Reductions in the cost of genetic testing and improvements in what we know about what it tells us produce obvious benefits; if you know you are likely to have some particular medical problem, you may be able to take precautions against it. But they also have at least one potential downside.
The more is known about the chance of bad things happening to us, the less able we will be to insure against them.
A solution to this problem that is sometimes proposed is to permit individuals to have their genes tested but forbid insurance companies to require testing as a condition of insurance or to use the information it produces. The problem with that is adverse selection. If the customer knows his risk and the insurance company doesn’t, high risk and low risk customers are charged the same price, making insurance a good deal for the former and a bad deal for the latter. Insurance companies, realizing that most of those who choose to buy their insurance are bad risks, will charge accordingly, driving more of the low or average risk customers out of the market. In the limiting case, insurance is bought only by high risk customers, at a high risk price. A famous description of the problem is Akerlof’s article “The Market for Lemons.”
If we allow both insurance companies and their customers to make use of genetic information, then both high risk and low risk customers can buy insurance, but at different prices. The risk of having genetic variants that make you more likely to suffer some expensive medical problem is uninsurable, although you can still insure against the risk that, given those genes, the problem will actually appear.
The theoretical analysis of the problem is straightforward; interested readers can find one version in Chapter 6 of my Law’s Order. But the theory does not tell us how large the problem is. That depends on empirical facts, in particular on how much the information provided by genetic testing affects the expected cost of insuring someone.
As it happens, I recently came across a datum relevant to that question, as a result of having my own genes tested by 23andMe, a company that does mail order genetic testing. It turned out that I had a genetic variant that implied a moderately increased risk of meningioma, the second most common type of brain tumor.
The information came a little late to be useful. Last summer, while I was part of a group on World of Warfare, one of the other players noticed that I had stopped responding. He called the house. My son took the call, came into my office, and found me half conscious on the floor. The diagnosis at the local hospital was meningioma, a benign (i.e. non-cancerous) tumor inside my skull but fortunately outside my brain. It was large enough to put pressure on my brain, so required surgery. I got surgery, all went well, and I am now fully recovered, aside from a visible scar and a tendency of my scalp to itch.
According to 23andMe, 35,000 Americans a year are diagnosed with meningioma, and in most cases the tumor is small enough not to require surgery. Assume that 10,000 of those, like my case, do, making the annual probability for a random American 1/30,000. Further assume that the average cost is $100,000. That’s the right order of magnitude—I saw the figures for what it cost my insurance company, but don’t have them ready to hand at the moment. The average cost to the insurance company of that particular risk is then about $3.
Finally, assume that my “moderately increased risk” means twice the average risk, which seems if anything a high guess. It follows that in a world where insurance companies had and used that data, my medical insurance would cost me, or my employer, three dollars a year more than in a world where the data was not available.
There are, of course, lots of other risks that my health insurance insures against. For some my genetics are presumably favorable, for some unfavorable. It would require much more information than I have to estimate how much the cost of insurance would vary from one person to another if all of that information was available and used. But at least the single datum I happen to have suggests that the effects might be small.
David Friedman is a law professor at Santa Clara University. He blogs at Ideas.