Continuous Preferences Can Cause Discontinuous Choices: An Application to the Impact of Incentives on Altruism

Implementing incentive systems can sometimes backfire in practice: experimental evidence and folklore both suggest that offers of explicit rewards can expose surprising discontinuities in behaviour. This Paper models two such discontinuities that have been claimed by psychologists and experimental economists to constitute important exceptions to the standard economic theory of human motivation. The first (?type discontinuity?) is the observation of a discontinuity in the distribution across population types of values of the willingness to accept payment in return for performing certain (?civic?) actions, such as giving blood or performing public service. It is claimed that this distribution is bimodal, even discontinuous: many people have a zero WTA, many have a large positive WTA, but nobody has a small positive WTA. The second (?payment discontinuity?, also known as ?crowding-out?) is that people who are willing to perform certain actions for free will refuse to perform them for a low price, even if they subsequently agree to perform them if the price is raised enough. Civic virtue may, on this view, be crowded out by the introduction of explicit incentives; people may stop doing things they were previously prepared to do without reward. The Paper shows that both phenomena may be observed as a result of individuals? acting in a first period of public service in the knowledge that the terms of their action signal their type, and their type will affect a process of assortative matching in a second period. Type Discontinuity, but not Payment Discontinuity, is observed in a signaling game in which individuals announce the prices at which they will perform a civic action. Payment Discontinuity, but not Type Discontinuity, is observed in a screening game in which individuals have only a binary participation decision available to signal their type. The proportion of individuals participating when rewards are zero can be higher than when rewards are positive but small.