My brother in law grows weed plants in his manure pile.
Canadian concepts on redlining are clearly different than the US
For example, donāt ice the puck!
I know this is CAS-specific, but just an FYI, weāre looking at some of these issues at the Academy as it crosses all areas of actuarial practice.
Iām one of the people on this committee:
https://www.actuary.org/committees/dynamic/DSAC
and weāre investigating a lot of these problems in how they cut across multiple fields and not only P&C. Because theyāre showing up everywhere now, and the regulators are aware of this, too.
One of the problems with the ābig dataā work is that there are lots of variables that are correlated and even if we never see variables that weāre not allowed to price/underwrite on, we may capture that.
Above, you referenced some old texts re: rate differences by geography from about 100 years ago.
Hereās a map based on the 2020 census:
The main difference, perhaps, is some movement north and west, and perhaps less concentration in the SE as more non-black people move in.
And hereās mortality related to kidney disease related to chronic hypertension (I did not pick this at random, obviously)
Iāve noticed this with certain geographic patterns in mortality for certain age-adjusted rates for causes of death, not only with respect to race, but just knowing things like where I-95 is.
Itās a good idea to be aware of long-term historical contexts of your inputs, and that can include redlining practices if that affects your inputs (or outputs).
Iām not assuming CAS is going to ignore any of this (nor most P&C actuaries), if only for the reason that regulators are not going to let you.
This is why the Academy is trying to be active here, because some of the regulations being passed might cause some big problems with respect to underwriting/rating in general.
Didnāt find the materials that I was looking for, but the CAS library has a ādedicatedā section to Race and Insurance Pricing.
However, I have copies of the material I recalled from the CSPA Exam. In many of the following, the material isnāt explicitly insurance, but I donāt think itās too difficult to see how the issues can cross over to various insurance topics.
-
Attacking discrimination with smarter machine learning (Summary illustrationāmaterial for the CSPA exam; however there is a more detailed paper here.)
-
Algorithms Need Managers, Too (Not quite what Jerry asked for; but this does a very good job, IMO, of highlighting the reason why machine learning is so prone to unintended biases.)
-
A paper published by the UK govāt titled āData Science Ethical Frameworkā apparently isnāt available any more (I still have a pdf downloaded, however). But Data science ethics in government has material that is fairly close.
Another resource to consider wasnāt a āfreeā resource. Itās Weapons of Math Destruction by Cathy OāNeil. Chapters 6 and 9 were on the CSPA syllabus. This might actually attend to what Jerry was looking for as well since it also includes āhard dataā to illustrate the topics/issues discussed.
Seems like there might be a āfreeā course from The Institutes on Ethics:
Ethical Decision Making in Risk and Insurance
This is part of the revised requirements for the CSPA designation.
Given meepās post (and interest), Iāve tweaked the thread title to be more generic and moved this to the āGeneral topicsā section.