Gene pools and risk pools
Why ignorance might be bliss
A few thoughts about health care, genetics and (wait for it …) post-Keynesian economics spurred by an incredibly interesting paper which crossed my desk the other day, sadly not available in electronic form.
The standard argument about genetic testing an health insurance is well known, and I’m assuming that all my readers have a reasonable intuitive grasp of it. A market for health insurance can’t operate if people can have genetic tests but are allowed to keep the results secret from the insurance companies, because if this happens, the only people who will want to buy insurance are the bad risks. Because, at every price of insurance, the people who choose to buy represent an adverse selection from the overall risk pool, no market-clearing price for insurance exists. One can dress this model up with the maths to prove it, but the intuition is there in the old fashioned actuarial distinction between insurable and uninsurable risks. Private information destroys insurance markets.
So, the debate is usually carried out in terms of the ethical dilemma of forcing people to share their test results with the insurers, and then what happens to the (presumed large) group of people who get priced out of the market? Usually, this is discussed in terms of the subsidy implicitly provided by the genetically healthy to the genetically inferior under current insurance arrangements (note that, for the purposes of this analysis, single payer schemes are also “insurance”), and the debate is more or less unpleasant and uncomfortable depending on how squeamish you are about Social Darwinism, interference with the market, or the disappearance of the health insurance industry. I have a lot of problems with arguments on all sides of this debate, but I’ll keep them for another time, or this already huge post will assume elephantine proportions.
What I want to advance is the cheering thought that things might be even worse than that; given sufficient amounts of genetic information about susceptibility to expensive illnesses, health insurance might become absolutely impossible to provide for anyone at all, no matter how fine their genes. Why? Let me give you a famous Alfred Hitchcock story:
Hitchcock would occasionally say that what he really wanted to do was to make a film about the Titanic disaster. People would be slightly bemused by this ambition; after all, everybody knows what happened to the Titanic, so what attraction would the story have to the Master of Suspense? “Ahhhhh”, Hitchcock replied “but they don’t know when it happens!”
Healthcare costs are like that. Let us say that, genetically, my thyroid gland is a racing certainty to swell to the size of a balloon and require some expensive surgery. It could happen tomorrow, or it could happen in thirty years. Nobody knows what suddenly triggers these dodgy genes into action. But, from a financial point of view, this makes a hell of a difference. For one thing, from a present value point of view, it makes a hell of a difference whether it happens tomorrow or in thirty years’ time; specifically it makes the difference of thirty years’ investment returns, assuming that the insurance company is being run on an ongoing combined ratio (premiums/costs + claims) of 100%. But more importantly, it means that you can know all that there is to know about my genes — stronger than that, you can know all that there is to know about the chances attached to every possible health outcome for me, and price insurance to me accordingly — and still have significant variance in your underwriting profit & loss account. In fact, somebody has done the maths and, on plausible assumptions, even if it were possible to precisely measure every conceivable factor contributing to an individual’s health care spending, it would still only be possible to predict about 20% of the variation in annual expenditure. And since insurance companies have to maintain their solvency on a year-to-year basis as well as on an actuarial long term basis, this residual variance in their cashflows matters.
And it matters a hell of a lot, quite possibly, for two reasons. First, when start refusing to insure or quantity-rationing a particular group (the genetically inferior), the mathematics of risk-pooling start working against you. The dispersion around the mean of your costs and revenues is likely to be greater as a percentage of the mean, and you are more likely to get a “tail” outcome in any particular period. The genetically inferior might be money-losing propositions in the long term at the pooled rate of premiums, but they contribute a lot to the year-on-year stability of costs and revenues through their sheer weight of numbers. Second, if you are selecting particular groups to offer more attractive terms to and therefore sell more insurance, you are most likely increasing the correlation of your risks. If there is a particular genetic type that is over-represented in your portfolio, then you are exposed to the risk of a particular set of environmental conditions which is associated with medical problems in that type (say there was a set of people who were largely genetically invulnerable to everything except melanoma; you’re bearing the risk of a hot summer). If you’re correlating your risks and reducing the size of your pools, then you need to hold more capital against possible tail outcomes, and this can jack up your costs to the point at which it’s uneconomic to provide insurance in the first place. But on the other hand, you can’t go back to the old pooled equilibrium where the genetically fit are cross-subsidising the unfit, because that’s no longer a stable market; you’re open to a competitor taking away your fit customers.
It’s a genuine problem, and it comes about, not as the result of anything intrinsic to do with genetics or health, but because of the fact that there is time and uncertainty in the model (which is the only real connection to post-Keynesian economics; note that this was all worked out in a completely ergodic framework). It’s as well to remember that a model which works fine in static time on an actuarial mean-variance basis, will have to be implemented in historical time, and one in which liquidity in the short term matters (liquidity is present as a constraint in this model through the need of the insurance firm to maintain short-term accounting solvency as well as long-term economic solvency).
Solutions? Sorry, don’t really have one, unless one seriously thinks that the genie of genetic screening can be pushed back in the bottle. I’d note, however, that the engine of most of these “problems of asymmetric information” (in this case, the adverse selection problem which makes the pooled equilibrium solution with private information untenable) is usually an embedded option. In this case, it’s the option of the insured party to choose whether or not to buy insurance. Since you can’t force them to buy the product, they will only do so when it’s to their advantage, and this turns out to be enough to knock down the existence of the market. I speak as a member of a health insurance scheme (the National Health Service) which doesn’t have the property that you can refuse to buy it if you don’t want it, and would humbly suggest that something along these lines might help the insurance industry out of what might turn out to be a nasty hole.
After all, shallow pools are the safest for swimming, but the most dangerous for diving into.