vulnerable to costly signal manipulation, for example by hidden buybacks. That is, prizes can be
manipulated by the innovator secretly purchasing the good so as to make it seem that the market
size is larger than it is. We show that, if the costs of these buybacks are small relative to the costs
of the innovation, any mechanism that induces innovation must necessarily induce patents.
Finally, we consider the patent-buyout mechanism proposed in an influential paper by Kre-
mer (1998). This mechanism can be used to implement the general method of using market signals
described in this paper when there is no signal manipulation. Kremer (1998) argues that an auction
exploiting information that competitors and, more generally, the market have in determining the
terms of the buyout can induce an efficient level of innovation without the costs of monopoly. We
show that Kremer’s auction is susceptible to manipulation by the innovator. Specifically, suppose
that the innovator can use one or more accomplices to participate in the auction. We show that Kre-
mer’s auction leads to inefficient outcomes. If the auction designer can exclude the innovator or his
accomplices from participating in the auction, the mechanism does indeed yield efficient outcomes.
Thus, the desirability of Kremer’s mechanism relies on the ability to preclude the manipulation of
the mechanism by the innovator.
Our main contribution is to show that the desirability of the patents as a mechanism to induce
innovation relies crucially on the ability of the innovator to manipulate signals. If such manipulation
is relatively easy, patents are necessary. If manipulation is costly, patents are harmful. In terms of
applications and designing mechanisms in practice, our paper implies that we should be cautious
about adopting proposed new mechanisms. Such mechanisms require consideration of how to make
them manipulation proof. Such manipulation could occur through bribes, buybacks, and, in the
context of auction-like mechanisms, the use of accomplices as bidders.
Our analysis has direct practical and policy implications. For example, following work by Kre-
mer (1998) the Advanced Market Commitments (AMC) plan has been set up to provide incentives
for Pneumococcal vaccine development with the active participation of a number of governments
and non-governmental organizations (see Experts Group Report 2008). This plan proposes to subsi-
dize, at deeply discounted prices, the vaccine manufacturers who sell vaccines which protect against
diseases to developing countries. The mechanism makes the amount of the subsidy a function of
the number of doses of vaccine sold by the pharmaceutical company. The mechanism is intended
to allow vaccine manufacturers to recoup the cost of innovation while ensuring that vaccines are
sold at the marginal cost of production. Our main result shows that this mechanism is vulnerable
to manipulation. Consider, for example, our analysis of costly signal manipulation via hidden buy-