Daniel McFadden

Sick Insurance: Adverse Selection and Regulation of Health Insurance Markets

Category: Lectures

Date: 26 August 2011

Duration: 31 min

Quality: MD SD

Subtitles: EN

Daniel McFadden (2011) - Sick Insurance: Adverse Selection and Regulation of Health Insurance Markets

When heterogeneity in consumer tastes and needs, and in cost and quality of products, are publically observable, markets can price, sort, and match these variations, and product choices made by consumers yield demand signals that foster efficient resource allocation

Thank you. Heterogeneity and consumer taste and needs and the cost and quality of goods and services are ubiquitous features of resource allocation. But as long as these are publicly observable, markets can price, sort and match the variations efficiently. For example for insurance which is what I'm going to concentrate on today, you can have an efficient insurance market where there is no private information that allows people on one side of the market to improve on the prediction of expected benefits based on public information. If you have those circumstances then you can have competitive underwriting in which the prices of insurance contracts are set at their actuarial value plus a competitively determined load and that market will operate efficiently. And incidentally an essential aspect of a well-working insurance market of that form, is that issues like the ability of insurers to re-price contracts or to cancel coverage if a buyer has a bad experience, are causes that will themselves be negotiated in the competitive market as contract terms and they will be appropriately priced. Now an example of an insurance market which does not satisfy these conditions is the market for health insurance. And there are a number of reasons, it's really a perfect storm of problems of failure, of market failure. There are various sources for this, one which I'll talk about subsequently is that there may be social sentiment for exposed equity or fairness and those can conflict with what would originally be a perfectly legitimate and efficient market outcome. That is to say you can have an efficient market outcome which is socially unacceptable and the response to that may be a factor to upset the market. Perhaps appropriately but nevertheless it's a problem to deal with. Other issues in this market arise from asymmetric information. And they are the classic problems of adverse selection. Where people are denied coverage or offered actuarially very unfair contracts because of the possibility that they are bad risk, moral hazard where coverages becomes very expensive because providers lack incentives to minimise treatment cost or withhold unproductive treatments. It's part of a principal-agent problem. Performance risk in which insurers evade paying benefits by making contract enforcement costly. A general problem in contract theory where you have asynchronous operations of the transaction. And finally signalling and screening as a device used by market participants to avoid either being lumped in together with high-risk peers or having to pay benefits to high-risk people. Let me say first something about the problem of social unacceptability of market outcomes, that's particularly an issue for health insurance. So consumers and society have difficulty refusing or withholding treatment after bad health outcomes are realised. There's a social sentiment for exposed equity and this can conflict with competitive underwriting. A good example is if someone had kidney failure, they are very expensive to treat, but social sentiment is that as a human right they are entitled to treatment at some affordable cost. Now the problem of conflicts between socially acceptable, what's socially acceptable and market outcomes could probably be taken care of if consumers could acquire unconditional lifetime health insurance ex ante, before their health status is ever determined. And if they chose not to buy health insurance they commit irrevocably to the consequences of having limited ability to pay. But neither of those conditions are met in reality. A consequence of that is that it's often the practice in health insurance and in other areas to forbid competitive underwriting on the basis of some criteria with the objective of trying to mitigate market outcomes that are deemed unfair. For example it's often in health insurance, a rule which prohibits discrimination on the basis of gender or on the basis of pre-existing conditions. Now from an economic point of view if the regulation requires these kinds of non-discrimination rules and that results in coverage for high-risk consumers at less than their actuarial value then the unintended consequence is that unless the market is subsidised, the premiums for the low-risk consumers are going to have to rise above their actuarial value, they will be penalised. And there is effectively an income transfer from the low-risk to the high-risk consumers. And this in turn can induce a form of adverse selection, it's now not coming from asymmetric information but simply from social regulation to try to ensure fair outcomes. Very difficult to have the inter-personal transfers required to achieve this level of fairness in a system that does not in itself introduce a substantial market distortion. So the type just mentioned. So very careful regulation is needed to try to deal with these issues of social fairness. Asymmetric information is the most common reason cited for failures in insurance markets and particularly health insurance markets and they are these classic triumvirate adverse selection, moral hazard and performance risk or counter-party risk. There's a long history in economics and pre-economics of attention to these sources, this is not, this is a very, very old subject and to partisans of market solutions who think carefully about it. You need to think seriously about whether you're in a market situation where these problems can be overcome or whether something of substance has to be done to overcome. There's a lot of other literature on markets under incomplete or asymmetric information. And I'm not going to spend time going through this but if you flash some names there you might recognise some of them. One comment, I'm talking about health insurance here today specifically but many insurance markets and markets in which, basically which depend on contracts, share many of the same problems. Labour markets and markets for financial derivatives which are insurance contracts have problems that are quite similar to those that arise in the health insurance market. I'm going to remind you of some price theory that I believe everyone has seen in their first graduate class in economics and these are some diagrams from Rothschild-Stiglitz, I would be cautious about putting up diagrams but in this case you've all seen these before so I just want to remind you quickly of what happens in an insurance market with asymmetric information. So here's a diagram with contingent commodities, the consumption of healthy or consumption of sick, and there's an endowment point in which the person will have less resources to consume and will incur a loss, if they're sick the loss is in red here, loss L, a 45 degree line that corresponds to equal consumption in either state. The budget line in this case is the locust of actuarially fair trades, that is to say insurance contracts which will break even for the insurer. And when individuals are risk adverse, expect a utility maximisers, the slope of their indifference curve at the 45 degree line gives the odds of being sick. And if you have actuarially fair insurance there is an optimal contract, in this case located at a little arrow from full insurance, you'll have full insurance, so that would be the point in which this particular consumer would locate. Very standard analysis. This works perfectly well if there's more than one class of consumers, you could have robust consumers, frail consumers, frail consumers have a higher probability of getting sick and if they were, if their risk class were observable then the market would simply offer 2 insurance contracts, F and R, each would get full insurance and at premiums which reflect the actuarially fair cost of insuring against their losses. Now suppose that the risk classes of the consumers is not observable, so that when an insurer offers a contract, he does not know ex ante whether the buyer is robust or frail. In this circumstance one possibility is that all consumers, both frail and robust would buy a contract and in this case you may get an actuarially, a locus of actuarially fair contracts. Which in this diagram is labelled the all-in break even contracts. And that's simply an average of the break even budget lines for the 2 classes, the location of the average will depend on how commonly people are robust or frail. In this diagram I've labelled what I call a candidate pooling contract. This would be a contract which is break even for a competitive insurer, if it was in fact purchased by all people. Detail on that, this is just a blow up so you can see a little better what's going on. And that particular contract, pooling contract, P, is located at a point of tangency between the robust consumer's indifference curve and the actuarially fair line or the break even line for pooled contracts. And that contract P then has the property that robust consumers would choose not to move. The higher risk consumers, the frail consumers would like to buy more of at that price but that's not, in this case not being offered. In the market there would be no incentive for an insurer to offer a broader coverage because only the frail would buy it and it would not break even. But this candidate cannot survive as a Nash equilibrium, the reason is that they are blocking contracts. In this case it's noted in this diagram by a little blue dot. And it's a contract located in the area which is better for the robust but worse for the frail. So what would happen is in the existence of the contract P, if an insurance company now offered the blocking contract, it could entice away all of the robust consumers from which it could then break even or better and leave behind all of the frail consumers at P in which case P would no longer break even. And equilibrium would break down. So this is the original Rothschild Stiglitz argument that a pooling equilibrium cannot exist. Another possibility that they considered and that occurs in this markets is a so-called separating equilibrium where each risk class will have its own policy and the policies now are positioned in commodity space in such a way that the high-risk consumers will not be tempted to try to misrepresent themselves as robust consumers and move to the contract being purchased by the robust group. And in detail you get a diagram which looks like this, the frail consumers are getting full insurance at a high premium, actuarially fair premium, the robust consumers are getting partial insurance and with the location of that being such that R is not a utility improvement for the frail consumer so they have no pause in it to move away from F to R. Now in the diagram as drawn here note that the indifference curve of the robust consumers does not intersect the all-in break even contract. And if you do a little analysis as Rothschild and Stiglitz does do, this, the result is that F and R in this case is a stable Nash equilibrium. However, and this is just a question of the proportions of consumers, frail and robust consumers and the nature of the difference curves of the robust consumers, you could also have a situation in which the indifference curve of the robust consumers through the separating contract R cuts the break even contract line. In which case there's now a blocking pooling contract. Now the pooling contact cannot itself be an equilibrium, we already know that. But it's enough to block the separating equilibrium and in this case no Nash equilibrium in this market will necessarily exist. So the conclusions of that classic study are that when consumers know their risk class and insurers cannot distinguish risk classes and in a competitive market any contract which is defined by its premium and its coverage that can be offered, then either there's a separating Nash equilibrium in which robust consumers are only partially insured or there's no Nash equilibrium at all. It's also known from literature about the same time that if the risk classes are very closely spaced or in the worst case a continuum of risk classes, then there is no Nash equilibrium, it will always break down. It is possible that there are Non-Nash or high-order of conjecture equilibria in these markets, that is a result of Wilson but I'm not going to concentrate on that. So what I do in the paper that I'm discussing today is ask the following question, if you have an insurance market like the health insurance market with asymmetric information where the market either unravels or achieves a very inefficient separating equilibrium what kinds of market organisations or regulations can be used to mitigate the effects of asymmetric information and achieve stability and relative or some second order efficiency. As a general problem, this is a problem of applied mechanism design and you can consider 3 kinds of elements, consumer, insurer and/or market interventions to control either a selection. Other problems that don't show up in Rothschild Stiglitz but are present in this market, incentive designs to control moral hazard from consumers and from providers of health services. And finally capital and conduct requirements to control performance or counter-party risk. I'm going to, because I have limited time, I'm going to give you one extremely simple regulatory scheme which solves the adverse selection problem. Suppose that you regulate this market by restricting the kinds of coverage that contracts offered in the market can provide. Specifically a considered fixed coverage regulation. It says that if there's a loss capital L associated with sickness then the only contracts that will be allowed by the regulators in the market will be ones that cover a specified share Theta of that loss. Once a coverage level is specified, the contract terms are specified, then the premiums can then be determined competitively in the market. So consider a market like that, suppose you have competitive insurers, suppose you have free entry and exit of insurers and initially, I'm going to go to the best case first. Suppose you have fixed coverage at a level that would obtain the same pooled contract that previously we said could not be a Nash equilibrium because it was blocked. And now if we just go back to the case of the pooled equilibrium we'll see that this simple regulation is enough to restore Nash equilibrium within this regulated market. The contracts that are now allowed effectively, the contract that's allowed is a contract which corresponds to P, competition would allow other contracts which are identical to that in terms of the loss covered but would vary in premium. And those correspond to a 45 degree line which is just a fine translation of the 45 degree line. So along that line labelled fixed coverage, various premiums, those are the contracts that could exist in this market. It's very easy to see, elementary to see that P is in this case, has to be a Nash equilibrium, it has to exist, effectively the intersection of the all-in-breakeven contract line, actuarially fair contracts and the fixed coverage line determine the Nash equilibrium. And with a little, just a little geometric argument you can easily convince yourself that there are no longer any possible blocking contracts. So in this case there is no adverse selection by restricting the kinds of contracts that can be offered. You manage to corral all consumers to purchase this, blocking contract that was shown here, I'm sure the contract that was shown here that cannot be blocked was not one with full coverage but it could have been. One could actually have full coverage and in that case the robust consumer would actually prefer a little less, the frail consumer would prefer a little more. But with no ability of the insurers to modify contract terms or offer alterative contract forms. There is no possibility that they could block. Now one has to be, if one were to use this device, one has to be rather careful about where one locates the fixed or minimum coverage level. If you have a lower level of minimum coverage than at P you get back into situation where you can have non-existence of equilibrium or you can have a separating equilibrium. In that case the robust consumers are either at the coverage limit or you're actually back into the case of a classic Rothschild-Stiglitz separating equilibrium. But the main point here is essentially in this case by regulating the kinds of products that are offered in the market one is able to bring adverse selection under control. And in fact it's very difficult although there are many, many ways to regulate an insurance market, it's very difficult to avoid the problem of adverse selection, at least with a continuum of risk classes of consumers, without some kind of regulation of product form. Essentially of the kinds of contracts that are, on which competition is permitted. I'm going to, I have about 3 minutes left, I'm going to just give you a quick list of some of the kinds of regulatory techniques that are used, I'm not even going to describe how they work, some work well, some work less well. My research program is to try to determine exactly which ones work best and it's kind of the insurance market equivalent of determining the optimal voting rule, the question is among all the forms of regulation, are there regulatory devices which are in some sense second best. On the consumer side you can have coverage mandates, you can require people to buy insurance, you can subsidise their premiums perhaps through using vouchers. If they have low income you can subsidies their premiums. And for the control of moral hazard you can have co-payments, you can have what are called reference price benefits which say that the, for any given condition the insurance pays the least cost effect available therapy and anything above that the consumer is on their own, that has some very nice automality properties from the standpoint of getting the incentive structure for consumers right. And you can have various kinds of prohibitions on over insurance, something that's really necessary when moral hazard is a problem and something which is markedly missing from the derivatives market, the market for financial insurance where people were able to greatly over-insure with disastrous results. You can regulate insurance, contract control, fixed or minimum coverage, that's the case that I already discussed and various kinds of contract entry restrictions. You can have premium regulation, generally does not work very well. You can have limits on administrative load, also a little bit invasive but may be necessary. You can have operating subsidies. That is to say a lot of the cost of the benefits or some portion, share of them are provided through public funds, not through premiums. And those can be effective and the reason that such operating subsidies can be effective is that with sufficient subsidy then the premium that any consumer sees will be at least as favourable as the actuarially fair contracts. So they have no reason not to buy the insurance and the result will be that you will get pooling equilibrium with adequate subsidies. You can have enrolment controls which have the effect of eliminating wasteful screening and signalling activities, you can have rescission control, that's to say non-cancellation policies. You can have risk adjustment which is an ex ante way of trying to correct an insurer's benefit obligations for the risk characteristics of its clientele. Experience adjustment which does the same thing, ex post. There are principal-agent problems and infinite horizon problems in all of these. There's reinsurance which is essentially another way of subsidising insurance companies particularly against high-risk people. To control moral hazard you can use capitation, what's called accountable care incentives which is to reward on the basis of outcomes, not on the basis of procedures. You can have capital requirements and controls on benefit management to try to control counter party risk. You can also intervene at the level of the market, you can reorganise these markets as single payer or government managed markets and in effect that's a direct and perhaps heavy handed way of being able to do management of contract form, restricting the contracts that are offered. You can have dual public private systems in which the public system is charged with being an exemplar or a competitive limit on an otherwise private market by setting reference contracts and premiums. And finally a device that's actually interesting for economists to study, you can use various kinds of incentive compatible bidding mechanisms as part of an industry subsidy to try to avoid some of the problems, adverse incentive problems there. One final comment and that is that there are a number of behaviour considerations in insurance markets that have to do with what consumers think and what they believe about their risk. Individual risk perceptions may be inaccurate, there are very few market mechanisms that would discipline individuals if they make mistakes about their own risk. One is they misperceive their average risk, second they misperceive the dispersion, possible dispersion risk, how much risk there really is. Both of those things can really foul up the operation of an insurance market even if there were no other problems of asymmetric information and so forth. If individuals cannot get it right, markets can go south. There's actually an important comment here which has much broader application, which is that our whole theory of efficient markets and the rational in economics for the use of markets is predicated on the assumption that consumers can do their part of the bargain, can respond in their own self-interest, can respond to incentives and can give signals through demand and through voting with their feet that are necessary to drive that market to efficient allocations. If consumers fail to do these things then you can get bad results from a market. That doesn't necessarily mean that the government can do it better, but it does mean that private markets are not doing it well. This is all I'm going to have time for today, perhaps there'll be a few more slides this afternoon. I got you up to the late '70's in this problem but work goes on and there are very interesting economic issues in applied mechanism design in this area and I would encourage you to, those of you who have some interest in this topic to read in particular about the US market for prescription drug insurance which is a controversial innovative and surprisingly successful attempt to use mechanism design theory to design a private market that does a reasonably good job in research allocation in the health area. Thank you very much.


When heterogeneity in consumer tastes and needs, and in cost and quality of products, are publically observable, markets can price, sort, and match these variations, and product choices made by consumers yield demand signals that foster efficient resource allocation. These conditions hold, roughly, for a broad swath of economic activity, allowing lightly regulated private markets to successfully approximate allocative efficiency. However, in health care systems around the globe today, participants do not necessarily see the big picture of lifetime health costs and quality of life, and in many systems the incentives that consumers and providers face do not promote efficient allocation of health care resources. Information asymmetries are the fundamental source of difficulties in health insurance markets and in efficient provision of health services. Additional factors contributing to poor performance of health markets include (1) government regulation that is intended to protect the disadvantaged and promote equity, but creates incentives antagonistic to allocative efficiency, (2) inefficient provider organizations and non-competitive conduct, sometimes sheltered by government policies, and (3) behavioral shortcomings of consumers in promoting their own self-interest, including inconsistent beliefs regarding low-probability future events, myopia, and inconsistent risk assessment.
The seminal contributions to economic analysis of Kenneth Arrow, George Akerlof, Joe Stiglitz, Mike Spence, Mike Rothschild, and John Riley establish that when there are information asymmetries between buyers and sellers, adverse selection, moral hazard, and counter-party risk can result, causing markets to operate inefficiently or unravel. Asymmetric information between buyers and sellers, or market regulations that restrict competitive underwriting and force common prices for disparate products, can induce adverse selection. Moral hazard occurs when effort to avoid risks cannot be observed by sellers and stipulated in insurance contracts, and buyers have less incentive for risk-reducing effort when some of their potential losses are covered. When the productivity and cost of medical interventions is not known to all parties, then buyers and third-party-payers may not make informed decisions on therapies. Counter-party risk occurs when sellers evade payment of benefits for losses, or fail as agents to respect the interests of the consumers who are their principals. Adverse selection of buyers with high latent risk or low risk-reducing effort, or sellers with high counter-party risk, make insurance less attractive to buyers, and may cause insurance markets to unravel. Administrative overhead will induce less than full insurance. By itself, this does not make insurance market outcomes inefficient, but increasing returns to scale in administrative costs may lead to an inefficient concentrated market.
In principle, the problems of asymmetric information can be overcome by government operation or regulation of health services; in practice, there remains a major mechanism design problem of designing incentives that handle the asymmetries; e.g., “single payer” systems permit additional levers of control, but information asymmetries cause principal-agent problems even in command organizations. Legal mandates and regulations can make adverse selection worse. Government policy on private health insurance markets often reflects a social ethic that individuals should not be denied health care because of inability to pay, expressed for example in requirements that hospitals admit uninsured patients with life-threatening conditions, and a social ethic that insurance contract underwriting should not be based on risk factors such as gender, race, and pre-existing conditions. When these requirements are not publically financed, they are implicit taxes on insurers and providers that are at least in part passed through to consumers as higher premiums that increase the effective load for low-risk consumers. Both the higher loads and the prospect of public assistance as a last resort reduce the incentive for consumers to buy insurance and to pay (or copay) for preventative care.
The United States has, more than any other developed country, relied on private markets for health insurance and health care delivery. These markets have performed poorly. Denials and cancellations, exclusion of pre-existing conditions, and actuarially unattractive premiums have left many Americans with no insurance or financially risky gaps in coverage. Administrative costs for health insurance in the United States are seven times the OECD average. These are symptoms of adverse selection. Delayed and inconsistent preventative and chronic care, arguably induced by incomplete coverage, have had substantial health consequences: the United States ranks 25th among nations in the survival rate from age 15 to age 60. This impacts the population of workers and young parents whose loss is a substantial cost to families and to the economy. If the U.S. could raise its survival rate for this group to that of Switzerland, a country that has mandatory standardized coverage offered by private insurers, this would prevent more than 190,000 deaths per year.
Given the damage that information asymmetries can inflict on private market allocation mechanisms, the obvious next question is what regulatory mechanisms can be used to blunt or eliminate these problems. This involves examining closely the action of adverse selection and moral hazard, and the tools from principal-agent theory and from regulatory theory that can blunt these actions. There is an extensive literature relevant to this analysis that can be focused on the regulatory design question. Less well investigated are the impacts of consumer behavior, particularly mistaken beliefs. This paper examines these issues, and studies the impacts of regulations intended to promote equity and efficiency. More practically, this paper investigates these issues with reference to the private market in the United States for prescription drug coverage for seniors, introduced in 2006 and subsidized and regulated as part of Medicare.