Improving the contingent valuation method: Implementing the contribution game

Abstract

This paper reports the use of laboratory experiments to investigate the properties of a public good provision mechanism as it is employed to elicit individual measures of willingness to pay for a collective good. The major finding is that contingent valuation method surveys based on the contribution game mechanism are incentive compatible, thereby reducing strategic bias; permit the use of open-ended valuation questions, avoiding the statistical problems encountered when using truncated data and the potential problems associated with focal price effects; and encourage the investigator to supply considerable information concerning the public good and the users (including potential users), overcoming many of the cognitive limitation problems noted in the literature. The contribution game setting is not capable of generating a completely bias-free data set. However, the design is capable of producing an identifiable subset of the data that is not biased. These results show that the received sentiment against the use of biased survey data may be misplaced when one is able to identify a subset of the data that is not biased. This paper is, to our knowledge, the first to demonstrate, by example, that zero-mean error predictions can be obtained by systematically choosing a subset of observations from biased data rather than by the traditional econometric approach of transforming the entire set of values.