HYPOTHETICAL BIAS: THE MITIGATING EFFECTS OF CERTAINTY QUESTIONS AND CHEAP TALK

This paper reports on an investigation of hypothetical bias and approaches to identifying and mitigating the bias. The split sample design includes an actual donation treatment, a contingent donation treatment with a follow-up certainty question and a contingent donation treatment with cheap talk. Studies comparing contingent values to actual payments consistently find that respondents report higher willingness to pay in a hypothetical payment situation relative to an actual payment situation. While the existence of hypothetical bias has been confirmed in such studies, less attention has been focused on the nature and causes of hypothetical bias. Previous research (Champ et al. 1997, Champ and Bishop 2001) suggests that a small and potentially identifiable group of respondents to a contingent valuation survey may be responsible for the hypothetical bias. The goal of this research is to use data from a carefully designed field study to identify the hypothetical bias associated with a contingent donation survey and identify the respondents responsible for the bias. Identification of the survey respondents responsible for the hypothetical bias allows us to examine the attitudes and socio-demographic characteristics associated with hypothetical bias. The data are based on a field study administered via a mail survey with three treatments. One treatment involves asking a group of study participants to actually make a donation. This treatment serves as the "benchmark" for identifying hypothetical bias associated with the two other treatments. The second treatment involves a contingent donation (CD) question with a follow-up question that asks respondents how certain they are about their response to the CD question. This information can be used to identify the respondents responsible for the hypothetical bias (Champ et al. 1997, Champ and Bishop 2001). While the second treatment can be considered an "ex-post" approach, the third treatment is an "ex-ante" approach in that before the contingent donation question a script is included in the survey that explains the nature of hypothetical bias and encourages respondents to answer the contingent donation question as they would for an actual donation solicitation. The script was developed to be similar to the cheap talk scripts used in contingent valuation surveys administered via the mail by Lusk (2003) and Aadland and Capland (2003). Previous studies that have employed a cheap talk script have found that inclusion of the cheap talk script diminishes hypothetical bias. However, we do not know if the respondents to responding positively to a CD question after a cheap talk script are similar to the respondents who respond positively to an actual donation solicitation. This study allows for such an investigation. The study involved development of a mail survey with three main sections: (1) The description of the good. A program through the International Crane foundation to purchase radio transmitters to help scientists track a newly established flock of whooping cranes was described in detail. (2) The willingness to donate question. In all three treatments, the question was single bound dichotomous-choice question. Five offer amounts were selected based on the results of an open-ended pretest. In the actual payment treatment, an actual donation was solicited and respondents who chose to donate were asked to send a donation check back with the survey. (3) The last section of the survey included questions about previous knowledge of and experience with whooping cranes, attitudes toward reintroduction of whooping cranes and environmental programs in general and socio-demographic questions. The survey instrument has been pretested and modified based on the results of the pretest. The final survey will be administered in mid-January. Preliminary results will be available in spring 2004