NBER WORKING PAPER SERIES THE WELFARE EFFECTS OF PUBLIC DRUG INSURANCE. Darius Lakdawalla Neeraj Sood

Transcription

1 NBER WORKING PAPER SERIES THE WELFARE EFFECTS OF PUBLIC DRUG INSURANCE Darius Lakdawalla Neeraj Sood Working Paper NATIONAL BUREAU OF ECONOMIC RESEARCH 15 Massachusetts Avenue Cabridge, MA 2138 October 27 We are grateful to the National Institute on Aging for funding (1R1AG2194. For any helpful coents and suggestions, we thank seinar participants at the 27 NBER Suer Institute Health Care session. The views expressed herein are those of the author(s and do not necessarily reflect the views of the National Bureau of Econoic Research. 27 by Darius Lakdawalla and Neeraj Sood. All rights reserved. Short sections of text, not to exceed two paragraphs, ay be quoted without explicit perission provided that full credit, including notice, is given to the source.

2 The Welfare Effects of Public Drug Insurance Darius Lakdawalla and Neeraj Sood NBER Working Paper No October 27 JEL No. H2,H51,I11 ABSTRACT Rewarding inventors with inefficient onopoly power has long been regarded as the price of encouraging innovation. Public prescription drug insurance escapes that trade-off and achieves an elusive goal: lowering static deadweight loss, while siultaneously encouraging dynaic investents in innovation. As a result of this feature, the public provision of drug insurance can be welfare-iproving, even for risk-neutral and purely self-interested consuers. In spite of its relatively low benefit levels, the Medicare Part D benefit generate $3.5 billion of annual static deadweight loss reduction, and at least $2.8 billion of annual value fro extra innovation. These two coponents alone cover 87% of the social cost of publicly financing the benefit. The analysis of static and dynaic efficiency also has iplications for policies copleentary to a drug benefit: in the context of public onopsony power, soe degree of price-negotiation by the governent is always strictly welfare-iproving, but this should often be coupled with extensions in patent length. Darius Lakdawalla The RAND Corporation 1776 Main Street Santa Monica, CA and NBER darius@rand.org Neeraj Sood The RAND Corporation 1776 Main Street Santa Monica, CA 947 and NBER neeraj_sood@rand.org

3 I Introduction Patents encourage innovation by awarding inefficient onopoly power to inventors. This leads to the failiar trade-off between inducing innovation, and ensuring the efficient utilization of invented goods. Public prescription drug insurance provides a way out of this dilea, because it helps decouple the price consuers pay fro the price innovators receive. By subsidizing coinsurance for drugs, public insurance encourages utilization, but without necessarily coproising innovators prices, profits, and incentives for research. As such, public insurance can siultaneously proote static and dynaic efficiency, which are often at odds. The social value of publicly provided drug insurance is typically thought to derive fro its insurance value, and the value of care provided to less affluent groups. However, its static and dynaic efficiency effects iply that public drug insurance is valuable to risk-neutral, selfinterested consuers. We deonstrate this point theoretically, and the data suggest that it is of considerable quantitative iportance. Even though Part D is not a particularly generous insurance plan, it generates an annual value of $3.5 billion in static deadweight loss reduction, and $2.8 billion in additional innovation. The total value of $6.5 billion is within 13% of edian estiates for Part D s social cost. Part D is nearly welfare-iproving, even ignoring the aspects of the progra typically thought to generate its entire social value. Taken as a whole, the Part D legislation like any public prescription drug insurance progras addresses ore than just insurance for prescription drugs. While prescription drug insurance per se enhances welfare, the auxiliary provisions of the Part D progra rest on shakier conceptual foundations. Two provisions in particular bear on its efficiency effects. First, the original legislation forbade the governent fro using its newfound buying power to negotiate 1

4 prices. This was otivated by concern that price-negotiation could lower pharaceutical profits and dapen innovation incentives. Second, and in stark contrast, the legislation also placed liits on the ability of innovators to gae the patent syste by acquiring extended patent protection. Many pharaceutical copanies use a variety of strategies to extend onopolies and block the entry of generic copetitors; the Medicare Modernization Act placed explicit restrictions on such behavior. The prohibition on price-negotiation is likely to be welfare-decreasing, even in a dynaic sense. 1 In particular, soe degree of price-negotiation is always welfare-iproving. The arguent is a failiar application of the second-best principle: it is welfare-enhancing to distort an efficient argin slightly, as long as it reduces distortion along another, inefficient argin. While patents distort consuer utilization, the innovator is nonetheless able to axiize his static profits with respect to price. Initially, therefore, sall deviations fro the optial onopoly price have no ipacts on onopoly profits, or on incentives for innovation. In contrast, they strictly lower the degree of deadweight loss fro under-utilization by consuers. As a result, soe degree of price-reduction enhances the efficiency of utilization, but has no deleterious effects on innovation, at least on the argin. The recent proposal in Congress to begin price-negotiation ay thus rest on sounder econoic footing than the original legislation. The case for or against patent-gaing is not nearly as clear-cut, but there is an iportant interaction between the two provisions. Soewhat surprisingly, the recent push to negotiate 1 While the current Congress has passed legislation that could end up overturning the prohibition on pricenegotiation, it reains in place as of this writing. The Medicare Prescription Drug Price Negotiation Act of 27 (H.R. 4 requires the Secretary of the Departent of Health and Huan Services to negotiate prescription drug prices for Medicare Part D starting in 28. This has not yet been signed into law. 2

5 prices actually undercuts the econoic case for liiting patent-gaing. The key is the wellknown result that the dynaically optial patent is long, but narrow, in the sense that it awards a negligibly sall per-period profit over an infinitely long period of tie (Gilbert and Shapiro, 199; Kleperer, 199. There is an iportant analogy between prescription drug policy and this principle of optial patent design. Patent-gaing can be thought of as a de facto eans of increasing patent length (albeit at soe social cost. In contrast, price-negotiation reduces profits at a given point in tie, and thus serves to narrow patent width. Since it is welfare-iproving to siultaneously lengthen patents and narrow arket power, it is siilarly welfare-iproving to couple patent-gaing with price-negotiation. This is precisely the opposite of the way the original Part D legislation was configured. In this paper, we analyze the welfare econoics of the Medicare Part D progra, with special attention to its provisions for drug insurance, price-negotiation, and patent-gaing. We begin with soe relevant background aterial, and then present our analysis in three parts. Section III presents a siple theoretical odel that deonstrates the welfare effects of Medicare Part D, as well as the approach to calculating the. Section IV uses the theoretical odel to quantify the welfare effects of the progra. Finally, Section V analyzes how the passage of Part D affects optial innovation and procureent policy. II The Medicare Part D Progra Design and Benefits of Medicare Part D Medicare Part D subsidizes the costs of prescription drugs for Medicare beneficiaries and was introduced by the passage of the Medicare Prescription Drug, Iproveent, and Modernization Act of 23 (MMA. Beneficiaries can obtain the Medicare Drug benefit through two types of 3

6 private plans: beneficiaries can join a Prescription Drug Plan (PDP for drug coverage only, or they can join a Medicare Advantage plan (MA that covers prescription drugs (MA-PD. Beneficiaries are required to ake preiu payents to obtain Part D coverage. However, preius are highly subsidized. Medicare Part D covers roughly 75% of the costs. Medicare Part D establishes a standard drug benefit that Part D plans ay offer. The standard benefit is defined in ters of the benefit structure and not in ters of the drugs that ust be covered. In 27, this standard benefit requires payent of a $265 deductible. The beneficiary then pays 25% of the cost of a covered Part D prescription drug up to an initial coverage liit of $24. Once the initial coverage liit is reached, the beneficiary is subject to another deductible, coonly known as the Donut Hole, in which they ust pay the full cost of drugs. When total out-of-pocket expenses on forulary drugs for the year, including the deductible and initial coinsurance, reach $385, the beneficiary then reaches catastrophic coverage, in which he or she pays a 5% coinsurance. In practice, Part D plans ight deviate fro this standard benefit but they ust offer coverage that is equivalent to or better than the standard benefit in actuarial ters. The law also stipulates that eployers sponsoring prescription drug coverage for retirees can receive a federal subsidy if the coverage is at least actuarially equivalent to the standard Medicare drug benefit. Eployers would receive a 28% subsidy to their portion of the individual retiree s drug costs between $25 and $5,. Finally, Medicare Part D also provides ore generous insurance and additional subsidies to low-incoe beneficiaries. Currently, dualeligible (eligible for both Medicare and Medicaid beneficiaries constitute the ajority of the beneficiaries receiving low incoe subsidies as they are autoatically enrolled in Part D plans. 4

7 Role of Price Negotiations One of the controversial features of the MMA was it did not allow Medicare to negotiate prices directly with pharaceutical copanies. Many critics regard this as poor stewardship of tax dollars, while those in favor argue that price-negotiation could dapen innovation incentives by lowering pharaceutical profits. However, this original legislation ight be overturned by the Medicare Prescription Drug Price Negotiation Act of 27 (H.R. 4. This bill, passed earlier this year by the House of Representatives, would require the Secretary of Health and Huan Services to negotiate directly with anufacturers to lower covered Part D drug prices, beginning in 28. This would reverse the MMA prohibition against federal negotiation of drug prices. The status of the bill reains uncertain, as of this writing. In April of 27, an attept to force a Senate vote on the bill was blocked (Marre, 27. For his part, President Bush has consistently proised to veto it if passed (Espo, 27. MMA and Patent-Gaing Patent-gaing refers to activities of pharaceutical copanies to extend onopolies and block entry of cheaper generics in arkets for blockbuster drugs near patent expiration. The ost coon tactic is to file a lawsuit against a generic copetitor for infringeent of a patent on the original product. 2 Under the Drug Price Copetition and Patent Ter Restoration Act of Another tactic is to file ultiple patents with the FDA after a potential generic entrant applies for FDA approval, and then to sue the new entrant for the infringeent of these later patents. Such lawsuits have triggered ultiple but staggered 3-onth delays. For exaple, the anufacturers of Paxil, a blockbuster antidepressant, received 5 staggered 3-onth extensions delaying the approval of generics by 65 onths. Still other tactics include the conduct of clinical trials in children, for which the Hatch-Waxan act guarantees six onths of arket exclusivity, or introducing new versions of a patented product that differ only in dosage, appearance or indications for use. For exaple, Eli Lilly (the anufacturer of the antidepressant Prozac, introduced Sarafe, a new drug cheically identical to Prozac but colored pink and lavender instead of green. Sarafe got ore than 2 years of arket 5

8 (coonly known as the Hatch-Waxan Act the filing of any lawsuit no atter how frivolous triggers an autoatic 3-onth delay in the introduction of generics. 3 An FTC study found that brand nae copanies sued the first generic entrant and triggered the autoatic 3- onth extension for 72% of the drugs analyzed in the study. The study also found that, for cases resolved in the courts, generic entrants prevailed 73% of the tie (FTC, 22. The MMA iposed liits on patent-gaing by stipulating that brand nae copanies could not get ore than one 3-onth extension for lawsuits filed against generic entrants. The law also allows the generic applicant to assert a counterclai to de-list a patent related to the brand nae drug. This further discourages branded copanies fro such patent-extension strategies. III The Welfare Effects of Stand-Alone Public Drug Insurance We first analyze the welfare effects of extending insurance, without any additional provisions. This analysis distills the key welfare effects of drug insurance alone. It deonstrates one of the unique features of such insurance: its potential to lower deadweight loss and raise onopoly profits, siultaneously. It can achieve this outcoe by partially decoupling the consuer s price fro the revenue earned by the onopolist. exclusivity, because it was approved for treating preenstrual depression, anxiety and irritability (whereas Prozac was approved for ajor depression, and an additional 6 onths extension because it was tested in clinical trials for children (Angell, Drug Price Copetition and Patent Ter Restoration Act of 1984, Pub. L. No , 98 Stat (codified at 21 U.S.C. 355 (1994 and 35 U.S.C. 271(d-(h (1994 ("Hatch-Waxan Act". 6

9 Static Iplications The provision of drug insurance can reduce deadweight loss, because co-payents below the onopoly price increase utilization by consuers. Define D ( p as the deand function, P (Q as inverse deand, MC as the constant arginal cost of production, and p as the equilibriu onopoly price. The social surplus generated by copetitive provision of the good is given by: D( MC SS c P( Q dq MC * D( MC (1 In the absence of insurance, deadweight loss in the branded pharaceutical arket is social surplus under copetition, inus social surplus under onopoly: D( p DWL SS c P( Q dq MC * D( p (2 Now suppose that the governent offers prescription drug insurance. Specifically, suppose the governent covers the share ( 1 σ of the arket price, and leaves the consuer with the coinsurance rate σ. If the governent continues to pay the onopoly price for pharaceuticals, the actuarial cost of the insurance is p 1 σ D( σp. Fro a purely static point of view, this ( cost is siply a transfer fro the governent to the pharaceutical industry. 4 The welfare effects eerge fro the change in quantity induced by this policy. Of course, Part D could ipact the onopoly price charged for pharaceuticals. The governent ay use its newfound onopsony power to negotiate prices downward, or 4 Later, we include the deadweight cost of public funds, which is so far absent. 7

10 onopolists ay exploit the subsidy to consuers and choose to raise prices. We discuss these possibilities later. For now, we analyze the welfare effects of the progra, given soe arbitrary post-part D onopoly price, ' p. Deadweight loss is a function of what the governent pays onopolists, and offers consuers in ters of co-insurance. This relationship can be expressed as: DWL( p ', σ D( MC P( Q dq MC * D( MC D( σp ' P( Q dq MC * D( σp ' (3 Ultiately, what atters for deadweight loss is siply σ ' p, the price faced by consuers. The welfare effect of lowering the consuer price either by lowering the co-insurance or by lowering the price paid to onopolists is given by: ' ( p MC ddwl ' = D p ' p ' '( σ σ (4, σ dσp Fro a static point of view, lowering the price paid by consuers always lowers deadweight ' loss, as long as σ p > MC, or that consuers continue to face a price that is at least as large as arginal cost. Epirically, this assuption sees to hold for Medicare Part D. Marginal cost is typically estiated to be 2% of the branded drug price (Caves, Whinston, and Hurwitz, 1991; Grabowski and Vernon, 1992; Berndt, Cockburn, and Griliches, In contrast, the average coinsurance rate under Medicare Part D is currently about 62%, well above arginal cost. 5 5 This nuber was calculated by the authors using the Medical Expenditure Panel Survey (MEPS data described later. Appendix A presents a foral proof that the anticipated percentage change in co-insurance rate (along with the elasticity of deand is a sufficient predictor of the change in utilization, even with a nonlinear benefit design. 8

11 There is one reaining loose end: to show that Part D will always raise consuption, once we account for the behavioral response of innovators. Part D ay induce innovators to raise their onopoly price, but never by so uch as to reduce quantity below its initial level. For a percentage subsidy s, a fir with constant cost faces the profit function: ( 1 + s p( q q MC * q (5 This has the first-order condition MC p' ( q q + p( q =, (6 1 + s which akes clear that a subsidy is equivalent to a reduction in arginal cost. As such, subsidies will always raise quantity. Dynaic Iplications The original intent of Medicare Part D was to provide drug insurance without affecting prices paid to innovators. Earlier, we showed that drug insurance iproves static welfare by lowering deadweight loss. We now show that this original aspect of Part D induces ore innovation, and increases dynaic social surplus. Let I denote industry investent in research, and let g ( I denote the probability of discovery with g ( I > and g ( I <. In other words, R&D investent raises the probability of new drug discovery, but in a concave fashion. Suppose the innovator enjoys a patent onopoly for T periods after the discovery and will ake zero profits thereafter. If the fir discounts the future at the rate r, it invests in research in order to axiize the present value of expected profits: 9

12 T I g I rt Π( = ( e p dt I π ( σ, (7 By the envelope theore, stand-alone drug insurance raises the expected profits of innovators, dπ because <. It will also induce ore innovation. The privately optial level of innovation dσ is given by: g' ( I = T e rt 1 π ( σ, p dt Π 1 ( σ, p, T (8 The arginal product of research is the reciprocal of onopoly profits, Π and, by extension, of patent length (Nordhaus, Therefore, since reductions in co-insurance raise profits, they ust also stiulate innovation. Define I ( Π as the level of investent induced by onopoly profits pat surplus can be written as: Π. Expected social S( T, p T, σ rt rt g( I pat ( Π e SScdt e DWL( p dt I pat ( Π σ (9 The arginal value of introducing stand-alone drug insurance is given by: S σ σ = = T Π '( ( 1 + '( < 1 T rt rt rt Iπ σ g I e SScdt e DWL p dt g e DWL p pdt (1 1

13 The ter in curly braces is strictly greater than unity, because total social surplus fro the innovation ust be strictly larger than the innovator s profits. 6 Therefore, the first ter is negative. This easures the value of drug insurance as a stiulant to innovation. The second ter is negative, because p > MC, iplying that deadweight loss will rise with a higher price. This is the value of insurance in itigating deadweight loss. Notice the iportant presuption that the profits of innovators do not exceed social surplus. Clearly, this condition always holds in a copletely private arket, even one afflicted by oral hazard in insurance provision (Lakdawalla and Sood, 26. Intuitively, consuers would never voluntarily pay ore than their consuer surplus for a drug in a spot arket, and they would never pay ore for an insurance policy than the expected value of its covered treatents. Public subsidies for eployer-provided health insurance ake it theoretically possible that profits could exceed social surplus. However, given the estiated rate of surplus-appropriation by innovators (see the discussion beginning on page 25, this would require extreely large transfers. It is even less likely aong the elderly population, where prescription drug insurance was relatively uncoon. 7 However, another ore controversial question concerns whether there is currently too uch or too little innovation, or equivalently, how uch innovators ought to be able to appropriate. In the standard odel, innovators ought to appropriate the full value of social surplus, which is 6 In theory, distortions like subsidies for health insurance could result in profits being higher than social surplus. However, we later docuent epirical evidence confiring social surplus is larger than profits. 7 As of 23, 6% of the aged (65+ population had no drug insurance, or insurance that was less generous on average than the standard Part D benefit. 11

14 ipossible in the absence of price-discriination. Soe econoists have pointed out that patent races, public subsidies, and other iperfections can alter this result, soeties substantially. Others have ephasized the extreely low rates of social-surplus appropriation by innovators. 8 Resolving this controversial question lies beyond the scope of this paper, but we can interpret our analytical results, regardless of whether innovation is too high or too low. If, as in the standard case, innovation is too low, Part D has a direct welfare benefit, without any auxiliary provisions. This is the analysis presented above. If in fact there is too uch innovation, the direct effects of Part D on innovation reduce social welfare. However, this adds additional value to pricenegotiation or siilar easures to liit, or even reduce, the profits of innovators. In this case, a Part D progra coupled with price-reductions that hold innovator profits constant would be strictly welfare-iproving. 9 The rest of the analysis is presented fro the point of view of the standard odel, that there is too little innovation; the possibility of over-innovation is discussed further in Section V, when we consider the benefits of price-negotiation. Public Financing and Deadweight Cost In the analysis above, we abstracted fro the costs of public-financing. When the governent has access to an efficient lup-su tax echanis, it does not atter whether insurance is publicly or privately financed. Clearly, the deadweight costs of public financing play a substantial role in the optial policy configuration. However, in this section, we show that they 8 For contrasting views in the context of pharaceuticals, see Garber, Jones, and Roer (26, copared with Philipson and Jena (26. In a broader context, see Shapiro (27, copared with Nordhaus (24. 9 Such a schee would also be feasible, since innovators have already revealed their willingness to operate at today s profit levels. 12

15 do not change the basic conclusion that soe public financing of drug insurance is welfareiproving deadweight costs erely change the optial level of financing. When deadweight costs of taxation rise, the optial degree of public financing falls, but soe public subsidy for drug insurance is always optial, because of its value as a deadweight-loss reduction device. Paying ore oney out of the public treasury, as opposed to private pockets, incurs social cost. This has two effects. First, providing insurance to the uninsured becoes less beneficial, on balance. Second, it now becoes strictly costly to attract the currently insured to a public plan with the sae co-insurance rate. In the absence of deadweight costs, public subsidies for insurance preia represent a pure transfer to such people. With deadweight costs, this transfer iposes net costs on society. In spite of these costs, soe degree of subsidization reains optial. Intuitively, providing a sall subsidy to the uninsured always provides soe positive benefit, because of the reduction in deadweight loss. Moreover, the cost of crowd-out (i.e., attracting currently insured consuers to a less generous public plan is initially zero, because in an unsubsidized equilibriu, they strictly prefer their ore generous private plans. This intuition can be seen ost easily by calculating the optial degree of subsidy, given a public co-insurance rate σ and a private co-insurance rate σ p. Specifically, suppose that there are N initially uninsured consuers, and I privately insured consuers. Without loss of generality, we assue that the insured consuers are all identical, and all have policies with coinsurance rate σ p. 13

16 Define s as the share of the consuer s preiu that is publicly financed. Define N (s as the nuber of uninsured consuers choosing the public drug plan at the preiu subsidy s, and define I (s as the nuber of insured consuers doing the sae. 1 The social arginal cost of public funds is μ : for exaple, if raising $1 of revenue introduces $.5 of deadweight loss, we say that μ =. 5. Finally, D ( p is per person deand at the price p. We assue this is unifor across people, but relaxing this assuption leaves the analysis unchanged. The social cost of a publicly financed drug benefit offering co-insurance rate σ is thus: ( 1 p [ N( s I( ] DWC = sμ D( σp σ s (11 + The total reduction in deadweight loss for all consuers is given by: D( σp N ( s P( q dq + I( s D( p D( σp D( σ p p P( q dq (12 The optial degree of public financing axiizes deadweight loss reduction, net of social cost, according to: ax s N( s D( σp D( p P( q dq + I( s D( σp D( σ p p P( q dq sμd( σp ( 1 σ p [ N ( s + I( s ] (13 The first-order condition for the optial subsidy can be written as: 1 Both I and N also depend on the co-insurance rates σ and σ p, but since we regard these as fixed, we do not explicitly consider the. 14

17 N D( σp s D( p P( q dq + I s D( σp D( σ p p P( q dq sμd( σp (1 σ p [ N + I ] μd( σp (1 σ p [ N + I ] s s (14 To show that the optial value of s exceeds zero, it suffices to show that the arginal return to s is strictly positive when s =. Without subsidies, neither insured nor uninsured consuers will choose the public drug benefit; otherwise, they would have chosen such an insurance policy in the private arket. Therefore, evaluated at zero, the arginal return to public financing is given by: N D( σp s D( p P( q dq + I s D( σp D( σ p p P( q dq > (15 If MC σ σ < p, both ters are strictly positive. If p σ > σ, the first ter is strictly positive, and the second ter is zero. To see this, observe that I ( =. Intuitively, insured consuers will not switch to a less generous plan, absent a strictly positive subsidy. Therefore, since I (s will be uniforly zero for all s below soe strictly positive s * >, its derivative ust be zero at s =. 11 s Two other points follow fro the first-order condition for public financing. Not surprisingly, the optial degree of public financing is lower when the arginal cost of public funds μ is higher. In addition, the degree of public financing is higher when the deadweight loss fro onopoly is higher. Indeed, for copetitive arkets with p = MC, there are no static grounds for the 11 More forally, there exists soe s * > such that insured consuers are exactly indifferent between the public and private plans: a strictly positive subsidy is required to copensate the insured consuers for their partial loss of * coverage. Therefore, I ( s =, for all s s ; this iplies that I ( =. 15 s

18 public financing of health insurance. Arguents would need to be ade on the conventional bases of insurance value, altruis, or erit goods. IV Calibrating the Welfare Effects In this section, we calculate in a back-of-the-envelope fashion the welfare ipacts of Medicare Part D. We consider both the static and dynaic benefits of increased drug consuption and the associated increase in pharaceutical innovation induced by Medicare Part D. We also estiate the social costs of financing this benefit due to deadweight loss fro increased taxation. We exclude dual eligibles fro the analysis as they already receive generous public prescription drug insurance fro Medicaid. The introduction of Medicare Part D does not substantially change the generosity of insurance for dual eligibles; all it does is transfer insurance fro Medicaid to Medicare Part D. The net benefits of Part D can be calculated according to the following equation: NB = Enrollees * ( PC Static Benefit + ( Dynaic Benefit ( Deadweight Cost (16 In words, the net benefit of the progra is equal to: the nuber of enrollees ultiplied by the per capita static benefits to those enrollees, plus the dynaic benefit of the progra, inus the deadweight costs of financing. The first ter ebodies the utilization effects on enrollees, the second the innovation effects, and the third the social cost of funding preius and eployer subsidies. Appendix B provides a suary of the calculations involved in calibrating the static benefits, dynaic benefits and dead weight costs of Part D. A ore detailed exposition of this calibration is provided below. 16

19 Part D Enrollees and Non-Enrollees We need to estiate both the nuber of enrollees, and the nuber of non-enrollees eligible for the eployer subsidy payent. The latter group affects the financial cost of the progra, if not the static welfare benefit. We used data fro the 23 Medical Expenditure Panel Survey (MEPS to estiate these quantities. Methods The Medicare Modernization Act classifies an individual s private prescription drug coverage as creditable or non-creditable. Broadly speaking, creditable coverage is private insurance that is actuarially at least as generous as Medicare Part D. Individuals without creditable coverage are required to enroll in Part D within 63 days of being eligible, or face a late enrollent penalty if and when they enroll. Those with creditable coverage are not subject to this penalty provision. Moreover, if a fir provides a Part D-eligible eployee with creditable coverage, that fir is entitled to a 28% subsidy on their portion of the individual retiree s drug costs between $25 and $5,. In our calculations, we assue that there are no costs of switching to Part D. Therefore, everyone without creditable coverage (i.e., less generous than Part D switches to the progra. We also assue that people with creditable coverage refuse to switch. 12 Evidently, therefore, we need to identify individuals with and without creditable coverage. To do so, we use actual 12 This assuption has less transparent foundations. Clearly, people with very generous coverage fail to switch, but the effects for people with arginally ore generous coverage are theoretically unclear. On the one hand, the subsidy of the Part D preiu suggests that soe ay switch to Part D. On the other hand, eployers could pass along the subsidy they receive for privately insuring a worker who is still eployed, which ay fully counteract the value of the preiu subsidy. We adopt the siple assuption of no-switching, because it best atches actual enrollent data, as shown in Table 1. 17

20 prescription drug expenditures observed in the MEPS, 13 and calculate the average co-insurance rate generated by the individual s current plan (or lack thereof, and by the standard Medicare Part D benefit. Those whose average co-insurance rate is lower under Part D are classified as having no creditable coverage, and vice-versa. Appendix A presents a foral proof that the average co-insurance rate is a sufficient predictor of value for an insurance plan, even one with a nonlinear benefit design. Estiates Using the MEPS sapling weights we estiate that, excluding dual eligibles, 36 illion beneficiaries would be eligible for Medicare Part D. Next, for each MEPS respondent we then calculate the average coinsurance for prescription drugs under 2 scenarios: (1 their current coverage; or (2 enrollent in the standard Medicare Part D benefit, described earlier. Based on this analysis, we identify individuals with and without creditable coverage. We estiate that 59% of eligible respondents, or roughly 21 illion beneficiaries, have no creditable coverage. We assue all these individuals enroll in Part D. The reaining 15 illion already have ore generous insurance copared to the standard Part D benefit and are assued not to enroll in Part D. Based on inforation about the source of coverage in MEPS we estiate that roughly 26% of beneficiaries have creditable insurance fro an eployer or union, which would then receive the eployer subsidy instituted by Medicare Part D. Finally, 14% of beneficiaries have creditable insurance fro other sources such as 13 Notice that we calculate the effective co-insurance rate fro current spending. This is equivalent to calculating the first-order welfare effect of switching to Part D. Individuals with a positive first-order welfare effect will benefit, and vice-versa. 18

21 Veterans Adinistration, Indian Health Service and state pharaceutical assistance progras. The eployer subsidy is not paid in these cases. As a validity check, we copared our enrollent estiates to the actual enrollent rates reported by the Departent of Health and Huan Services (HHS, both of which appear in Table 1. The HHS estiates show that, as of January 27, and excluding dual eligibles, 36 illion beneficiaries were eligible for Medicare Part D. Of these, 21 illion were estiated to have no creditable coverage prior to Part D. 14 The reaining 15 illion had creditable coverage fro eployer/union or fro other sources such as Veterans Adinistration, Indian Health Service and state pharaceutical assistance progras (Kaiser Faily Foundation, 27. These nubers are quite siilar to the estiates we derived fro analysis of the MEPS. Table 1: Prescription Drug Coverage Sources Aong Medicare Beneficiaries. HHS Estiates MEPS Estiates Population (il Percent Population (il Percent Coverage Type Enrolled/No creditable Coverage 21 59% 21 59% Creditable Eployer Coverage 1 28% 9 26% Others with Creditable Coverage 5 13% 5 14% Total 36 1% 36 1% 14 Of the 21 illion beneficiaries, 17 illion enrolled in Part D and the reaining 4 illion continued to have no creditable coverage. We assue that, over the long-run, these reaining 4 illion respondents will switch into the ore generous coverage afforded by the Part D progra. If not, this incoplete take-up rate would lower the welfare benefits of the progra. 19

22 Static Benefits The next step is coputing the static benefit enjoyed by enrollees in Part D. Using a linear approxiation to deand, the benefit associated with a particular change in price and quantity is 1 siply the size of deadweight loss reduction triangle, or ( Δp( Δq one half ties the 2 reduction in price, ties the increase in quantity. This benefit can be equivalently written using the anticipated percentage changes in price and quantity, along with the elasticity of deand. Appendix A presents a foral proof that the anticipated percentage change in co-insurance rate and elasticity of deand are sufficient predictors of change in utilization, even with a nonlinear benefit design. Assuing that the price paid to the anufacturer does not change, 15 soe siple algebra yields the equivalent forulation of the static benefit for an enrolling consuer: 1 ( σ ND σ D ( σ ND σ D SB = e OOP 2 σ ND σ ND ND (17 The ters σ D and σ ND are the average share of price paid by the consuer with and without Part D, respectively. OOP ND is the out of pocket prescription drug expenditure of the consuer under the status quo, and e is the elasticity of deand. To calculate the static welfare ipact of Medicare Part D, we need epirical estiates of: (1 The percentage change in price to the 15 This assuption atches the available epirical evidence. For exaple, Lichtenberg and Sun (27 use data fro a large retail pharacy before and after introduction of Medicare Part D and find that Medicare Part D had a negligible ipact on overall prices paid to anufacturers but a significant ipact on prices faced by consuers. 2

23 consuer induced by Medicare Part D, (2 the elasticity of deand for prescription drugs, and (3 the out of pocket costs of purchasing prescription drugs. Percentage Change in Price We calculate the percentage price change that enrolling consuers would enjoy if they took up the progra by coputing for each elderly consuer in MEPS the difference in average coinsurance between: (1 status quo insurance, and (2 an insurance plan with the features of the standard Medicare Part D benefit described earlier. We calculate this percentage change in price for each MEPS respondent estiated not to have creditable coverage. Respondents with creditable coverage are assued not to enroll in Part D and thus experience no change in price. Based on these calculations the average percentage change in coinsurance for those without creditable coverage due to the standard Medicare Part D benefit was estiated to be 3.1%. Price Elasticity of Deand Long-run generic prices (assued to be equal to arginal cost are approxiately 2% of the prices charged for the corresponding on-patent drug (Caves, Whinston, and Hurwitz, 1991; Grabowski and Vernon, 1992; Berndt, Cockburn, and Griliches, Thus we assue that the ark-up on pharaceutical prices is roughly 8%. The standard theory of onopoly would then iply, based on a 8% ark-up by onopolists, a price elasticity of uninsured deand around 1.25, or the inverse of the arkup. However, the above elasticity is not the relevant one for insured patients who face copayents, rather than anufacturer prices (Chandra, Gruber, and McKnight, 27. Thus, for the insured consuers we use elasticity estiates that rely on changes in patient cost sharing aong the insured elderly population. For exaple, Chandra, Gruber and McKnight (27 estiate the 21

24 price elasticity of prescription drugs aong the elderly by studying a policy change that raised patient cost-sharing for retired public eployees in California. Their estiates of price elasticity range fro.5 to We take the idpoint of their range, and assue that the price elasticity is 1. aong the insured elderly population. Based on these estiates of the elasticity of deand and percentage change in price the average percentage change in nuber of prescriptions for those without creditable coverage was estiated to be 34.5%. Out-of-Pocket Costs The out of pockets costs of purchasing drugs are available directly fro MEPS. For those without creditable coverage, out-of-pocket costs are estiated to be $1,32. Results Based on these estiates, we estiate the aggregate static benefit of Medicare Part D to be $3.5 billion or $99 per eligible beneficiary. There is wide variation in the per capita benefit enjoyed by beneficiaries depending on insurance coverage, or lack thereof, prior to the introduction of Medicare Part D. Since insurance coverage is highly correlated with incoe, the poor enjoy greater benefits than the rich. For exaple, we estiate the per capita benefits to be $116 for those with incoes less than $15, per year, $94 for those with incoe between 15, and 5,, and $46 for those with incoes greater than $5,. Siilarly, beneficiaries in poor self-reported health and those with higher prescription drug costs also enjoy greater benefits. Those reporting their general health to be poor enjoy per capita benefits of $19, those 16 They also do not find significant variation in price elasticity of prescription drugs by age, incoe, and health status. 22

25 reporting good health enjoy per capita benefits of $19 and those reporting excellent health only receive $51 in per capita benefits. Dynaic Benefits Since Medicare Part D likely increases pharaceutical copany profits, it has the dynaic benefit of inducing additional innovation. We can estiate the value of this induced innovation just as we estiated the static value of the progra. 17 First, we aintain the assuption (inherent in the original MMA legislation that Part D continues to forbid price-negotiation, and that pharaceutical firs will continue to receive the onopoly prices set before Medicare Part D ( p. However, firs do experience an increase in deand for their products due to the reduction in price for consuers after the introduction of Medicare Part D. Step 1: Change in Pharaceutical Revenues For a given consuer, the percentage change in total drug expenditures is equal to the percentage increase in the quantity of drugs consued, ( σ ND σ D e, which is calculated as above. σ ND Assuing anufacturer prices have been so far unaffected by Part D, the percentage change in quantity is equal to the percentage change in revenues for innovators. The average percentage change in price for all eligible beneficiaries was estiated to be 17.9%. This estiate cobined with the elasticity estiates iplies an average percentage increase in drug expenditures of 2.5%. 17 We are proceeding under the standard assuption that there is too little innovation, because innovators cannot capture full social surplus. In Section V, we discuss how Part D could be configured for axial welfare benefit, if in fact there is over-innovation in the status quo. 23

26 Step 2: Creation of New Cheical Entities The increase in pharaceutical revenue will induce ore R&D and innovation. The nuber of new drug introductions induced by Part D will depend on the elasticity of new drug introductions with respect to pharaceutical revenues. Aceoglu and Linn (24 estiate that the elasticity of non-generic drug approvals with respect to revenues is roughly 3.5. We use this elasticity and the estiate of change in pharaceutical revenues to calculate the percentage change in the nuber of new cheical entities (NCEs. The baseline rate of NCE introduction is assued to be 32 NCEs per year. This is the average nuber of NCEs introduced per year during the period 1995 to 24 as reported in the FDA Orange Book. Applying the estiated percentage change to this baseline level yields the absolute nuber of new drugs projected fro the passage of Part D. Step 3: Innovator s Private Value of New Cheical Entities The next step is to copute the annual private value of these additional drugs to their innovators. In general, it is quite difficult to copute the expected value of the arginal drug directly, because it is hard to identify the arginal drug and just as hard to identify expected value. However, it is easier to calculate the actual arginal cost of bringing an additional drug to arket. Theory suggests that this arginal cost ought to be equal to the expected arginal private value of an additional drug. 18 Di Masi et al. (23 estiate that the arginal research and developent cost of bringing an NCE to arket is $939 illion in year 26 dollars. To annualize this cost, we use a standard epirical estiate of the annual cost of capital in the 18 Grabowski et al (22 provide epirical evidence that the theory is consistent with the data in the pharaceutical industry. 24

27 pharaceutical industry, of 12% per year. 19 Therefore, the annualized arginal cost of bringing an NCE to arket is expected to be (.12*($939 = $113 illion. This then yields our estiate of the annualized arginal expected private benefit. Step 4: Marginal Social Value of New Cheical Entities The last step is to infer the arginal social value, which theory predicts will exceed the private value to the innovator (although see the caveat in footnote 17. To estiate the social return, we ust estiate the fraction of social surplus captured by the innovator. Several estiates are available fro the literature. Based on data fro 1948 to 21, for exaple, Nordhaus (24 estiates that innovators capture just 2.2% of the total present value of social returns to innovation. In a pharaceutical context, Philipson and Jena (26 use data fro over 2 published studies of healthcare innovations to estiate the distribution of surplus-appropriation by pharaceutical innovators. They find that the edian producer share of social surplus is 17%, the first quartile is roughly 1% and the third quartile is roughly 25%. To be conservative, we assue that innovators are able to capture as uch as a quarter of the social surplus fro pharaceutical innovation. This paraeter yields an estiated social rate of return on pharaceutical R&D investents of 48% per year, four ties the estiated private return. This suggests that the annual social value of the arginal drug is equal to ($113 illion/(.25 = $452 illion. 19 The latest estiates of the cost of capital by industry are available online at Based on these data, we estiate the cost of capital for the pharaceutical industry to be 12% per year. This estiate of cost of capital is siilar to estiates of private rate of return on R&D investents in the pharaceutical industry (Grabowski, Vernon, and DiMasi,

28 Gross Static and Dynaic Benefit of Part D Using the ethods described in Step 1 we estiate that Medicare Part D would increase pharaceutical sales by $15 billion per year. Given baseline pharaceutical sales of $275 billion in the US in 26, this corresponds to a 5.4% increase. The 3.5 innovation elasticity fro Aceoglu and Linn (24 then iplies that the nuber of new drugs per year would increase by roughly 19%, or 6.1 NCEs. Earlier, we calculated an annual social value of $452 illion for the arginal additional drug, which yields a gross dynaic benefit of $2.8 billion annually. Cobining the dynaic with the static benefit yields a gross risk-neutral welfare benefit for Medicare Part D of $6.3 billion annually. Deadweight Costs of Financing Medicare Part D It reains to copare the aggregate benefits of Part D with its social cost. The progra itself is just a costless transfer. However, since it is publicly financed, there are deadweight costs associated with its financing. The actuarial cost of the Medicare Part D insurance for a beneficiary who enrolls in the standard Part D plan is siply: ( σ ND σ D s Total _ Exp(1 σ D + (1 σ D Total _ Exp e (18 σ ND s is the subsidy provided by Medicare Part D and is estiated to be 75%. The first ter in the square brackets is the actuarial cost of the benefit under the initial deand for pharaceuticals and is siply the total cost of prescription drugs ties the plan share of costs. The second ter is the actuarial cost of the additional deand induced by Medicare Part D and is equal to the 26

29 change in total drug costs ties the plan share of costs. As discussed earlier, the percentage change in total drug costs, ( σ ND σ D e, is siply the percentage change in price to consuers σ ND induced by Part D ties the elasticity of deand. All the above quantities can be easily estiated fro available data. Using data fro the MEPS, we estiate average total drug costs for MEPS respondents without creditable coverage who are likely to enroll in Medicare Part D to be $1,537. As discussed earlier, we can also calculate the price change consuers would enjoy if they took up the progra; price-elasticities of deand for pharaceuticals (taken fro the literature then iply the associated increase in drug costs. Siilarly, the costs of providing the eployer subsidy (28% of costs between $25 and $5 can be estiated easily using data on prescription drug expenditures for those with creditable eployer/union provided insurance. The results show that for those receiving the preiu subsidy, Medicare costs are $893 per enrollee and for those receiving the eployer subsidy, Medicare costs are $584 per enrollee. These estiates are siilar to those obtained by HHS. For exaple, the average preiu for a Part D plan was roughly $27 per onth or $327 per year. Since Medicare subsidizes preius by 75%, Medicare costs of providing the preiu subsidy equal $985 ($327*3 per year. Siilarly, HHS estiates indicate that the cost of eployer subsidy was $549 per beneficiary receiving the subsidy (Kaiser Faily Foundation, 27. The true social cost of the progra is the deadweight cost associated with paying the actuarial cost out of public funds. While there is debate in the public finance literature on the agnitude of deadweight loss, we use a conventional estiate that each additional dollar spent on 27

30 Medicare Part D generates 3 cents of deadweight costs due to increased taxation (cf, Jorgenson and Yun, Based on these estiates, we estiate the deadweight costs of financing Medicare Part D to be $7.2 billion per year, 87% of which is covered by the risk-neutral benefits of the progra. This analysis reveals that conventional estiates of deand, dynaic benefit, and deadweight loss yield the surprising result that Part D insurance is nearly break-even for a society of riskneutral and self-interested consuers. V Drug Insurance and Innovation Policy The previous section characterized the direct welfare benefit of Medicare Part D. The passage of a public prescription drug insurance progra welfare-iproving or not also has iportant indirect effects, by changing the socially optial rewards for innovation, and thus the optial public policy toward innovation. We consider the optial configuration of innovation policy, with a focus on two policy instruents in particular. First, the governent could potentially use its bargaining power to influence prices paid to innovators. Second, it can increase or decrease de facto patent length by regulating patent-gaing. We show that using both these instruents in conjunction with public drug insurance can allow the governent to achieve first-best utilization and innovation. Of course, there ay be significant inforational or political constraints to pursuing the first-best policy. Therefore, we go on to show that it is always strictly welfare-iproving to negotiate prices down to soe degree, and we derive the conditions under 2 Different authors have suggested that the nuber could be as high as $1 of deadweight loss for each $1 of public spending (cf, Feldstein,

31 which the governent should tighten or loosen patent enforceent. As is often the case, the exact degree of price negotiation or patent enforceent requires the estiation of the key tradeoffs outlined below. The First-Best Benchark In addition to the co-insurance rate, suppose the governent can also set a price paid to innovators, and a de facto patent length. Since the governent is not perfectly free to dictate ters to innovators, define now solves: R Π as the iniu profits innovators will accept. The governent S R ( T, p s. t. Π g, σ = ax Π R T rt rt T p g I pat Π e, SScdt e DWL p g, σ ( ( ( σ g dt I pat ( Π (19 This has the first-order conditions: I I I π π π Π Π Π T p σ g' ( I g' ( I g' ( I e e e rt rt rt SScdt SScdt SS dt c T T T e e e rt rt rt DWL( σp DWL( σp DWL( σp g g g dt 1 g( I e dt 1 g( I dt 1 g( I rt T T e e DWL( σp rt rt g + λπ dtdwl' ( σp dtdwl' ( σp g g T σ + λπ p g = p + λπ σ = = If the governent knows all the paraeters of the proble, and can freely adjust its policy instruents, it can always achieve the first-best outcoe. Given the conventional result that first-best innovation requires profits equal to total social surplus, the faily of first-best policy solutions satisfies: 29