Did the varying behaviour of European real estate markets since the onset of the crisis have as much to do with approaches to valuation as underlying property fundamentals? Ian Cullen looks at some of the latest IPD research for an answer
The financial crisis that started in 2007 had the initial and immediate effect of destroying investor confidence in the markets, which in turn precipitated recessions across most mature western economies. The combined effect of both these events triggered a series of dramatic price falls in European investment property.
It is worth noting at the outset that this sequence of events was radically different from that of the last cycle in the late 1980s and early 1990s. At that time many economies moved more or less deeply into recession without the precursor of a financial crisis. The fall in GDP hit the occupational demand for commercial property and the resulting fall in rents then translated into a fall in capital values.
The reason why this difference is important is because real estate markets normally respond to the two basic drivers of value - investor confidence and underlying economic strength - independently.
In this article we shall focus upon European property investment markets and examine the different ways in which they have responded to the latest combination of market drivers. This should enable us to explore the extent to which these responses have been accurately and sufficiently recorded in the process of professional investment valuation and to pinpoint where that process has missed either the scale or the timing of extreme cyclical behaviour.
Investment Property Databank (IPD) has published a pan-European index for the past five years - coinciding with the recent crisis - and has reported the investment performance of the 15 most mature European property investment markets for close to, or in some cases over, a decade. For most markets this reporting has been on an annual basis, reflecting the valuation regimes adopted in those markets. In order to address the question of the precision of market valuation, a number of synthetic quarterly indices has been created, each one covering as close to 10 years as is technically possible.
These have been constructed by utilising all available valuation evidence, even when not synchronised to the calendar year-end, as well as precisely timed cash flow data, and by utilising interpolation techniques to estimate values at each quarterly period. These series permit close comparison of real estate market cyclical responses to the financial and economic drivers of performance over the past four or five years.
Figure 1 indicates the quarterly estimated performance of 11 of the most significant western European property markets. It shows how dramatically they have varied, at least in terms of the volatility of their responses to seismic economic changes, with UK returns falling by over 30% in a two-year period, while German returns have moved very little.
The volatility differences are easy to read from this graph. It is less easy to infer the full scale of the return damage suffered or the extent to which the shapes of these responses were finely synchronised or decoupled.
It is relatively straightforward to test for the varying scale of the damage. Figure 2 simply accumulates the losses across each market from whenever the quarterly return peak was reached. It clearly documents the pain suffered in the UK market (the Irish series has been omitted as it would dwarf the others - with 50% losses - and it is by a significant margin the smallest European market). What is more striking is the absence of any period of negative returns in five (identified with dashed lines on figure 2) of Europe's top 11 property markets. They all experienced some measure of falling values, and for Germany that pattern has prevailed throughout the decade, but in the most robust five markets a resilient income stream offset these falls.
To test for synchronisation between the quarterly return series is trickier. But with the new quarterly measures it proved possible to correlate each of the series with the other 10 and then cluster them on the basis of the varying levels of correlation. Thus, at the first stage, the two markets with the highest pair-wise correlation were linked together and correlations then recalculated between this new grouping and all remaining markets to successively construct clusters of similarly performing markets. Since the purpose of this exercise has not been to create a market-weighted ‘balanced European portfolio' but to examine similarities and dissimilarities, at each stage at which markets are clustered, the combined returns were created on an equally weighted basis.
Figure 3 summarises the stages in this clustering process. It demonstrates a very high degree of performance similarity among the southern European markets, starting with France and Spain and then quickly adding Portugal and Italy (even though the latter two did not fall into negative territory) to the mix. This southern European cluster then stretches up into the Nordic region, picking up the Netherlands en route. So we have a group of eight markets stretching from south-west to north-west Europe exhibiting a common pattern of market performance at a correlation level still in excess of 0.7.
At this stage a new and very different cluster is formed - by grouping Germany with Switzerland - at a correlation of just 0.6. Figure 4 demonstrates the ‘shape' independence of these three remaining groups after the cumulative averaging of the most westerly markets (apart from the UK) and the separate combination of the more easterly German/Swiss return profiles. Finally, in order to add the much more volatile and noticeably ‘decoupled' UK market to the dominant western European group led by France and Spain, the correlation is forced down to well below 0.5.
However, the deep-seated independence of the German/Swiss group is dramatically revealed at the final stage of the analysis, because the very last correlation score between the two remaining groups - the French/UK and the German/Swiss - forces the correlation to actually fall below zero.
What all this suggests is an exceptionally high degree of independence among at least two significant groupings of European markets in their patterns of response to what was undeniably a highly and painfully synchronised financial and economic crisis.
The impact of valuation methods
An obvious question raised by this analysis is that of the variation in approaches to valuation adopted across Europe. Are the recorded differences between the market responses as much to do with approaches to valuation as they are to do with underlying property market fundamentals? While this is a tempting initial response, it is quite possibly missing the point, due to the varying characteristics of the longer-term economic environments within which the European markets were delivering returns.
The point is probably better made by focusing upon the two extreme cases: the UK and Germany. They have exhibited the greatest difference in volatility and the lowest return correlations across all 11 markets. Each has developed apparently very different approaches to property investment valuation, with the open market approach of the UK, driven strongly by comparable evidence, contrasting with the German Verkehrswert approach, which emphasises investment worth and which is in turn driven by medium-term sustainable income.
This conclusion would be at best an over-simplification, due to the nature of the economic crisis and the very different ways in which those two markets entered that crisis. The UK and Germany also happen to have been the most and least sensitive to the movements in investor sentiment during the first six or seven years of the decade, with UK prices rising continuously in a generally weak - or at best stable - macro-economic environment, and German prices continuously drifting downwards in an equally weak environment without the benefit of any measurable investment pressure. So, when the financial bubble burst in June 2007, the highly inflated UK property market was immediately and painfully affected, while the German market that had never experienced such inflationary pressure was not affected.
France, Spain, Portugal, Italy and much of Northern Europe lie between these two extremes: hybrid markets which to a greater or lesser extent did experience property inflation of the UK yield compression variety, but in heavily diluted doses.
Can we test the sharpness of valuation?
Given that the two most extreme responses to the crisis were provided by markets with ostensibly the most different valuation regimes it becomes important to try and test for the sharpness and accuracy of valuation. IPD has been undertaking such tests for many years, using very simple measures of the proximity of individual valuation results to after-the-event market tests implied by the completion of a sale. The four most liquid European investment markets - the UK, France, Germany and the Netherlands - have been subjected to such testing for five to six years.
The valuation accuracy studies have only been completed to December 2009, although the historical comparisons now go back right to the beginning of the decade. These studies consistently indicate that no individual national market appears capable of reducing the average spread between a valuation and subsequent market transaction below approximately +/- 10%.
However, over the last seven to eight years we have noticed a convergence across the four major markets upon a spread within a 9-15% range. These results are calculated on an equally weighted basis. If they are value-weighted to accord more importance to the larger asset results, the spread drops to 10-13%. And on this value-weighted basis, the German spread over the last four to five years has been only approximately 10-15% worse than that recorded in the UK.
In other words, while there are difficulties in pursuing highly accurate property valuations, these appear to have been shared across all the major European markets. Differences between those markets in accuracy scores do not seem to explain much of their differing patterns of performance.
The impact of transaction illiquidity
The valuation accuracy tests do not directly address the key issues of volatility and value at risk (VaR), which have become the central concerns of European regulators and investors over the past two years - and so have become inevitably more central to investment strategy formation.
In the context of extreme financial uncertainty and long-term economic stagnation, the problem of fully capturing not just the synchronisation of markets but also their genuine capacity to lose value over relatively short periods is bound to increase.
We have therefore started to extend this work recently by creating hybrid indices that attempt to combine the regular valuation evidence with the extra data that flows from the direct trading of investment assets. These hybrid indices are best described as transaction-linked indices (TLI).
This work is still in its early stages but is designed to address the central problem of trading-linked volatility. The theory is that regardless of how well synchronised national market indices are with the basic underlying cyclical pressures, the fact they are derived from a judgemental process informed by an illiquid underlying market process is likely to mean the understatement of volatility. By ‘transaction-linking' the valuation series, we can both assess the extent to which the valuation process understates volatility and provide both regulators and investors with an evidence-based means of informing their risk judgements.
The early results of this work demonstrate that the addition of trading evidence has the consistent effect of ‘de-smoothing' the underlying valuation records, as indicated in figure 5.
However, the scale of this extra trading linked volatility seems to vary significantly from market to market. Of the European national markets we have tested, only the UK, France and the Netherlands have generated relatively robust transaction-linked models, and only the UK has been updated to the end of 2010.
Figure 5 demonstrates that the increased volatility of transaction linking has been much lower in the UK than in the other two markets. The standard deviation of the quarterly price movements in the UK is only 15% greater than that of the professionally valued price movements. At the other end of the spectrum, the standard deviation in the Netherlands, after transaction linking, is a factor of four greater than that of the professionally valued price series.
The all-important VaR has a very similar relationship to the underlying standard deviations. While the VaR derived from a pure valuation index in the UK hardly deviates from an index constructed with the benefit of transaction linking, a risk multiple of three or four times higher is required in both France and the Netherlands to accommodate the extra trading linked volatility.
So what conclusions, if any, can be drawn from this work regarding the genuine differences between investment property markets?
It is too simple to say these differences are merely artefacts of their valuation procedures. But, equally, it would be naive to continue assuming that every facet of a national market's performance is perfectly captured through its valuation process.
The truth, as always, probably lies somewhere between the extremes. The still unfolding crisis among European economies was unique and the effects upon property markets were genuinely different. If valuers had told us these markets were all reacting identically to a global crisis they would have been wrong. If at the other extreme they had claimed their work always ‘marked to market' to the full extent of its sticky movement they would have been equally wrong.
The research reported above was supported by the IPD Solvency II Review, sponsored by INREV, ABI, BPF, BVI, EPRA, IPF and ZIA
Ian Cullen is co-founding director at IPD