Respondent uncertainty in contingent valuation: the case of whale conservation in Newfoundland and Labrador

Abstract

In this article we investigate the issue of respondent uncertainty in contingent valuation studies while estimating the willingness to pay for a whale conservation program off the coasts of Newfoundland and Labrador. We use data from a phone survey administered to a sample (N = 614) of adult Canadians, proposing a policy consisting of subsidizing and enforcing the use of acoustic devices that would reduce the likelihood that whales become entangled in fishing nets. A follow-up question asked respondents how certain they were about their answer to the main dichotomous-choice question, which allows us to investigate how the treatment of uncertainty affects value measures. A mean willingness to pay of about $81/year per respondent is estimated when accounting for the degree of certainty with which respondents expressed their willingness to pay. We also analyse payment vehicle effects using a split-sample approach whereby some respondents were asked a dichotomous-choice question about a tax contribution while others were asked about a voluntary donation instead.

Keywords:

• contingent valuation
• whales
• preference uncertainty
• dichotomous choice
• payment vehicle
• willingness to pay

JEL Classification::

• Q21
• Q26
• Q51
• Q57

Acknowledgements

We would like to thank Wayne Ledwell for valuable input on the problem of whale entanglement; Lars Hallstrom, Van Lantz, Malgorzata Winiszewska, Mykyta Vesselovsky, Jennifer Smith, Vernon Smith and Larry Felt for their suggestions on how to improve the questionnaire; Arne Rise Hole for unpublished materials that helped us with the econometric analysis; and Diane Dupont, Deborah Febres and one anonymous referee for their comments on previous drafts. The survey effort was made possible by funding provided by the Applied Research Fund of The Leslie Harris Centre at Memorial University of Newfoundland.

Notes

¹ This subsection borrows substantially from the recent surveys of the issue contained in Shaikh et al. (2007) and Akter et al. (2008).

² Shaikh et al. (2007) found evidence of the validity of this assumption in the case of estimating the willingness to accept a tree planting program data.

³ In our analysis below, however, we used both DK responses and a follow-up question.

⁴ See Carson et al. (2003) for details about the Exxon Valdez contingent valuation survey.

⁵ The questionnaires were approved by Interdisciplinary Committee on Ethics in Human Research at Memorial University.

⁶ The following proportions apply to the population (sample): male – 48% (46%), female – 52% (54%); household income intervals (in $2005): less than $30 000 – 26% (22%), $30 000–$50 000 – 21% (20.5%), $50 000–$70 0000 – 17% (19%), $70 000–$90 000 – 13% (13%), more than $90 000–24% (27%); age intervals: 19–24 years old interval – 9% (9%), 25–34 interval – 17% (16%), 35–44 interval – 20% (20%), 45–54 interval – 21% (23%), 55–64 interval – 15% (18%), more than 64 years old – 18% (15%).

⁷ To ensure randomness, only the person in a household who had the nearest birthday was interviewed. A series of age and gender quotas were also pursued.

⁸ This proportion happens to be remarkably close to its counterpart in Loomis and Larson's (1994) study of the valuation of whales in Califonia (35%).

⁹ As the bulk of the literature suggests (e. g. Bateman et al., 2002).

¹⁰ Two respondents who had refused the offered bid suggested, when asked about their reasons to refuse, that they would have been willing to pay a lower amount. Therefore, those respondents information was recoded with a ‘yes’ and an appropriately adjusted bid, based on this additional information.

¹¹ See Royston (2008).

¹² Apart from helping when it comes to using the results of the study for benefit transfer studies.

¹³ Or chooses the option ‘don’t know’.

¹⁴ The results in Lyssenko and Martinez-Espineira (2009) suggest that correcting for endogeneity increases econometric efficiency and substantially corrects the bias affecting the estimated coefficients of the experience variables, by isolating the decreasing effect on option value caused by having experienced the resource.

¹⁵ In general STATA's standard post-estimation methods cannot be directly applied to multiply-imputed data. Methods relying on likelihood comparisons (lrtest) are not applicable because multiple imputation does not involve calculation of likelihood functions for the data (Galati et al., 2008).

¹⁶ In recognition of this, some authors proposed to ask respondents to assume, in experimental settings, that the good valued could be bought only during the experiment itself and not later nor elsewhere (Blumenschein et al., 1998, 2001; Johannesson et al., 1998).

¹⁷ For example we built a variable with only three values: 1 for ‘not sure at all’, 3 for ‘completely sure’ and 2 for ‘anything in between’.