There has been a lot of discussion about the declining fortunes of American retailers. Much focus is on sales declines from internet competition – the subject of a prior blog (2/2019). The question of how the decline impacts various shopping centers is complex. Some centers, or “types” of centers have store mixes that make them more or less vulnerable to e-commerce. Within the mall category, some centers were built at great locations originally, while others were squeezed into more marginal sites. Some have had capital infusions and renovations. So are there patterns in the data relating the likelihood of a center’s demise to its location and characteristics?
We recently examined the inventory with a sample of 710 malls existing in 68 MSAs in 2010. In this case, “malls” refers to the combined “regional” and “super regional” categories. We then looked at the same malls in 2020, using exact geo-coding to be sure the building cluster originally identified as a “mall” in 2010 is the same cluster in 2020. That said, these malls were anything but static over the decade. Of the original 710 malls, 64 were reclassified out of the category – generally to a smaller “community center.” There were 45 newly built malls and 39 new malls as a result of reclassification into the group (through expansion and upgrading).
The most important statistic is that 139 (20%) of the original malls did not exist in 2020. A follow up survey revealed that these “deaths” had a wide range of stories. Some were closed, some were torn down, some were converted to other uses besides a shopping center. That leaves 507 malls that survived within the same mall category although often having undergone alterations and expansions.
So, what characterized the 139 malls that figuratively “died” relative to the 507 that survived? To answer this question, we estimated a range of statistical models, from a simple Linear Probability Model (LPM) to a more complicated “Probit” estimation. All this was necessary because we were trying to predict a dependent variable that is “dichotomous” – 0 if the mall survived, 1 if it died. Remember that failure here means not being open in 2020, having been there 10 years prior.
Here is what we learned, with the overall failure probability being 21%:
Malls that are 1 million sq. ft. larger in Net Rentable Area (NRA) have a 22% lower probability of closure over the decade.
Having undergone any renovation between 1990 and 2010 lowers subsequent risk by 10%.
A 2010 vacancy rate that is 20% higher increases closure risk by 4%.
2 million sq. ft. or more of competitive mall NRA within a 5-mile radius raises risk by 4.5%
Of course, many other factors go into determining mall failure for which we simply have no data: the makeup of the anchor tenants, management expertise, etc. But investors might want to take a closer look at their portfolios to see how they stack up against these more basic criteria.
 Interested readers are referred to the CBRE Working paper “Which Malls Close”, by William Wheaton and Morgan Fleischman, July, 2020.
 All results are “statistically significant” at 2% confidence levels. These are Probit estimates. LPM impacts are larger but not guaranteed to be between 0 and 1.