Evaluating Health Insurance: A Review of the Theoretical Foundations

Abstract

Over the last 50 years, the theoretical basis of the value of health insurance has transformed itself from one based primarily on a gain from risk avoidance and a welfare loss from the additional medical care purchased by those who are insured, to one based primarily on the welfare gain from the additional medical care purchased by those who are insured. This transformation reflects (1) an increasing realization of the importance of health care in the demand for insurance, and (2) an increasing recognition that insurance, even insurance that pays off by paying for care, acts to transfer income from the healthy to the ill. This article traces the development of the theory of the value of health insurance using the above two themes as the basis for its organization.

Introduction

Private health insurance is a contract to transfer income or wealth from those who buy insurance and remain healthy, to those who buy insurance and become ill. That is, redistribution is the essence of the private health insurance contract.

Economists commonly evaluate economic behavior on the basis of efficiency and have the powerful tool of marginal analysis for doing so. In analyzing economic behavior, efficiency considerations have been shown – both theoretically and empirically – to be important sources of welfare. Economists, however, have been reluctant to venture into the evaluation of behavior based on redistribution or equity because they do not have similarly powerful tools. In order to be able to thoroughly evaluate economic behavior on the basis of equity, it is necessary both to make interpersonal utility comparisons and to be able to evaluate these utilities from a social welfare perspective. Both of these tools, however, are missing from the economists’ toolbox.¹

In the case of insurance, however, economists are partially able to bridge the gap between efficiency and equity by using expected utility theory. Instead of evaluating the effect of insurance on the utility of those who become ill separately from those who remain healthy, expected utility theory assumes that a single consumer can imagine what it would be like to be both healthy and ill, and can evaluate the utility in these two states from an ex ante perspective. The utilities from these states are then weighted by the probabilities (perceived or actual) associated with being in these states, in order to determine the total effect of insurance on the typical consumer. So, instead of insurance being thought of as a redistribution of income or wealth across consumers, from a theoretical perspective, it is typically viewed as a redistribution of income or wealth across health states for the same consumer. This allows what is in essence an equity issue to be transformed into and analyzed as an efficiency issue.

In order to analyze insurance as an efficiency issue, however, it is still necessary to specify the differences in the marginal utility across health states. In the earliest models of the demand for health insurance, the transfer of income or wealth across health states was analyzed using the assumption that utility is a single-argument, continuous and twice differentiable function of income or wealth, and that the marginal utility of income or wealth is declining. In these early models, health care spending was regarded as a loss of income or wealth, and marginal utility of income increased with illness solely because of that loss.

From this beginning, it is possible to view the evolution of the theoretical literature on the evaluation of health insurance as the confluence of two trends. First, there is an increasing realization that health care, and the utility that it generates, plays an important role in determining the value of health insurance. Initially, health care is essentially missing from the model, only entering the demand for health insurance as the source of financial loss. In contrast, in the most current model, access to health care, and to the utility derived from it, plays a central role in the private demand for health insurance and in determining its value to society.

The second trend is the increasing recognition of the role of health insurance in redistributing income from the healthy to the ill, and of the subtle way in which this redistribution is manifested by health insurance. In initial models, health care price and income elasticities were implicitly assumed to be the same for all consumers, both healthy and ill, but in the most recent model, the response of a healthy person to a change in the price of medical care is assumed to differ fundamentally from the response of an ill person, and similarly for a change in income. In other words, from an efficiency perspective, the price and income elasticities for a given individual consumer are not constant, but are state dependent.

Similarly, in initial models, when health insurance paid for the care of an ill beneficiary it was viewed as a price decrease for the insured consumer, whether ill or healthy. However, in the most recent model, paying for care is viewed as a price decrease that is effective for only the ill, and as such, constitutes a redistribution of income from the healthy to the ill. As a result, the welfare implications for moral hazard in the most recent model differ dramatically from those of initial models.

In this essay, the history of the theoretical thinking about the value of health insurance is reviewed, using these trends as the organizing themes. For each of a series of four articles, the central contribution is summarized and the limits of the analysis's applicability in the real world are discussed, using the U.S. as an illustration. The seminal work of Friedman and Savage begins this review.²

Friedman and Savage (1948)

While the history of the analysis of demand for insurance can be traced as far back as Daniel Bernoulli's 1738 paper which first suggested the concepts of utility and diminishing marginal utility (in trying to derive a reason for the low willingness to pay for the gamble associated with the St Petersburg paradox), the modern analysis of the demand for insurance derives from Friedman and Savage's famous paper.³ According to this analysis, utility is assumed to be increasing with income or wealth at a decreasing rate. The consumer, faced with the prospect of losing a pre-specified amount of income or wealth by chance, will choose to purchase insurance because expected utility is greater with insurance than without it.

Formally, if y₁ is the original income or wealth level and a loss of (y₁−y₀) occurs with a probability of π because a person becomes ill and must spend that amount of otherwise disposable income or wealth (henceforth, just “income,” the “or wealth” will be understood) on health care, then expected utility without insurance is:

If y^* is the expected value of the income given this health care spending, such that the actuarially fair premium payment is (y₁−y^*), then expected utility of insurance with a payoff that covers the entire loss is:

If the consumer's utility function exhibits the standard concave or “risk averse” functional form, then insurance produces a higher expected utility than no insurance.

A consumer who maximizes expected utility, therefore, would choose insurance over no insurance. As the expected income levels are the same with or without this fair insurance contract, the gain in welfare was attributed to “choosing certainty in preference to uncertainty”.⁴ Thus, the demand for insurance and the welfare gain associated with it were related to the efficiency derived from certainty. In this model and many subsequent applications of it that were intended to evaluate health insurance, the gain from certainty, or equivalently, the gain from the “avoidance of risk,” became the sole source of value to be recognized by economic theory.

In order to apply this efficiency gain to health insurance, the consequences of becoming ill had to be limited to the loss of existing income represented by health care spending, a loss that was analogous to the loss of existing wealth under fire and casualty insurance. That is, spending on health care when ill was treated in the same way as the loss of wealth that would occur if a consumer's house burned down or if a consumer had an accident that totally demolished his or her automobile. Thus, health care was modelled as paying a certain pre-determined amount for health care if the consumer became ill, and the gain from health insurance was limited to the certainty of paying a premium equal to the expected cost of that spending, rather than enduring the uncertainty of incurring this pre-determined loss or not.

This perspective ignores a number of important aspects of the health insurance contract. First, consumers receive health care in return for their health care spending, and this health care is often very valuable because it affects their health and lives. This is in contrast to a fire or casualty insurance where, say, a fire simply destroys something of value.

Second, this health care is predominantly consumed by those who are ill, that is, by those whose health status has changed in important ways, compared to their health status when the contract was purchased. This change in health status is associated with the demand for additional services – health care – that would not otherwise have been demanded at all. These services represent net additions to the individual's aggregate demand for commodities. For example, without a diagnosis of coronary arterial disease, a consumer might demand housing services and food, but with such a diagnosis, he might demand those commodities plus a coronary bypass procedure. As a result of this increase in demand for the health care services associated with illness, the consumer's marginal utility of income increases when the consumer's health status changes. In other words, it is not simply because health care spending reduces income available for purchases of consumer commodities that the marginal utility of income increases. It is instead because the change in health status increases the marginal utility of income available to be spent on medical care itself.⁵ Moreover, a transfer of income from those with low marginal utility of income to those with a higher marginal utility of income because of illness (even though it is expressed as the expected-utility equivalent) results in a net welfare gain that is exactly the same as the gain typically associated with the redistribution of income from rich to poor.

Third, the assumption of a predetermined loss (that is either paid for by the uninsured consumer or the insurer) ignores that insurance represents a contract where income is transferred from the healthy to the ill, and that consumers with more income are likely to purchase more health care when they become ill, compared to those with less income. Thus, the prospect of insurance paying for care represents an effective increase in income to the consumer, who in turn responds by purchasing more health care because health care is a normal good.⁶ Indeed, the income transfer in insurance may permit the purchase of health care that is far beyond the consumer's budget constraint and access to this health care makes insurance very valuable.

Overlooking the effect of the income transfer on medical care consumption appears to be due, in part, to the specification of the model. The model is conventionally specified as a choice between certainty and uncertainty. The purchase of insurance is interpreted as implying that consumers “prefer a certain loss to an uncertain loss of the same expected magnitude.” Indeed, this or similar language is often used in health economics texts to explain why consumers purchase insurance. This choice specification, however, obscures the underlying transfer of income from healthy to ill, and has distracted analysts from appreciating the role that income transfers play in increasing medical care consumption.

Indeed, in the experimental studies associated with prospect theory,⁷ where subjects are given exactly this choice, a majority invariably prefer the uncertain loss to the certain loss of the same expected magnitude, exactly the opposite to the conventional explanation. In one study, a head-to-head comparison was made between (1) an insurance policy that cost $50 and covered a 25 per cent chance of losing $200 and (2) a choice between a $50 loss and a $200 loss that occurred 25 per cent of the times.⁸ A majority (65 per cent) of the 208 subjects preferred to purchase insurance when confronted with the choice framed as in situation (1), but a majority (80 per cent) preferred the uncertain option, favoring the 25 per cent chance of a $200 loss, when confronted with the choice framed as in situation (2). Thus, not only does this specification obscure the income transfer within insurance, but the words used to explain the demand for insurance are exactly contradicted by experimental evidence.

In the standard theory of the consumer, the purchases that the consumers make are typically viewed by economists as quid pro quo transactions: the consumer pays a price in dollars and in exchange, he or she receives a good, service, contract obligation, etc. The reason that consumers make a purchase is that the value of the commodity, determined by the consumer's maximum willingness to pay, exceeds its price and generates a surplus for the consumer. A specification of the demand for insurance that is consistent with the quid pro quo transaction is simply to find the difference between equations (1) and (2):

Collecting terms associated with the same states of the world yields:

Thus, under a quid pro quo specification, the net gain in utility from purchasing the insurance contract is the utility gained from the income transferred to the consumer if ill, net of the utility lost from paying the premium if healthy. If utility were measured in dollar equivalents, this would be analogous to an expected consumer surplus.⁹

The explanation of why consumers purchase insurance differs depending on the specification of the model. With the conventional specification (equations 1 and 2), the demand for insurance was interpreted as a “demand for certainty,” or a demand for the “avoidance of risk.” As a result of this specification, “risk,” conceptualized as what is avoided when equation 2 is chosen over equation 1, represented the focus of and the important motivator in the decision to purchase insurance. That is, insurance eliminates risk. However, if a quid pro quo specification (equation 4) had been adopted, then the demand for insurance would have been a “demand for an income transfer in the event of illness.” This would have placed the emphasis on the income payment when ill (or when the bad state of the world occurred) in explaining why consumers purchased insurance. A quid pro quo specification would have led analysts to recognize the importance of the redistribution of income in insurance contracts, and the potential effect of this increased income on increasing health care purchases. And, by collecting utilities associated with health states, the changing nature of the utility function in response to changes in health state would have been easily recognized.

Pauly (1968)

As part of the debate surrounding the passage of Medicare and Medicaid in the U.S., Kenneth Arrow argued that if health insurance were provided at actuarially fair premiums to consumers who maximized expected utility and who possessed “risk averse” utility functions, the case in favor of insurance would be overwhelming.¹⁰ In his paper, Arrow held that the gain from insurance owed solely to the certainty (or risk avoidance) that was described by Friedman and Savage's (1948) model and because of this, insurance generated a welfare gain.

In response to this paper, Mark Pauly wrote what was to become one of the most influential health economics papers to appear during the last third of the 20th century.¹¹ This paper questioned the conclusion that a welfare gain was eminent with health insurance because of the presence of moral hazard. Pauly recognized that if insurance pays off by reimbursing for any health care spending that occurs and the consumer knows this, then the insurance payoff is equivalent to changing the price of medical care to zero. As this price change is artificial – the true price is still the marginal cost of producing the health care – insurance generates a welfare loss that is associated with the additional health care consumed when insured. Pauly noted that this welfare loss might be so large as to swamp the gain from certainty, resulting in a net change in utility that “could well be negative”.¹²

Martin Feldstein embraced this model as his own, and in still another seminal paper, presented an empirical calculation of the net welfare implications of the level of insurance coverage in the U.S. at the time.¹³ He calculated that the level of insurance coverage in the U.S. was represented by an average coinsurance rate of about 33 per cent – including the contribution of those who were uninsured at the time. He concluded, however, that this rate of coverage was “excessive” and overall produced a “very substantial welfare loss”.¹⁴ He suggested that the coinsurance rate should be raised to 67 per cent or higher in order to reduce the welfare loss from health insurance.

With Pauly's model,¹⁵ health care itself entered into the welfare calculations for health insurance. In contrast to Friedman and Savage¹⁶ who viewed the health care associated with health insurance as simply generating a financial loss, the additional health care in Pauly's model had value. This value, however, was less than the cost of producing this health care, and as a result, generated the oft-cited “moral hazard welfare loss” associated with health insurance.

While the utility from health care now appears in this model of the demand for health insurance, any recognition of the redistribution associated with insurance is missing. Specifically, Pauly's demand curve does not distinguish between those who are ill and those who are healthy. The demand curve analysis assumes that consumers would generally respond to the change in the price of health care just as they would respond to any a change in price of any other consumer commodity. Pauly's model does not recognize that healthy and ill consumers might respond differently to a change in the price of medical care. For example, a healthy consumer would not consume a coronary bypass procedure, regardless of how low the price becomes, because of the disutility associated with medical care when healthy. In contrast, a consumer who has been diagnosed with coronary blockage may respond to the price decrease by purchasing a bypass procedure that he would not have been able to afford at a higher price.

Although Pauly's model might be appropriate for certain types of discretionary health care that do not require being ill as a precondition – cosmetic surgery, drugs to enhance sexual function, or designer prescription sunglasses – it does not apply to health care that is consumed by those who are seriously ill. Most health care spending appears to be devoted to paying for the more serious types of health care. For example, the Medical Expenditure Panel Survey found that the top 10 per cent of health care spenders in the U.S. accounted for about 72 per cent of all health care spending.¹⁷ It is likely that almost all of this spending represents hospital procedures – trauma, organ transplants, resections of lesions, etc. – for those who are seriously ill, and not the sort of discretionary procedures that Pauly and Feldstein apparently had in mind.

Moreover, once it is recognized that the demand for specific health care procedures differs by health state, the redistribution of income from healthy to ill must then be brought into the model. And once this redistribution is recognized, then it must also be recognized that at least a portion of the additional health care consumed when insured is due to this income transfer. Thus, the problem with Pauly's model is that the moral hazard is not necessarily a response to a change in price alone. As insurance also represents a redistribution of income, a portion of the additional health care consumed by the insured is due to this income transfer. As a result, health insurance paying for health care could either be modelled as a price reduction, or as the vehicle by which income is transferred from those who are healthy to those who are ill, or a combination of both.

Although some have claimed the contrary, to my knowledge the interpretation that an artificial price change can be a vehicle for transferring income appears not to have precedence in economic theory.¹⁸ Instead, economic theory has uniformly viewed the establishment of artificial prices as changes in incentives and sources of inefficiency. To paraphrase the personal correspondence from a distinguished insurance theoretician, when you have the hammer of incentive theory, every problem with an artificial price change looks like a loose nail.

The models of Friedman and Savage¹⁹ and Pauly²⁰ dominated the thinking about the value of health insurance in the U.S. during the latter half of the 20th century and represented the accepted paradigm through which health insurance in the U.S. was viewed and evaluated. The dominance of this paradigm is evidenced by the number of studies that based their empirical calculations and conclusions regarding the value of health insurance in the U.S. on these theories alone.²¹ A number of these studies were based on data from the RAND Health Insurance Experiment, which was also based on this paradigm.²² The dominance of this paradigm is also evidenced by the large number of U.S. health economics texts that presented the theory of the demand for health insurance as (1) an efficiency gain from the avoidance of risk, balanced against (2) an efficiency loss derived from moral hazard, and therefore as consistent with these two models.²³ Only recently have a few of these texts presented an alternative, more comprehensive theory of the value of health insurance.²⁴

Subsequent to the publication of the Friedman and Savage, and Pauly papers,²⁵ a number of other models appeared in the literature addressing the demand for insurance and health insurance. A number of these made important contributions. For example, Mossin presented the case that the demand for insurance fell with wealth.²⁶

With regard to the health portion of this literature, however, Richard Zeckhauser's paper is especially noteworthy.²⁷ While state dependent utility can be traced back to Eisner and Stotz,²⁸ Zeckhauser appears to have been the first to apply state-dependent utility to health insurance.²⁹

Zeckhauser also implicitly demonstrated that when insurance pays off by making a lump sum transfer to the beneficiary, health insurance consumes more health care than if uninsured. However, the most enduring contribution of his 1970 paper (and the aspect that most influenced subsequent work) was, perhaps, his conceptualization of optimal health insurance as a “tradeoff between risk spreading and appropriate incentives,” as the title of his paper makes clear. This conceptualization, however, is consistent with the welfare gain and loss described by Friedman and Savage, and Pauly.

While Zeckhauser recognized the importance of health state on determining behavior with insurance and thereby better captured the value of the health care, his paper did not recognize that a portion of the incentives within standard (coinsurance) insurance were in fact “appropriate” because they generated health care purchases whose value exceeded the costs. Moreover, his model did not explicitly include the income transfer, and without recognizing the income transfer, it is difficult, if not impossible, to understand the welfare implications of the insurance price reduction.

As I stated in the introduction to this paper, the intent of this somewhat idiosyncratic review of the literature is to follow two threads through the literature: the realization that the value of health insurance is closely related to the value of health care, and the recognition of the income transfer in insurance and the role of the insurance price reduction in accomplishing this transfer. By following these threads, the theoretical basis for evaluating health insurance is expanded from (1) the risk avoidance gain, net of (2) the moral hazard loss, to include in addition (3) the gain from access to otherwise unaffordable care (and the simultaneous reduction of the portion of the moral hazard welfare loss that is attributed to the income transfer), and (4) the external benefits that this access to health care bestows on other members of society. To remain consistent with this intent, this essay considers only 2 other models.

De Meza (1983)

The type of health insurance where a beneficiary is paid a lump-sum amount upon diagnosis of a health condition is often referred to as “contingent claims” insurance. In an important paper, de Meza considered the issue of why people purchase contingent claims health insurance, rather than save for a “rainy day” (that is, build up cautionary savings to pay for health care) or borrow when ill.³⁰

Even though de Meza presents a model of contingent claims health insurance (and not conventional health insurance that pays off by paying for health care and, as a result, by reducing the price of care), his paper makes a number of insightful contributions to understanding conventional insurance. His model can be summarized as follows. First, he assumes utility is state-dependent: utility depends on the spending of disposable income on other goods when healthy, but it depends on the spending of disposable income on both medical care and other goods when ill. Next, he makes the standard assumption that when ill, disposable income is allocated so that the marginal utility of spending on other goods equals the marginal utility of spending on medical care, and then assumes that because of these additional demands, the marginal utility of disposable income (optimally divided between medical care and other goods) when ill exceeds the marginal utility of disposable income when healthy. In this way, he is able to finesse the specification of a utility function when ill, but at the same time, incorporate the realism that the value of insurance is related to the value of health care on which the additional income generated by insurance is spent, and this health care is generally very valuable to an ill person.³¹

De Meza then establishes that medical care spending when insured is greater than medical spending if financed by either saving or borrowing. He does so by showing that the cost of medical care spending if financed by insurance is smaller in terms of present consumption foregone than the cost of medical care spending if financed by saving or spending. This is because, in order to spend an additional $1 when ill if financed through saving, $1 of present consumption must be foregone, but in order to spend an additional $1 when ill financed through insurance, only a fraction of $1 of present consumption must be forgone because not everyone becomes ill at the same time. Thus, the consumer with insurance pays into the insurance pool only a small portion of the amount that the consumer is paid if ill. As a result, the optimal disposable income when ill is larger when using insurance to finance health care, compared with using saving or borrowing to finance it, and as a result, spending on medical care is also larger.³²

Although de Meza considers the demand for insurance compared to the demand for cautionary savings or borrowing, rather than the more conventional comparison of the demand for insurance compared to simply being uninsured, his model contains two other important insights for understanding the demand for health insurance. First, he “materializes” the mechanics of the insurance contract by emphasizing the fact that any payout when ill originates in premium payments when healthy. That is, although expected utility theory is used, the mechanics of the income transfer are made explicit in his model.

Second, and perhaps most important, the income effect of the insurance payoff is also modeled explicitly. That is, the reason why medical care spending is greater when insured is because disposable income is greater due to the income transfer, and this implies that medical care spending will be greater as well. As a result, moral hazard–the consumption of additional medical care when insured–can occur because of an income effect.

Subtle differences in specification can often have enormous theoretical implications. There is, perhaps, no better illustration of this than the difference between how de Meza and Pauly specify health care spending when healthy.³³ De Meza makes the assumption that health care spending is $0 when healthy. This is a reasonable model for many specific illness/treatment pairs. For example, if becoming ill is operationalized as contracting appendicitis for which medical care is an appendectomy, then “ill” results in medical care being consumed and “healthy” clearly does not.

Pauly, in a 1983 response to de Meza, took issue with the idea that there would be any income effect at all with health insurance. Pauly's argument is based on the equally realistic assumption that health care spending when healthy would be some small amount, say, $100. If ill, spending would be an additional $1,000, or $1,100 in all, in Pauly's example. If the probability of illness is 0.1, then the actuarially fair premium is $200, because total spending for every 10 consumers is (9 × $100+1 × $1,100=) $2,000. Under a contingent claims policy where the insured consumer was paid either $100 or $1,100, depending on whether healthy or ill, respectively, any income effect would be negative for the healthy consumers (since they paid $200 for the insurance premium but received only $100 in payoffs). Assuming a certain constant propensity to spend income on medical care that does not depend on health state, the reduction in spending by the healthy would exactly cancel the increase in spending by the ill. Thus, under these assumptions, health insurance cannot generate any additional health care spending due to income transfers because the income gains equal the income losses. This assumption, however, leads to the unreasonable conclusion that the massive redistribution of income from healthy to ill consumers, represented by the widespread application of private contingent claims insurance, would not result in any increase in health care spending at all.

As mentioned, de Meza assumes the propensity to spend income on medical care is zero when healthy, consistent with the absence of health care spending when healthy, and becomes positive only when ill. This latter small change in specification leads to the conclusion that contingent claims insurance generates additional health care spending only among the ill. To make this assumption more realistic, it would simply be necessary to assume that the propensity to spend income on health care is greater when ill than when healthy.

Pauly's 1983 comment also established the state of the art regarding the understanding of moral hazard at the time. He suggested that his original discussion³⁴ pertained only to moral hazard for relatively minor illnesses and health care expenditures, such as “routine physicians’ services, prescriptions, dental care, and the like” but that “[t]he relevant theory, empirical evidence and policy analysis for moral hazard in the case of serious illness has not been developed. This is one of the most serious omissions in the current literature”.³⁵

Despite de Meza's insights and Pauly's admission, the theory that dominated health economics research and health policy continued to hold that (1) health insurance was demanded because consumers preferred certain losses to uncertain ones of the same expected magnitude, but that (2) the additional health care consumed because of health insurance made consumers worse off. Empirical calculations of the value of health insurance during this period continued to suggest that moral hazard so dominated welfare calculations that health insurance at current coverage parameters made consumers worse off.³⁶ As is evidenced by the RAND Health Insurance Experiment and many other similar studies, health economics in the U.S. was dominated by the issue of empirically determining the extent that health insurance led to high and rising health care costs in part because of the theoretical connection between the growth of health insurance coverage in the U.S. and the additional (presumably inefficient) health care spending.³⁷

Also because of the widespread acceptance of this theory, most of health policy in the U.S. – from the implementation of cost sharing of the 1970s to the adoption of managed care of the 1980s and 1990s to the recent interest in consumer-driven health care and health savings accounts – was based on the presumption that in an insurance-dominated environment, too much quantity of health care was being consumed. The portion of the costs of medical care that were attributable to the high prices of medical care were actually seen as favorable because the high prices reduced the size of the moral hazard effect.³⁸ The RAND Health Insurance Experiment made it clear that greater cost-sharing would reduce the quantity of health care demanded.³⁹ As the health of those who responded to cost-sharing by reducing health care consumption appeared to be largely unaffected by the reduction in care the Health Insurance Experiment also seemed to confirm the theoretical presumption that the extra health care consumed with insurance was not very valuable.

Nyman (2003)

An alternative theory has been suggested that is based on two tenets.⁴⁰ First, a health insurance contract is a quid pro quo transaction. That is, when consumers purchase a health insurance contract, they are paying a premium if healthy for an income transfer if ill, as equation (4) above illustrates.⁴¹ Therefore, the demand for health insurance is a demand for an income transfer if ill, rather than for risk avoidance. The premium if healthy is the price of this contract and the net gain can be conceptualized by the expected consumer surplus.⁴² In this conceptualization, the financial loss from medical expenditures is no longer the focus and is replaced by a change in health status.

Second, the price reduction in health insurance no longer represents a movement along the demand curve, but is instead a vehicle for transferring income from those who remain healthy to those who become ill. This is based on the realization that for most health care, a change in price to zero would not entice healthy consumers to purchase it. For example, what healthy consumer would purchase a coronary bypass procedure simply because the price dropped to zero? As a result, for most health care, the price reduction is effective only for those who are ill, and as such, it represents the vehicle for transferring income to them.

It is also based on the realization that the response by an ill person to an insurance price reduction without the income transfer would leave the consumer on their original budget constraint. That is, if everyone who held insurance became ill and as a result, there was no transfer of income from healthy to ill, an insurance contract would simply represent a contract that entitled the beneficiary to pay a lower price for health care. The cost of this lower price and the beneficiary's response to it, however, would be entirely paid for “up front” by the premium, and would not move consumers off their original budget constraint.

In comparison, with health insurance, only a portion of consumers become ill in any given period. Thus, although the price of health care with insurance decreases, it is only because of the transfer of income that the consumer is able to consume beyond his or her budget constraint when the price falls. Thus, in health insurance, the price decrease is actually the vehicle by which income is transferred from the healthy to the ill, and so, a portion of the response to the price decrease is due to this income transfer.

The core implication of this theory is that moral hazard – the additional health care consumed because of the price reduction that is used to pay off the health insurance contract – is made up of a portion that is due to the income transfer and a portion that is due to the pure price effect, that is, due to using a price reduction to transfer income. The former is efficient and welfare increasing, and the latter is inefficient and welfare decreasing. As a result of this, the welfare implications of health insurance are substantially different from those of the conventional theory. Not only is the welfare loss smaller because inefficient moral hazard represents only a portion of the total moral hazard, but also the remaining portion of the moral hazard no longer represents a welfare loss, but instead, a welfare gain. The substitution of a gain for a loss in the welfare calculations causes a dramatic increase in the value of health insurance. Intuitively, much of moral hazard represents the purchase of additional health care that consumer would not otherwise be able to afford, and this increased access to expensive and often life-saving health care generates an important welfare gain that has not been recognized in previous theories.

For example, consider the example of Elizabeth who has just been diagnosed with breast cancer. Without insurance, Elizabeth would have purchased a $20,000 procedure to excise the cancer from her breast. She would have considered purchasing an additional $20,000 procedure to correct the disfigurement caused by the first procedure, but this breast reconstruction would be too expensive without insurance, given the competing claims on her resources. As a result, she would only purchase the first procedure if uninsured.

Fortunately, Elizabeth had purchased a health insurance contract for $4,000 that pays for all her care. With this insurance, she purchases (1) the cancer excision procedure costing $20,000, (2) the breast reconstruction costing $20,000, and (3) an extra 2 days in the hospital to recover costing $4,000. Moral hazard is represented by the additional consumption of health care when insured, represented here by the additional $24,000 in spending for the breast reconstruction and 2 extra days in the hospital.

To determine the efficient and inefficient portions of moral hazard, it is necessary to decompose the additional spending into the portion generated by the income transfer and the portion generated by the pure price (substitution) effect. As the insurer has paid $44,000 out of the insurance pool for Elizabeth's care while she has paid in only $4,000 in premiums, a transfer of $40,000 in income has occurred. What would Elizabeth have done if instead a cashier's check for $44,000 had been written to her upon diagnosis of breast cancer and the transfer had been a lump sum amount?

With her original income plus this additional amount, assume that Elizabeth would have purchased (1) the excision of the cancer and (2) the breast reconstruction, but not (3) the extra 2 days in the hospital. If so, efficient moral hazard is represented by (2) the $20,000 breast reconstruction, because Elizabeth could have purchased anything of her choosing with her original income plus the additional $40,000 in income (from the $44,000 cashier's check minus her $4,000 premium), but chose to purchase the breast reconstruction. That is, her willingness to pay with the additional $40,000 now apparently exceeds the purchase price of the breast reconstruction procedure, $20,000, therefore, the purchase is efficient. On the other hand, the 2 extra days in the hospital would not have been purchased, so their purchase under the policy that pays for all care represents inefficient moral hazard and a welfare loss for the conventional reasons.

Nyman has suggested that the welfare gain from the income transfer effect is the dominant reason for the purchase of health insurance.⁴³ Indeed, it so dominates the welfare calculations that it results in a welfare gain from moral hazard that is about three times the costs, according to calculations based on findings from the literature. That is, the main reason for purchasing insurance under this theory is the valuable additional health care that is purchased, not an aversion to risk.

This theory explicitly takes account of the transfer of income from those who remain healthy to those who become ill. It assumes that reduction in income represented by the premium payments does not reduce the number of cancer excision procedures among those who remain healthy–an obvious point, but critical to the idea of a differential (income) effect from the income transfer. Therefore, there is a net increase in health care consumption because of the transfer of income from the healthy to the ill.

The theory also explicitly takes into account the fact that for most medical procedures, the price decrease is effective only for those who are ill. No healthy consumer would purchase a procedure as serious and unpleasant as a breast reconstruction or a course in chemotherapy just because the price has dropped to zero. Therefore, the price decrease is the vehicle by which income is transferred to the ill, and to a certain extent represents a shifting out of the (Marshallian) demand curve, rather than a movement along the demand curve.

This theory has a number of implications that have been discussed in more detail elsewhere.⁴⁴ First, the price of insurance under conventional theory is the loading fee, but under the new theory, the price of insurance is the entire premium. Second, the inefficient portion of moral hazard represents the transactions costs of a contract that uses a price reduction to transfer a certain amount of income to a consumer who becomes ill. Third, in contrast to an exogenous decrease in the market price, the price decrease in health insurance must be purchased. Therefore, instead of decomposing the exogenous price decrease into a single Hicksian income effect, the price decrease in health insurance contains two income effects: (1) the effect of the income transfer described earlier, and (2) the effect of having to pay a larger premium in order to obtain an insurance contract with a lower coinsurance rate. Fourth, the financial “loss” that is salient in the conventional theory is replaced by a change in health state and a desire for additional consumption (the health care) in the ill state. Fifth, the demand for a health insurance contract is derived from the transfer of income to the ill state and the additional health benefits that are generated by the resulting increase in health care purchases. Sixth, an important positive demand-side externality is the benefit that others derive from an ill person being able to gain access to the needed care, care that such persons would not otherwise have been able to afford to purchase if they were not insured. This external benefit is a market failure and represents the prime reason for government intervention either to subsidize the purchase of health insurance or to provide it to all its citizens. This gain is missing from studies that evaluate health insurance on the basis of the risk avoidance gain, because it is assumed that with or without insurance, the needed care would be purchased.

Conclusions

This paper holds that the redistributive nature of health insurance has been obscured by the requirement that the demand for health insurance be modelled in expected utility terms. As a result of embodying the demand for health insurance in the expected behavior of a single consumer, early models of the demand for health insurance did not adequately distinguish between the behavior of those who became ill and those who remained healthy. And the subtle way in which the transfer of income from the healthy to the ill is manifested in health insurance–as a price reduction that is effective only for those who are ill–was missed.

Moreover, the role of health care in the demand for health insurance also evolved from something that generated only a welfare loss, to something that generated a net welfare loss, to something that generated a net welfare gain. Indeed, empirical studies suggest that the welfare gain derived from access to the additional health care that is generated by health insurance now represents the dominant explanation for the demand for health insurance.⁴⁵ The certainty- or risk-related benefit on which the earliest theory was based in its entirety is now estimated to provide only a negligible fraction of the gain from the additional health care.⁴⁶

Very little empirical work actually underpins the newest theory. Evidence is lacking regarding the portion of moral hazard that is efficient and inefficient, and we have little evidence that can be used to determine the welfare effects. Ideally, a randomized trial, similar to the RAND Health Insurance Experiment, would be done to determine these magnitudes. For example, those who do not now have insurance could be randomized into three arms: (1) remaining uninsured, (2) becoming insured with insurance that pays for all care, and (3) becoming insured under an equivalent contingent claims insurance, where the consumer is paid a cashier's check upon diagnosis in an amount equal to what the insurer would pay if the insurer simply paid for all care. Comparison of the amount of spending in arm (3) to spending in arm (1) would determine efficient moral hazard, and comparison of spending in arms (2) and (3) would determine inefficient moral hazard. Clearly, there are practical, budgetary, and ethical issues in designing such a study.

In the meantime, studies could be done that would indirectly determine the relevant magnitudes. For example, Koç has shown that moral hazard effects differ for those who are healthy compared to those who are ill.⁴⁷ Also, Nyman and Barleen have estimated the health benefit derived from moral hazard in the case of supplemental health insurance in Brazil.⁴⁸ Other similar studies are underway.

In summary, this paper describes an evolution of the theory of the value of health insurance that suggests that health insurance is much more valuable than previous theories have held. Health care represents a substantial portion of the economies of developed countries and of many developing countries as well. Understanding the welfare associated with the principle mechanism for financing and gaining access to this health care – private and public health insurance – would seem to be an important undertaking, based on the size of the health care sectors alone.