Get Better Results With Improved Property Type Segmentation

What seems to be best data practices can become a way to get stuck in the mud. They can become comfortable and an automatic choice, getting reinforced through how software vendors typically handle data.

A recent article by MSCI Research Executive Director Bryan Reid, Senior Associate Fritz Louw, and Executive Director Will Robson in The Journal of Portfolio Management looked at how the usual combination of property type and geographic location can be useful and yet remain limited in how they can identify and explain performance variation.

“For many real estate investors, property-type and geography segmentations are the primary lens through which they measure and manage their portfolios,” the authors wrote. “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.

“In an analysis of over 26,000 UK properties between 2002 and 2022, however, the authors find that traditional property-type/geography segmentations explained an average of just 20% of asset-level total return variation. Testing six potential real estate style factors in a cross-sectional multifactor model, they were able to explain an additional 8% of asset-level variation, suggesting that real estate factors could play a role in helping investors manage their portfolios more systematically.”

The additional six factors they tested were size, yield, leasing, growth, momentum, and volatility “in a cross-sectional multifactor model.” Not all the factors proved significant in their descriptive powers. And the usual ones of property type and geography weren’t necessarily as useful as might be expected.

“The results from our baseline regressions show that the explanatory power of the standard property-type and geography segmentations varies over time between 2% and 55% but has averaged just 20% over the entire analysis period,” a significant variation.

“Unsurprisingly, the explanatory power of standard property-type and geography segmentations has been weakest during periods where there has been greater homogeneity in the total return of the segmentations and increased during periods where there has been greater divergence in sector total returns.”

The larger multifactor model explained an additional 8% of the variation. “Of the six factors tested, yield, leasing, and momentum showed the most promising results, but the results for size and growth were weaker and volatility was in-between,” they wrote.

The authors noted that this wasn’t a definitive assessment, as there are many others that could be examined. It also could be that markets, however defined, could vary in which factors are most descriptive of variation in value, performance, or some other metric.