How Many People are Really "Uninsurable"?

In a recent Econtalk, Ed Dolan attempts to make the argument that a fully free market in health-care is not feasible. He claims that some people are completely uninsurable and cites some (very) rough figures on the distribution of health care cost on individuals. I think this is a bad argument, and for reasons I've stated several times before.

The first problem is that health care spending is very, very asymmetrically distributed, and it goes by basically a--some people call it the 80-20 rule, that 20% of people account for 80% of health care spending; and in fact the top 1% account for about 10% of health care spending. So, the result of that is that there are a lot of people for whom it is true that their health care spending needs exceed their income. In fact, exceed their entire lifetime income in a certain number of cases. Now, of course, it's also true that if your house burns down, the cost of rebuilding your house exceeds your income; and we solve that through insurance. But, health care needs are increasingly uninsurable, because in order to be insurable, a risk has to be fortuitous--it has to be due to random chance. But an increasing number of health care risks are predictable on the basis of pre-existing conditions or things that are determined, testing that's determined before you are born. So, we have this combination of catastrophic risks--which are risks that exceed your ability to pay, sometimes even on a lifetime basis, not just on current income--and we have uninsurability.

The problem with the claim that 20% of people account for 80% of the costs is that it's not grouping things the right way. I've seen this done in a very sloppy way, but as an actuary I'm used to seeing it done the correct way. The wrong way to do this is to line everybody up in order of their actual medical spending, pick out the top 20%, and calculate their share of total medical spending. The skewness of actual spending tells you nothing whatsoever about whether a risk is insurable. By this standard, fewer than 1% of homes account for 100% of fire-related homeowners insurance claims. About 1% of drivers account for 100% of bodily injury liability claims in auto insurance. Of course, all of these risks are actually quite insurable, and there are thriving insurance markets in these areas. The distribution is skewed, but not nearly as much as this presentation implies. (This Incidental Economist post by Austin Frakt has some figures, which are in the range of what Dolan says.) This greatly exaggerates how many people are truly "uninsurable."

The correct way to do this is to rank people by their expected costs, then take the top 20% (or whatever quantiles you're doing this for) and calculate their share of total medical spending. Expected cost is calculated based on some statistical model applied to a large database of historical costs, using predictors like age, gender, general health, and treatment history. For most kinds of insurance, annual premiums are determined based on these results. Doing the calculation this way, grouping by expected rather than actual costs, you don't see such extreme values. The top 1% of auto risks (measured by their expected costs, which is reflected in their policy premiums) might be responsible for, say, 5% of claims. Even these very bad risks can usually find insurance.

There are a lot of people who are predictably unhealthy: they have chronic conditions or genetic predispositions to something. But within this population, in a given year some of them will require a lot of care and others won't. Some will have a medical episode of some sort: an infection, a fall, a complication. It will result in their consuming a lot of health care in that year and possibly for a several years in a row. But the question of which person in the high-risk population actually incurs these costs is still a random guess. Contra Dolan, these costs are fortuitous (meaning random, not the colloquial meaning of "fortunate"). They are imminently insurable. If Dolan can show that health costs are as skewed as his presentation implies after doing it the right way, I might concede that there should be a government risk pool for the very highest cost (again, expected cost) individuals.

This is disappointing, because Dolan even mentions house fires and states why they are insurable. He should have had the tools to see why his "skewness" argument doesn't make sense.

Contra his argument on pre-existing conditions, those risks are also insurable. They key to solving this problem is getting people to insure against the risk of acquiring a pre-existing condition ahead of time. If you sustain an injury that requires years of expensive treatment, that's a "pre-existing condition" if you attempt to switch health carriers. But if you injure someone with your automobile, your insurer at the time of the accident will cover the costs of that injury (up to the policy limits), even for care that happens years later. Health insurance could (and should) cover injuries the same way that auto insurance does. It's just as feasible that we could sell insurance that pays out if you are diagnosed with some kind of chronic condition (and my post lists several ways of doing this). Dolan mentions the possibility of testing for conditions prior to birth. Those conditions are insurable, too. Expecting parents can purchase a policy that will cover a child's genetic conditions, so long as they purchase it before any genetic testing is done. This could (and should) even be the default setting for a family health insurance policy, so they don't have to worry about buying it at exactly the right time (after the child is conceived but before any neonatal care is done). This way, the presence of a genetic condition is fortuitous, from the point of view of the insured and from the point of view of the insurer. All of this admittedly requires overhauling of our existing health insurance markets, but it's doable. What I'm describing is basically how other kinds of insurance work. Life insurance and chronic illness insurance policies often have long coverage terms, (very) high face-values, and affordable premiums. Covering the prospect of a pre-existing condition would be no different.

Dolan says some other things that I found off-putting. Russ Roberts mentions that people spending their own money would make wiser health spending decisions. They wouldn't blow their children's inheritance on last-ditch hail-Mary medicine. ("You wouldn't impoverish your children" he says.) Dolan stops him bluntly and somewhat rudely: "Stop right there. That's false." False? Always false? It's sometimes false? Is there a careful experiment or handy dataset? Oh, yeah, there is. Oh, and there's more. Roberts is basically right here, according to the best randomized controlled studies we have, the RAND health experiment and the Oregon Medicaid Experiment. People who have to pay more out of pocket (because they have less insurance coverage) consume less health care, and they don't experience a deterioration in health outcomes for it. Dolan is a health economist. He should know that.

Of course, I'm fixating on the negative. I'm specifically nit-picking at Dolan's errors (they are errors in my opinion, anyway). Don't let that put you off listening to the full podcast. It was a very interesting discussion.