Demand For Health Care

OXFORD REVIEW OF ECONOMIC POLICY, VOL 5, NO. 1

THE DEMAND FOR HEALTH CARE AND HEALTH INSURANCE TIMOTHY BESLEY All Souls College, Oxford

I. INTRODUCTION This paper reviews the economic theory of the demand for health, health care and health insurance, the main object being to exhibit explicitly the links between the three. The analysis of demand plays a central role in modem economic analysis. It is central since it embodies some key tenets of the 'neoclassical* approach to micro economics. Rational agents maximize a utility function, defined upon goods and services subject to a budget constraint. For many purposes, this conception is found to be satisfactory and as such, has been widely applied.2 An area in which the basic framework seems immediately less appropriate is in the realm of

health (for example Culyer, 1969 has argued). None the less, there have been notable attempts at examining the demand for health as a good among others, such as cheese and carrots, in an agent's utility function. This approach is associated most of all with the work of Grossman (1972).3 From the demand for health is derived the demand for health care. If health is regarded to be what yields utility directly (i.e. health care has no independent effect on utility) then the demand for health care is purely derived. But an agent's health state is subject to random shocks (such as the contraction of a disease), and to this end insurance is an issue. In many systems (and for reasons which we shall attempt to explain below), insurance is paid as a function of health expenditures; tying together the demand for health insurance and health care. It is with these demands and in particular their interdependencies

1

I am grateful to Paul Fenn and Alistair McGuire for their comments on an earlier version. However, they should not be held responsible for the contents of this paper. 2 For a comprehensive account of the modem theory sec Dcaton and Muellbauer (1980). J For a simplified presentation, sec McGuire, Henderson and Mooncy (1988).

0266-903X/89$3.00 ©OXFORD UNIVERSITY PRESS AND THE OXFORD REVIEW OF ECONOMIC POUCY UMITED

21

OXFORD REVIEW OF ECONOMIC POLICY, VOL. 5, NO. 1

that much of this paper is concerned since they have important implications for the incentives to which health insurance gives rise! In the realm of policy, a proper understanding of them must precede measures designed to improve efficiency in health care provision and its finance. Prior to addressing the main issues, I shall discuss a possible objection to proceeding via using conventional demand analysis at all. This is the view that health is such an important good that it cannot be traded off against other goods which we consume. How should we make sense of this claim? One reason why this may arise is because preferences between health and other goods are lexicographic4 which, if true, prevents the construction of a utility function in the conventional way. However, assuming this degree of centrality for health seems too strong and quite contrary to any (albeit casual) evidence. People regularly undertake activities which worsen their health both now and in the future (e.g. smoking) yet presumably do so because they are prepared to trade a certain kind of pleasure against health. Most activities in which we engage have some implications for aspects of our health either now or in the future yet this aspect is rarely uppermost when we engage in them. Hence it does not seem so reasonable after all to claim that preferences for health are lexicographic. This does not deny that health may have a high value as compared with other goods, but this can be captured by the consumer's preferences and hence in the demand for health. In view of this, we will use some elements of standard demand theory as the cornerstone of an analysis of the demand for health, health care and health insurance. There are a number of policy issues of interest in studying these themes and we shall take many of these up again towards the end of this paper. In most economic analyses, three types of motivation for State intervention can be found: arguments from the fact that the free market allocation is inefficient, arguments that it is inequitable and arguments thatindividuals are not the best judges of their own interests. All three of these can be found in discussions of health and health care. This paper

4

reviews a particular framework for considering and weighing up these issues. Whilst in many ways simplistic, it has the virtue of putting them in sharp relief. Theoretical reasoning provides a tool for clear thinking about complex issues. The parameters of interest for policy makers are diverse and no amount of theorizing can provide a panacea for the problem of policy design. In much discussion in health economics, the focus has been upon the question of whether State provision of health care is desirable. Asking the question does not sit happily with the framework for analysis which much economics adopts and sits uncomfortably with the trichotomy of reasons for State intervention suggested in the last paragraph. While economists are interested in the State versus nonState provision question, perhaps because this has often been central to the political agenda, we shall proceed here in terms of a more traditional economic analysis. Questions about the role of State provision are deliberately kept in the background. The framework of demand theory which we use here usefully permits these questions to be relegated. The structure of this paper is as follows. In the next section we examine the demand for health. In section three we look at the demand for health care, whilst section four considers insurance questions. Section five uses some of the earlier insights to reflect upon alternative systems of health care provision.

II. THE DEMAND FOR HEALTH The demand for health is modelled much as the demand for any other commodity. Each individual maximizes utility subject to a budget constraint Contrary to the simplest model of demand, health is not a good which can be bought and sold directly. Rather, health is produced using inputs in the form of health care. Hence a further constraint on individual utility maximization is imposed in the form of a technology for the production of health. Typically this technology will depend upon a whole

A preference for a good x is said to be lexicographically prior to that for a good y when more x is preferred whatever value of y is chosen. Hence states are evaluated in terms of the amount of x before the amount of y is brought into consideration. The term 'lexicographical' draws the analogy with letters in a dictionary—all the a's are ordered before any of the b's.

22

T. Besley

set of a particular individual's characteristics. The model we are proposing here is not special to health demand. These models are generally known as 'household production models' and these have been examined by Becker (1965) and Gorman (1959/80) among others. In this more general case a household is thought of as caring about goods such as 'meals' into which food, energy and time are inputs. In the context of health, viewing the demand for health as being constrained by a technology is useful since it enables one, schematically, to divide the influences into those due to tastes and technologies. The latter we have already discussed. By the former we simply mean an individual's preferences. This kind ofmodel was analysed in the context of the demand for health and health care by Grossman (1972). His analysis considers the dynamics of health care demand, Le. how one invests in goods which improve one's health at different points in one's life cycle. He emphasizes that the gain from investing in one's health has two components. The first component is the consumption gain, i.e. the fact that being healthy enhances most things that one does. The second is an investment gain which constitutes a gain in one's lifetime earnings. This model has also been applied empirically; most notably by Grossman (1972) and byWagstaff(1986).

III. THE DEMAND FOR HEALTH CARE As a first approximation, health care is often divided into two categories; preventative and curative. Whilst undoubtedly overlapping in some respects, there are pure cases. For example taking an aspirin for a headache is a purely curative type of treatment whilst a purely preventative treatment is something like a vaccine. Both of these can be set in the context of a demand for health. Curative medicine contributes directly to one's health state and hence can be seen as being motivated by the demand for health. On the other hand, preventative medicine is protection against future possible illnesses. By purchasing it, one affects the probability of becoming ill in the future and hence it is the future demand for health from which a demand for preventative medicine is derived.

In the purest case, a demand for curative health care arises when an agent wishes to demand more health immediately. Most often this arises in the face of a 'shock' to one's health state, i.e. if it is diminished for some reason (for example by contracting an illness). A central constraint faced by an agent in obtaining more health is the health technology described in the previous section. After one's health state has fallen there are two possible effects to be accounted for. First, it may not be technologically possible to restore one's health to previous levels since the required technological means are unavailable. This arises for example if one has a limb amputated. However, in many instances one's day to day functioning might be restored despite certain injuries having been experienced. Hence, whilst one's state of health is in a strict sense irreversible it does not follow that one's functionings are permanently impaired. Second, even if this were not the case, the individual concerned may not choose to gain utility via an increase in health but rather chooses to substitute towards other goods and services because of the change in relative prices. In many health care systems, the role of tastes is muted by the fact that insurance schemes are in force. These alter the price of health care to the consumer and in some cases provide health care free of charge. We postpone full consideration of insurance and its impact upon the demand for health care until later in the paper. Whilst possibly diluting the role of tastes, it would be a mistake to think that it filters them out of the picture entirely. Even if health care itself is provided free of charge, the concomitant costs in consuming it might be substantial. For example, time off work and spent queuing may be substantial, or there may be travel costs involved in visiting a health facility. Furthermore, consumers have different degrees of aversion to submitting themselves for medical treatments. Hence, even if direct costs of purchasing health care are met by a third party, tastes may still play an important role in determining an agent's course of action in the face of ill-health. A third essential determinant of a consumer's reaction to an illness is his or her informatioa Better informed people may be inclined to treat them-

23

OXFORD REVIEW OF ECONOMIC POLICY, VOL. 5, NO. 1

selves rather than seeking specialist advice. Alternatively it may be those who are best informed who are inclined to seek specialist advice since they harbour less suspicion about putting themselves into the hands of health professionals. In many instances tastes and information are hard to disentangle. Knowing more may change my evaluation of an object, thereby changing my tastes.3 A further important factor determining the interaction between the demand for health and health care arises from introducing physicians into the picture. They play an important role in providing information and prescribing treatments. In doing so, however, they are constrained by technologies and to some degree by tastes. Patients have some say in what types of medical treatment they are prepared to submit to. Furthermore, individuals may choose whether or not to visit a health facility in the first place. Modelling the interactions between the tastes, technologies and professional judgements which result in health care choices is a challenge for economic theory at a number of levels.6 First, it seems implausible to model physicians as if they were entirely self-interested, yet one does not wish either to model them with no self-interest7 One wishes to model altruists not angels. Secondly, one has to recognize the 'power' available to physicians from the extreme asymmetric information which characterizes their dealings with most patients. Thirdly, one wishes to find some role for a patient's tastes. Neither a bargaining approach nor one based upon a principal agent8 relationship seems to capture all of the relevant subtleties of the physician/ patient interaction, although either might cast some light upon the problem. Nevertheless, understanding the effects of medical insurance depends crucially upon obtaining some understanding of such interactions. Furthermore, as we suggest below, introducing insurance means that the interactions

5

are complicated further since the insurance company becomes a further actor whose behaviour influences the demand for health care. There are a number of studies of the demand for health care which do not use the demand for health model of Grossman (1972). These studies are reviewed in detail by McGuire, Henderson and Mooney (1988), Table 9.1. The variable explained varies from patient-days to admissions, to simply medical treatments. The way in which the price of health care is measured also varies somewhat. Once insurance is brought into the picture the price of health care faced at the margin may become a function of the amount of care consumed. For example, this is true if there is a deductible, i.e. initially patients must pay their entire medical costs. Despite great differences in technique, data and explanatory variables, most studies arrive at the intuitively agreeable result that the demand for health care is inelastic. However, there does seem to be some responsiveness by individuals to price. Hence individuals do seem to be prepared to trade off health care against other goods given some possibility. Recently, the evidence of Manning ef a/.(1987) from the Rand health insurance experiment has reinforced this conclusion. Confronting individuals with different coinsurance rates, they found that agents did seem to respond to copayments in the way in which demand theory predicts. They conclude that their 'results leave little doubt that demand elasticities are nonzero and indeed that the response to cost sharing is nontrivial'. (Manning etal. (1987), p. 267.) Our discussion so far makes it clear that the demand for health care is not like the demand for many other commodities. On the other hand, one does not wish to divorce it too far from conventional analysis. Demanding other professional services, such as

Some of the issues concerning the role of information and evaluation in the provision for ill-health arc taken up in Bcsley (1989). 6 A challenge first elucidated in Arrow (1963). 7 For an interesting discussion on the morality and reality of the issues! see Gorovitz (1982). • A Principal Agent model is one in which a leader (a principal) sets an incentive scheme for a follower (the agent) subject to the actions of the follower being unobscrvable to the leader. For a detailed analysis of this problem, see Holmstrom (1979). Some, for example Evans (1985), have remarked that the frameworkneeds modification in the context of thedem and for health care, since to some degree the principal and the agent do not have conflicting objectives. Whilst this is correct as a point of detail, it would be a mistake to attach too much importance to it unless one believed that the principal and agent had perfectly consonant interests. The Principal-Agent framework remains of interest as long as there is some difference in the interests of physicians and patients.

24

T. Besley

accountancy or legal advice, has many similar characteristics. There is no reason at all to believe that the conception of the process as a demand is inappropriate. It is more a question of extending demand analysis than of abandoning it There is room for the economist's value theory, if it is desired, in the realm of health care demand. In view of this, we proceed to an investigation of health insurance.

IV. THE DEMAND FOR HEALTH INSURANCE The main function of an insurance contract is to reduce the risk faced by the person who buys it, Such contracts typically operate in terms of an agreement by the insurance company to pay something to the insured in the event of a particular outcome, in exchange for the payment of an insurance premium. An essential feature of a workable insurance contract is that the outcome which triggers the payment be observable to both parties to the contract. In the case of health insurance, the two parties to a contract are the patient and the insurance company, and there are two conditions upon which insurance payments can be made to depend; one is the state of health and the other is expenditure upon health care. Many insurance contracts also require there to be a third party whose role it is to assess whether a particular insurance claim seems reasonable, in the light of the state upon which insurance is paid out. In the case of health this would involve an insurance claim being assessed by a medically qualified individual with knowledge of the claimant's medical condition. For insurance based upon the latter, the role of the insurance assessor may be very important. If insurance is written upon an agent's state of health, then an insurance contract typically pays a certain amount to the insured if a particular disease is experienced. In effect, the British NHS operates on such a basis. There is a payment in kind, in the form of health care which is determined as a function of one's health state. For example, if one needs a

hip replacement the NHS 'pays out' in the form of a hip replacement operation tailored to the needs and particular circumstances of the patient in question. Similar contracts are also observed in the USA. The Medicare programme has recently instituted a system of Diagnostic Related Groups (DRGs) under which a hospital is paid a lump sum depending upon the disease in question and the treatment which it requires. Health Maintenance Organizations (HMOs), although forms of private insurance, operate in a fashion analogous to the NHS. We shall discuss these and DRGs further below. At this point, it is essential to indicate that the question of what type of insurance system should be operated, is quite separate from the question of whether the State should be responsible for running the system. In fact, there are two distinct areas in which the role of the State can be important. The State may take on the role of insurer or the role of provider of medical care or both. However, there is no necessary connection between public provision of health care and insurance. It is better to organize the discussion around the structure of insurance contracts rather than around public versus private provision. Whilst in theory disease contingent contracts may be first best9 in the economic sense, in practice they may fall well short. A major problem is that in practice such contracts cannot practically be tailored to meet all individual needs. This has, indeed, been the major source of criticism of the DRG system which fails to discriminate between more or less complicated forms of disease. An NHS style system, in which payment is in kind, can be more flexible, yet again does constrain some aspects of patients' choice. For example, food and other aspects of the environment in which one is treated are predetermined. A second type of health insurance contract to which we give consideration bases insurance upon reimbursement of health expenditures. Instead of looking behind expenditures to the underlying condition for which treatment was received, such

' In a world of diversified 'individual risks', optimal insurance, in the absence of any incentive problems, is that which sets the insured's marginal utility of expenditures equal to a constant no matter what his health state. Intuitively the optimality of this is seen by realizing that if it did not hold, then an agent could raise his utility overall by transferring resources from states where his marginal utility was at present high to those where it is low (assuming that marginal utility is decreasing in expenditure: the assumption of risk aversion).

25

OXFORD REVIEW OF ECONOMIC POLICY, VOL. 5, NO. 1

contracts reimburse the insured for expenditures incurred in the course of receiving treatment This makes use of the fact that health expenditures are generally cheaply and easily observable to the patient and to the insurance company. This type of contract is often observed in practice: it is still the dominant variety of insurance in the US A and in the private sector in the UK. In view of our discussions of sections two and three, these contracts are particularly interesting since they tie health insurance directly to the demand for health care and hence to the demand for health. In practice such contracts have a simple structure.10 The insured first of all faces a deductible, i.e. the first part of any claim is disallowed. After expenditures have exceeded this level there is reimbursement at a fixed rate (possibly 100 per cent), up to some upper limit, beyond which no coverage is given. A schedule of this form is given in Figure 1. Economic theory suggests that there is no reason why the best contract should be of this form, and there has been much discussion of optimal nonlinear insurance schedules.11 In practice the costs of calculating such schedules and the paucity of the information upon which such calculations could be based mean that economic agents are often most content with a simpler structure. The deductible health insurance contract can be Figure 1 oald

npnflnra

I detotitfc

10

Bppff hnul

explained in a number of ways. Arrow (1963) has shown that without moral hazard, an optimal insurance contract has an agent completely insured above a deductible limit However, deductibles may also serve a role in limiting the demand for health care if it is deemed to be excessive, since it discourages repeated visits to a physician. The Rand Corporation's health insurance experiment has presented important evidence on this issue which is reported in Manning et al. (1987).12 Deductibles in insurance contracts may also be explained by the presence of administrative costs13 in processing insurance claims. Each claim has a fixed cost of processing and verifying the claim, i.e. a cost which is independent of the magnitude of the claim. Hence very small claims could not profitably be processed by an insurance company. With a deductible the size of claims is restricted to a level where all claims exceed their fixed administrative cost Both types of insurance scheme which we have considered, Le. reimbursement insurance and disease contingent insurance, are lacking in an important sense when viewed as means of compensation for those who are unhealthy. Both are structured around providing support for the cost of medical care. Neither, in practice, provides compensation for the fact the illnesses are unpleasant and impose psychic costs. In this sense most forms of health insurance that we observe are incomplete contracts. Further aspects of incompleteness derive from the fact that other costs of ill-health, such as lost earnings, time spent queuing for treatment and travel costs, are not covered In some cases, separate insurances cover these contingencies, for example insurance against lost earnings in the event of sickness. However, one's overall impression of observed systems of health insurance, whether one looks at public or private insurance schemes, is of incomplete coverage against the effects of ill-health. In this paper we shall hereafter concentrate upon insurance focused upon coverage of medical expenses. We turn shortly to a discussion of the incentives which result from having an insurance scheme based upon reimbursing health care

For a full discussion of shape of such schedules and some of their implications see Keeler et al. (1977). Sec, for example, Holmstrom (1979). The key paper in the field, however, is by Mirrlees (1971). 12 Coinsurance ratheT than deductibles may be a more effective means of containing 'excessive'health expenditures. The are also compared in Manning et al. (1987). 13 See Shavell (1977) for a development of a formal argument 11

26

T. Besley

expenditures. Prior to this some more general issues, which form an essential backcloth, will be introduced.

metric for evaluating health outcomes. The reader should bear in mind that this way of proceeding is not without its critics.

(i) Efficient and Equitable Allocations of Health Resources

In our analysis, we will, however, proceed, for the time being, under the assumption that individual choices over health vis-a-vis other goods provides a sovereign criterion for evaluating the worth of these outcomes for normative purposes.

In most economic contexts normative questions about the allocation of resources are neutral with respect to the type of resource which is being considered. Normative questions are typically of two kinds: is the allocation efficient (normally in the sense of Pareto), and is the allocation equitable? These questions are normally raised with respect to allocations of all goods rather than with respect to health care in particular. Where health care is at issue, this neutrality is often rejected since some attach special significance to egalitarian distributions of health. As a general notion, this was introduced by Tobin (1971) as specific egalitarianism. Principally this view rejects the application of the principle of consumers' sovereignty to health allocations.14 If this is done, both equity and efficiency characterizations must be modified. Another reason for wishing to treat health differently has also emerged in the influential work of Sen on welfare economics.15 He has emphasized the poverty of the traditional economists' approach to evaluation based upon Utilitarian accounts of value. The allencompassing utility function in terms of which outcomes are typically evaluated merges together different concepts. For instance, while the statements that 'I desire x\ 'I would be happy with x' and 4I value x' are quite distinct (at least their meanings seem broadly different), the framework for evaluation used in economics fails to make them distinct. Sen advocates an approach to evaluating outcomes based on 'basic capabilities'. Here is not, however, the place to expound this view, with which the present author is sympathetic. The interested reader is referred to Sen (1986). None the less, the demand for health framework tends to go hand-in-glove with the view that individual utility is the appropriate

We shall focus here primarily upon efficient rather than equitable allocations of health care, although much of what we have to say has a bearing upon issues of equity. To say that an allocation of resources is Pareto efficient implies that nobody in the economy could be made better off without some other individual being made worse off. One of the central theorems in micro-economic theory says that all perfectly competitive equilibria are Pareto efficient However, this theorem assumes that there are no externalities and that the market structure is complete. It was Arrow (1964) and Debreu (1959) who furnished us with the insight that the creation of markets for each 'state of the world' could preserve the efficiency of competitive markets in the face of uncertainty.16 In practice markets are incomplete and hence competitive equilibria are in general not efficient in theory. Economists have found it useful to refine the notion of Pareto efficiency to that of constrained Pareto efficiency. This looks at allocations which are Pareto efficient given the market structure in force (the constraint being the markets in being). (See, for example, Diamond (1967) and Stiglitz (1982) who discuss stock markets in this context.) Demands for health will none the less arise in each state of nature. Unless the equilibrium is constrained Pareto efficient then the demands will diverge from their efficient levels. Hence an increase or reduction (equally, a tax on or a subsidy to) in the demand for health care could be associated with a Pareto improvement17

14

Musgrave (1959) refers to goods where we do not accept individual valuations as sovereign as 'merit goods'. The idea is developed formally in Bcslcy (1988a). 15 Sec, for example. Sen (1986). 16 Malinvaud (1972) refines this model to the case of 'individual risks' which is more appropriate for the analysis of health care. 17 See Greenwald and Stiglitz (1986) for an extensive discussion of constrained Pareto efficiency and its theoretical implications for government intervention.

27

OXFORD REVIEW OF ECONOMIC POLICY, VOL. 5, NO. 1

The importance of this tale emerges in trying to establish a benchmark for what one means by the efficient level of health care provision. It is often argued, as we observed above, that the demand for health care is increased by private or public insurance which leads to a socially excessive use of health inputs. While being unable to provide a castiron refutation of this one should at least be cautious in accepting i t Consider, for example, the case in which there is no insurance at all, then almost nobody would be able to afford to have certain kinds of expensive surgery, because of credit market failures. The advent of health insurance might greatly increase the demand for surgery. However, there is no a priori reason to regard the latter situation as necessarily less efficient Indeed, compared with the initial position, it could yield a more efficient allocation of resources, if the economy were to progress towards the constrained Pareto optimal allocation of resources. Once we are in the second best, one cannot maintain that anything which increases the demand for health care from its pre-insurance level results in 'excessive' demand for health care. Judgements of this variety require a more cautious approach. A problem to which health insurance is subject, arising if individuals who are purchasing insurance are different in a way which affects the manner in which they demand health care, yet is unobservable to those from whom they purchase insurance, is that of adverse selection. This has been much analysed in the literature (see, for example, Pauly (1974)). In practice such problems are diminished when health insurance schemes operate for large groups of randomly selected individuals. This is likely to be true of schemes which operate in large firms, for example. Many health insurance schemes operate on this basis. Whilst conceding that the consequences of adverse selection in health insurance may be of importance, we shall not make this a focus of this paper. Instead, we shall concentrate on the problem of moral hazard to which we shall now turn.

to have been grouped under the singletitleof' moral hazard' problems. We prefer to keep them separate. In the immediately ensuing discussion, we will throw caution to the wind and assume that initially, we are in a Pareto efficient situation. Hence all moral hazards are effectively 'bads', i.e. lead to losses in efficiency. (a) Insurance and Preventative Care This variety of moral hazard problem arises when an insured individual alters his behaviour so as to affect the probability of incurring a particular level of health expenditures. For example, having health insurance may incline agents away from preventative medicine towards the now (relatively cheaper) curative medicine. Alternatively, in the knowledge that one is insured, one may pursue more dangerous sports (such as hang-gliding) or increase health damaging pursuits (such as smoking). Whether these effects are likely to be observed in practice is uncertain and it seems unlikely that one might regard this as in any way the Achilles' heel of health insurance schemes. After all, given the intrinsic unpleasantness of illness, it seems unlikely that anyone would be inclined to trade cure for prevention to any great extent because insurance is available. None the less, as the direct analogue of the moral hazard problem discussed in the context of fire insurance: that once insured I may be less inclined to guard against a fire occurring, it is important to mention this potential moral hazard problem.

(ii)Health Insurance and the Demand for Health Care: Moral Hazard

(b) Physicians and Insurance When patients are insured, then physicians may be inclined to change their behaviour also. They may prescribe more expensive and complex treatments than they would otherwise, since patients are to some degree insulated from the financial consequences of this. It seems unlikely that but for having insurance, many patients would have the means to have a brain scan or to receive certain kinds of operation: a fact which would greatly constrain physicians' prescriptions.

Under reimbursement insurance there may be incentives to increase consumption of health care. These arise under three main headings, which tend

It is for this reason that insurance companies often play an important role in the diffusion of new medical techniques. Physicians' behaviour is typi-

28

T. Besley

cally constrained by what forms of treatment are permissible for insurance purposes since otherwise they may be too expensive for many patients to undertake. The advent ofnew medical technologies therefore presents a problem for physicians. Until they are legitimized as objects of insurance coverage, they cannot be regularly prescribed. Beyond the direct sphere of physicians' choice, there are other tendencies towards the inflation of costs due to health insurance as the 'non-health care' costs of seeking treatment are inflated. There is a tendency towards provision of more luxurious surroundings which is passed on into costs and is ultimately reflected in insurance premia.18 Arrow (1963, 1968) has emphasized the importance of trustworthy behaviourin ameliorating this problem: ' One of the characteristics of a successful economic system is that relations of trust and confidence between principal and agent are sufficiently strong so that the agent will not cheat even though it may be "rational economic behaviour" to do so.' This accords with our discussion of the role of physicians in the demand for health care.

care is lower at the margin under reimbursement insurance. Referring to Figure 2, we can examine some consequences of this.20 First, a fall in the price of health care from q to (l-a.)q induces a rise in the demand for health from htoh'. The triangle ABC then represents the 'deadweight loss' resulting from such a Figure 2 price of health care

(Hi demand for health care

h1 (c) Insurance and the Price of Health As we mentioned above, reimbursement insurance works by designating a particular rate of coverage at the margin.19 This rate of insurance determines a subsidy to health care. The insured's cost of health care is reduced by an amount proportional to the rate of insurance. This subsidy in turn induces an increased consumption of health care and in accordance with the demand for health model, health. To see this, we can return to the model laid out in section II above. When insurance is available at a rate a then health care with a cost of q per unit before insurance now costs (l-a)<7 per unit. Hence, the price of health

health care

move, i.e. loss in consumers' surplus. It arises since in otherwise competitive markets a subsidy on one good means that allocative efficiency is impaired (i.e. resources are no longer allocated in a Pareto efficient way). The size of the triangle ABC is inversely related to the price elasticity of demand for health care (see Feldstein, 1973). This can be seen by noting that, for a given subsidy, the triangle analogous to ABC is larger, the flatter the demand curve, i.e. the more elastic is the demand for health care. Hence the 'deadweight loss' of having reimbursement health insurance increases as the elasticity of demand for health care increases.21

11

For e x a m p l e , there m a y b e a m o v e towards more treatment in private rooms with luxuries such as televisions and telephones. T h e s e arc not 'health care costs' as w e c o m m o n l y understand them. M o r e borderline cases are things such as the installation of computerized record keeping—costs of which m a y be recouped from patients (and which may thereby be insurance financed) but which m a y or m a y not improve their medical conditions. " This has been recognized and discussed in Pauly ( 1 9 6 8 ) and Feldstein ( 1 9 7 3 ] . 20 Our presentation follows that of Feldstein (1973). 21 This argument is generalized in B e s l e y (1988A) to allow for cross elasticities and for different goods having different risk profiles.

29

OXFORD REVIEW OF ECONOMIC POLICY, VOL. 5, NO. 1

Against the loss implied by losing the conditions for allocative efficiency^one must weigh the gains from improved risk sharing. Consumers of health care gain since they are to some degree insulated in their health expenditures. Note that the effect discussed here is quite different from that discussed in the previous sub-section on physicians and insurance. Physicians need play no part in the effects discussed in this section. They would occur no matter who was choosing health care provision. (iii) Risk Sharing vs. Incentives in Health Insurance The previous discussions make it clear that insurance conditioned upon expenditures on health care may result in increases in demand for health and health care. Note, however, that one cannot say a priori that such increases are detrimental. As we argued in section IV(i), in the absence of insurance we are not in a 'first best' world. Indeed it might seem reasonable to suppose that the demand for health care would be heavily constrained were there no insurance at alL In part, this might be overcome if credit markets worked efficiently since then it might be possible for agents to borrow to finance medical treatments. However, credit markets also tend to work imperfectly: lenders cannot distinguish between borrowers who are honest and dishonest, and are unable to monitor what their borrowers do with borrowed funds, which may be invested in a risky activity making repayment less likely.23 The existence of imperfect credit markets therefore enhances the value of insurance. The main issue in the provision of health resources therefore becomes that of trading risk sharing possibilities against some possibly adverse incentives to consume increased amounts of health care. In much recent economics it is fashionable to appeal to asymmetries of information, i.e. one party in a contract knowing more than another, as the main

22 23

30

constraint upon overcoming the incentive problems that we discussed in the previous section. If an insurance company could monitor an agent's risk-taking behaviour (e.g. his propensity to go hang-gliding), then it would be possible to write an insurance contract which stipulated a maximum number of hang-gliding hours, for example. Similarly, if the insurance company could monitor in detail the activities of providers of medical care then it could stipulate precise modes of treatment which counteracted the incentives under insurance. There are two main constraints upon this. First, monitoring and information acquisition is itself very costly, sometimes prohibitively so. Second, it ignores the ethical and social framework surrounding medical care. The nature of the relationship between physicians and patients is intrinsically confidential and one based on trust. It is unthinkable that any third party should be allowed to play a very detailed monitoring role. Mandates from insurance companies concerning modes of treatment could not be regarded as acceptable for ethical reasons. Altogether, informational asymmetries and the social framework of control prevent the problems of incentives to consume excessive medical care from being solved directly. Instead, indirect solutions must be entertained. One such has recently emerged in the USA amidst some measure of controversy. This is the Diagnostic Related Group as mentioned above. Introduced for Medicare, this method of control pays out an amount for treatment which depends upon what illness a patient is suffering from. However, the categories are broad and neglect finer details. Such a scheme represents, in some ways, a break with the notion of reimbursement insurance, and a move towards disease contingent insurance. It can be viewed in the main as a response to the tendency of medical expenditures to increase under conventional reimbursement schemes. It fights against the tendency of physicians to prescribe expensive drugs and technologies. Moreover, it limits the inflation of medical costs resulting from the subsidy to medical care which

W e are assuming here that the economy is otherwise first best. This argument is invalid if there are other distortions. These are the adverse selection and moral hazard problems respectively.

T. Besley

health insurance grants. It therefore limits the increase in the demand for health and health care in the face of insurance. However, this innovation has proved controversial, in the main because of the perceived encroachment upon the autonomy of physicians. For many, it presents an unacceptable intrusion into the ability of physicians to prescribe the correct treatment for a particular patient, rather than being constrained to prescribing what would be right for the patient on average.24 In the next section we rum to the question of regulation of health insurance and its consequences for equity and efficiency.

V. HEALTH CARE PROVISION AND REGULATION A number of implications for health care provision and regulation flow from the above. Most obviously there is the possibility that insurance based upon reimbursement of medical care may induce incentives towards increased consumption of medical care. Policy to counteract these incentives without subverting the whole institution of reimbursement insurance might be envisaged. First, the government might offer subsidized preventative care to help counteract any incentives away from this. Second, it might regulate the behaviour of health care providers directly in a broad way, i.e. regulating charges for medical treatment and to some extent the prescription of advanced and experimental treatments. A third solution might involve discouraging people from taking out high rates of coverage. A tax on health insurance which was negatively related to the degree of co-insurance might have such an effect More crudely, a straight tax on health insurance or at least a removal of its tax advantage might help in controlling some of the incentives since individuals would choose less coverage.25 Arrow (1963, 1968) has emphasized the role of non-economic relations of 'trust' in combating

unfavourable incentives and has doubted the efficiency of relying on purely economic incentives. For example, he mentions reliance upon 'the professional ethics of physicians not to prescribe frivolously expensive cost of treatment, at least where the gain is primarily in comfort and luxury rather than in health improvement proper*. This intemalization of moral principles might go some way towards meeting the incentive problems which attend reimbursement insurance, but it seems unlikely that they could completely obviate the need for external intervention. This is especially true if one has reasons to doubt that physicians are the best judges of restraint in providing medical care, with the incentives which the Hippocratic oath entails. Other solutions involve movements away from reimbursementinsuTance as an institution. Among these, we have discussed Diagnostic Related Groups which make insurance payments a function of the medical condition. An alternative is to move to a system of payment in kind. In a certain sense the Health Maintenance Organizations which have grown up in the USA are such an example. This breaks with the third party structure which normally characterizes health insurance market and merges the provider of care with the insurer. The NHS is effectively such an institution which despite early prognostications26 has not been subject to the large inflation in medical expenditures witnessed by the predominantly reimbursement system in the USA. These systems can be thought of as anathema to the demand for health conception, which emphasizes sovereignty of consumers and their demands. Breaking this link provides the central philosophical difference between reimbursement systems and those which focus on more direct controls of treatment and prescription. These different perspectives should cut across political divisions since the debate has nothing intrinsically to do with the State as a provider of medical care or of insurance. The key issue is the importance which one attaches to the contracts reached between physicians, patients and insurance companies. If these contracts approximate demands from rational consumers who value health

** The trend of medical malpractice litigation in the USA may by itself have enforced a trend towards uniformity of treatment for certain diseases. For a survey of these issues see Danzon (1985). 25 This issue of tax policy towards health insurance and health care in the USA has recently been analysed in Pauly (1986). 76 Sec, for example, Buchanan (1964).

31

OXFORD REVIEW OF ECONOMIC POLICY, VOL. 5, NO. 1

properly, and are, furthermore, regulated for any inefficiencies which may arise, then it is natural to use a reimbursement scheme which is likely to be responsive to individual demand. If the underlying motivation for such a system is rejected, then the choice of a system which pays in kind and is less inclined to serve individual demand seems more likely to be palatable. Comparisons on this basis are made less clear cut by the fact that the remedies for inefficiencies which insurance might engender become intertwined with the question of what one regards to be the best health care system according to one' s acceptance or

rejection of the demand for health as a normative framework. A centrally provided or an HMO system may be more efficient even if one is inclined to attach normative significance to the demand for health reached in private contracts between physicians, patients and insurance companies. None the less, it is comforting that economic theory provides no grindstone for political axes. State versus non-State provision for health care is a second order question. It is questions of efficiency and equity and one's acceptance or rejection of principles of consumer sovereignty in health care demand which are the desiderata of economic analyses.

REFERENCES Arrow, K. J. (1963), 'Uncertainty and the Welfare Economics of Medical Care', American Economic Review, 53,941-69. — (1964), 'The Role of Securities in the Optimal Allocation of Risk-Bearing', Review of Economic Studies,3l,9l-6. — (1968), 'The Economics of Moral Hazard: further comment', American Economic Review, 57,537-9. Becker, G. (1965), 'A Theory of Allocation of Time', Economic Journal, 75,493-517. Besley, T. J., (1988a) 'A Simple Model for Merit Good Arguments', Journal of Public Economics, 35, — (19886), 'Optimal Reimbursement Health Insurance and the Theory of Ramsey Taxation', forthcoming in the Journal of Health Economics. — (1989), 'Ex Ante Evaluation of Health States and the Provision for Ill-Health', forthcoming in \heEconomic Journal Buchanan, J. (1964), "The Inconsistencies of the National Health Service', Institute of Economic Affairs Occasional Paper No 7, London. Culyer, A. J. (1969), 'The Nature of the Commodity Health Care', Oxford Economic Papers. Danzon, P. M. (1985), Medical Malpractice: Theory, Evidence and Public Policy, Harvard University Press, Cambridge, MA. Deaton, A. S. and Muellbauer, J. N. J. (1980), Economics and Consumer Behaviour, Cambridge University Press, Cambridge. Debreu, G. (1959), "Theory of Value', Cowles Foundation Monograph, Yale. Diamond, P. A. (1967), "The Role of the Stockmarket in a General Equilibrium Model with Technological Uncertainty', American Economic Review, 57,759-76. Feldstein, M S. (1977), "The Higher Cost of Hospitals and what to do about it', Public Interest, 48, Summer. Gorman, W. M. (1959/1980), 'A Possible Procedure for Analysing Quality Differentials in the Egg Market', Review of Economic Studies, 47, 843-56. Gorovitz, S. (1982), Doctor's Dilemmas, Oxford University Press, New York. Greenwald, B. and Stiglitz, J. E. (1986), 'Externalities in Economies with Imperfect Information and Incomplete Markets', Quarterly Journal of Economics, 101,229-84. Grossman, M. (1972), The Demand for Health: a theoretical and empirical investigation. National Bureau of Economic Research, New York. Feldstein, M. S. (1973), "The Welfare Loss of Excess Health Insurance", Journal of Political Economy, 81,25180. Keeler, E. B., Newhouse, J. P. and Phelps, C. E. (1977), 'Deductibles and the Demand for Medical Services: The Theory of a Consumer Facing a Variable Price Schedule under Uncertainty', Econometrica,, 45,641-55. McGuire, A., Henderson, A. and Mooney, G. (1988), The Economics of Health Care: an Introductory Text, Routledge and Keegan-Paul, London.

32

T. Besley

Malinvaud.E. (1972), 'The Allocation of Individual Risks in Large Markets' Journal ofEconomic Theory, 4,31228. Manning, W. G., Newhouse, J. P., Duan, N., Keeler, E. B., Leibowitz, A. and Marquis, M. S. (1987), 'Health Insurance and the Demand for Medical Care', American Economic Review, 77,251-77. Mirrlees, J. A. (1971), 'An Exploration in the theory of the Optimal Income Tax', Review of Economic Studies, 38,175-208. Musgrave, R. A. (1959), The Economics of Public Finance, McGraw Hill, New York. Pauly, M. V. (1968), The Economic and Moral Hazard: a comment', American Economic Review, 51,531-7. — (1974), 'Overinsurance and the Public Provision of Insurance', Quarterly Journal of Economics, 88, 44-62. — (1986), 'Taxation, Health Insurance and Market Failure in the Medical Economy', Journal of Economic Literature, 24, 629-75. Sen, A. K. (1986), Commodities and Capabilities, North Holland, Amsterdam. Shavell, S. (1977), "Theoretical Issues in Medical Malpractice' in Helms, R. (ed.), The Economics of Medical Malpractice, American Enterprise Institute, Washington D.C. Stiglitz, J. E. (1982), "The Inefficiency of Stock Market Equilibrium', Review of Economic Studies, 4 9 , 2 4 1 61. Tobin, J. (1971), 'On Limiting the Domain of Inequality', Journal of Law and Economics, 13,263-77. Varian, H. (1978), Microeconomic Analysis, Norton, New York. Wagstaff, A. (1986), 'The Demand for Health: Some New Empirical Evidence', Journal of Health Economics, 5,195-233.

33