Jay Bhattacharya’s selective libertarianism: on COVID and insurance rating
A recent paper finds that drivers who are not vaccinated against COVID are substantially more likely to be involved in serious auto accidents than vaccinated drivers.
In response, Jay Bhattacharya, an author of the Great Barrington Declaration and a prominent opponent of lockdowns and vaccine mandates, tweeted that the study “should not be used by automobile insurers as a basis to discriminate against the unvaxxed.”
Well, why not? One argument Bhattacharya makes is that the “result can’t be interpreted as a causal link between vax status and accident probability” and “Good or bad health could alter monthly premiums, whether any known mechanism links health status to bad driving.” But many variables used to rate auto insurance are predictive of accident risk but not causally related to accident risk. Credit scores are used by insurers in many states, and are highly predictive of accident risk, but the relationship between a poor credit score and a higher accident propensity is almost surely not directly causal. Presumably, people with bad credit scores have some other characteristic that also lead them to have an elevated accident risk, but the mechanism is unknown. Geography also plays a large roll in auto insurance rating, but again the mechanism linking certain areas with higher accident rates is unclear and may not be directly causal.
I suspect that vaccination status would be much less predictive of accident risk once a full complement of common rating variables is included in a statistical analysis. But the bare fact that the mechanism linking vaccination status to accident risk is unclear and presumably not causal is not a generally accepted reason to prohibit using it as a rating variable.
Bhattachya continues:
The issue is philosophical, not statistical. Do we really want to permit discrimination against covid unvaxxed individuals? Why not account for natural immunity, which is surely also correlated?
I say we reject such irrational discrimination on principle.
Ok, so let’s discuss “philosophy”. Reasonable people can argue about whether and when health status or (in this case) personal choices related to health should be allowed to influence insurance premiums. But Bhattacharya claims to be a classical liberal or libertarian – at least when it suits his anti-lockdown, anti-vax mandate agenda. Yet, when the possibility arises that insurance companies might charge unvaxxed people more for insurance, Bhattacharya opts for moralistic opposition to freedom of contract. It sure seems like he’s just engaging in unprincipled identity politics. The unvaxxed are his people, so he throws his libertarian principles under the bus to protect them from the possible free market consequences of their choices.
Disclosure: for what it’s worth, I worked in the auto insurance industry for many years.
“…people with bad credit scores have some other characteristic that also lead them to have an elevated accident risk, but the mechanism is unknown…”
[Could not possibly be that poor judgement and uncontrolled impatience might lead to excessive risk taking that causes increased undesirable outcomes both in using credit and in driving.]
Maybe I am just basing that on my late wife’s posthumous $114K in credit card debt along with her abnormally high record of auto accidents including two total losses, both of which were her fault.
OTOH, my wife was fully vaccinated for Covid, but was sick and missed her appointment for the multi-valent booster. Then she died before it could be rescheduled. As a heavy smoker over her entire adult life with obesity and chronic sinus allergies, then even she could do the math. It was more difficult for her to imagine losing at online gambling or making a dangerous mistake while driving distracted.
Why do people believe that their ignorance gives them any sort of rights? People should be given a choice: get vaccinated or get a shot of live smallpox.
Ooh, I like that.
One argument Bhattacharya makes is that the “result can’t be interpreted as a causal link between teenage status and accident probability”
FIFY
There is no Constitutional right to drive an automobile. There is no Constitutional right to auto insurance.
Correlation ≠ causation, for sure. But if I’m in the insurance business, and I bet against boys age 16-24, I’ll make more money than I’ll lose.
As I’ve posted elsewhere, libertarianism is the political philosophy of teenage boys and arrested development. This is exactly the sort of sophistry that passes for reason among adolescent males.
Vaccination or lack thereof is a measure risk tolerance. Risk tolerance can be correlated to risk taking behavior. It is not discrimination to consider the probability of your insured taking a risk that incurs a cost you would be liable for. When you consider all the things that go into your insurance rating, the incremental value for most of them is small. That is what actuaries are for.
Completely agree that it looks like a situation that maintenance modelers would recognize as “competing weibulls”. Strongly feel there will be little interest to fight to include this factor. Several headaches and probably little overall risk model improvement. If the vaccine was a “one and done” proposition, think it might be more tempting.
I’m wondering is how the researchers got medical records of over 11 million people to start the study. My auto insurer has not asked me about my medical conditions ever. There are conditions that get into your driver’s license, but apart from corrective vision, I don’t have any such disclosures. I very much doubt auto insurers see enough payback here to want to try it. The mechanism isn’t medical, but attitude/behavior. But data like traffic citations gets at that without trying to get medical records or eventually having to establish a policy on what qualifies in a situation of recommended multiple boosters and natural exposure. You can pick at Bhattacharya’s line of thought, but think he ends up right where insurers will end up: don’t try it. I think it won’t even be too popular with health insurers any time soon. Once you decide to put it in your risk model, you need to really model it. I bet most will still clear for 10 years or so until there is much more data.
is covid vaccination nearly an irrelevant distraction to take your eyes off the prize? take your eyes off the real remedy for excessive accidents?
forget covid vaccinations, but change all the two-way roads and streets to one-way streets and Roads!
Simplicity is the meaning of life.
~~ Occam’s first postulate ~
Judging from the paper’s abstract and some skimming, it was based on Canadian health records available in coded form for health research. There were about 11 million subjects, about 84% of them vaccinated. They searched the records for traffic accidents requiring medical care and then back checked for COVID vaccination. They found a raw relative risk of 1.72 which worked out to an adjusted risk of 1.48. That is still quite high. The effect of not being vaccinated was similar to that of having sleep apnea, somewhere between alcohol abuse and having diabetes. The increased risk was primarily explained by events when driving at night. Among the implications: emergency responders should take COVID precautions when dealing with traffic accidents.
As for Bhattacharya, it is all about whose ox is gored.