Patents are a dreadfully inefficient way of encouraging new product invention. First, granting a monopoly will create deadweight loss, and patents functionally create 14 to 20 year monopolies. Second, monopoly price is not clearly linked to the consumer value a new good creates, so even ignoring deadweight loss, social and private incentives are not aligned. Third, patents distort inventive activity away from complements to already-patented products. Having government direct research by, for example, offering prizes to inventors of various new products is suboptimal for informational reasons.
In a well-known 1998 QJE, Michael Kremer proposes patent buyouts to alleviate some of these concerns. A patent holder could offer to auction his patent (or simply to retain it as in the status quo). Other firms bid in a standard second-prize auction. After the auction, the government randomizes. With probability p, the government pays a markup over the winning bid, and with 1-p sells to the private firm. This randomization ensures both that rival firms have an incentive to reveal their private information about the value of the patent, but also that, because of the markup, the returns to research are higher than under the status quo, and higher in a way that positively incentivizes research.
Kremer notes that when a patent has common value, but the seller has private information about its value, then the government markup can allow trade to happen. With no government markup, clearly no one with private information can sell: every potential buyer will know that if the seller is willing to sell, then he must have private information that the patent is worth less than the sales price. But if with some probability, the government pays a markup, then the seller will go to auction anyway hoping to get the markup. [A side note: there appears to be a mistake in this section, in the example in IV.A. Markup or not, a bidder in a common value second-price auction surely needs to shade his bid in equilibrium, right? Assume that the markup is such that the patent-holder always auctions. Let there be some large (approaching infinite) number of bidders, with signals of x+[-1,1] uniform about the value of the patent, where x is the true common value. Conditional on winning, my expected signal is x+1, and my expected payment is also x+1 (as the number of bidders goes to infinity), which gives me profit of -1. The paper appears to me to conflate winner’s curse with some sort of No Trade theorem…though I’m more than happy to be corrected on this point!)
There is an unanswered question lingering in this paper: would we be better off just having no patents at all? Government payouts are distortionary since raising tax revenue is distortionary, so a buyout system needs to provide a much better welfare outcome than a no-patent system in order to be worth it. There are business surveys that suggest huge percentages of patents are taken out for strategic reasons other than gaining market power in the specific product: for instance, in the technology sector, it is frequent for firms to trade licenses on their patents to each other after downstream products are created, and to exclude new firms who have no such patents. A voluntary patent buyout system does not solve these problems. I’m also less sanguine than Kremer about regulatory capture. He suggests that history mainly consists of government expropriation of patents rather than government handouts to favored firms. I think that the recent history of copyright law is utterly the opposite: both the US and other countries have regularly legislated handouts to the powerful.
Finally, one argument that Kremer’s system avoids. Many (most?) inventions would be created even in the absence of patents, because the inventor is able to make up the R&D costs in other ways, such as first-mover advantage. To the extent that the patent itself is not valuable to firms other than the inventor, Kremer’s buyouts would not cost the government anything, since no firm would bid more than zero for the patent.
Later this evening, I’ll discuss a new paper about how we ought to choose the government markup if we were to implement such a system.
http://dash.harvard.edu/bitstream/handle/1/3693705/Kremer_PatentBuyouts.pdf?sequence=2 (Final QJE version at Harvard’s very useful ungated depository of published papers by their faculty)