Watching Glenn Loury and Virginia Postrel’s bloggingheads discussion from a couple of weeks ago, I was struck by Glenn’s claim (at about 47:00) that bans on health insurance companies’ exclusion of pre-existing conditions constituted redistribution rather than insurance. After all, said Glenn, insurance is about protecting yourself from risks that might arise in the future, not bad things that have already happened; when an insurance company excludes a pre-existing condition or charges a higher premium for someone whose health is bad, it’s merely acting like an insurance company. Why, he asked, should insurance companies be expected to act otherwise?
That led to a bit of email back-and-forth, reproduced below without edits and at Glenn’s suggestion.
Kleiman to Loury
I watched your bloggingheads with Virginia Postrel, and disagree with you about health insurance. The problem is that, unlike having your house burn down, a really serious illness is likely to generate a flow of costs extending beyond the one-year duration of an insurance contract. Employer-based health insurance gets around that by spreading risk across employees; that may be an irrational basis for providing health insurance, and it causes all sorts of labor-market rigidities, but it solves the adverse-selection problem. Without such a solution, or the equivalent in the form of community rating, an individual health insurance market won’t spread most of the risk that households rationally want to spread.
Following Dick Zeckhauser, when I teach about income redistribution I teach it as providing insurance against the risk of being poor, which is risk all of us would insure against behind the veil of ignorance but which we can’t insure against because, by the time we’re born, we already know a lot about the actual risks we face. You can still analyze the problem using insurance categories: as usual, there’s a tradeoff between risk-spreading and moral hazard.
Loury to Kleiman
On the ‘insurance’ question, I don’t think we disagree.
Social insurance can be conceptualized as an implicit contract to indemnify people against adverse ex post social outcomes that would have been agreed to by rational individuals behind a ‘veil of ignorance,’ in a hypothetical ex ante situation. Whereas, conventional insurance is an explicit contract enter into by individuals facing statistically independent risks which achieves a more efficient allocation of risk-bearing overall.
My point, though, is about how to implement a social compact that might be rationalized as ‘insurance’ for people behind the veil. A single-payer system would implement this compact in one way; another, and more complicated way is the current reform proposal, which works through conventional insurance contracts, but burdens them with the commitments of social insurance. This institutional framing leads silly things being said – like, insurance companies ‘discriminate’ against people with pre-existing conditions – through the pricing and coverage dimensions of the contracts they offer in the marketplace. Well, of course they do. Otherwise, they’d not be in the business of assessing and pricing of risks – which is to say, they’d not be selling ‘insurance.’
I also agree that one needs a broad pool of potential risks in order to implement either type of insurance: social, or conventional. But, whereas the goals of conventional insurance (efficient allocation of spreadable risks) are fostered by the segmentation of these risk pools into as many differentiated sub-categories as possible (the better to identify and price the anticipated risks facing any given insure), the goals of social insurance (the better to indemnify ‘losers’ in life’s lottery against the already known consequences of their misfortune) are frustrated by such differentiation. Segmenting the risk pool is what private insurance is all about, and should be if the goal is economic efficiency. Aggregating the risk pool is what social insurance is all about, and should be if the goal is social equity.
To which I would add only one thing: Â Though it is, indeed, a little bassackwards to run a social insurance program through private insurance companies, health insurance needs to be social rather than private. So if the choice is between regulating private health insurance to make it more “social” and leaving the health insurers to run their business according to the logic of underwriting, we ought to choose the former. There are good arguments for single-payer, but there aren’t the votes for it. Â And we have long done things in this country (telephone service, electric power distribution) via regulated private firms that other countries do, or used to do, via public agencies.
So I’ll take second-best rather than sticking with what we have now.
That description of "conventional insurance" seems a bit utopian to me. Sure, it's the way you might want things to work in a world where every market had many competing insurance companies and all of their risk models and payout histories were readily accessible to prospective customers for analysis. But it's nowhere near the way that conventional insurance appears to work, whether medical or otherwise. When you have property-insurance companies using house-by-house topographical, criminal and other data to price their products, you're not really spreading risk, you're concentrating it. Same thing when you look at how pre-existing conditions are used, either in setting prices/rejecting applicants or in rescinding policies once issued.
In the real world, one or two firms dominate any given market, their data is secret, and they have enormously more information about potential costs than customers do. So we're no longer really dealing with a risk-spreading calculation, we're dealing with a monopoly revenue-maximization calculation. (One sign of this is that insurance companies don't do much intertemporal risk-spreading — it's easy to imagine longterm policies that, like mortgages, would commit both side to mutually beneficial behaviors, but churn seems to be a more successful business model.)
There's ultimately a bigger question here: can "conventional insurance" survive the information age? The more information that's available (purchasing-record feeds, anyone? british-style traffic and CCTV data?) the more closely individual risks can be parsed, leading to both cherry-picking and adverse selection, depending on who has a temporary information advantage. In the absence of a more or less voluntarily imposed partial veil of ignorance, it's hard to see how ever-increasing information and analysis won't lead to a death spiral.
I'm extremely late to this post, but hope I won't be lost in the 'tubes for all that. Just out of college, I worked as an actuary, computing required reserves for a variety of lines of insurance – workers comp, auto, commercial liability, etc. Characteristic of all these is that the insurer pays out on the basis of when the insurable incident occurred, not when the charge came due. So payouts for a given batch of policies might extend for years beyond the expiration of those policies, especially in the case of commercial liability.
It has always seemed to me that if health insurers want to exclude pre-existing conditions, then at the very least, they must be held to pay for all claims, whenever charged that arise from a given diagnosis. That is, it seems to me that the diagnosis is what ought to trigger insurance coverage rather than the 17th visit to the chemo lab.
This doesn't bear on the distinction between private and social insurance, but does seem pertinent to any discussion of how private health insurance works.