Voting Manipulation in Referendum Contingent Valuation: Experimental Evidence with Induced Value

This work attempts to measure the direction and size of misrepresentation in the answers to referendum type contingent valuation (CV). In a laboratory experiment simulating a simple public project, participants are asked to reveal, by means of a CV procedure, their induced Willingness To Pay (WTP) for this project. The results of the experiment indicate that, at least in a symmetric environment and a credible project, the modal behavior is truthful revelation of induced WTP, followed by under-statement, leaving over-statement as the least common behavior. Estimation of a panel data model indicates that declarations are strongly correlated to, and on average smaller than, the induced WTP. Estimation of a Markov chain model shows that whether a player declares above, at, or under her induced WTP is not significantly related to that WTP, but is strongly related to that player’s previous period declaration. A variation of the experiment indicates that one cause for this behavior is the absence, in CV, of a rule stating how the players’ declared WTP are going to be used to decide if the project is implemented. † I would like to thank Professors Jordi Brandts, Michael Creel, Theodore Groves, and Giuseppe Munda for helpful comments. Partial financial support for this project was provided by the Spanish DGCICYT (PB94-0663-C03-01). Instructions (English and Spanish versions) and experimental results are available from the author upon request.