Contingent valuation is sometimes presented as applicable to any environmental good, unlike other valuation methods which can be used only in particular circumstances. Recent thinking suggests a narrower role.
A symposium in the Journal of Economic Perspectives (1) offers three views of contingent valuation, a widely-used method for valuing non-market environmental goods. Richard Carson argues in favour and Jerry Hausman against, while Catherine Kling and her co-authors take a more nuanced position. A careful reading, however, suggests much common ground.
Contingent valuation is often criticised as subject to hypothetical bias, that is, when people are asked how much they would pay to help conserve an environmental good, their responses tend to overstate their true willingness to pay. This can be tested where willingness to pay can be measured directly, and all three papers accept that many studies have found upward bias in such situations. For Hausman, such findings support his general criticism of contingent valuation, but Carson and Kling argue that they are of little relevance.
Central to Carson’s and Kling’s position are the concepts of consequentiality and incentive-compatibility. A question about an environmental good is consequential if respondents care about the good and believe that their responses may influence the actions of an agency towards the good. Where a question is consequential, people’s responses can be viewed as a form of economic behaviour, revealing something about their preferences. A survey question is incentive-compatible if a truthful response is an optimal strategy for respondents who perceive the question to be consequential. Questions may use a variety of elicitation formats (open question, binary choice, multinomial choice, etc), and Carson and others have explored which are incentive-compatible and which are not (2).
Carson and Kling argue that good contingent valuation studies are consequential and incentive-compatible, whereas findings of hypothetical bias relate to studies that fail to meet those conditions (3). They accept that studies which do not meet the conditions will not yield reliable valuations. So all the authors agree that contingent valuation studies that do not meet the consequentiality and incentive-compatibility conditions will not give good results. For Hausman this is because contingent valuation studies in general are subject to hypothetical bias (and other problems described in his paper). For Carson and Kling it is because a study that does not meet the consequentiality and incentive-compatibility conditions is not a proper contingent valuation study, and is therefore subject to hypothetical bias.
How large is this common ground? There is, I suggest, a large class of actual and potential contingent valuation studies that lack consequentiality. Consequentiality surely depends in part on the circumstances surrounding a study, such as whether the government is considering a policy towards the good being valued, and will take account of the results of relevant studies in developing its policy? A researcher can tailor a study to such circumstances if they exist, but is unlikely to be able to bring about such circumstances if they do not already exist.
This point is not clearly brought out in Carson’s paper because he uses forms of words that focus on communicating to respondents that their responses may influence policy, and on respondents understanding that that is so (4). Communication and understanding are important, but what this seems to neglect is whether it is actually the case that the survey responses may influence policy. Perhaps Carson takes this as read, as such an obvious presupposition of consequentiality as not to need stating. Another interpretation, however, is as allowing the possibility of what might be termed a pseudo-consequential study, in which people are led to believe that their responses may influence policy when this is not actually the case. It is perhaps not surprising, therefore, that Hausman (quoting Harrison) criticises proponents of contingent valuation as seeking ways to make people feel that their responses matter by tricking them into believing things that are not true (5).
If contingent valuation requires consequentiality, and not merely pseudo-consequentiality, then its scope for application differs in a fundamental way from that of other valuation methods. A researcher can choose an interesting subject for, say, a travel cost study, and realistically expect that good research design, good data and sound analysis will lead to a reliable valuation. But if a researcher chooses a subject for contingent valuation simply because it looks interesting, then the study is unlikely to be consequential and unlikely to yield reliable results, however carefully it is done. Only if a contingent valuation of a good is commissioned by an agency whose actions may affect that good, or in a limited range of other circumstances, is it likely to be consequential and therefore to have the potential to give reliable results.
Notes and References
1. Journal of Economic Perspectives, Vol 26 No 4, Fall 2012 including: Kling C L, Phaneuf D J & Zhao J From Exxon to BP:Has Some Number Become Better than No Number pp 3-26; Carson R T Contingent Valuation: A Practical Alternative When Prices Aren’t Available pp 27-42; Hausman J Contingent Valuation: From Dubious to Hopeless pp 43-56.
2. See especially Carson R T & Groves T (2007) Incentive and Informational Properties of Preference Questions Environmental and Resource Economics 37(1) pp 181-210.
3. Carson, in (1) above p 37. Kling, in (1) above p 11.
4. Carson, in (1) above p 30 (4 lines from bottom) and p 31 (line 7).
5. Hausman, in (1) above p 45.