GSFch10 - ‘ I‘l‘I_E"‘Iql‘l‘v_l‘l‘lql...

Info iconThis preview shows pages 1–12. Sign up to view the full content.

View Full Document Right Arrow Icon
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 2
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 4
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 6
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 8
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 10
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 12
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ‘ I‘l‘I_E"‘Iql‘l‘v_l‘l‘lql Iv ID Iv IF I. A's ‘1 q geaaaaaaaeauauua. :13_<B_<B<B<B<B<B1n ‘ I i ‘ 1 l- 1 1 fl ffisasaeawaemea Hal _< «_B1B<B<B mpetiect lntotmniion: Simmer-induced Demand and Smot Ateo Vomtions E Supplier-Induced Demand (SID) I Models of SID \ The Identification Problem E Evaluation of SID I Small Area Variations (SAV) H The Physician Practice Style Hypothesis SAV and the Social Cost of Inappropriate Utilization I Evaluation of SAV perfect agency. In particular, we saw that problems of information asymmetry affect the health insurance industry with serious consequences. This chapter describes two problems of information that have generated even more academic and policy interest. The phenomenon of supplier-induced demand (SID) occurs when physicians abuse the agency relationship with their patients in order to generate demand for personal gain; this is made possible because physicians are more fully informed than patients. The phenom— enon of small area variations (SAV), the frequently wide variations in per-capita rates of surgery, physician visits, and hospitalization among otherwise-similar market areas, may arise even when the physician acts as a perfect agent. In this case, it is imperfect informa- tion on treatment outcomes that creates a distinctive problem leading to welfare losses. in chapter 9, we examined various applications of asymmetric information and im- CHAPTER 10 Imperfect Information: S upplier—Induced Demand and Small Area Variations 205 SID and SAV have become major policy issues, and this chapter is devoted to both problems. The two parts are best read together to understand the similarities and the distinctions. However, the bodies of literature have developed separately and are pre- sented here so that they can be studied separately. Because no single hypothesis has generated more interest and controversy among health economists than supplier- induced demand, we begin with this issue. Supplier-induced Demand (SID) The idea behind SD is that health care providers have and use their superior knowl- edge to influence demand for self-interests; that is, they take advantage of the “infor— mation gap.” The conditions for SID also are made possible through the physician’s dual role as advisor to the patient and the provider of services. For many, these propositions have intuitive appeal and seem consistent with anecdotal evidence. Nevertheless, as we will see, it is difficult to develop and test formal models of SID. In a world where physicians and patients were equally well informed or where physi— cians always act as well~meaning agents who are concerned only with patient welfare, SID would not occur. However, it is commonly argued that providers sometimes abuse their influence to generate demand—actions that can subvert the market’s normal function. On this point, Uwe Reinhardt (1989) went so far as to declare: The issue of physician—induced demand obviously goes straight to the heart of probably the major controversy in contemporary health policy, namely, the question whether adequate control over resource allocation to and Within health care is best achieved through the demand side . . . or through regula- tory controls on the supply side (p. 339). Reinhardt questions whether health care markets can function according to the stan- dard demand—supply model. In the standard model, consumers are sovereign and the volume and kinds of goods and services produced respond to independent consumer demands. If, however, consumers’ demands are strongly influenced by providers, mar— ket forces may not restrain prices, limit the consumption of care, and allocate resources to their best uses. If SID is a major problem, Reinhardt argues that regulation should be a policy option. A COMMON REPRESENTATION OF SID The origins of SID can be traced to Shain and Roemer (1959) and Roemer (1961) who found positive correlations between short-term general hospital beds per 1,000 popu- lation and hospital days per 1,000 population. The relationships held across states as well as across upstate New York counties. The apparent phenomenon was interpreted as “a bed built is a bed filled,” and this observation became known as Roemer’s Law or the Roemer Effect.l 1Roemer (1961) also reported on a natural experiment where a sudden increase in hospital beds in one county, with no changes in other factors, led to a sharp increase in utilization rates. The terms Say’s Law or Parkin— son’s Law for hospital beds as well as the availability effect have been used to describe the Roemer Effect. RT III In ormation The principle of Roemer’s Law was thought by some to apply as well to physician services, and the general term identified as supplier-induced demand (SID) seemed a natural extension of the same basic idea. The underlying notion can be understood as an agency problem in that the reliance of the patient on a physician gives the physician a degree of discretionary influence. Not all health economists, however, accept the SID proposition; they are divided into two camps. Those who believe that SID is an important phenomenon are labeled as the “broads” (Bs); those who challenge the phenomenon are called the “narrows” (N5). Figure 10.1 illustrates a commonly used model to explain SID, as well as to distin- guish between the Ns and the Bs. Here, D1 and 51 represent the initial demand and short-run supply of physician services with an equilibrium at Q1. Suppose the supply of physicians increases to 52, ceteris paribus. The Ns would use standard analysis to predict that the price will fall from P1 to P3 and that the quantity demanded will increase from Q1 to Q3. Total spending will increase or decrease depending on the elasticity of de- mand. For services with an inelastic demand, such as physician care, total spending Will decrease. Also, because more physicians are dividing up a smaller pie, the typical physi- Cian’s share of patients and earnings also will diminish. According to the Bs, however, an increase in supply leads directly to an increase in demand through the SID effect. Physicians are motivated by what otherwise would be a loss of patients and earnings to use their discretionary influence to shift demand. That is, demand will not remain independent and stable at D1 but will be shifted out. Where it increases to a level such as D2, the equilibrium price declines to P2 with equilibrium quantity Q2. However, if the increased demand is sufficiently large, such as shown by the shift to D’z, the equilibrium price may even rise above the initial level to F; with equilibrium quantity Q’z. Thus, as shown through Figure 10.1, the new equilibrium fee may be lower or higher under this view of SID. The case where the fee becomes higher came to be called the Reinhardt fee test of inducement. Under this test, evidence that fee increases follow in- creases in supply is interpreted as unequivocal support of the SID proposition.2 Much of the early empirical support for SID, however, came not firom evidence re- garding fees but from studies showing correlations (either simple or through multiple regression analysis) between physician availability and utilization rates. Although this evi- dence is consistent with SID, it also is consistent with standard analysis, as can be seen in Figure 10.1. Note that the equilibrium quantity of health care is larger after the supply curve shifts outward, regardless of whether or not the demand curve also shifts outward. Unless the study carefully distinguishes between shifts in demand versus a movement along a de- mand curve, the evidence of positive correlations does not support one theory over another. Furthermore, even evidence that determines that demand has shifted somewhat following increases in supply is consistent with traditional economic arguments. An ex- planation of a demand shift is plausible under standard theory when the opportunity cost of patient travel times is a significant component of total cost. When increases in physician availability reduce travel times or waiting times, one would predict that the demand will increase. Furthermore, if increases in availability also lead to improve- ments in the quality of care, this too will shift the demand curve. zRecall though from chapter 9 that the Pauly/Satterthwaite consumer information model offers an alterna- tive explanation of higher fees. i i, l r i r \ r i I. i . CHAPTER 10 Imperfect Information: Supplier—Induced Demand and Small Area Variations 207 P 81 P'e Price of P1 services P2 P3 Quantity of services Models of SID The representation of SID in the familiar context of a competitive market raises im- portant questions that must be answered by a better working model. What is the mo- tive to induce demand? Why would providers induce demand to a certain degree at one time and to a lesser degree at another? Why do providers not simply induce fully all the time? What is the mechanism through which they induce demand? How does induce- ment change in the face of increased supply? To answer questions like these, economists have deVeloped several models of SID. Price Rigidi’ties and SID One approach that can provide a logical explanation of demand inducement within the context of a competitive market model is to argue that prices tend to be rigid; that is, price does not adjust quickly to changes in demand or supply. To elaborate, suppose that prices in Figure 10.1 do not drop from their initial equilibrium at P1 as supply increases. With the new supply curve, 82, there is now an excess supply at P, if physicians continue to provide the original quantity of servicesAn excess supply means that at least some physicians are not on their supply curves providing the quantities they would like to sell at the existing price, and thus they would clearly have a motive for inducing additional quantities if that is possible. i The ability to induce demand depends on the agency responsibility that gives physicians opportunities for inducement. However, the motive to use this power and the extent to which it is used depend on the relative gains from the additional earnings versus the costs of the additional inducement activities. Aside from direct costs, time RT III Information costs are incurred because it may take time to persuade patients to consume more care. There is also a potential loss of patients who detect that the physician is not acting in their best interests. Thus, while the demand may ultimately be shifted to the right, there is no way of theoretically determining just how far to the right. It depends upon the relationship be- tween the incremental earnings and incremental inducement costs. If the incremental costs of inducement are relatively large because of serious damage to a physician’s repu- tation, then we would expect little or no inducement. If it is easy to persuade patients to consume more services and professional and ethical constraints are minimal, then the demand shift can be Substantial. We suggest though that demand will never be shifted to a level such as D'z Where there would be an excess demand at the price P1. Given the costs of inducement activity, providers have no reason to shift demand to the point where the quantity demanded exceeds the quantity providers are willing to supply at P1.3 ' THE TARGET-INCOME MODEL OF SID The target-income model often is used to explain the rapid increases in physician fees experienced in the 19605 and 19705. These increases occurred despite rapid increases in physician availability. Under the target income hypothesis, instead of leading to lower fees and incomes, increases in physician availability led to higher fees in order for earn- ings to be maintained. Consumer ignorance provides physicians with discretionary in- fluence to manipulate demand when, for example, fees are controlled. Physicians also have an untapped degree of monopoly price power that could be used to raise fees and incomes to some target level. Evidence of physician behavior under fee controls, especially under the Economic Stabilization Program introduced in the early 1970s, has been cited to support the tar- get-income hypothesis. The imposition of fee controls that limited the annual increases to 2.5 percent is often offered as a natural test of inducement. ‘ Under conditions of price controls, several investigators found that utilization tended to increase—evidence consistent with the target-income model. Similar support comes from studies of Medicare utilization in Colorado" In 1977, Colorado restruc- tured its Medicare reimbursement rates by virtually freezing rates for urban physicians while providing large increases for nonurban physicians. Rice (1983) found that the rel— ative decrease in fees for urban physicians led to increased surgical and ancillary ser- vices, as Well as to more intensive medical and surgical care (measured by the number of standardized units per service). Despite strong initial interest in target-income pricing, extreme versions of this ap- proach have been largely discredited within the academic community. Significant prob- 3Some previous work on price rigidities in the physician services market found evidence of SID (Stano et al., 1985; Cromwell and Mitchell, 1986). Despite these contributions, however, price rigidities and disequilibrium models have not received wide attention in literature so that the corresponding picture remains incomplete. Further, the support has been weakened by empirical evidence for the hospital services markets that is con- sistent with a standard demand—supply model rather than with various disequilibrium models that incorpo- rate inducement (Hay and Anderson, 1988). 4See Rice and Labelle (1989) for a review of this work. McGuire and Pauly (1991), however, distinguish be- tween inducement effects following fee changes and income effects on physician behavior. CHAPTER 10 Imperfect Information: S upplier~Induced Demand and SmallArea Variations 2.09 lems arise in explaining how the target income is established and why physicians do not consistently exploit their monopoly power by always pricing at profit-maximizing levels.5 Note also that the target-income concept is similar to often-heard claims that large corpo- rate firms have unexploited monopoly power to be able always to pass on tax increases or increases in other costs in order to maintain profit margins. Such claims, which would oth- erwise require these firms to price below profit-maximizing levels, are largely unsupported by empirical evidence. Furthermore, if physicians had the ability to target price, it would be reasonable to predict that physicians would have little concern about increases in costs, such as increases in malpractice insurance premiums, because they could always pass the increased costs on to consumers in the form of higher prices. Such, however, is not the case. THE DISUTILITY OF DISCRETION MODEL To understand the STD literature as a progression of ideas, we should reconsider one of the problems identified with the target-income approach. In particular, we will direct our attention to the problem of accounting for the chosen target. Why are physicians satisfied with a given target income? How can we explain the equilibrium inducement level and income level within the model itself? A major innovation in this regard was introduced by Evans (1974). Evans repre- sented the physician as a utility-maximizing provider whose utility function is expressed as follows: U= U(Y,W,D) where Yis net income, Wis hours of work, and D represents the discretionary influence used to augment demand. The physician has monopoly power so that he or she faces a negatively sloped demand curve that shifts inward as the number of physicians in- creases. A key feature of the model is the assumed increasing marginal disutility of aug- menting demand. According to Evans, physicians prefer not to induce demand and find greater incremental displeasure with greater inducement. This displeasure must be counterbalanced by the gains to income that inducement provides. When competition tends to reduce income, the physician may increase inducement to help offset this. By placing limits on the extent to which demand will be increased, this feature ad- dresses a weakness of earlier SID models that had no apparent brakes on physicians’ abilities and willingness to increase demand. In the Evans framework, a physician will augment demand to the point where the marginal utility of the additional income equals the marginal disutility of the added work plus the marginal disutility of the dis- cretionary influence. Graphical Representation of the Evans Model Several essential ideas of the Evans model can be made clearer with the aid of a graph. To do this, let us suppress the variable W, the work effort of the physician. Work effort was important to Evans’s empirical tests, and thus we refer the interested reader 5Rizzo and Zeckhauser (1998) draw on economic theory to resurrect the possible importance of income tar— gets to physician behavior. See also Rizzo and Blumenthal (1996). 1T III Information to the original article. In this simplified model, let the physician’s utility depend solely on income, Y, and discretionary demand, D. In Figure 10.2, the indifference curves represent the physician’s preferences. The indifference curves slope upward because one of the two “goods,” that is D, is really a “bad.” Can you see why they must slope upward? In order to remain on the same in- difference curve, the physician must be compensated for engaging in a higher level of D. As is the usual case, however, higher curves are preferred. To make the example straightforward and the measure of discretion concrete, let discretionary demand, D, be measured in units representing the number of induced units of service. If these induced units are sold at a constant profit rate of "lT, and if the maximum profit level with zero inducement is 'er0, then the physician’s total profit (in- come) can be described by Y: ero + 'n'D. This equation is a linear equation with an intercept of erD and a slope of “IT. This linear equation represents the attainable com- binations of Y and D available to the physician, and thus it represents the physician’s income possibilities. Figure 10.2 can now be used to describe a disutility of discretion equilibrium. In the figure, the physician seeks to reach the highest indifference curve attainable, and this outcome occurs at the point of tangency shown as point A. Furthermore, the figure il- lustrates the physician’s response to increased competition. Let increased competition be represented as a reduction in the average profit rate the physician receives from pro- viding units of service. A lower profit rate, 11", results in a budget constraint with a flat- ter slope and a lower intercept. The new equilibrium occurs at point B. In this example, increased competition has led the physician to choose a higher degree of inducement (i.e., DB > DA). Discretionary influence CHAPTER 10 Imperfect Information: Supplier-Induced Demand and Small Area Variations 211 Limitations of the Model This simplified version accounts for inducement and income endogenously, one of the benefits of Evans’s original model. The figure also illustrates a drawback of the origi- nal model—the fact that the model is ambiguous in its predictions of the effects of in- creased physician availability on the level of D.To see this, experiment with Figure 10.2 by drawing other indifference Curves representing physicians with different preferences. You will find that a physician might alternatively reduce the inducement level, D, when the budget constraint rotates downward. That is, tangency can be achieved at point B’ with a lower level of D on the indifference curve represented by the dotted curve. Econ- omists summarize this ambiguity by saying that the comparative static properties of the model are not determinate. Using a different utility function, Ellis and McGuire (1986) note that imperfect agency may, in some circumstances, lead to an undersupply of care. Another concern with the approach stems from the fact that physician producers are firms. In conventional economic theory of the firm, economists have found that the profit- maximization model often yields strong resultsAn alternative SID approach would deter- mine whether aprofit—maximization model yields stronger predictions in the context of SID. A PROFIT-MAXIMIZING MODEL Before describing such a framework, begin with the standard profit-maximizing condi- tions for a firm with a negatively sloped demand curve. This is shown in Figure 10.3 (at the intersection of the marginal revenue curve with the marginal cost curve) to produce the equilibrium price—quantity combination, P1,Q1. One of the key distinctions between monopoly and competition is that, at the monopoly equilibrium, price exceeds marginal cost. In competition, to the contrary, price equals marginal cost. The gap between price and marginal cost in imperfect competition, much like the price rigidities case for price takers, creates an incentive to enhance demand. Why? If the seller can sell more at a price that is higher than marginal cost, profitability in— creases. Of course, alternatively, if there had been perfect information and perfect agency, the seller could not thus improve its position. The Parallel Between lnducement and AdVertising On the other hand, with imperfect information and the ability to induce increases in demand, sellers often can improve their profits‘by inducing customers to buy more output. This idea is similar to theories of more typical commodities where sellers have an incentive to introduce advertising and product promotion in order to increase de- mand. However, these activities also raise costs so they should be undertaken only to the point where their marginal revenues equal marginal costs. The new equilibrium is illustrated in Figure 10.3 by the price—quantity combination P2“? As noted, the decision to induce, under this theory,has elements that are similar to stan— dard advertising/product promotion decisions. The physician expects revenues to increase, but costs in the form of reputation loss and time costs also have to be recognized. Follow- ing Stano’s (1987) representation of STD within this context, several observations follow from the advertising analogy.6 First, the incentive to advertise will be linked to the gap 6Although it appears that the utility approach introduced earlier has been abandoned, in order to Simplify the analysis, the various costs associated with foregone leisure and imperfect agency can be incorporated into the cost function. LRT III Information Price of services MR1 Quantity of services between marginal revenue and marginal cost.A small gap, which arises under a relatively elastic demand, reduces incentives. If the typical physician has a high degree of monopoly power, the gap will be large. Empirical evidence indicates that while the overall elasticity of demand for physician care is low, the elasticity of demand facing individual providers is considerably higher.7 That is, the market for physician services is monopoliStically com- petitive. Further, with suggestions that the market has become more competitive,8 the fore- going approach predicts that the incentive to induce demand is likely to be A second important consideration follows in that the optimal level will depend on the relative costs and gains of manipulating consumer demand. Previous discussion of agency suggested that consumers are not as ignorant as is often assumed. Ethical and professional considerations restrict physician activities, and, notably, various contracts between principal and agent have evolved to limit physicians” informational advan- tages. These include the monitoring feature of the traditional physician—patient rela— tionship, as well as other arrangements introduced by third—party payers and managed care contracts. The efiects of the various constraints on physician behavior are to raise costs of imperfect agency and thereby reduce the extent of SID. Finally, if we draw the advertising analogy closely, we also can expand somewhat our concept of physician inducement. That is, we can recognize some forms of induce- ment by physicians that are no more or less harmful to the patient than are some forms 7See Langwell and Nelson (1986) for a review of the evidence. 5See P. Feldstein (1986) and Newhouse (1988). CHAPTER 10 Impelfect Information: Supplier—Induced Demand and Small Area Variations 213 of advertising to the general consumer. Certainly, if inducement takes the form of a physician providing misinformation, recommending procedures or risks not suitable for the patient, or recommending excessive levels of care, then inducement is correctly characterized as a problem of imperfect agency. This is understandably the form of SID that is worrisome to policy analysts. However, just as advertising may sometimes work its demand—enhancing effects through its informational function, so sometimes may physician inducement take the form of providing enhanced information to the patient. The Identification Problem* Despite extensive literature on the subject, the difficulties of developing and testing the- oretically consistent models of SID have limited the applicability of much of the em— pirical work. The problem is that in much of the empirical work, the SID effect cannot be econometrically identified. The identification problem in econometrics was ex- plained in chapter 3, and reference to that chapter may be helpful. THE IDENTIFICATION PROBLEM: AN ALGEBRAIC EXAMPLE The problem will be illustrated next with the help of an algebraic example. For those readers who wish to forego the algebraic derivations, we summarize as follows. Begin with demand and supply curves such as those illustrated in Figure 10.1. The identifica- tion problem is exemplified by the following fact: If we construct an algebraic model with no SID, the resulting equation that is commonly estimated by researchers will look much the same as the equation derived from an algebraic model with SID. This fact is suggestive of the problem faced by the econometrician in that SID often cannot be dis~ tinguished from no SID by the statistical evidence. To confirm this point, following Folland and Stano (1990),we begin with an algebraic demand—supply model without SID as represented by the following linear equations: QD = a0 + alP + azY + Lil (10.1) and Q3 = [)0 + b1P + bZX + b3lVlD + Hz (10.2) where QD and QS represent quantities demanded and supplied and P represents price. Exogenous demand variables such as income and education are represented in the de- mand equation by Y, and while exogenous variables in the supply equation, such as the price of inputs, are represented by X. The variable MD represents the availability of physicians. From these two equations plus the condition that the quantity supplied, QS, equals the quantity demanded, QD, we can solve for the equilibrium quantity: Q = Co + 01X + CZY + c3MD + v (10.3) The coefficients in equation (10.3) are complicated combinations of those in equations (10.1) and (10.2), as will be discovered when one performs the algebra (see exercise 5). When the econometrician estimates the parameter c3 and discovers it to be greater than zero, he or she can conclude correctly that increased physician availability (that is, in- creased MD), has a positive effect on equilibrium utilization. However, this cannot be taken as evidence of SID; recall that by assumption we started with a model with no SID. *For more advanced students. [II Information For SID to exist would mean that physician availability affects demand directly; that is, it would mean that MD appears directly in the demand equation. To illustrate, sup- pose that SID does exist. In that case, a model of demand with SID would be as follows; 0 = a0 + a1P + azY + a3MD + u1 (10.4) Here the availability of physicians (that is, the variable MD) appears directly in the de- mand equation as a demand shift variable. Upon solving algebraically equations (10.4) and (10.2), the resulting equation with SID looks just like equation (10.3) without SID, Furthermore, the econometrican usually will not be able to use the resulting esti- mates to determine whether or not SID ein'sts. As Auster and Oaxaca (1981) have shown, this identification problem is not resolved easily, even when more sophisticated estimation methods are used. This finding has been reinforced through an innovative approach involving the demand for childbirths. Dranove and Wehner (1994) show that a statistical method (two-stage least squares) often used to estimate the inducement ef- fect produces “evidence” of inducement for childbirths eVen though one would not ex- pect to find inducement here. The apparent anomaly arises because that method does not deal properly with identification issues. Dranove and Wehner tested this issue in an unusual and revealing way by begin- ning with the “assumption” that obstetricians per capita (OB GYN) have the pOWer to induce demand for childbirths and asked “suppose we test this hypothesis using the typical approach in the SID literature—the two-stage least squares technique.” This technique, like other regression techniques, requires that the equations be correctly identified. Applying U.S. county data, the researchers first showed that OBGYN and child- births per capita were positively correlated (0.33, p < 0.01), a result to be expected through the ordinary long run operations of supply and demand.They then applied the two-stage procedure, as it is typically used in the SID literature, to measure the “pure SID” effect, an effect that is logically impossible to occur in the experimental circum- stances. Their estimated SID elasticity for obstetricians per capita was 8 percent, seem- ing to ludicrously imply that an influx of obstetricians per capita of one standard deviation (a standard deviation was 0.064 obstetricians per 1,000 population) would in— duce an additional 7 percent in childbirths per capita! As a result, We conclude that much of the reported evidence of SID is unreliable. An account of the evidence must appeal to studies that have in some way successfully avoided the identification problem. THE EVIDENCE ON SID One way of overcoming the identification problem is through the study of the effects of fee rigidities or price controls. These studies, some of which Were reviewed earlier, gen— erally indicate that the volume of services increases after controls are imposed; that is, some SID effect is found.The magnitude of the effect, however, tends to be modest. Another way of overcoming the identification problem is to circumvent it. In this regard, Hay and Leahy (1982) formulated a simple proposition. If imperfect agency leads to a higher volume of services provided, we ought to see those with more com- plete information consuming fewer services. The authors, in fact, found that after con- trolling for other socioeconomic factors including health status, medical professionals and their families are at least as likely to visit physicians as others who are presumably less well informed. CHAPTER 10 Imperfect Infonnation: Supplier-Induced Demand and SmallArea Variations 215 The contributions by Rossiter andWilensky (1983, 1984) are also noteworthy.These authors distinguished between patient-initiated visits, presumably not subject to physi— cian control, and what may be subsequent physician-initiated visits.The authors attempt to identify the extent of SID by fecusing on the relationship between the physician- initiated component and variables representing physician self-interests. In particular, if SID exists, increases in physician availability would be expected to increase physician- initiated services, ceteris paribus, but would not affect patient-initiated visits. The au- thors found exactly this pattern; however, the estimated SID effect elasticity, although it was statistically significant, was small—less than 0.1. This means that a 10 percent in- crease in physician availability produces an inducement effect of less than 1 percent. In the context of Dranove and Wehner’s findings, even these magnitudes are suspect. In a similar study of Medicare beneficiaries, Escarce (1992) found that greater availability of surgeons increased initial contacts with surgeons but had little effect on the intensity of subsequent services. The higher utilization resulting from an increased supply of surgeons is thus associated with patients’ preferences for access to care or re- ferrals rather than to SID. Elsewhere, Grytten and colleagues (1995) also distinguished betWeen patient and primary physician-initiated visits under the fixed-fee schedules of Norway’s public health care system. The investigators did not find any evidence of either SID or ra- tioning of physician-initiated visits, though there was some indication of inducement for laboratory tests. More recently, Shrenson and Grytten (1999) studied the efiects of in— creased competition in Norway. Increases in the number of physicians in an area, as ex— pected, reduced the number of patients seen by a typical physician. However, contrary to the SID hypothesis, Norwegian physicians did not compensate for their falling in— comes by increasing services for patients who remained. OTHER LIMITATIONS ON SID TWO other contributions provide theoretical grounds to suppose that there are inherent 7 limits on the physician’s ability to influence demand. Rochaix (1989) recognized that a patient is not fully dependent on any physician and in fact forms prior expectations about his or her own condition. If the physician’s suggestions deviate sufficiently from those expectations, the patient may not consent to the proposed care, but rather seek another opinion. This ex ante monitoring constrains physicians’ behavior. The author demonstrates that it takes only a small number of informed patients to constrain physi- cians to act as “better agents.” Similarly, Dranove (1988) proposed that patients form expectations about the severity of their conditions. In the inducement decision, a physician is concerned about the loss of business were he or she to develop a reputation as overly aggressive. Among various factors, Dranove showed that improvements in patients’ diagnostic skills reduce the gains from inducement. Evaluation of SID Our partial review of the SID literature has selected from a wide range of contribu- tions and ideas. The whole of the STD literature shows that the intuitively simple idea of physicians inducing demand actually has many, often interesting complexities. Pauly (1988b) and Phelps (1986) have even suggested that the complexities inherent in RT HI Information the SUD—imperfect information controversy mean that the issue may never be settled completely. Elements of competition, agency relationships, and consumer search place sub- stantial limits on the ability and willingness of physicians to generate increases in de- mand. Research contributions that lend support to SID are often unable to overcome theoretical or statistical problems that cloud their interpretation. Often reports of SD) confound the normal workings of markets with the perception of extraordinary work- ings. SID must first be separated from cases where the supply of physicians increases but no shift of demand occurs. This problem in econometric analysis is the “identifica- tion problem” to which two stage least squares analysis is typically addressed. Two—stage analysis is itself prone to errors and misinterpretation. The best two stage analyses, in any case, suggest a very small SID efiect, if any. Other approaches also tend to minimize our estimates of 81]). Therefore, it seems safe to suggest that health care markets often can be studied us- ing models that postulate stable demand functions even though there may be some de- gree of SID. In other words, consumers are not nearly as helpless as many might imagine in that they take steps to protect their self-interests when those interests are threatened. It follows that it would be reckless to argue market failure, at least on the basis of SD), and to dismiss policy analysis based on standard models as ineffective. Area Variations (SAV) Our work on asymmetric information and agency relationships emphasized the effects of a patient’s information disadvantage, as well as the physician’s dual role as provider and agent. We have not yet, however, considered the likelihood that physicians them- selves are not perfectly informed about the outcomes and relative effectiveness of a1— temative treatments. For conditions with alternative treatments, it is easy to imagine physicians having different levels of information or disagreements about the most ap- propriate care. Are the wide geographic variations in the per-capita utilization rates for many medical and surgical procedures related to this information problem, and what is the significance of these variations? The study of medical decision making and small area variations (SAV) developed in medical research. Economists became increasingly interested in SAV once they rec- ognized its policy implications, as well as the potential role of information in explaining this phenomenon. Whereas SID occurs in theory due to the information gap between physicians and patients, difierences in treatment patterns across small areas may occur because of physicians’ uncertainty and ignorance over the best medical practice. Thus, even if physicians acted as perfect agents, the consequences resulting from their own in— formation gaps must be examined. Much of the modern interest in SAV stems from the pioneering work of John Wennberg and colleagues (1973, 1977, 1982), who studied New England hospital mar- kets. Their evidence of wide variations across small service markets has been corrobo- rated by many others. Nevertheless, the sources for these variations, as well as their importance to public policy, remain controversial. Why do women in one New England town undergo hysterectomies at more than twice the rate of another, apparently simi- i. i g. CHAPTER 10 Imperfect Information: S ztpplier-Induced Demand and Small Area Variations 217 Rate per 100,000 population at risk 1000 6 New England 900 O Non/vegian counties El Districts of West Midlands 800 RHA England 700 600 500 400 22:4 that: i M. t Hernia Appen- Cholecys- Prosta— Hyster- Hemor— Tonsil- repair dectomy tectomy tectomy ectomy rhoid- [ectomy ectomy lar, New England town? As Figure 10.4 shows, the phenomenon is not limited to New England or even to small areas Figure 10.4 illustrates some early findings for seven common surgical procedures in regions of New England, the United Kingdom, and Norway. The use rates, which have been corrected for age and gender differences across populations, vary widely for some procedures, as shown by the ranges in Figure 10.4. Wide intercountry differences also exist in the mean rates. Despite many possible sources of the variations, most of the controversy has fo—. cused on the role of physician practice style and the question of whether the observed variations are evidence of unnecessary or inappropriate care. The resolution of this con- troversy is clearly crucial for policy. As we will show, if the variations can be reduced through scientific research on patient outcomes, substantial welfare gains for society can be achieved. The Physician Practice Style Hypothesis SAV analysis often focuses on service utilization rates in small, contiguous hospital mar- - kets with populations that are relatively homogeneous. This method helps rule out dif— ferences in population characteristics (e. g., health status, education, and income) as the sources for the observed variations. Because initial studies of this type suggested that LRT III Information population characteristics do not account for much of the observed variations, some re. searchers proposed that SAV is caused in part by supplier—induced demand. This ex- planation further links SID and SAV. Others posit a distinct source of the variations through the “practice style hypothesis.” Wennberg (1984) argued that much of the observed variation is closely related to the degree of physician uncertainty with respect to diagnosis and treatment.When there is little consensus about the effect and value of a medical procedure, a wide range of physician practice styles is considered to be within the bounds of accepted practice. The physician’s habits, beliefs, and patterns of practice are seen as his or her “style.” Some analysts have further suggested that physicians within a given market tend to “follow the pack” and eventually share a common set of beliefs. These beliefs, and therefore practice styles and utilization rates, often will differ across markets. MODELING PRACTICE STYLE To develop and test a rigorous model of practice style is a more difficult challenge than simply to postulate this hypothesis. One approach introduced by Folland and Stano (1989) modeled differences in practice style through differences in physicians’ beliefs about the true production function of alternative procedures to treat a particular con- dition.Within the language of health economics, the practice styles of physicians relfect their imperfect information about the technology of producing health (i.e., the produc- tion function). Figure 10.5 shows three different production functions for a patient with uterine leiomyoma (fibroids)—a condition often leading to hysterectomy, which is a high variation surgical procedure. In reality, there is a true production function but it is not known by the physician. The physician must conjecture or estimate what the effects of each treatment step Will be. Health status MC Medical care CHAPTER 10 Imperfect Information: Supplier-Induced Demand and Small Area Variations 219 Suppose that the horizontal axis, MC, measures the medical care labor and capital combinations that are used to treat the patient. Moving along the horizontal axis rep- resents a progression of care from simple office visits and surveillance to surgery. Dif- ferences in practice style are shown by differences in beliefs about the true production function. Physician 2 with the production function 52 is much more optimistic about the effects of medical care on the patient’s outcome or health status (HS) than Physician 1 with production function $1. In principle, both physicians could be “misinformed,” and the true production function might be one such as 5*. Which physician is more likely to recommend more services including hysterec— tomy? Under the practice style hypothesis, those like Physician 2 who most strongly believe in the efiectiveness of hysterectomy will be more likely to recommend surgery. Differences in practice styles, therefore, provide a possible accounting for interarea dif— ferences in utilization rates. If physicians of type 1 and 2 are not distributed‘similarly across markets, then the hysterectomy rates will vary. Of course, wide variations in prac- tice styles within markets can be consistent with little variation in average practice styles across markets, and therefore in the predicted per-capita rates of utilization. This possi- bility marks one of the deficiencies of studying the practice style phenomenon through population-based utilization data. The production function approach to practice style also helps to show that a seem- ingly high surgery rate in a market does not automatically represent unnecessary care, particularly where there is ambiguity or incomplete scientific information about the true production function. Furthermore, as medical science uncovers better information about the true production function (assume it is S*), efforts to disseminate this infor- mation will help reduce unnecessary surgery among Physician 2 types, but it also will promote hysterectomies among those who were previously overly pessimistic about their effectiveness. Finally, our approach indicates why it can be difficult to distinguish SID from SAV. Areas with a high proportion of providers like Physician 2 will have high surgery rates, tempting some observers to conclude that SID is prevalent, Whereas the' true source is practice style. It also can be true that physicians in those high utilization areas are merely acting as good agents by supporting their patient beliefs and preferences for surgery. Here the variations are due to patient preferences rather than SID or SAV. TESTING THE PRACTICE STYLE HYPOTHESIS How well does the practice style hypothesis hold up empirically? The basic problem with even well-defined theoretical concepts of practice style is that the empirical re- searcher has no apparent objective measure of practice style. In the absence of an op- erational measure, evidence for the hypothesis will necessarily be indirect. Such indirect evidence is available, and it comes from several sources. Education, Feedback, and Surveillance Some studies show that education, feedback, and surveillance programs directed at physicians can alter their behavior and thus presumably their practice styles.9 One study found that an informational program significantly affected the tonsillectomy rates in 13 9Those discussed here are Wennberg et al. (1977), Dyck et a1. (1977), and Nyman et a1. (1990). RT III Information New England areas. Another similarly found that the rate of unjustified hysterectomies dropped by two-thirds subsequent to a review program introduced in the Canadian province of Saskatchewan. But even though intuitive reasoning suggests that monitor- ing should change behavior, the effects are not always large. For example, reviews to identify inappropriate claims submitted by physicians under Medicare only marginally reduce the rates at which inappropriate claims are submitted. Comparing Utilization Rates in Homogeneous Areas A second empirical approach to investigate the practice style hypothesis is illus- trated through an early but influential study by Wennberg and Fowler (1977). The au— thors extensively studied utilization rates and population characteristics for six areas in Vermont. After finding that morbidity and many other socioeconomic variables were not statistically significant, they concluded that the differences in use rates probably are due largely to practice style differences across areas. Though practice style differences can contribute to differences in utilization across areas, the inference that interarea difierences in utilization rates, even in apparently similar areas, are due largely to difierences in practice style is incorrect. Green and Becker (1994) performed a SAV analysis in which they compared two demographically similar towns in Michigan where acute cardiac ischemia (AC1) discharge rates differed by a factor of three. Large difierences in admission were due more to differences in the proportions of patients with ACI who went to their hospital emergency department rather than to differences in physician decision making. Area variation does not neces- sarily mean variation in practice style. Multiple Regression Approaches It is generally desirable to examine interarea utilization through statistical tech- niques that control for the other factors simultaneously. The inclusion of many or con- ceivably all relevant measurable variables in the analysis suggests another indirect approach to the practice style issue. As more variables are added to the analysis, one learns better how much of the observed variation can be explained by standard socio- economic variables without resorting to the hypothesis of practice style. In cases where the variations are largely explained by such variables, the potential role of practical style is small and much of the mystery of the variations phenomenon is eliminated. In con— trast, an inability to explain the variations in terms of the usual variables leaves open the possibility that practice style and other unknown factors are responsible for much of the variation. Practice style is then essentially the residual when other impacts have been explained. Multiple regression studies using standard socioeconomic variables explain sub— stantial amounts of the interarea variations, especially when they focus on aggregated utilization rates (commonly representing a weighted average of many procedures) in larger geographic areas.10 Multiple regression analysis at the small area level for indi- vidual procedures, as opposed to aggregates over large numbers of procedures, has gen— erally been less successful in explaining the variations. Nevertheless, Phelps and Parente mSee Folland and Stano (1990). Folland and Stano (1989) also show the conditions under which the resid- ual variation from multiple regression can be taken as an upper bound for the contribution of practice style. They showed that practice style is relatively unimportant for large aggregated sets of commodities. CHAPTER 10 Imperfect Information: S upplier—Induced Demand and Small Area Variations 221 (1990) found that standard demand and supply variables typically account for between 40 percent and 75 percent of the variation in their study of 134 separate diagnostic cate- gories. Elsewhere, Escarce (1993) found that 43 percent of the variation in cataract surgery rates for the Medicare population is explained by socioeconomic variables. This still leaves a potential 57 percent that may be due to practice style. However, after purg- ing the surgery rates of any possible practice style influence, Escarce found that the vari- ation in cataract surgery rates was reduced by only a small amount. These contributions indicate that practice style could be important for some pro- cedures, but the contribution of other variables should not be overlooked. As Escarce demonstrates, it also would be wrong to attribute all or most of the unexplained varia- tion in utilization rates to practice style. For many procedures, conventional economic and demographic influences may well dominate the as—yet-undetermined effects of practice style.11 SAV and the Socal Cost of Inappropriate Utilization The most important and controversial issue in the SAV literature is the proposition that substantial variation in utilization rates across market areas is an indication of in— appropriate care. It is tempting to reach these conclusions (even though our previous discussion warns against this) and to extrapolate savings under the assumption that high—use areas experience the rates found in lower—use areas. For example, Wennberg and colleagues compared utilization rates in Boston and New Haven.12 The authors ar— gued that the cities are demographically similar, but because Boston has much higher utilization rates, substantial reductions on health care spending in Boston are possible. The potential savings from improving the quality of care and reducing excessive uti- lization in particular has led to a national focus on research into patient outcomes. In— dicative of this priority, the US. Congress in late 1989 created the Agency for Health Care Policy and Research. The agency’s primary missions are “to determine what works and to develop practice guidelines and standards to assess and assure quality of care.” The concept of consumers’ surplus shows how welfare losses due to variations in uti- lization rates can be measured (Phelps, 1995).The basics are illustrated in Figure 10.6, where the true marginal benefit curve for medical care involving a procedure X is given by the curve MB*.This curve reflects the benefits as perceived by the patient if he or she were to have full and complete scientific knowledge. We emphasize that the marginal benefit curve is not objectively determined by medical science but reflects patient valu- ations over the health and other consequences of the procedure. The efficient, fully in- formed rate of use is thus R2, the rate at which MB* crosses the marginal cost curve, MC. With underutilization, such as at R1, a gain in consumers’ surplus could be achieved by increasing use of medical care up to R2. The gain in consumer surplus is shown by tri- angle A, measured under the marginal benefit curve and above the marginal cost curve. This area represents the inefficiency or welfare loss to society from underutilization. In \ 11This result is illustrated by Coyte and colleagues (1997), who found that more than 90 percent of the varia- tion in knee replacement (KR) rates in a large Canadian province was explained by demographic and re- gional variables. Surprisingly, the reported prevalence of arthritis and rheumatism was not a statistically significant determinant. The authors recognize that the remaining variation could be accounted for by many factors including “clinical perceptions of the appropriate rate of KR.” l2Wennberg, Freeman, and Culp (1987);Wennberg et al. (1989); Fisher et a1. (1994). ART HI Information ' Marginal benefits and costs i i v t Rate of utilization of intervention X excessive utilization e the welfare loss of inefficiently a similar fashion, one can measur at rate R3 by triangle B. Despite the snnple and welfare losses of practice vari of these costs. Phelps and Parent ’ ts to the ' lication of consumers surplus concep I SE31: has proved difficult to generate rehable estimate: e (1990) and Phelps and Mooney (1992) approac e13 ' ‘ ‘ a1 benefit curves. ‘ 'on analys1s to estimate true margin . _ the pmblem by usmgli‘ligredihwn criticism in several respects. Regressmn always pr:ll However, the approac as ' 'dual. How much of the resrdu ‘ ' a re51 . diets the dependent variable 1mperfectly,leavmg ctors’ sum as Pauent is due to inappropriate use V ‘ ex ante Furthermore DIaIlO e (1995) has pointed out that pattents also make errors 7 tru " ' ' ' ‘tional ' Will dehver ex post. This addi ' ' e value that medical care ‘ . ' Whe'n eStclinturamtlig 3111: estimates further. Most economists stud g the issue probably “n01se” s yin . . . would agl‘ee While we are Sure Inappropriate use OCCUIS Commonly a. nd 15 Oft'ell 1y edon ha Sima S alCSS mp COSt , W 0 Ct VCCO 6 t6 t ta 0 t 3 tot O t of a PIOPUatC Cale . . . 1 om_ Regardless of the empirical challenges, many leading scholars in the medica c ' ‘ t of benefits, as op- gmz‘ ‘ f usmg an economic concep I ' ow reco e the importance 0 _ . . t e are m g mumty 11 fit easnres determined strlctly by medical scrence, and h y S posed to hem m ‘ ' Differences between econorru - - ' t dec1stons. ‘ ' to involve atrents m treatmen . 3 eat to be physmlal‘ls hprs over SAV Primtlce Stylev and appropnate car; thlis P th t and medical researc e ’ the health semces literature a ‘ ‘ win in narrowing. In particular, the consensus IS gro g aluafion (Chapter 4) are the COSt—efieCfiveneSS anaIYSis agglOthertEIiii: flirgiltgl‘lgffliii‘efiects on social welfare. ' ' trea e best means to identify and r i e average rate across markets is loss from the top 25 procedures ___— 1z’After eliminating legitirnat the correct rate,Phelps and exceeded $6 billion in 1987. . . . th of variation, and by assurnmg that 3223:1213; (1992) estimated that the total welfare CHAPTER 10 Imperfect Information: Supplier-Induced Demand and Small Area Variations Clinical Decision Making and Patient Preferences David Eddy is a piOneer in studying the role of information, medical science, and prefer- ences in clinical decision making. Eddy (1990) breaks the treatment decision into two main components—evaluating the outcome of al- ternatives and then comparing and choosing from among the options.14 The first component is largely a scientific one determined primarily through clinical studies of patient outcomes. However, for many patient conditions, even a fully informed physician Will face the problem of inadequate scientific information. The second component, comparing the alternatives, involves judgment and patient preferences. Eddy emphasizes that “it is the patient’s preferences that should determine the decision” (p. 442). A failure to make appropriate decisions can arise from failures in both stages. Clearly, Evaluation of SAV a physician’s inadequacies in education and training, as well as deliberate attempts to mis- represent outcomes (e.g., to engage in SID), will distort information needed at the second step. Similarly, if patients misrepresent or physicians misinterpret their preferences, or if inadequate information is communicated to patients, the second step will be flawed. The medical community is increasingly being alerted to the role of the patient through the development of new information technol- ogy to improve the decision-making process. For example, Wennberg and colleagues have I developed an interactive videodisc that helps patients with prostate cancer choose fiom among several treatment optionsTo help with the decision, patients answer questions de- signed to measure their attitude toward risk and their ability to tolerate discomfort. 223 Many medical and surgical procedures have Wide variations in use rates. Some of the variation involves little or no welfare loss, and health economists caution against the no- tion of a scientifically correct use rate for all areas. Eficient rates will vary to the extent they reflect the confluence of patient preferences and other economic factors with sci- entific information. Nevertheless, physician uncertainty and practice style may lead to unwanted or un- necessary variations in care. The degree to which they contribute to the variations is not completely understood. Despite the emphasis on practice style, some SAV researchers believe that SID and the availability of hospital resources are also important. Others have found less obvious sources including some associated with purely statistical phe- nomena and inadequacies in data-gathering methods.15 1“'Eisenberg’s (1986) work on physician decision making and practice patterns is also a major contribution and 42:.“ valuable reference. 15For example, Diehr and colleagues (1990) have demonstrated that, due to random variation, the observed variation across small areas can be high even when the underlying rates across the areas are the same. Legnini et al. (1991) showed that information about new technologies contributes to increasing variation, as well as increasing use of these technologies. Schwartz et al. (1994) found that variations in hospitalization rates using more than one year of data are substantially reduced. 0W kRT III Information Regardless of the relative contribution of practice style to interarea variations, it is likely that uncertainties in the clinical benefits of many procedures create substantial welfare losses through overuse, underuse, and misuse. The growmg recognition of m- formational inadequacies as major barriers to better care has led to massrve private and public efforts to determine “what works.” More formally referred to as evrdence-based medicine, this search involves the systematic collection and evaluation of clinical data. With its strong emphasis on cost-effectiveness analysis, many scholars are opumlstic about the role of evidence-based medicine in helping health plans improve the quality and efficiency of their care. L Supplier-induced demand (SID) refers to the phenomenon of physicians deviating from their agency responsibilities to provide care for their self-mterests rather than their patients’ interests. _ I . SID represents one of the major intellectual and pohcy controverSIes in health eco- nomics. Its origins can be traced to Roemer’s Law, which predicts that a “bed built is a bed filled.” . i 3. SID is typically represented graphically by increases in demand followmg increases in supply, although there are theoretical flaws with this representation. . I 4. The target-income hypothesis is a related idea that suggests that physrcrans use their discretionary advantage to achieve a target level of income. . Evans theorized that the utility of the gain in income has to be balanced by the disu- tility of inducement and the disutility of the work in providing care. This idea es- tablishes theoretical limits on the level of inducement. 6. The disequilibrium approach to SID resulting from price rigidities establishes a gap between demand and supply and a motive to induce increases in demand. ' I 7. Another approach develops the analogy between inducement and advertismg. It predicts that inducement will diminish as competition reduces the gap between a physician’s marginal revenue and marginal cost. I ' 8. The identification problem, which refers to the difficulty of determining the extent of SID, has plagued much of the empirical literature on the subject. . 9. The empirical work that has overcome or sidestepped the identification problem has found evidence of SID. However, the magnitudes tend to be much smaller than was believed originally by supporters of the hypothesis. . 10. Although conventional economic analysis oversimplifies the physician services market, the standard competitive and monopoly models provide useful insightrnto the effects of changes in demand and physician availability on prices and quantities. 11. The small area variations (SAV) phenomenon refers to the wide interarea varia- tions in the per-capita use rates found for many medical and surgical procedures. Such variations have been found in the United States and many other countries. 12. The practice style hypothesis, which refers to physician uncertainty in diagnosis and treatment, has been proposed as the major explanatory factor for the variation. 13. Practice style can be modeled through differences among physicians in their behefs about the marginal benefits of alternative treatments. ‘ 14. Most of the evidence for the practice style hypothesis is indirect, coming from stud- ies showing the following: 5») 5n CHAPTER 10 Imperfect Information: Supplier-Induced Demand and SmallArea Variations 225 - changes in practice patterns following physician education and monitoring; - wide variations in utilization rates across small, homogeneous areas; - high, unexplained residuals in multiple regression analyses of interarea utiliza— tion rates. 15. The SAV phenomenon is often interpreted to suggest that considerable amounts of unnecessary care are being provided and that major savings are possible through elimination of that care. 16. Substantial problems remain, however, in identifying inappropriate care and prop- erly accounting for the role of patient preferences. Discussion Questions 1. \000 10. 11. 12. Suppose that insurers monitored all health care payments to determine Whether the services were appropriate. Would you expect to see more or less tendency to- ward SID? . Figure 10.2 shows how increased competition can lead to a higher degree of in- ducement at point B. This suggests that providers try to induce more usage to com- pensate for lower profit margins. Suppose, however, that the physician ends up at point B’, where there is less inducement. What types of preferences would lead to reduced inducement in response to less profits? . How is the possibility of SID related to the price elasticity of demand for services? Does increased demand elasticity increase or decrease the possibility of SID? . If prices are relatively rigid so that they do not adjust rapidly to changes in supply, explain how SID can be manifested. . What are some criticisms of the target-income approach to SID? In the target-in- come model, what determines the physician’s target income? Would target incomes differ between physicians in practice? Would some physicians choose instead to be income maximizers? . What forces limit a provider’s ability and willingness to engage in SID? In the profit-maximizing model of SID, what are the costs to the physician of inducement? Give examples of inducement costs. . Assume that SID is prevalent and that few forces limit or check this phenomenon. What are some of the implications for policy? . In what ways is inducement similar to advertising? In what ways is it different? . How do the medical care consumption patterns of physicians and their families dif- fer from those of nonphysicians? If physicians may be considered well informed, what does this suggest about the SE) hypothesis? Does it matter for this point that physicians tend to have larger—than-average incomes? Discuss. What is the SAV phenomenon? What are some economic forces that can help ex- plain SAV? What are some demographic and other considerations? How might physician uncertainty lead to SAV? What evidence supports the practice style hypothesis? What possible criticisms can you propose for the idea that areas with high utiliza- tion rates for a procedure indicate the provision of unnecessary care? Suppose you observed that there are no interarea variations in the use rates for some procedures. Can this be taken as evidence that good health care is being pro- vided? If not, what other kinds of information would be useful? 111 I\) SJ.) Information . If one cause of SAV is lack of information about various procedures, would you ex- pect SAV to have increased, decreased, or stayed the same over time? Why? As new technologies become available more rapidly for given procedures, would you expect an increased or decreased amount of small area variation? Suppose large variations occur in average use rates per person within small areas, as well as between small areas. Does this necessarily mean that the practice style hypothesis is incorrect? Suppose there are zero or small variations in average use rates between small areas. Does this necessarily mean that the practice style hy. pothesis is incorrect? Discuss. Roemer’s Law states that “a bed built is a bed filled.” Use demand and supply analy- sis to show how Roemer’s Law may be an example of the false cause fallacy Where cause and effect have been reversed. Hint! Assume an increase in demand for hos- pital beds under conditions Where beds can be supplied at a constant unit cost. Explain why the standard demandesupply framework in which prices adjust imme- diately to a new equilibrium is inconsistent with SID. Explain why the indifference curves in Figure 10.2 are positively sloped. If the use of discretionary influence were not a “bad,” What would the indifference curves in Figure 10.2 look like? Where would the equilibrium solution along the physician’s income locus be found? . Set QB in equation (10.1) equal to Q5 in equation (10.2) and solve for P, showing that it is a linear function of the remaining variables. Substitute your solution for P back into either (10.1) or (10.2) and solve for Q = QD (or QS) to find equation (10.3). Alternatively, solve for P in each of the equations by putting P on the left-hand side. These are known as the inverse demand-and- supply equations. Then eliminate P by setting the two equations equal to another and solve for Q. Consider the evidence that fee controls or price ceilings tend to be followed by uti- lization increases. Why doesn’t this evidence necessarily support either the target income pricing or SID hypotheses? Hint: Consider a physician with some degree of monopoly power and, for simplicity, constant marginal cost. Examine the profit— maximizing price and quantity combinations before and after a price ceiling is in- troduced to determine what will happen to utilization and spending. ‘ In comparing SAV across diseases and diagnoses, would more complicated diseases suggest greater or lesser variation? . Using Figure 10.6, what are the welfare gains to consumers in going from R1 to R2? What are the costs of the increased consumption? Show that the net gain corre- sponds to the area represented as A. Consider the approximation of the welfare loss due to interarea deviations from the correct rate of care. All else equal, which pr0cedures would yield the largest wel- fare losses—those with low price elasticities or those with high price elasticities? Why is this the case? P A R T I V : INSURANCE AND ORGANIZATION OF HEALTH PROVIDERS ' ‘s‘uqv‘a‘aaaflql-E‘ J“ J!“ (9 av iv U! DID! a caucuses: mean Ev tunic H ageqaqaqaqeqeqa —=>——=—-—u»u— IllaEllau‘nan‘agtfldn’tlma‘nun‘flan wivawdulwlvlaavl Raem5memnanaemene «BaaqEqbqawnquwu .-u—u—-—u—_____- u :1 .- rgonzoten nsuronce orkets Loading Costs and the Behavior of Insurance Firms 3 Employer Provision of Health Insurance and the Tax System E Who Pays for Health Insurance? E Employer—Based Health Insurance and Job Mobility The Market for Insurance The Uninsured—An Analytical Framework Conclusions surance. Insurance is an arrangement that allows risk-averse individuals to re- duce or eliminate the risks they face. Consumers buy insurance to replace the uncertainty of a large loss or major expenditure with the more certain prospect of regu— lar premiums. Profit—seeking firms supply insurance. In an idealized insurance market, the premium as a percentage of the loss resulting from an event will approach the prob— ability of the event’s occurring. We concentrated fundamentally on the impact of insurance on individuals. In this chapter, we focus on the insurance market and the behaviors of firms within that mar- ket. Within the context of the employer—provided health insurance that is so common in the United States, we establish who pays for health insurance. We continue with an ex- amination of employepprovided insurance and job mobility. We then look at the tradi- tional community-rated health insurance (where individuals or groups all pay the same premium) and show how that market has changed. An analysis of the uninsured con- cludes the chapter. Chapter 7 introduced the concept of insurance with a primary focus on health in- 227 ...
View Full Document

This note was uploaded on 02/08/2012 for the course ECON 320 taught by Professor Chan during the Spring '11 term at SUNY Albany.

Page1 / 12

GSFch10 - ‘ I‘l‘I_E"‘Iql‘l‘v_l‘l‘lql...

This preview shows document pages 1 - 12. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online