Search the archive.

Historical evidence presents a valuable opportunity for exploring key features of this debate. The framers of U.S. policies were aware of the options that had prevailed in the colonial period and in Europe, but explicitly rejected the use of “premiums” or prizes in favor of property rights in patents. The patent system was market-oriented, offered open access to creative individuals regardless of their social status and background, enabled strong enforcement of such rights, ensured useful disclosure, and promoted extensive markets for technology. The empirical evidence on the early patent system in the United States suggests that patents and their effective legal enforcement played a substantial role in influencing the rate and direction of inventive activity in a country that would become the world’s leading industrial nation. Patent institutions were not perfect, but as Demsetz points out, their imperfections did not necessarily imply the superiority of any other system.86 Perhaps the most telling evidence comprises the endogenous diffusion and adoption of the distinctive U.S. rules and standards towards property rights in patents by other countries who wished to emulate its industrial achievements. In contrast, the majority of organizations that had specialized in granting prizes for industrial innovations ultimately became disillusioned with this policy, and the practice of bestowing technology awards declined among both private and public institutions. As observers noted in the nineteenth century, industrial prizes faltered in part because of their lack of market orientation, and even the democratic nature of economic institutions in the United States could not overcome such drawbacks in administered prize systems.87 Judges had to combine technical and industry-specific knowledge with impartiality, but even the most competent personnel could not ensure consistency; decision making among panels was complicated by differences in standards, interpretation, capture, and risk aversion. Such difficulties tended to lead to haphazard decisions or were often overcome by simply giving the award to the person or firm with the most established reputation. Juries were not immune to the effects of outright bias, capture, cognitive dissonance, lobbying, and “marketing.” Prizes tended to offer private benefits to both the proposer and the winner, largely because they served as valuable advertisements, with few geographical spillovers. Winners of such awards were generally unrepresentative of the most significant innovations, in part because the market value of useful inventions would typically be far greater than any prize that could be offered by private or state initiative. Even prestigious, well-meaning, and amply funded institutions such as the RSA failed to develop truly valuable inventions. A systematic assessment of the role of incentives for innovation in the nineteenth century therefore highlights the advantages of market-oriented policies that economize on information, especially in the decentralized determination of price, value, and “winners.” Market mechanisms also bypassed many of the high transactions costs attendant on negotiating, monitoring, and contracting with applicants and winners. This is not to say that administered inducements are never effective, especially in the context of such market failure as occurs in the provision of tropical medicines or vaccines, where significant gaps might exist between private and social returns. However, in distinguishing between the numerous ingenious theoretical prize mechanisms that have been proposed, such transactions costs need to be recognized and incorporated. In particular, governance issues and the potential for rent seeking and corruption should be explicitly addressed, especially in countries where complementary institutions and political control mechanisms are weak or nonexistent. For, the historical record indicates that the evolution of the institution of innovation prizes over the past three centuries serves as a cautionary tale rather than as a success story.

Notwithstanding their virtues, sale-triggered patent buyouts for federal funding recipients are a form of limited government intervention in the pharmaceutical industry. Their impact is accordingly limited. Using patent buyouts, the government could only set prices for its approved therapeutics, which would take years to acquire or develop. In most instances, acquired investigational candidates would fail in clinical trials, dashing the hopes of patients eager for affordable treatments.253 Moreover, a promising government-backed investigational candidate could chill early-stage R&D within a therapeutic class because successful approval of the candidate would constrain class pricing. For example, if the government successfully marketed a diabetes drug and charged a competitive, rather than supracompetitive, price, researchers might abandon alternative diabetes drug projects; the existence of a lower-cost product in the diabetes market might constrain pricing, with consumers choosing the lower-cost drug unless the alternative drug were markedly superior. Whether to continue R&D would be a strategic decision for potential competitors because many promising candidates fail and some competitors may have already sunk significant resources into similar products. Additionally, patent buyouts require significant upfront expenditures, providing return on investment at best a few years later when early adopters of new R&D-related tax incentives decide to spin off therapeutic businesses. Given campaign realities for legislators elected to two- or six-year terms, delayed gratification may mean no gratification.

Buyouts could adversely impact the market. If the bought-out patent is similar to other products still under patent, then those close substitutes would suddenly be competing with low cost products in a competitive market. Anticipating such a situation ahead of any buyout, investors may reduce their focus on these areas.42

Perhaps the most obvious approach to assessing the social value of an invention is to measure use of the product by consumers. This might be accomplished either by forcing some consumers to pay for the product or to consider purchases by consumers of the product once it is placed in the public domain. Guell and Fischbaum (1995, p. 225), who propose patent buyouts for pharmaceutical products, suggest that the pharmaceutical be sold initially in a test market. Shavell and van Ypersele (2001) suggest that the government wait until the product is placed in the public domain and the government observes the goods sold. This approach runs into at least three significant difficulties. First, there is not an identity between a patented invention and a consumer product. Abramowicz (2003, p. 143). (There often is an identity, however, between a copyrighted work and a consumer product, so this approach may be easier to implement in the copyright context.) Often, many innovations are inputs into a single product, and an innovation may be an input into many different products. Hopenhayn et al. (2006, p. 1047). A further complication, identified by Chari et al. (2012), is that producers may have an incentive to manipulate market signals about the value of their inventions, for example by bribing other producers. Chari et al. conclude that the relative strength of patents and reward system depends on how costly it is for producers to manipulate market signals. Second, measurement can be difficult even when there is an identity between product and contribution. One problem, identified by Coase (1946) and applied to the patent reward context by Duffy (2004a), is that buyouts can cause misallocations of resources even where a good is used. For example, government transportation may lead consumers to live inefficiently far from a city. The fact that consumers take advantage of the transport does not mean that they value it at the cost of the production. The relevant question is how much consumers “would have demanded the good if government policy were different and they had expected to pay the full cost of the product.” (p. 43). Duffy points out that we should like to know the most that consumers would pay for a product, but “the government will have no data to answer that question because consumers are not paying full costs and never expected to do so.” (Duffy, p. 43). Meanwhile, manipulation by producers can cause distortions. Roin (2014, p. 1037 n.174, pp. 1059-60) observes that producers may sell goods below marginal cost to inflate apparent demand, not only thwarting value measurement but also leading to inefficient overuse. Third, even measurement of use may face administrative complications. For a pharmaceutical, whose use is probably considerably easier to measure than many other inventions’, we would need to be able to determine total manufacturing by generic firms, perhaps not an insurmountable problem but one requiring some degree of coordination. Sometimes, once use is known, it may be possible to estimate consumer value, subject to the objections above, but often, the government may take shortcuts, such as assuming benefits based on clinical trials. Hollis and Pogge (2008, pp. 30-31) point out that this approach may be flawed, because clinical trials may feature the consumers most likely to benefit from an invention.

Patent buyouts do presuppose a functioning patent system, so in that sense they are not alternative to it. With this scheme the government offers to buy awarded patents, at their private value, say, and then puts the new technology in the public domain. In principle this model could eliminate some shortcomings of the patent system, including the ex post inefficiency due to the monopolistic pricing of patented innovations. But, clearly, a successful implementation of this system relies on the (possibly unwarranted) presumption that the government has the required kind and level of information.

An alternative approach, or perhaps sometimes a complementary one, is to use a market mechanism to estimate the value of intellectual property forfeited by its owner in exchange for a buyout price. Of course, such a system can function only as an adjunct to the existing intellectual property system. The mechanism in the literature that has received the most attention is offered by Kremer (1998), who suggests allowing an optional system in which an inventor can submit an invention for an auction to the public. With some small probability, the auction is consummated; the rest of the time, the invention is placed in the public domain. (p. 1147) The virtue of this system is that the government can harness private information. The government pays the patent owner an amount based on the bids when the auction is not consummated. The system, however, invites a number of objections. A patentee might make side payments to bidders. Collusion can even include implicitly; a company might bid often on another company’s intellectual property in the hope of reciprocation. Kremer suggests combating this by paying some multiple of the third-highest bid. (p. 1158) The effectiveness of this solution depends on the number of plausible bidders. More generally, an auction might not be competitive, especially for exceptionally valuable intellectual property. It may not be worthwhile to undertake the expense of researching the value of intellectual property if there is only a small probability that the sale will be consummated. Clancy & Moschini (2013, p. 229) notes also that potential bidders must have sufficient information to make meaningful evaluations. In addition, truly random assignment will lead to some patents being kept in private hands. “The government must be willing to allow the cure for AIDS to remain in private hands if random chance would have it.” (Duffy 2004a, p. 48)

The second-price auction, which is of crucial importance to the effective operation of patent buyouts, is potentially vulnerable to collusive behaviour between the patent holder and auction participants (Kremer, ). Patent holders have incentives to pay auction participants to make a bid that is higher than their true valuation of the patent to increase buyout prices (Kremer, ). Most of the purchased patents are placed in the public domain, whereas only a small fraction of them would actually be sold to the highest bidders. Hence, on the one hand, the bribed bidders would face a low probability of having to pay the patent authority. On the other hand, the patent holders would be confident of getting an inflated price. However, Kremer () points out several mechanisms for preventing collusive behaviour, for example, sealed bids, punishing colluding firms, or rewards for whistleblowers, among others.

Because patent sales are voluntary, the private value of the patent acts as a floor on the patent buyout price. If the mechanisms for limiting collusion discussed in the previous subsection were thought inadequate, there are several ways that governments could establish ceiling prices, and thus reduce the risk of paying vastly inflated sums for patents. 1. A waiting period of several years could be required before patents were bought by the government, and ceiling prices could be set as a multiple of annual revenues prior to the patent buyout. A waiting period would also make it easier for bidders to assess the value of patents, and would further guarantee that inventors would have incentives for marketing and development. Of course, setting ceiling prices as a multiple of prebuyout revenue would lead firms to artificially boost sales, for example by offering hospitals discounts on other drugs in exchange for purchasing the patented drug, or even by paying outright kickbacks. Tied sales would have to be prohibited. A second ceiling could be established based on the prebuyout price times postbuyout consumption. 2. The amount paid for patents could be capped by total sales of the drug following the patent buyout, times an administrative estimate of the social value of the drug per dose or per patient. International estimates of the cost in disabilityadjusted life years of various diseases are already available. These could be combined with information on drug effectiveness from FDA trials to estimate a ceiling price per dose. Setting these ceilings requires administrative discretion, but the associated rent-seeking may be limited by the fact that setting a ceiling only requires a fairly transparent decision about the social value of the drug per dose. Public interest groups could monitor attempts to set outrageous prices per dose more easily than they can monitor whether the NIH is subsidizing pure science in the guise of developing an AIDS vaccine or whether national laboratories and breeder reactors are being created for pork barrel or scientific reasons. The historical record suggests that the political economy problem in setting patent buyout prices is more that governments have an incentive to expropriate inventors ex post than that inventors wrangle huge sums for unimportant products. 3. It also might be worth considering a mechanism like that in Figure II, with a ceiling price based on the actual profits obtained from the new drug. Inventors who wished to participate would have their patents randomized. The patent would be randomized to the high bidder with some probability p, and placed in the public domain with probability 1 2 p. Inventors would be paid only if the patent was randomized to the high bidder. In this case, they would receive (M/p) min(bid,p), where M is the markup, bid is the estimated value of the patent based on the bids, and p is the realized revenue (or ideally profits) from the drug. The expected payment received by the inventor is thus M min (bid,p).16 Inventors could try to manipulate the ceiling price by bribing the high bidder to boost sales artifically through tie-ins with other products. The government would have to monitor bidders carefully to minimize such tie-ins. While tie-ins will cost the government money, they will not necessarily reduce efficiency, since by increasing sales they may counteract the static distortions created by monopoly pricing. The majority of patents would be put in the public domain so the issue would not arise, and in any case, it is likely to be difficult for firms to greatly increase demand artificially for drugs covering serious sharply defined diseases, such as cystic fibrosis. It is easy to dream up scenarios under which people could evade rules designed to discourage collusion, but it is also important to remember that many institutions which are theoretically vulnerable to collusion operate relatively well. For example, peer-review is highly vulnerable to collusion, yet the National Science Foundation and the National Institutes of Health seem relatively effective. Moreover, limited collusion is not necessarily that harmful. Collusion itself is not a problem; deadweight losses due to collusion are a problem, and there is little reason to think that these deadweight losses would exceed the deadweight losses due to insufficient original research, monopoly-price distortions, and the diversion of effort to ‘‘me too’’ research in the absence of patent buyouts. Even if collusion raises patent prices above their social value, the social value of inventions may be approximated better by the collusive price than by the existing patent system, under which private incentives for developing new inventions are likely to be less than half the social value of the inventions. Finally, if implicit collusion were expected to significantly raise patent prices, then the markup over the private value could be reduced accordingly. For example, if the optimal markup was three (as seems plausible), and if collusion were thought to raise prices by up to 50 percent, then the government could simply offer a markup of two. The auction mechanism described in this paper may be appropriate for many pharmaceuticals, but it would not be appropriate in industries where markets are too thin for auctions, or patents are not an effective means of protecting inventions. In such industries the government could simply offer to buy out patents for an amount equal to postbuyout sales times an administratively determined estimate of the average consumer surplus per unit of the good consumed.

The auction mechanism is potentially vulnerable to collusion, since patent holders would have an incentive to bribe auction participants to bid high. The bidder would only have to pay the government with probability p. However, with 100 percent probability, the patent holder would receive an inflated payment. It is impossible to eliminate collusion, but, as subsection VII.A explains, a variety of mechanisms could be used to minimize collusion. Subsection VII.B discusses how the prices paid by the government for patents could be limited by ceiling prices based on actual sales of the patented pharmaceutical. VII.A. Mechanisms for Preventing Collusion The government could make collusion more difficult using standard procedures such as requiring bids to be sealed, punishing companies and individuals found guilty of collusion, and rewarding whistle-blowers. This would make collusion more difficult and more dangerous. Several additional methods specific to patent buyouts could also be used. 1. The government would base the price it offers the inventor on the third highest bid. The original patent holder would therefore have to bribe three companies instead of one to ensure a substantial increase in the buyout price. This should significantly increase both the difficulty of collusion and the chance of detection. 2. The agency purchasing patents could have authority to call the bluff of suspected overbidders by reducing the markup and selling to the high bidder without randomization in a set percentage of cases. For example, suppose that based on the other bids, and any knowledge of the industry, the government’s best estimate of the patent’s value was p. If a bidder offered p 1 x and the government suspected collusion,15 the government could offer to buy out the patent at p 1 $1, and then require the suspected colluding bidder to purchase the patent at its bid of p 1 x. The government would make a profit of x 2 1 from the attempted collusion. 3. The government could develop lists of suspect bidders by checking whether winning bidders made money, since systematic overbidders would incur big losses. 4. To prevent inventors from forming front companies and having them submit high bids, bidders would have to provide information on any ties they had with the inventor. Bidders who lied about financial ties with the inventor would be subject to prosecution. 5. Bidders could be required to pay a licensing fee or deposit allowing them to participate in a number of auctions. This would make it unprofitable for patent holders to set up dummy companies simply to bid on their own patents. Excluding bidders will be costly if there are only a few potential bidders, but there will often be many potential bidders, because it is straightforward to manufacture most drugs, as evidenced by the fact that there are often many different producers of generic drugs. (For example, fourteen firms produce oral albuterol sulfate.) Any system of patent buyouts that relies on auctioning patents should focus on drugs that are easy to produce, rather than those which require complicated manufacturing facilities. 6. To prevent inventors from developing a reputation for buying back their patents at inflated prices, inventors would be prohibited from buying back the patent from the winning bidder, or making other payments to bidders. Preventing these side payments might be one of the most difficult aspects of preventing collusion. At least in the early years of any program of patent buyouts, the government agency administering the program would presumably have a budget for patent buyouts, and would not be able to afford all the patents that were available for purchase. One option would be to allow the agency to choose which patents to purchase, so that if the price for a particular patent was too high, it could decline to purchase the patent and cancel the randomization. This would tend to reduce collusion because raising the patent buyout price would reduce the chance that the patent would be bought out.

The patent system as now implemented also causes substantial welfare losses because consumers who would buy the product if it were priced at somewhere nearer production cost do not buy it at the monopoly price.12 The welfare loss caused by this is called by economists the “deadweight loss” (DWL) of monopoly pricing, since there is a pure loss to society when consumers do not obtain a product which they value more than the cost of producing it. Guell and Fischbaum (1995), using highly aggregated data, claim that the scale of deadweight loss in the US drug market is on the order of $3bn- $30bn annually; in a more detailed paper (1997) the same authors estimate deadweight losses of $5bn on $8bn of sales, which indicates very large DWL for the market overall.13 Baker and Chatani (2002) construct a very rough estimate for DWL of $5bn - $20bn annually for the US. Globally, the DWL is certain to be many times this figure, because in many markets, drug insurance is unavailable and so consumers are more price-sensitive

Patent system provides little incentive to develop orphan drugs

It has long been acknowledged that patents, while incentivizing innovation, fail to lead to the first best outcome for society because they rely on the distortion-creating incentives of monopoly (Nordhaus, 1969; Wright, 1983; Shavell and van Ypersele, 2001). The monopoly structure generally results in too little innovation (dynamic loss) and in too-high pricing (static loss) relative to the social optimum. One area in which these issues may be particularly salient is global health. Many medicines, including life-saving drugs, are inexpensive to manufacture once innovated, but patents associated with their innovation can generate high prices which limit access to these technologies (Stiglitz and Jayadev, 2010; Quigley, 2015). In addition, the incentive for patent-driven investment in innovation may be too small relative to what is socially optimal, particularly when the global burden of disease falls heavily on poor populations. For example, the latter has been argued to be a major contributing factor to the low private investment in HIV/AIDS vaccine research relative to the disease’s high global health burden (Kremer and Snyder, 2006).

It is well known that monopoly exploitation of innovations under the patent system can reduce the benefits or “surplus” available to society from an innovation. This inefficiency is tolerated because the monopoly profits create an incentive to innovate, and in the absence of innovation, even less social surplus is created. Underlying this trade-off between high prices and innovation is the understanding that willingness to pay a high price for a good indicates high value. The greater the value to consumers, the higher the price the innovator can charge, and the greater the profits. This means, in turn, that the incentive for innovators is to undertake research that is valuable to society, since such innovations will earn high rewards. If rewards are not proportional to the social value, then the patent system cannot work well: it will lead firms to invest in innovations which have little social value, while ignoring avenues of investigation which could be of immense social value. Unfortunately, pharmaceutical markets are among the least well functioning of all markets, undermining the connection between value and reward. Doctors prescribe medications based on their beliefs as to the best medicine, somewhat influenced, presumably, by the extensive detailing and advertising focused on them. Since doctors do not pay for the medicines they prescribe, price is not an important component of their decision-making process. Consumers are typically ignorant of possible choices, and of the differences between various therapies and medicines and how these would relate to their own physiology, and may be paying only a fraction of the price of any medicine, or may pay nothing at all. The other part will be paid for by an insurer – possibly government or a private company – which has limited influence over the medicine prescribed.6 In these circumstances, price is a relatively unimportant strategic variable for competition between drugs – detailing of doctors may be more important. In addition, prices in many countries are regulated by government. The result is that prices may be either too high or too low compared to the ideal market (one in which consumers are informed about the choices they make and then bear the full cost of those choices). Since prices in pharmaceutical markets do not necessarily reflect value to consumers, profits are not likely to be proportional to the social value of an innovation. There are four types of problems which arise here, which I will discuss in turn. First, the pricing of pioneer drugs may bear no particular relationship to value. Second, “me-too” drugs may be able to generate large profits even though they offer little or no therapeutic advantage over prior therapies. Third, firms may find it very profitable to develop minor modifications to their own existing drugs, as a sort of evergreening strategy. Fourth, profits from developing and showing new uses for non-patented compounds will be small and may not support investing in clinical trials to demonstrate efficacy.

Pharmaceuticals are a natural area to try patent buyouts because markets would be relatively competitive in the absence of patents, patent protection is effective, monopoly markups are large, drugs are nondurable, “me too” inventions are widespread, and considerable information is generated during FDA trials so potential bidders could make informed bids. Moreover, because many pharmaceutical patents are valuable, the administrative costs of the system are likely to be small relative to the benefits of patent buyouts. Once FDA approval has been granted, little new development is typically required. Finally, buying out pharmaceutical patents is likely to have benign distributional consequences, whereas buying out, say patents for improved yachts will not. Financing pharmaceutical research through patents places the financing burden on disease victims. If disease incidence is random, and not fully insurable, people will prefer ex ante to insure themselves by funding the research out of general tax revenue. The system could initially be applied to treatments for a few specific diseases considered to be particularly important or particularly subject to problems resulting from the patent system. Orphan drug legislation provides a precedent for establishing special rules for drugs designed to treat particular diseases.

A standard way of eliciting information on the value of indivisible goods, such as patents, is through auctions. Figure I shows how an auction could be used to determine the price at which the government would offer to buy out patents. Under the mechanism, the market value of patents would be determined through a sealed-bid second-price auction,7 and the government would then offer to buy patents at this private value times some constant markup which would reflect the typical ratio of social to private value. Most of the patents that the government bought would be placed in the public domain. However, in order to give auction participants an incentive to reveal their true valuations, a small proportion of patents, chosen randomly, would be sold to the high bidder. Patent holders would have the right to accept or reject the government’s offer. Although the government might require a waiting period following either patenting or FDA approval of new drugs before it would buy out patents, patent holders would be free to postpone patent buyouts. Inventors who wished to sell their patents to the government would be responsible for paying for the administrative costs of the auction. Based on the empirical estimates of the social return to innovation discussed above, it seems likely that the government should offer to buy patents at a markup of at least twice their estimated private value. This will not match the social value of inventions, but it is likely to be a better approximation than the private value of patents, which is what inventors receive in the absence of patent buyouts. Under a sealed-bid, second-price auction, auction participants will bid their expectation of the patent’s value, given their information, conditional on their making the winning bid. It will be efficient for the government to estimate the private value using information from the entire distribution of bids, rather than only the highest bid, since there is no reason to throw away the information provided by the other bids. If the government knew the prior distribution of valuations, it would be able to aggregate the information of all bidders to estimate the private value of the patent. In practice, the government does not know the bidders’ prior distribution of valuations. Therefore, it might be best for the government to use a simple rule, such as offering the original patent holder some multiple of the third highest bid.8 The next sections consider the operation of the patent buyout mechanism in several different environments: with inventor cost or informational advantages; with complementary or substituting inventions; and with the threat of collusion. As discussed in Section VIII, this analysis suggests that pharmaceuticals may be particularly well suited to patent buyouts. However, before considering these more complicated cases, it is worth first reviewing the advantages of patent buyouts in a perfectly competitive, collusionfree environment. 1. The markup would raise private incentives for original research closer to the social benefit created by the invention. 2. Deadweight losses due to monopoly pricing would be eliminated if patents were put in the public domain. 3. Since monopoly profits would be eliminated, researchers would not have excessive incentives to invent substitutes for existing drugs to steal profits.

Mitigating patent rights as a way to improve welfare fits with current concern over the patent system. Concerns about the potentially deleterious effects of patents have been voiced in both academic and policy debates. In a prominent book, Jaffe and Lerner (2004) conclude that in the U.S. patents have become “sand rather than lubricants in the wheels of American progress.” Some have even gone so far as to recommend more draconian reductions in permissible patenting (Boldrin and Levine, 2008; Bessen and Maskin, 2009). The buyout system replaces some of the rents that are obtained through monopoly rights with a prize. If the planner cannot discern the quality of innovations, pure prizes are difficult to implement because the value of the prize cannot be tied to the surplus generated by the innovation as the demand is unknown by the policy maker. Even if the planner could get good information about the number of units sold at a given price, for instance by observing units sold under perfect competition, this is insufficient to construct the inframarginal values of consumers, which is essential to estimating the full value of an innovation. In our model, the hybrid mechanisms can generate greater innovation incentives and increase social welfare if the policy maker has less than perfect information about market demand.

Incentives for marketing and development are likely to be enhanced by patent buyouts. Suppose that the private value of a patent is p(E), where E denotes expenditures on development and marketing. In the absence of patent buyouts, patent owners will invest in development until p/E 5 1. Patent owners who expect to sell their patent at a markup of M will invest until p/E 5 1/M. Inventors would be free to delay patent sales to first undertake development and marketing if they wished. Some opportunities for further development may appear only after patents have been sold. For example, technological advances in related fields may open up new opportunities for development after the patent has been bought out. Patent buyouts will strengthen incentives for this further development if it can be patented or appropriated in some other way. This is because the market for a complementary invention will be larger if the original invention is sold at marginal cost. Moreover, the developer of the complementary invention will not have to split its value with the original inventor or take the risk that unresolved patent disputes with the original inventor will block new complementary products. Although patent buyouts increase inventors’ incentive to conduct development prior to the buyout, and increase others’ incentives to conduct development after the buyout, they reduce inventors’ incentives to conduct unpatentable development after patent buyouts. For example, pharmaceutical firms have much less incentive to test for new uses for generic drugs than for patented drugs. In practice, there may be other ways of appropriating investment in marketing and development. For example, some aspirin manufacturers have sufficient market power that even without patent protection they advertise aspirin’s effectiveness in preventing heart attacks. The substantial technological improvements to the Daguerreotype process and the cotton gin after the patents were bought out provide at least some suggestion that patent buyouts do not generally greatly discourage further technological development. In any case, pharmaceuticals typically need little new development after they have been approved by the FDA.9

In this paper we have examined the problem of a social planner aiming to maximize consumer welfare and innovation incentives while observing prices and quantities practiced by the innovator over time. We have shown that in our model information about market outcomes may allow the planner to generate more welfare than a traditional patent system through patent buyouts. There are a number of historical experiences in which governments bought patents out. The most famous example of patent buyout took place in July 1839 when the French government purchased the patent for the Daguerreotype photography process. The inventor, Luis Jacques Daguerre, was not able to find buyers for the process, but obtained the support of a politician that convinced the government to acquire the patent and put the rights in the public domain. Within a short period of time the process spread around the country to become the technology standard in photography (Kremer, 2001). In recent academic and policy debates pharmaceutical patent buyouts have been suggested as a strategy to improve health in low income countries. For example, Banerjee et al. (2010) propose that a Health Impact Fund compensate drug manufacturers that sell in low income countries at marginal cost. They suggest that the compensation to a given manufacturer would depend on use of the drug and evidence of realized health benefits.20 Our paper provides two main insights into the design and application of such buyout schemes. First, the planner may find it beneficial to collect market data before the buyout and use them to estimate the surplus generated by the innovation. In practice, surplus may be estimated through structural econometric models that allow policy makers to estimate the primitives of consumer preferences and to generate out of sample predictions (Cho and Rust, 2008). Such estimates can provide useful guidance in the determination the buyout compensation for the innovator. Second, the planner should consider the welfare cost associated with collecting price-quantity observations. As long as local variation in market outcomes can be exploited to learn about the global properties of the demand, prices close to marginal costs minimize the loss in consumer surplus.

Previous sections have examined the potential of patent buyouts to supplement our current system of promoting innovations through patents and direct government support of research. The government could offer to buy out patents at their private value, as revealed by an auction, times a markup designed to cover the difference between the private and social values of inventions. This mechanism involves more government discretion than the current patent system, but substantially less discretion than government funding of research through the NSF or NIH. Patent buyouts could potentially increase incentives for original invention closer to their social value; reduce incentives for wasteful ‘‘me too’’ research; and eliminate monopoly pricing distortions. On the other hand, patent buyouts could also cause a number of problems, including collusion to raise buyout prices. When new institutions are proposed, there is a natural tendency to focus on their potential risks and shortcomings. However, it is also important to recognize that existing mechanisms of encouraging innovation have serious flaws. In an 1851 editorial (cited by Dutton [1984]) urging that patents be abolished, The Economist wrote that the granting of patents ‘‘inflames cupidity, excites fraud, stimulates men to run after schemes . . . begets disputes and quarrels betwixt inventors, provokes endless lawsuits [and] makes men ruin themselves for the sake of getting the privilege of a patent, . . .’’All this is true, and yet it seems clear that the world is much better off with patents than without them. The same may be true for patent buyouts. Since it is difficult to gauge the effects of patent buyouts based on theory alone, and since large-scale patent buyouts would be risky, it might be useful to first try patent buyouts on a limited basis. Such a trial could help determine whether inventors would sell their patents, and whether bids would be substantially greater than realized profits or revenues, as would be the case if collusion were severe. This would help policy makers judge whether patent buyouts should be abandoned, redesigned, or used more widely. Pharmaceuticals are a natural area to try patent buyouts, since markets would be relatively competitive in the absence of patents; patent protection is effective; monopoly markups are large; drugs are nondurable; ‘‘me too’’ inventions are widespread; and considerable information is generated during FDA trials, so potential bidders could make informed bids.17 Moreover, because many pharmaceutical patents are valuable, the administrative costs of the system are likely to be small relative to the benefits of patent buyouts. Once FDA approval has been granted, little new development is typically required. Finally, buying out pharmaceutical patents is likely to have benign distributional consequences, whereas buying out, say, patents for improved yachts will not. Financing pharmaceutical research through patents places the financing burden on disease victims. If disease incidence is random, and not fully insurable, people will prefer ex ante to insure themselves by funding the research out of general tax revenue. The system could initially be applied to treatments for a few specific diseases considered to be particularly important, or particularly subject to problems resulting from the patent system. Orphan drug legislation provides a precedent for establishing special rules for drugs designed to treat particular diseases.

As with prizes in general, high specificity can be possible in models that include a condition to leave innovations unpatented or eschew monopoly pricing, so product risks are reduced and again, only results would be rewarded. With patent buyouts designed to spur future innovation, specificity is still considerable but less than perfect. It might reduce the cost for other researchers to produce follow-on innovations, who otherwise would have needed to pay for rights to use the patented knowledge. If the buyout price is set between private and social value (and social value can be accurately estimated) then this is a good investment of resources. If the patent holder is the low-cost producer (which is plausible, given their intimate knowledge of the invention), then they would likely still wish to sell, as they would receive the mark-up over private value and be well positioned in the market.38 Patent buyouts would decouple rewards from access by reimbursing inventors separately form sales. Lower prices would then result, which would broaden access to the invention. Monopoly distortions would be eliminated and, since researchers into complementary inventions would not need to contract to split rents or to seek licensing agreements, complementary innovations might burgeon. This happened with the French government patent buyout for the daguerreotype;39 photography flourished. Buyouts could be very effective when applied to frontier technologies, therefore, as follow-on innovations are often necessary for these to deliver their full benefits.40

Patents are a useful but imperfect reward for innovation. In sectors like pharmaceuticals, where monopoly distortions seem particularly severe, there is growing international political pressure to identify alternatives to patents that could lower prices. Innovation prizes and other non-patent rewards are becoming more prevalent in government’s innovation policy, and are also widely implemented by private philanthropists. In this paper we develop a model in which a patent buyout is effective, using information from market outcomes as a guide to the payment amount. We allow for the fact that sales may be manipulable by the innovator in search of the buyout payment, and show that in a wide variety of cases the optimal policy in our model still involves some form of patent buyout. The buyout uses two key pieces of information: market outcomes observed during the patent’s life, and the competitive outcome after the patent is bought out. We show that such dynamic market information can be effective at determining both marginal and total willingness to pay of consumers in many important cases, and therefore can generate the right innovation incentives in our model.