Aggregation Bias

What appears on the next page is a graph representing “aggregation bias” in data. Per Bakken’s report, he assumes that there are no differences between stand-alone retail properties (such as Walgreens) and strip-mall retail properties. He even uses a case study to “prove” that there are no differences in rents between strip and stand-alone retail properties. However, by looking at the graph on the next page, it is possible to find a stand-alone retail property (from Submarket B) that overlaps with a strip-mall retail property (from Submarket A). It is insufficient to claim that because one can find a limited number of “overlapping” property rents that there is only one “market” for retail properties. A proper analysis would require two dimensions of information: the “mean” for each submarket and the “standard deviation” for each submarket. With both a central tendency measure (mean) and a dispersion measure (standard deviation), only then can one claim that there is no difference between markets. The correct procedure would be to analyze retail rental data using, preferably, parametric statistics (such as a difference in means test—t-test); or one could employ a non-parametric statistic (such as a signed rank test—Wilcoxon) if the data do not conform to the parameters of a normal distribution.

Wisconsin Pharmacy Sales Data:

The mean sale price per square foot for these triple net lease pharmacy properties is $327 which is in the middle of the range, $291-350. The median sale price per square foot is $324 and that too is in the middle of that range.
Wisconsin Pharmacy Data and General Retail Data
Adding general retail data to the distribution we generate the following graph:

This graph depicts the situation described with an aggregation bias. The general retail data do not include any operating pharmacy sales, and the general retail data distribute in a distinctly separate manner than do the pharmacy sales data. There are no pharmacy sales prices per