This paper uses commonly available prevalence estimates to bound future incidence. The bounds rely on restricting the fraction of contacts between individuals of different infection statuses. It is argued that these bounds can be further tightened by restrictions of economic models of infectious disease that imply that uninfected individuals have larger incentives to avoid matching with infected individuals than do the infected individuals themselves. This implies that incidence predictions from canonical models of infectious disease are worst-case upper bounds, with the degree to which they overestimate new cases being monotonically related to this type of infection-dependent matching. Evidence in support of the economic type of infection-dependent matching is presented, by using data on the joint distribution of partners' HIV statuses in a random sample of couples from San Francisco in 1988-1989.