Time for another go at this piece of conventional wisdom, I think:
The time is fast approaching when we will have an inexpensive test that is capable of revealing a person’s genetic propensity to contract a broad array of chronic diseases. That means that we will be able to accurately assess the cost of medical treatment over their lifetime . . .
No. I am not going to definitively say that no such thing will ever be possible in any factual or science-fictional world, but the second sentence does not follow from the first. I will assert with truculence and jutted jaw that the derivation of commercially relevant and actionable insurance information from genetics is much, much more difficult than that, and furthermore, will tentatively and politely advance the possibility that to "accurately assess the cost of medical treatment over a lifetime" might end up being at least as much of a tough-nut as the socialist calculation problem.
As the rant linked above tries to point out, the big issue here is the "Titanic problem" (from Hitchcock's aphorism about it being possible to make a suspenseful film about the Titanic given that everyone knows that it sinks - they don't know when). Knowing genetic propensities to develop various conditions is just about one tiny baby step along the way to making a cost estimate of the sort described. Just reeling off the other forecasting fun from the top of my head:
1) How long will it take for the condition to develop? (ie, how many years' premiums will the punter have plugged in before he needs the disease)
2) How much will the condition cost to treat when he gets it?
3) How does your answer to 2) depend on your estimate of medical cost inflation?
4) How does your answer to 2) depend on your assumptions about technological progress in medicine - given that we are talking about thirty or forty year horizons at a minimum here?
5) How do your answers to 3) and 4) interact, given that technological progress can either facilitate the prolongation of late-stage care at massive expense, or make hugely debilitating conditions curable with a twenty-dollar appliance?
6) How do the genetic markers interact; what is the possibility that a marker for a profitable condition (ie, ideally, one that kills the insured stone dead with no warning after having paid a lifetime's premia) is correlated with an unprofitable one?
7) How will the customer's estimates of 1-6, plus your terms and premiums, affect his or her behaviour?
This looks totally impossible to me; 4) is the real killer I think. People who comment on this issue almost always do so on the basis of genuine progress by the doctors in diagnosis and prognosis, without usually considering the fact that medical risk to the insured has to be mapped onto financial risk to the insurer.
However, wonderful, wonderful mathematics provides an easy way on this one. The mathematics of risk pooling is the closest thing to a free lunch as you are ever going to get. Because, although the Titanic problem is massive, you can be sure close to certainty that the Grim Reaper is going to play that fuck-awful Celine Dion song to all of us, sooner or later. If you give up on price-discrimination, then what you are modelling is the average outcome for the average patient, and he's a much more tractable soul.
In general, standard terms and risk pooling are proven technology. They work, and they have worked for about 250 years, give or take. Risk pricing, risk management and discrimination have a much shorter history and have caused quite a few spectacular flameouts for their users in their short existence (note that this wasn't the case for the early days of life actuarial science - the compilation of the first mortality tables more or less did for the benefits-societies and became hegemonic precisely because they worked straight from the off).
I have a fairly deep respect for actuaries as a profession. They've had their scandals, their Equitable Lifes and so on, but consider this - for 250 years, they have been in charge of a very complicated and very important set of social institutions, and not once have they totally destroyed the thing they were put in charge of through intellectual arrogance and bad faith. Neither bankers nor socialists are able to make the same boast.
endnote: I think that a lot of people also get misled on this one by the observable fact that the insurance industry does in fact take a very strong interest in genetic markers and prognosis. In my opinion, this is not because they have some dream of hunting down the chimera of perfect price discrimination, but because they're worried about good old fashioned moral hazard; they're worrying about the patients having an information edge over them. If you have good information about the likelihood of something happening to you, you can load up on coverage and effectively use your insurance policy as a low-cost savings vehicle, to the considerable expense of your insurer. Genetic testing (and this ought to be obvious really given that the basis of insurance is risk-pooling) is a threat to health insurance, not an opportunity.
I think you're confusing the issue a bit on the threat/opportunity point (though the post overall is very well-taken). To say it in the language of your Fatty example, suppose instead that Fatty can't tell his own actuarial risk but that there's some hope of an insurance company developing a fatty detector. This is super-valuable for whoever has it first, since they can now price discriminate fatty out of their risk pool while keeping the Jim-Bobs in. The overall effect is a transfer of wealth from the dumber industry players to Fatty as no company without the fatty-detector can offer a competitive contract to the Jim-Bobs.
ReplyDeleteWhat's really going on here of course is that the insurance industry is fiercely competitive, so one man's price discrimination is another man's adverse selection -- you have to chase the chimera because someone else might catch it (see GEICO, Progressive).
"...but because they're worried about good old fashioned moral hazard..."
ReplyDeleteI guess you meant adverse selection and not "moral hazard"?
good point.
ReplyDeleteAlso, why do we expect technological progress in medicine to make its marginal cost of production go up? The world would be very different if this was true of electronics, transport, food, basic industrial products, textiles, so on.
ReplyDeleteIs this something to do with intellectual property or market structure? Interestingly, the only other industry I can think of that exhibits the same inverted Moore's law is armaments.
It's because it's so damnably nonergodic I think. Locally, all the cost curves are downward sloping, but the aggregate is subject to leaps when someone invents a new technology that is really expensive but does something that couldn't be done at all before - it's basically a problem of very rapid depreciation-by-obsolescence.
ReplyDeleteIt might be similar in consumer electronics (although your other examples are deffo right) - each individual item has been massively subject to declining cost curves, but I bet that the aggregate cost of the electronic devices in my house has gone up massively.
Interestingly, the only other industry I can think of that exhibits the same inverted Moore's law is armaments.
ReplyDeleteAnother one: Theatre and the performing arts. Apparently Broadway tickets have had above-inflation price rises for the past fifty years; part of the explanation is that it's remarkably hard to improve actors' productivity.
It's not just that the second sentence doesn't follow from the first. The first sentence isn't true either.
ReplyDeleteAlthough many people selling genomic research would like to forget it, risk prediction works by pooling as well. Even if we assume that there exists some function of the genome that is a good predictor of risk, we still have to estimate this function. We can't do that statistically from remotely feasible numbers of people unless the genomic information is actually pretty low-dimensional, which seems unlikely. That means we would need in silico models of sufficient accuracy to simulate the effects of mutations, which certainly are not just around the corner.