Pension funds across Europe have widely differing views as to the appropriate allocation of their portfolios to real estate. According to Mercer Investment Consulting, Austrian pension funds typically hold 1% of their assets in property, while Italian pension funds devote more than 20% to it.
Research from WM Company (using UK and Netherlands data) has also shown that while pension fund property weightings declined from the mid-1980s until the late 1990s, they have now started to rise.
Given that pension funds currently allocate on average about 6% of their portfolio to property, it is time to examine what the optimal allocation should be and how much further pension funds need to increase their exposure before reaching this level.
JPMorgan Asset Management’s recent white paper ‘How much of their portfolio should European pension funds allocate to real estate’ goes someway to answering this question.

The investment universe
As a starting point we investigated the size of the invested universe available to European investors. This is important, given that pension funds cannot hold more real estate than there actually is. Also, assuming that all investors are rational and all asset markets efficient, theory suggests that the actual allocation to different assets is also the optimal allocation.
In establishing the size of the invested universe, we distinguished between the concepts of invested and investible stock. According to DTZ, the total invested real estate stock at end-2003 was €1.7trn, which is 43% of the investible stock of €3.9trn.
We added the total invested real estate stock to the total European market capitalisation of equities and bonds (€6trn and €8.5trn respectively), and an estimate of the total assets in non-real estate private equity and debt, commodities, derivatives and cash.
These calculations gave us a total invested universe of €17.6trn, of which just under 10% is real estate. This allocation is significantly higher than the 6% average allocation that pension funds currently hold, suggesting that they are allocating inefficiently.

Assessing the characteristics of real estate
We then looked at real estate’s characteristics as an investment asset. This issue is clouded by well-known difficulties in measuring real estate performance. In addition to difficulties encountered due to lack of data, tax implications, liquidity profile of the income and capital components of real estate returns, there are the well-known problems of return smoothing by property valuers. We attempted to compensate for the latter by ‘de-smoothing’ the raw returns data.
We concluded that real estate has a high and stable income yield that seems to be well correlated with average earnings growth and total returns that are poorly correlated with other asset classes. These characteristics were key drivers of the conclusions reached in the next two sections of the research.

Optimal asset allocation
In order to establish the optimal asset allocation between various assets, we used the classic tool of mean-variance optimisation analysis. We did this for a range of return targets using data for cash, bonds, equities and de-smoothed real estate. The modelling process was run for all the European countries for which we could assemble sufficient data: the UK, France, the Netherlands, Sweden and Ireland.
Our results found that average real estate weightings over the possible return spectrum ranged from 8% to 35%, with an average weighting of 24%. Results differed by country and according to whether institutions were return maximisers, which always have real estate weighted at zero (allocating more to equities, which yield on average higher returns) or risk minimisers, which always have some allocation to real estate.
During the 1990s pension funds seemed to behave like return maximisers. This left some of them significantly underfunded once the equity markets crashed in 2001. As a result, and encouraged by legislative changes, pension funds have become more focused on matching their assets with their liabilities.
In the final section of the research, we looked at the role of real estate in matching pension fund liabilities. We found that property has a number of characteristics that in principle make the asset relatively well suited to liability matching.
In our liability matching analysis, we considered the results of two formal asset-liability modelling exercises conducted for the US and UK, both of which indicated an average real estate allocation of a little over 15%.
If interpreted conservatively, our results concluded that an average European pension fund should probably seek to hold between 10% and 15% of its assets in real estate. This implies that European pension funds currently hold only around half of their optimal real estate allocation, which is very significant.
Even if nothing else changed, substantial new investment inflows to real estate could be generated. Based on the Mercer figures quoted above, a shift to a 10-15% allocation would result in an additional inflow of net real estate investment from pension funds of between €150bn and €350bn, and if all investors adopted a similar reweighting strategy the net new investment required could be of the order of €1trn.