Reader Response

Jul/Aug 2022 Reader Response

“Sense and Sensitivity”

Dear Editor: 

 I was disappointed by the continued push for “equity” in the March-April AR. The cover story asked if rates are “fair.” Accidents and speeding tickets indicate poor driving, and charging more for them deters bad driving. However, only 78% and 75% of the respondents were willing to say that either is fair. Only 50% agreed that hard braking/sharp turning was fair. This approach may lead to charging everyone the same price — a clear contradiction to the ratemaking principles. Kyle Bartee’s letter stated: “Since most of us believe the exam process to be unbiased, one has to conclude that there is a bias in the funnel.” Bartee ignores other possible explanations, including the fact that some races are overrepresented in poor school districts. Even the “It’s a Puzzlement” column, usually apolitical, is pushing for an “equitable pass curve” at the same time we are being assured that we don’t need to worry about the CAS using multiple pass marks on our exams. This is all happening as the CAS just issued four poorly argued papers on “Race and Insurance Pricing” that all state that today’s differences in outcome are due to past discrimination and must be corrected. I feel that the CAS’s continued push for equal outcomes is putting all of the members in an awkward place, where it is difficult to charge actuarially sound rates. In this world, companies will compete on their marketing plans instead of their pricing plans as we’ve seen with U.K.’s

—Joel Atkins, FCAS 

Dear Editor: 

Your March-April 2022 cover story discusses the fairness of predictive rating variables. It failed to discuss the fairness of variables used by certain companies used to predict the price elasticity of insurance. I moved to a fairly nice area of Arizona, only to be subject to multiple double-digit price increases by a few different companies upon renewal despite a very good driving record. Upon discussion with the pricing actuaries of some of those companies, I was told that the companies look at where I live, the cars I drive, then intuit that I probably have a good job and not a lot of time to investigate switching companies if I get a large rate increase. The company’s models say I will likely tolerate a relatively large rate increase without shopping for new insurance. I know there has been discussion of this topic at some CAS meetings and have heard actuaries respond that “there is some tension between predictive models of elasticity and pricing laws,” to another actuary saying that “they only come up with models, and the underwriters set the rates.” The fact is that I consider such elasticity models to be in violation of insurance laws and actuarial principles of pricing since what someone is willing to pay has no relation to that person’s cost for the insurance companies. I suggest any company using such elasticity models reconsider their practices. 

—Steve Visner, FCAS, MAAA

Sample Bias Discussion

Dear Editor: 

Kyle Bartee’s letter made technically questionable use of the phrase, “sample bias.” In statistics, sample bias refers to non-representative sampling of a population. It can lead to mis-estimation of parameters. For example, the sample mean could be higher, on average, than the population mean. If you estimated the proportion of the U.S. population who like ice cream by polling patrons exiting the local ice cream parlor, you might get an overestimate due to sampling bias. However, when Bartee refers to sample bias, he is not discussing sampling at all. Rather his argument is that sample bias exists because the percentage of actuaries who are Black is lower than the percentage of Blacks in the overall U.S. population. But there is really no sampling involved, and hence no sampling bias. The technically accurate statistical statement is that the subpopulation of actuaries in the U.S. is not a random sample of the U.S. population. The same could be said about many occupations. Disproportional representation of various racial, ethnic and religious groups likely exists, but it does not prove or disprove any discriminatory bias was involved. Bartee’s use of sample bias terminology confounds the statistical sense of the term with the charge he is implicitly making that there is racially discriminatory bias in the CAS credentialing process or in the funnel leading to it. This use of “sampling bias” is a misuse of statistical terminology.  

—Ira Robbin, FCAS

Dear Editor: 

The March-April AR published a flawed, illogical letter to the editor from Kyle Bartee. Even more unfortunate, that letter was given a prominent position. Bartee talks about “sample bias” in the CAS membership. He states: “… the CAS membership can be considered a sample of the population where members come from.” He doesn’t specify what this population is but refers to the CAS website, where I could not find anything dispositive. The infographics he refers to show the current membership, for certain demographic categories. This is not a population from which the membership has been sampled; it is the membership. There is no sampling. The common approach of DEI advocates is to use the total U.S. population as the basis for sampling — a seriously flawed approach for CAS membership. It ignores prerequisites (e.g., education, aptitude and interest) and performance requirements (e.g., gaining sufficient pertinent knowledge and demonstrating that knowledge by passing exams). The prerequisites alone will probably eliminate a large majority of the U.S. population, which is hugely unlikely to occur in exactly the same proportion of every possible demographic characteristic. The performance requirements (who is taking and passing exams) could be viewed in terms of demographic characteristics; these are also likely to depart from the U.S. population. In summary, Bartee’s presentation is ambiguous, with weak logic and poor statistical analysis. The letter should be retracted. 

—Robert Finger, FCAS

Kyle Bartee responds: 

The CAS DE&I website has an imbedded video that shows the demographic distribution analysis (first video under “Highlights” section). I encourage you to watch the whole video, but the distributions are shown starting at 1m:35s and 5m:30s. Finger rightfully points out that comparing the demographics of the U.S. population to the CAS Membership does not consider prerequisite and performance requirements, but that point is explained with a comparison between the CAS Membership and U.S. Math Graduates (source: National Center for Education Statistics). While not perfect, using the math graduate demographics help control for the prerequisite and performance requirements. Since underrepresentation of minority groups in the CAS membership is more significant than that of math graduates, it implies that the CAS is losing out on potential talent that would otherwise enter the actuarial pipeline. Filling that talent gap is the whole purpose of the CAS DE&I initiatives, and it is being achieved by increasing awareness of our profession with the underrepresented groups and removing financial barriers for potential candidates with no other means of reimbursement. Those initiatives come from the Barriers to Entry study. I don’t see anything wrong with increasing awareness among groups that have never heard of an actuary because the goal of improving diversity has been a CAS goal for decades and aligns with what employers are seeking. Lastly, if the CAS is going to spend resources achieving that long-standing goal, they should also measure their successfulness, making sure those resources aren’t being wasted.

Note: Reader Response letters are the sole views of the letter writers and do not necessarily reflect the views of the CAS and its administration.