backs. Vaccine manufacturers have strong incentives to buy, or have accomplices buy, dosages of the
vaccine secretly. If such buybacks are easy to implement, the mechanism used by the AMC plan is
likely to yield highly inefficient outcomes. If hidden buybacks are privately costly, this mechanism
is likely to do well in stimulating innovations while avoiding deadweight losses. This mechanism is
also vulnerable to implicit or explicit bribes. For example, vaccine manufacturers are often large
pharmaceutical firms producing and selling a variety of products. Such firms can arrange for implicit
bribes in the form of discounts for other products (for example, antibiotics) they sell to countries
participating in the AMC in return for larger amount of vaccines purchased.
The most closely related paper in terms of the focus on the tradeoff between prizes and
patents is Hopenhayn, Llobet, and Mitchell (2006). They study a mechanism design problem of
cumulative innovation. They study the optimal reward policy when the quality of the ideas and
their subsequent development effort are private information. Scotchmer (1999) is another influential
paper that studies an optimal mechanism design problem with private information about costs and
profits on the side of innovator and shows optimality of patents. These two papers do not allow for
the use of market signals as a part of the optimal mechanism which is the focus of our paper.
There is a small literature on how information available on the market can be used in de-
signing rewards for innovation.3 Kremer (1998) is the most influential recent paper with a detailed
prize reward mechanism. As we have argued above, his mechanism is subject to the possibility of
manipulation. Guell and Fischbaum (1995) propose a mechanism which uses sales on a test market
for a relatively short period of time to obtain an estimate of the social surplus. Once such infor-
mation is received, the government extrapolates this information to obtain an estimate of the total
value of the social surplus if a good were to be sold on the total market. Then the innovator receives
a prize with the value equal to the estimated surplus. This proposal is certainly subject to market
manipulation. The innovator has strong incentives to increase the demand in the test market. In the
most plausible cases, if one assumes that the marginal cost of production is small compared to the
value of the innovation, and if one assumes that the monopolist can sell the good at zero price, then
this mechanism leads to extremely inefficient outcomes. Shavell and van Ypersele (2001) propose an
optional reward system in which they allow an innovator to either stay with the patent or choose a
buyout reward. Their mechanism has rewards only if the lowest social payoff is positive. If such an
assumption does not hold, patents are optimal. Boldrin and Levine (2001) in recent influential work
make entirely different technological assumptions for production of new goods. They argue that
3See Abramowicz (2003) for a review of a variety of proposals.