New research on asset-specific risk has important implications for real estate investment processes and strategy. Paul Mitchell explains
Research presented by Prupim's Paul McNamara and myself at the Nick Tyrrell Memorial Seminar in London last year has important implications for the ways in which fund managers price assets and structure their real estate portfolios.
The themes of the research would have appealed to Nick, with respect both to its thought-provoking findings and to how research can enhance its contribution to real estate investment processes.
Traditionally, real estate investment researchers have focused on markets and on portfolio and investment strategy, limiting their contribution on individual assets to the effect of the market and the local economic context. One reason for this is that a comprehensive understanding of the drivers of individual property returns has eluded researchers. This in turn has stunted the development of formal processes within many real estate investment management firms for pricing individual assets, whether research-based or not.
Despite a 25-year history of institutional real estate investment in the UK, individual asset appraisals remain primarily the responsibility of the investment and asset managers working at the coalface.
Other than the high-level research perspective, modus operandi often incorporate personal styles and biases which, while no doubt drawing on great skill and insight, can lack the systematic foundations and research-based assumptions of other asset classes.
Yet the importance of having the right assets in a portfolio is the key to investment success in real estate. Stock selection (that is, having the right - or wrong - assets) is the main determinant of portfolio outperformance and underperformance not only over one year but also the three-to-five-year horizon typical of many mandates.
It is also known from the research I undertook for the UK's Investment Property Forum (Alpha and Persistence in UK Property Fund Management, 2008, by Paul Mitchell and Shaun Bond) that this portfolio outperformance (or underperformance) on account of individual assets does not last. Skill in choosing assets typically does not persist.
Understanding the performance of individual assets and involving the investment researcher in decision-making therefore could potentially improve: (a) the contribution from individual assets; (b) portfolio performance, strategy and risk control; (c) investment decision-making processes, thereby making real estate more transparent for investors.
The research that Paul McNamara and I did for the Nick Tyrrell Memorial Seminar was based on the study of 200 assets held continuously since 2000 by various Prupim funds. The research not only drew on statistical analysis of the performance and characteristics of these assets but also 25 in-depth case studies of some of the most interesting properties. These case studies drew heavily on the expertise and knowledge of Prupim staff. This combination of statistical analysis and case studies provided unique insights into the nature of individual property performance and risk.
What was surprising from the research was the extent to which individual property performance was driven by the overall market. They plot the assets' total returns against their IPD segment (for example retail warehouses, City offices). In doing so, they not only emphasise the effect of the market but also highlight the subtle differences that can emerge across individual properties.
The first asset's relationship with the market is very evident - its returns closely track those of the IPD segment. This pattern applied to most assets. Less obviously, the second asset also closely tracks the market but in an overly sensitive way: its returns increase relatively quickly when the market is strong and fall relatively quickly when the market is weak. This makes it a high beta asset.
The asset-specific risk for these two properties is the difference between their total return and that implied by the market segment taking into account their different sensitivities (any alpha also has to be accounted for in the calculation of specific risk). For the period shown, the specific risk (measured as a standard deviation) for these two assets was a lowly 2% and 4.5% per annum, respectively. These were widely-spread patterns: the market was the dominant influence on the performance of most (70%) of the 200 properties during the 2000s, and the specific risk of individual properties was typically less than 10%.
The research also found little evidence of additional factors - such as size and yield - widely affecting individual property performances, although there were indications that multiple tenancies did have an effect. The absence of an effect from yield is notable given received wisdom; IPD market data, for example, reveal a big underperformance of high-yielding property over the past two to three years.
However, my research shows there is a range of performance among high-yielding property from good to very bad and that past yield is not always indicative of current yield or returns. It is clear that the performance of high-yielding properties is more complex than the IPD average reveals.
The strong influence of the market observed in most assets is to be expected at a time when general factors such as economic volatility and wildly fluctuating risk appetite weighed heavily on all real estate. However, this market influence on assets was much stronger than at comparable points in previous cycles. In line with this, correlations between individual properties over the past five years, averaging 0.7, are far greater than at any other point in the past, including the last extreme cycle in the late 1980s and early 1990s when they averaged approximately 0.3.
Specific risk then was also much higher (almost two-fold). These higher correlations and lower levels of specific risk have, as outlined later, important implications for diversification strategies. Something has occurred to make individual properties more attuned to market conditions. One source might, for example, be better valuation practice than during the last extreme cycle.
For a minority of assets, the market explained relatively little of the year-to-year variation in returns. Specific risk for these assets was much higher, typically around 15% and in a few cases more substantial. This high specific risk also seemed to take two broad forms.
Specific risk in the first asset is persistently high, while in the second asset it is manifested as a very substantial one-off, characteristic perhaps of the type of ‘fat tail' known to afflict some financial markets.
The 25 case studies were designed to reveal the sources of this high specific risk and the characteristics of properties exposed to such high risk. Each ‘abnormal' annual return (favourable as well as unfavourable) was assessed;
The absence of factors associated with the property's fabric and with locational events is notable; physical or locational obsolescence does not appear to manifest itself in a ‘lumpy' way. By contrast, the factors associated with high specific risk are predominantly associated with so-called ‘lease events'. In particular, the main reason relates to the variability in what is priced during the run-up to lease expiry, rather than the lease event itself.
An example is a medium-quality property asset whose re-letting prospects became much more favourable during a strong occupational market, only for them to be downgraded substantially when the UK entered recession. This type of phenomenon is generally observed in the UK for some types of property approaching lease expiry and it is manifested as the high beta.
However, what this example, and some of the other case studies suggest, is there is an additional layer of asset-specific risk in this type of property. It is also notable that downside risks like this example were most common when UK occupational markets and prospects deteriorated rapidly between 2008 and 2010.
Outturns to lease events, however, rarely turned out in the way expected and these outcomes were also an important source of high specific risk, as the table shows. Interestingly, more often than not, such outcomes proved to be better than had previously been expected.
A particularly fortuitous example of an unexpectedly favourable lease event relates to the property whose tenant exercised a break clause in the lease only then to have difficulty finding somewhere else and having to re-lease on much better terms for the landlord. This led to an abnormally large return for this property but, of course, there were also assets where lease expiries, tenant defaults and unfavourable re-lettings led to very poor outturns and returns, mainly in 2009 and 2010. Of course, there were also unfavourable outcomes.
The case studies also suggested that while poorer-quality, ‘secondary' properties were more prone to high specific risk, prime was just as badly affected when hit.
The research has important implications for real estate investment processes and strategy. It suggests that the investment researcher can play a greater role in appraisals of individual assets. First, most assets - with the notable exception of some approaching lease expiry - appear to have beta risk close to the market average and should have a corresponding hurdle rate. The researcher can contribute towards such assessments and the extent to which an asset is potentially high beta and requires a higher hurdle rate.
Second, there is great potential in correctly taking a view on the outcome of a lease event. As illustrated earlier, market sentiment in the run-up to expiry can swing rapidly and, when pessimistic, outturns can be better than expected. Research can contribute by taking a systematic view on likely outcomes and the extent to which these are priced by the market; in my view, this is an area where research resources can productively be re-orientated.
Considering the lessons for portfolio strategy, the generally increased correlations between individual properties imply that fewer assets than before are required for diversification. However, the research also suggests that portfolio risk will depend crucially on the types of asset held.
In particular, the types of asset unearthed in the case studies - typically experiencing high specific risk in the run-up to and on account of lease events - will generate much higher portfolio risk. This not only reflects their inherently high risk but also their abnormally high likelihood of extreme returns.
This means that portfolio construction and risk-control techniques, traditionally based on sectors and regions, should also have regard for the types of asset exposed to lease events, indicating a new focus on ‘time diversification'. Not only does the proportion of assets exposed to such shocks and their contribution to portfolio risks need to be controlled and quantified, but there also needs to be diversification across such assets to ensure upside as well as downside.
Finally, it is interesting to consider a possible fundamental difference between the UK and continental Europe. On the one hand, the UK is still characterised by relatively long lease terms where occupational value is tested only periodically, while in many parts of continental Europe properties are exposed to the letting market and to asset management much more frequently.
Overall, IPD market returns are a lot less volatile than UK-specific returns, partly on account of ‘smoother' valuations. But might it be the case that this lower volatility on the continent also reflects individual properties there being less exposed to the ‘concentrated', once-every-10-year re-leasing risk, which can have such a profound impact in the UK?
Paul Mitchell is an independent consultant and founder of Paul Mitchell Real Estate Consultancy