- Segmentations based on property type and geography have traditionally been the primary lens through which investors have viewed their real estate portfolios.
- However, in an analysis of U.K. data, we found that traditional segmentations explained an average of just 17% of asset-level total-return variation between 2002 and 2020.
- We tested five potential “style factors” and found they helped to explain an additional 9% of total-return variation. This suggests that factor frameworks could be a useful tool for investors looking to move beyond property type and geography.
For many real estate investors, property type and geography segmentations are the primary lens through which they measure and manage their portfolios. Whether it is defining allocations, constructing benchmarks, attributing performance, forecasting or modeling risk, segmentations built on property type and geographical classifications play an important role. However, in our analysis of over 25,000 U.K. properties between 2002 and 2020, we find that traditional property-type/geography segmentations explained an average of just 17% of asset-level total-return variation.1 Testing five potential real estate style factors, we were able to explain an additional 9% of asset-level variation, suggesting that real estate factors could play a role in helping investors manage their portfolios more systematically. Until now, any return variation beyond property type and geography has been deemed idiosyncratic and attributed to stock selection — the ability of a manager to find a good asset and manage it well. While the factors we tested did not all show conclusive results and are not meant to be viewed as definitive, the results highlight the potential for further development in this area. With many investors embracing proptech and alternative/big data, data volumes have increased, and factors may offer a framework that will help investors translate these increased data volumes into strategy and subsequently understand how impactful they have been in driving performance.
Applying a Factor Approach to Real Estate
In general, the term “factor” is used to describe any characteristic that can help explain the risk and return performance of an asset. In equity markets, factors have become a well-established tool in the investor arsenal. Factor modeling and factor investing stem from the capital asset-pricing model of the mid‐1960s, arbitrage pricing theory in the 1970s and Fama and French’s three‐factor model from the early 1990s. Early factors included size and value, but many additional factors have been added that help explain the variation in equity returns and risks.
To test the potential feasibility of asset-level style factors for real estate, we borrowed cross-sectional regression techniques employed in equities and constructed the five factors in the table below.
Factors in Our Model
|Leasing Profile||Floor-space occupancy rate and weighted average remaining lease term|
|Growth||Market rental-value growth|
Factor exposures are calculated 12 months prior to the total-return period. For example, for the December 2020 regression, the equivalent yield exposure is calculated as of December 2019. Leasing profile is only occupancy-rate exposure prior to September 2007. After this date it is an equally weighted combination of occupancy and remaining lease term exposures.
These factors were selected because they generally have some parallel in the equity space and because of additional a priori beliefs. For example, it is sometimes assumed that, because higher-capital-value properties are harder to buy, they may therefore attract an illiquidity premium.
The factors were then tested in a multifactor model on a sample of over 25,000 retail, office and industrial properties from the MSCI UK Quarterly Property Index between 2002 and 2020. Of the five tested factors, three (yield, leasing profile and momentum) showed promising results.
Rolling Annual Returns for Tested Factors
Sample Statistics of Tested Factors
|Factor||Mean |t|||% |t| > 1.96||Mean 12-Month Return (%)||R2 Gain (bps)|
Mean |t| is the mean absolute t-statistic for the factor over the 73 quarterly regressions. % |t| > 1.96 indicates the percentage of quarters for which the absolute t-statistic was greater than 1.96. Mean 12-Month Return (%) is the mean rolling annual factor return. R2 Gain is the increase in R2 in multivariate, cross-sectional regressions due to adding the factor when all other factors are present in the regression.
Perhaps just as important, when we look at the explanatory power of the regressions with the five style-factor variables, we see a notable improvement compared to baseline regressions against just the traditional U.K. property-type/geography segmentations.
Adding Style Factors Improved the Explanatory Power of the Regressions by an Average of 9%
Factors Have Potential to Help Explain Real Estate Performance
This analysis does not seek to offer the definitive assessment of real estate factors. Beyond the five factors we have tested here on data from a single national market, there are many more to potentially explore. What this analysis does demonstrate is that there are systematic drivers of risk and return beyond traditional property-type and geography segmentations and that the development of asset-level real estate style factors could help investors better understand and manage their portfolios.
1The standard segmentations we included in this analysis were: shopping centers; standard retail — South East; standard retail — rest of U.K.; retail warehouse; office – City; office - West End and Midtown; office — rest of South East; office — rest of U.K.; industrial — South East; and industrial — rest of U.K.