Section Three: Diversification and Stock Performance Drivers

Section Three: Diversification and Stock Performance DriversAn allocation to a sector will not guarantee the fund will deliver a return from its properties that match that of the sector average. Firstly, a number of holdings are required to diversify the specific risk. Secondly, there are stock characteristics that will systematically influence returns within sectors, such as income security, property quality and sustainability factors. This section will examine the data provided in the PFV Handbook to allow investors to estimate the likelihood of a fund performing in-line with its structure (diversification) or to tilt their portfolio exposure towards one or more stock factors.Diversification
As real estate assets are large and heterogenous, investors require additional data on concentrations of fund exposure to individual properties, tenants and lease expiry dates to ensure that they are likely to achieve the sector exposure they seek.An exposure to more properties, more tenants and a spread of stock characteristics that match the Index will reduce the expected performance differential between a fund and the Index.The expected performance differential can also be measured in absolute terms. For example, an investor may reduce absolute risk through an exposure to the most secure income streams and the least volatile segments of the market.Whether seeking to reduce relative or absolute risk, investors may actually seek to tilt their portfolio structure towards one or more of these risk factors, for example to developments, short leases or higher quality property to generate higher returns.Number of propertiesAcquiring an additional property will reduce return volatility by diversifying specific risk. However, the raw number of properties might be misleading if, for example, the fund has a small number of large properties and a long tail of smaller assets. The equally weighted equivalent (EWE) is a measure of the actual diversification power of the properties within a fund. The principal is simple: a fund with one small and one large property will achieve less diversification than two equal sized properties. The true degree of diversification on Balanced funds for example is closer to 40 properties (the EWE) than the average of nearly 70 holdings.In contrast, the degree of concentration risk is much higher in Specialist funds, with an EWE of less than 20. Of course, by definition, Specialist funds hold properties in an individual sector, so an exposure to an equivalent of 20 holdings will give a high degree of diversification within that sector.The average number of properties in Balanced funds has increased slightly since the GFC from slightly under, to slightly over, 60 properties. The EWE for Balanced funds had drifted marginally lower, to under 30, before rising again.The number of properties in Long Income funds has risen so quickly since their inception that they now exceed the degree of diversification in Balanced funds.Table 3.1 Average number of assets versus the equal weighted equivalent

^{Source: Authors own calculations using the PFV Handbook}As more properties are added to the portfolio, the risk reduction benefits diminish. Once funds consist of 20 or more assets the marginal diversification benefit of additional properties is small. However, the relationship between the number of assets and tracking error is still detectable: with portfolios with a higher number of properties typically having lower volatility. The same relationship holds for both the EWE and the total number of assets.Figure 3.1 Number of properties and EWE versus volatility of absolute return, 2007-2019

^{Source: Authors own calculations using the PFV Handbook}Similarly, for relative performance, the distinctive shape of diminishing diversification benefits from additional property is still visible.Figure 3.2 Number of properties and EWE versus tracking error, 2007-2019

^{Source: Authors own calculations using the PFV Handbook}Top 10% propertiesA different way of measuring diversification is to look at the concentration of portfolio value accounted for by the 10 largest properties. The greater the exposure of a fund to its largest properties, the higher the concentration risk and the more importance needs to be placed on these larger individual properties (stock selection). To aid such analysis the PFV Handbook lists the name, location and sector of each of the 10 largest properties⁶.Figure 3.2a tracks the average concentration of property exposure to the 10 largest properties by fund category. As Specialist funds have fewer properties, the proportion in the top 10 is inevitably going to be higher. Although the trend is far from dramatic, there has been a slight drift down in fund exposure to the largest 10 assets in the portfolio.

_________________
_{⁶If the UPRN (unique property reference number) were to be provided these can be mapped by external analysis tools, such as CoStar, Radius or Datscha, this would both aid further property specific research and promote the creation of additional datasets.}Figure 3.2a top 10 % property GAV

As for the results using the number of properties, funds with a higher exposure to their 10 largest properties have shown a greater spread of fund returns than would have been expected from their structure alone.Figure 3.3 Top 10 properties versus tracking error to predicted structure benchmark 2007-2019

^{Source: Authors own calculations using the PFV Handbook}
A higher proportion of fund value in a small number of properties will not efficiently diversify specific risk. To justify this extra risk an additional return (achieved perhaps through greater management focus) is required. However, there is no evidence that funds have achieved an additional return, with a negative correlation between the concentration of fund value in the top-10 properties and the Sharpe Ratio.Figure 3.4 Top 10 properties versus Sharpe Ratio to predicted structure benchmark, 2007-2019

^{Source: Authors own calculations using the PFV Handbook}What the analysis cannot tell is whether the results are influenced by the sector allocation. In other words, are the largest properties in a portfolio more likely to be in sectors which generated poor risk-adjusted returns (shopping centres for example) or have large properties, regardless of sector, generated poor risk-adjusted returns⁷?Tenant concentration riskA concentration of lettings to a particular tenant can also lead to a divergence in fund performance, most obviously if the tenant becomes insolvent, but also if the tenant’s credit rating is downgraded and this is reflected in the valuation.Tenant insolvency is a significant risk factor in a downswing, leaving the asset non-income producing at a time when rental values are falling and potentially also incurring capital costs to return it to a lettable condition. In an upswing, tenant insolvency can potentially be a fillip to performance, if a higher letting rent can be achieved.The PFV Handbook provides a list of the top 10 tenants and the proportion of rent roll they account for. Balanced funds have the lowest exposure to their largest 10 tenants, although this exposure is only marginally below that of Specialist funds (which have a lower number of properties).
____________
_{⁷To disentangle these relationships the property concentration risks would be required by segment.}
As the Specialist fund sector has grown, the exposure of individual funds to their largest tenants has also fallen. However, Long Income⁸ funds are significantly more exposed to individual covenants than other fund types. This exposure is more of a concern as the assets, by definition, have a high proportion of value in the lease itself, which may not be replicable if the current tenant defaults.Figure 3.5, Tenant concentration risk; top 10, % rent passing

^{Source: Authors own calculations using the PFV Handbook}The greater the fund exposure to the 10 largest tenants the greater the tracking error of the portfolio.Figure 3.6 Tenant concentration risk versus predicted structure benchmark 2007-2019

^{Source: Authors own calculations using the PFV Handbook}^{_______________

8}_{The inclusion of Long Income funds has not been accompanied by a change in the reporting template. What types of leases are within the different funds in the Long Income category? What is the proportion of inflation-linked versus traditional leases, are there income ‘strips’, and what is the proportion of ground-rents?}The lower Sharpe Ratios would lead to the conclusion that this higher volatility has not been coupled with higher returns.Figure 3.7 Tenant concentration risk versus Sharpe Ratio (2007-2019)

^{Source: Authors own calculations using the PFV Handbook}
Again, what the analysis cannot tell is whether the results are influenced by the sector allocation. In other words, is the effect simply because sectors with predominantly multi-let properties generate stronger risk-adjusted returns?Property specific characteristics
A fund with a large number of properties, of equal size and with no tenant concentrations would be expected to perform in-line with its portfolio structure. However, if all the properties had similar income, quality or size characteristics it is possible that such ‘style’ factors could also produce significant performance differentials.Vacancy ratesA vacancy rate is a glass half-full / half-empty performance driver. On the half-full side, a vacant unit will perform strongly if re-let and achieves an uplift in value. From a half-empty perspective, a unit that remains vacant will generate a negative income return due to empty rates and other vacancy costs.If the risks (re-letting probabilities) are priced to perfection, a collection of vacant units would theoretically be expected to outperform, on an absolute basis, a collection of let units. The additional return is required due to the higher volatility of the average return from all vacant units through a cycle (due to the fluctuations in occupier markets and investor risk tolerance). Investors may also demand a higher premium for the higher uncertainty of the individual property return.A concentration of vacancies in a particular fund would therefore be expected to produce a higher tracking error, underperform in weak market conditions and out-perform in a strong market, but with a similar Sharpe Ratio overall.Higher vacancy rates have indeed been associated with lower risk-adjusted returns, particularly for Balanced funds (correlation -0.5), during the period of weak market conditions.Figure 3.8 Vacancy rate versus predicted structure benchmark – 2007-2012

^{Source: Authors own calculations using the PFV Handbook}During the stronger market conditions experienced from 2013-2019, the impact of vacancy on risk-adjusted performance has been much less clear. Market performance during this period was not uniformly strong, with significant weakness in retail.Figure 3.9 Vacancy rate versus predicted structure benchmark – 2012-2019

^{Source: Authors own calculations using the PFV Handbook}Weighted average unexpired lease terms (WALT)Like vacancies, income security can boost fund performance in a strong occupier market but it is likely to be decretive to performance in a downswing. The weighted average unexpired lease term (WALT) is a measure of whether the fund is tilted towards more or less secure income. The authors have estimated the WALT of a fund using the lease expiry profiles provided in the PFV Handbook. The profiles published breakdown the proportion of lease expiries into 5-year intervals (e.g. 0-5 years, 5-10, 10-15, 15-20, etc.). The estimated WALT is the average of the mid-point of each interval (e.g. 7.5 years for the proportion of leases expiring between 5-10 years) weighted by the proportion of leases expiring in that interval.Lease terms in commercial property have been on a continuously downward trend for over a century, down from 125, to 42, then 25 years and now a 10-year lease with a five-year break has become almost aspirational. Unsurprisingly, fund WALTs have reflected this downward trend as previously long leases are replaced with shorter ones.Figure 3.10 WALT

^{Source: Authors own calculations using the PFV Handbook}During both strong and weak market conditions the impact of WALT on risk-adjusted returns versus the structure benchmark was random. This is likely to be a result of the sample of funds having very limited variability in their estimated WALT.Figure 3.11 WALT versus predicted structure benchmark – 2007-2012

^{Source: Authors own calculations using the PFV Handbook}Figure 3.12 WALT versus Sharpe Ratio – 2012-2019

^{Source: Authors own calculations using the PFV Handbook}Initial yieldSeveral funds mentioned an investment strategy focussed on an above average income return. The initial yield gives an indication of the future level of fund income return.Balanced funds across all time periods have had a higher net initial yield than Specialist funds. This likely reflects the difference in net reversionary potential with Specialist funds typically having higher reversionary potential. This illustrates one of the issues with using portfolio yields as a metric for quality. ERV per sqm would likely represent a better indication of asset quality.Figure 3.13 Average initial yield

^{Source: Authors own calculations using the PFV Handbook}Figure 3.14 Net initial yield versus Sharpe Ratio 2007-2019

There appears to be no clear relationship between portfolio yield and risk-adjusted returns over the full time period or between weak and strong market conditions.Please see the Appendix for data on reversionary potential, lot size, unlisted fund and joint ventures and listed holdings. The variables were not found to be significant in the tracking error model.Reversionary potentialCommercial real estate is typically valued on the basis of a projected cash flow. The current income may be below or above the levels if the properties were re-let today, this is known as the reversionary potential. If occupier markets remain strong, reversion is likely to be ‘collected’ at review or reletting and the income will rise. If fund reversion is negative, then fund income is likely to fall or at least lag that of more reversionary funds.Reversionary potential naturally tracks the trend in market rents and so it peaked before the GFC and then dived as rental values fell, before recovering to a plateau after 2016. Specialist funds have seen a recent dip due to their exposure to retail property. Balanced funds have remained at around 4% net reversionary potential, very similar to that in 2006/07.Figure 3.15 Net reversionary potential

^{Source: Authors own calculations using the PFV Handbook}The change in the sector composition of net reversionary potential confirms the association with occupier market conditions by segment: the cyclical nature of London offices, the strength in industrial and the changing fortunes of retail warehouse property which moved from highly reversionary to negative over the analysis period.Figure 3.16 Segment contribution to net reversionary potential, Balanced funds

^{Source: Authors own calculations using the PFV Handbook}This relationship was noticeable for Specialist funds where a positive relationship can be found. This reflects the diverging trends at the sector level with specialist industrial and London office funds exhibiting greater reversionary potential than retail funds.The reversionary potential of each fund is broken down further, splitting out reversionary rent, over-rented rent and vacancies but the segment breakdown is again by property type rather than the full segmentation.Figure 3.17 Reversionary potential versus predicted structure benchmark – 2007-2012

^{Source: Authors own calculations using the PFV Handbook}Figure 3.18 Reversionary potential versus predicted structure benchmark – 2013-2019

^{Source: Authors own calculations using the PFV Handbook}Lot sizeThe size of a property has an impact on both performance and liquidity. These influences can vary over time: sometimes tenant demand is focussed on larger units and sometimes on smaller space. Similarly, investors may sometimes favour an exposure to larger property and sometimes not.

The trend is dominated by the survivor bias in Specialist funds, with the funds that survived the GFC tending to focus on the larger retail lot sizes of shopping centres and retail parks.Figure 3.19 Average lot size

^{Source: Authors own calculations using the PFV Handbook}Unlisted funds and joint ventures (JVs)Some funds have investments in JVs or other unlisted funds. These investments provide an analytical challenge to ‘see-through’ these investments and estimate the exposure to the underlying sectors, leverage, and developments. Such investments may also have an impact on the ability of the fund to quickly alter their portfolio structure or to meet redemptions.From 2006 to 2009, Balanced funds held an average of nearly 9% of GAV in unlisted funds and JVs. This proportion fell steadily to under 2% by the end of 2016. Few Specialist funds have holdings of unlisted funds or JVs, but those that do have substantial holdings. The sharp change in 2019 is due to one fund leaving the sample. One Long Income fund has a small indirect holding.Figure 3.20, Unlisted/JVs, % GAV

^{Source: Authors own calculations using the PFV Handbook}ListedAs for JVs and other unlisted funds, a holding in listed companies, also impedes the analysis of the underlying fund structure. Listed property properties also tend to be weakly correlated with direct property over short time periods. Property company shares are typically held as a means of managing redemption requirements in the fund as an alternative to holding cash.Outside of the daily traded retail funds, the use of listed investments is very limited. For daily traded retail funds, the average level pre GFC had risen to over 9% of GAV before falling to close to zero, presumably to meet redemptions. Holdings rose again, but only to 4%, and fell sharply post the European Union referendum.Figure 3.21, Listed, % GAV **(no data was available for daily traded retail funds in q4 2011)

^{Source: Authors own calculations using the PFV Handbook}Tracking error model
The significance of the diversification and stock measures has been tested in explaining the tracking error between the actual and predicted return of each fund.

The predicted return is calculated from the fund segment weightings, including cash, and the return of that segment in the MSCI Quarterly Property Index. Cash is assumed to return that of 3-month Treasury Bills. The fund return is then adjusted for leverage by dividing by (1 – leverage). For example, if the predicted quarterly fund return was 2% and the fund had 50% leverage the predicted fund return rises to 4%.
By construction, the tracking error will be due to differences in the performance of the fund’s properties and the sector average.DiversificationAs anticipated, the higher the number of fund properties, especially if measured on an EWE basis, the lower the expected tracking error.
The proportion of fund value in the top-10 assets and top-10 tenants was also significant in the models. However, the sign was occasionally negative rather than positive, suggesting that this measure might be picking up differences in the quality of fund holdings (secondary industrial estates for example) rather than the influence of diversification.
A higher development exposure increased the tracking error of funds against their predicted return in several of the models, but the measure was not universally significant. This result cannot be regarded as definitive due to the low exposure to development of the analysed funds.
Property specific risk factors

The performance of vacancy rates, WALT and initial yield in the models were mixed. As the occupier markets were very mixed across the sectors from 2013-19 the analysis would be better undertaken at the segment level.Table 3.2 Factor regression model results

^{Key:

**Significant at the 5% level - *Significant at the 10% level}Summary

Data on fund structure alone does not allow an investor to estimate the expected risk and return on a fund. Firstly, investors require data on the degree of diversification and secondly on any differences in stock characteristics (intentional or otherwise).

Several diversification measures were considered: number of properties, the EWE and the proportion of fund weight in the top-10 properties. Permutations of these measures were all found to be significant in explaining the divergence in past fund returns from that expected from their structure alone.

As some sectors are associated with larger lot sizes, data on diversification should be provided at the sector level.

A fund strategy might focus on a smaller number of properties and forego some of the diversification benefit from more, smaller, properties. To be a success, the additional return from the extra management focus should out-weigh the higher return volatility due to the lower diversification. The evidence suggests that this additional return has not been achieved.

Excluding the Long Income funds, the estimated average lease length did not appear to impact either risk or risk-adjusted returns. This is likely due to limited variability in the WALT within the sample with most Balanced funds having a WALT of between 7-9 years.

Any analysis of the property specific / income factors is hamstrung by the provision of a single fund level metric. The fund vacancy rate for example may camouflage a low vacancy rate in some sectors and a high vacancy rate in others. Similarly, a fund describing its strategy as ‘high income’ may have a high initial yield due to the allocation to high yielding sectors or to properties with higher yields within individual sectors.

A third dimension is missing from the available data which is the quality of the property. Quality can be measured by rental value per sqm. Data on property quality, combined with income security, would allow investors to construct portfolios that differentiate by style. For example, a portfolio might consist of lower quality short lease regional industrials or high quality, secure income central London offices.