns, volatilities and Sharpe ratios (where risk free is defined by the government bond returns shown) across both the UK and globally for property, equities, gilts and realised inflation over the last five years. These show that both private and public property markets have convincingly outperformed gilts but versus equities, the results are mixed. Futhermore, it appears that direct global property is an excellent component of an asset portfolio - the extent to which this is affected by artificial smoothing is again an issue.
The longer-term performance of property investments has been dealt with in numerous papers and falls outside the scope of this article. However, even a cursory review of the above merits further investigation.
Example of property inclusion in a UK Pension Scheme Asset Liability Analysis
As with all defined benefit investment strategies, we consider property investment relative to the underlying liabilities the scheme is required to fund. We have done this for a £2bn (€2.9bn) UK pension scheme (Figures 3 and 4). We have set the starting asset allocation (Figure 3) at 60% Equities / 38% Fixed-Income and 2% cash. We are interested in the potential benefits of making an allocation to property - (basket of securities and direct investment) in increments of 5%. We then stochastically project the future path of both the assets and the liabilities, and examine the solvency position in 1,3,5,7 and 10 years' time. The assumptions we have used in the stochastic simulation are shown in Table 4. Any stochastic analysis is only as strong as the assumptions on which it is based. Volatilities and correlation are based on a history of annual returns for the indices shown over the period December 1989 to December 2006. For direct real estate we assume the volatility is an average of the volatility of gilt and equity returns due to inherent smoothing issue with IPD data as previously discussed. Risk premiums for the global real estate basket are derived from estimates by RREEF Alternative Investments - and reflect our current outlook for property markets. We assume a small positive correlation between inflation and property in the asset portfolio to reflect our view that direct property is a positively correlated asset with inflation - albeit an imperfect one. Our model simulates the fat-tails associated with real-world asset markets and volatility is modelled stochastically.
Static risks
We begin by summarising the static risks in the net asset liability position of the fund through a simple scenario analysis using heat maps (Figures 5 and 6). These demonstrate solvency exposure to independent movements in two variables, in this case, interest rates and inflation. The left-hand chart demonstrates the basic sensitivities of a fund with no allocation to real estate - as interest rates fall and inflation increases, the fund falls in solvency. The right-hand chart includes an allocation to real estate. By assuming some correlation of property to inflation, we effectively assume that it has some inflation ‘duration', or that volatility in its returns is driven by inflation. The effect of this duration is seen in the second heatmap by a slight increase in solvency for scenario increase in inflation versus having no property in the portfolio.
Stochastic simulation
We now simulate projected funding through time by running a stochastic simulation for rates, inflation, equity and property using 5,000 Monte Carlo paths for each asset over 10 years. The results of the simulation are summarised in Figures 7, 8 and 9. Figure 7 shows the distribution of funding ratio through time for portfolios with varying allocations between equities and real estate as shown. We see a slight decrease in funding through time for portfolios with higher real estate allocations - unsurprising given the slightly lower risk premium we have assumed for real estate versus equities.
Figure 8 shows the lower tail of the distribution of funding ratio after 10 years for the different portfolio with and without real estate allocations. It is no surprise to see that the inclusion of real estate in the asset portfolio reduces volatility in the funding ratio distribution. This is expected because of the lower volatility real estate versus the equity that is being replaced. Furthermore, the correlation of real estate with equities and rates means that the portfolio is diversified and volatility and tail risk is reduced.
This is demonstrated further by Figure 9, which plots a tail risk measure - 99% conditional variance - against time for the different portfolios. The final chart shows the kind of improvement in risk metric that a pension fund can gain from investing in an alternative asset which has a low volatility and is relatively uncorrelated with the most common components of an asset portfolio. The quantitative interpretation argues strongly for the inclusion of property as part of a diversified portfolio.
Although most pension schemes have allocations to property that are far lower than the high levels of the 1970s, there has been a recent increase in property allocations. Beyond the recent strong performance of the asset class, there have been important changes that have served to establish property more as a "mainstream" asset class.
The increasing transparency, liquidity and depth of the market has encouraged greater flows of capital to property, but it has also enabled the development of a range of increasingly sophisticated investment products. On the one hand, the "public" markets have grown dramatically, whether "public equity" markets of listed property companies or the fast-growing "public debt" markets. On the other hand, there has been the growth of new products, most notably property derivatives, but also other products such as mezzanine and hedge funds.
Joseph Hamilton is head of pensions advisory, Pretty Sagoo is vice-president, pensions advisory group, Deutsche Bank
This information is provided for educational purposes only and does not create any legally binding obligations on the part of Deutsche Bank AG and/or its affiliates. The opinions or recommendations expressed herein are those of the authors and are not representative of Deutsche Bank AG as a whole.
