I should go ahead and disclaim that I have a couple of projects at work that are eating up almost all of my time, a situation that is likely to continue until I go on vacation at the end of August. Therefore, please don’t be surprised if I’m a bit slow on some stories, and if my posting schedule continues to be somewhat erratic.
With that said, I encountered an interesting followup to the FTC’s release of their insurance credit scoring study:
Commissioner Pamela Jones Harbour, the only one of the five FTC commissioners to vote against releasing the report, said in a statement the research was poorly conducted and that its conclusions can’t be trusted.
“The data collection and analysis fell short of the commission’s gold standard for rigor and completeness, and did not reflect the agency’s best practices,” Harbour wrote. “Better alternatives were available and should have been utilized.
The article mentions that the FTC study was based on data drawn from a prior industry study, with the caveat that only five carriers participating in that earlier study agreed to have their data forwarded to the FTC.
If the mentioned industry study is the one I think it is…while I think that Commissioner Harbour’s concerns are reasonable, I wonder if they could have been eased somewhat by disclosure of the data selection standards from the earlier study, which I’d bet is disclosable by the folks who performed that earlier study.
One comment in particular in the article caught my eye:
Those working on the report, Harbour said, “immediately recognized that the original data set did not accurately reflect the racial and economic demographics of this country. Minorities and poor people were under-represented in the sample provided by the insurance industry.”
I hate to point this out, but aren’t minorities and poor people under-represented in the population of personal lines insurance to begin with? I mean, if you look at homeownership rates by skin color, or the socioeconomic biases in the makeup of uninsured drivers, it’s reasonable to expect such biases to come through into the study data.