• Not all adverse selection leads to a death spiral

    Health plans in competition are unlikely to attract the same types of consumers. Some plans will be more attractive to relatively sicker people, thereby experiencing adverse selection. Others will be more attractive to relatively healthier people, thereby experiencing favorable selection.

    Upon making that observation, it’s become fashionable in the health policy commentariat to make a nod to, if not build an entire argument, around the possibility of “death spiral” — that the plan(s) with the most adverse selection will enter into a downward spiral of premium hikes and loss of their (relatively) healthier customers, ending in smoldering ruin.

    This is, of course, a possibility. But it isn’t a foregone conclusion that adverse selection leads to a death spiral. Were that the case, then whenever plans competed it’d be a fight to the death. But one doesn’t have to look very far to observe plans co-existing and in competition for years. It happens in the Federal Employee Health Benefits Plan. It happens in Medicare Advantage. It happens in Medicare Part D. It’ll probably happen in some of the new Obamacare marketplaces.

    Apart from risk adjustment and other interventions that help plans survive adverse selection, the very nature of the market can also lead to inherent stability. Roger Feldman and Bryan Dowd explained two of them very nicely in a 1991 paper in the Journal of Health Economics. (I thank Dennis Shea for the tip.) They imagine an HMO in competition with a fee for service (FFS) plan. The former has a restricted network and lower premium, attractive to relatively healthier consumers (experiencing favorable selection). The latter has an unrestricted network and higher premium, attractive to relatively sicker consumers (experiencing adverse selection).

    We join the story, already in progress, as the FFS plan is losing market share to the HMO. Is a death spiral underway (cue ominous music)?

    Proponents of the ‘death spiral’ argument implicitly assume that the fair premium difference between the HMO and FFS rises as HMO enrollment share rises. They seem to forget that the HMO’s risk pool must become worse as its enrollment share increases. The only way the HMO can get a larger share is to attract FFS enrollees whose marginal risk is greater than the present average risk in the HMO. Thus the HMO’s average risk must rise. Whether this factor, by itself, stops the death spiral depends on the distribution of risks. For example, it can be shown than an exponential distribution of risks leads to a premium difference that widens as HMO enrollment share increases. The uniform distribution assumed in our earlier paper and explained above, implies a constant premium difference.

    The second critical factor is the strength of peoples’ preferences for FFS insurance []. Proponents of the ‘death spiral’ argument either overlook this factor entirely or implicitly assume that even the last FFS enrollee would switch plans for a fairly small out-of-pocket premium difference. To the extent that FFS enrollees are loyal to their plan, this assumption will not be accurate and the premium spiral will tend to be self-limiting. We have found empirical evidence that employees will not readily switch plans if their present plan allows unrestricted choice of health care providers and the alternative plan limits this choice. This would be the case in a firm that offers a FFS plan and a group, staff, or network model HMO.

    It’s worth mentioning that a fair amount of discussion about adverse/favorable selection these days isn’t between plans but between an entire market (an exchange) and options outside that market. That is, the question is often, will the selection into exchanges be more adverse than anticipated? Though exchanges have plans in competition, this question is analogous to imagining an exchange is just one big plan and consumers are choosing between it and nothing.* The cost of nothing does not increase as its market share grows (exchange participation shrinks), so at least one of the stabilizing mechanisms Dowd and Feldman describe doesn’t apply in this case.

    * Some individuals have other worthwhile outside options, like employer plans or non-exchange, individual market plans. If one considers these, the story becomes complicated. I’m ignoring them for that reason.

    @afrakt

    Share
    Comments closed
     
    • I recall having made a similar argument somewhere recently. The media’s “death spiral” really doesn’t exist in the academic literature. If you look at a very textbook case of adverse selection like Einav and Finklestein: http://economics.mit.edu/files/5810 then there generally is neither a “spiral” nor a total collapse of the insurance market. Only in the most extreme case where average cost exceeds willingness-to-pay for all individuals do we get a total collapse of the market–but such a case probably can’t happen because it requires that the first individual has absolutely no risk aversion, and subsequent individuals have very little risk aversion, which is unlikely. The reality is that adverse selection is self-limiting. We can’t even say that adverse selection leads to less than the optimal level of insurance, since there could be redistributive gains from the policies that cause it, for example. All we can say for sure is that adverse selection causes some amount of deadweight loss.

      On the other hand, the “spiral” part of the “death spiral” is a misinterpretation of the underlying model. Again, the textbook case of adverse selection is a static model–in equilibrium, inefficiently few people have insurance–the model says nothing about how we get there. Indeed, the only way you get a “spiral” is if you assume that insurers violate rational expectations and naively price their policies based on their previous period’s risk pool, rather than their expected risk pool.

    • What if one or more plans of a single insurer do suffer a death spiral though the other plans of the same insurer don’t. Does that make the exchange a failure or a success? It seems that the decision not to adopt federal (uniform) standards, thereby permitting a proliferation of plans with marginal differences, will reduce the overall risk to the insurer while increasing the risk to particular plans. I’m not sure what the consequences will be, but whatever they are, I suspect they will come as a surprise to a lot of experts who designed the exchange.

    • Austin, the 3 examples you gave of plans that survived death spirals —
      Fed Employees, Medicare Advantage, Medicare Part D — had this one thing in common: the government was paying most of the premium.

      In the states that had a death spiral in the individual market – New York, New Jersey, Kentucky for a time, Washington for a time — the market really did just about collapse. Individual policies in New Jersey were costing well over $1,000 a month, which is not absolute collapse but close enough in my book.

      That is because healthy people had to pay the entire premium, and they just would not do so during a death spiral.

      That is not a very elegant way of phrasing this trend, but I think it is accurate.

      With the ACA, the subsidies will make a death spiral move more slowly. But the individuals without subsidies will have a real problem.

      • FEHB isn’t a government subsidy in the same way as Medicare. The plans differ quite a bit from each other and their are different patterns of provider participation, as well. What’s critical here is that regulation enables a market to exist. Absent that, the distortion that’s endemic to health care markets takes over.

    • Whenever plans do compete it is a fight to the death. The reason you see plans co-existing for years is because they really don’t compete. The health insurance industry has always been highly regulated for half a century. It is a state sanctioned and protected cartel. The ACA is just the latest iteration of how that cartel is regulated.

    • I’d like to know if there’s anything in law that prohibits an Amerikan from buying Mexican insurance for treatment in Mexico? It seems to me a bundle could be saved by a young man who refused Obamacare and just got Mexican insurance.

      More sensible than Mexican insurance, of course, would be to refuse Obamacare and bank the savings for eventual self-pay treatment in Thailand, India, Brazil, Argentina, Mexico, Hungary and the Czech Republic, where great savings in medical care and drugs can be realized.

    • Another stabilizing element is the provider network. If the networks are materially different from one plan to another, switching among plans is reduced, enrollment is more stable. Inertia and status quo bias also play a role.

    • Adverse selection is probably not a problem if everyone has to buy health insurance, but when Washington State guaranteed coverage, but allowed people to wait until they got sick to buy it, there were serious problems. One of the tricks is that most people would actually rather be healthy than sick, despite the myriad economic theories to the opposite. (After all, one gets more from the system if one is suffering from ill health. At some point the marginal utility yadda yadda. Google it.)