Several recently published studies concerned with the validation of willingness-to-pay estimates derived with the contingent valuation (CV) method have utilized methods similar to experimental economics’ methodologies where experiments are conducted with groups of subjects and responses to both hypothetical and real valuation questions are obtained. It is often— but not always—the case that results from such experiments demonstrate significant differences between responses to the real and hypothetical valuation questions. Such differences are commonly interpreted as reflecting “hypothetical bias.” Concern with hypothetical bias has motivated a growing number of researchers to explore techniques that might eliminate such bias, thereby providing methods from which unbiased estimates for willingness to pay might be obtained with the CV method. One method used in efforts to derive unbiased value estimates relies on calibration techniques. For example, McKinley Blackburn et al. (1994) and Fox et al. (1999) use within-sample techniques where subjects respond to hypothetical and then real valuation questions. A calibration function is then estimated which relates differences in responses obtained in the two treatments to subject characteristics. William Schulze et al. (1999) explore similar techniques where hypothetical willingness to pay is “calibrated” by subjects’ self-reports of embedding. While calibration may prove to be a useful tool for ex post adjustments of stated preference values, its practicality appears to be limited by the extent to which a calibration function derived for one good can be used to calibrate hypothetical values derived for a different good. Indeed, Fox et al. (1999 p. 2) note that: “As our results show, calibration appears to be commodity-specific, and as such, [the calibration approach] must proceed on a case-by-case basis until enough evidence exists to reveal any systematic bias that can eventually lead to a general calibration function.” A second and very different approach to dealing with the hypothetical bias problem focuses on designs for the CV questionnaire that might directly induce subjects to provide responses to hypothetical valuation questions that correspond with responses observed when actual cash payments are involved. To date, efforts to implement this approach have relied on adding to the CV script brief requests that subjects consider budgetary substitutes in responding to the valuation question. Problems encountered with this particular approach are illustrated by Loomis et al. (1994) * Department of Economics, School of Policy Studies, Georgia State University, Atlanta, GA 30303. The authors acknowledge financial support for this work from the U.S. Environmental Protection Agency Contract No. R824710-01-0. Thanks are due to Nicholas Flores, Bill Schulze, and two anonymous reviewers for helpful comments. The authors also gratefully acknowledge the participation of Glenn Harrison in early stages of this research, and the assistance of Harold Ball, Kelly Brown, Kate Gardner, James Murphy, Tom Stevens, and Melonie Williams in conducting experiments. 1 See, as examples, John Loomis et al. (1994), Helen R. Neill et al. (1994), Cummings et al. (1995a), Thomas C. Brown et al. (1996), and John Fox et al. (1999). 2 Two studies report estimates of willingness to pay where hypothetical bias is not identified (David Brookshire and Don Coursey, 1987; Mark Dickie et al., 1987). V. Kerry Smith and Carol Mansfield (1997) fail to find hypothetical bias in an experiment based on willingness to accept. As is discussed later, we also fail to find evidence of hypothetical bias in one particular set of experiments. 3 The methodology used in Schulze et al. (1999) may be described as “adjustment” rather than “calibration” since “true” values (those observed when actual payments are required) which form the basis for calibration are not obtained in their experiments. 4 Similar results are commonly reported in the marketing literature. For example, Jordon Louviere (1996 p. 170) observes that: “I think it very unlikely that one simple and generalizable approach to ‘adjusting’ (stated preference) numbers will be found, despite clear evidence supporting strong monotonic links between stated preferences and actual choices in real markets . . . differences in product awareness, learning, etc. . . ., necessarily imply that no one magic constant can exist across all product categories.”

Unbiased Value Estimates for Environmental Goods: A Cheap Talk Design for the Contingent Valuation Method

The contingent valuation method: a review

A comparison of stated preference methods for environmental valuation

Participants in contingent valuation surveys and jurors setting punitive damages in civil trials provide answers denominated in dollars. These answers are better understood as expressions of attitudes than as indications of economic preferences. Well-established characteristics of attitudes and of the core process of affective valuation explain several robust features of dollar responses: high correlations with other measures of attractiveness or aversiveness, insensitivity to scope, preference reversals, and the high variability of dollar responses relative to other measures of the same attitude.

/pdf/economic-preferences-or-attitude-expressions-an-analysis-of-3bbagmdxei.pdf

Economic Preferences or Attitude Expressions?: An Analysis of Dollar Responses to Public Issues

https://environment.yale.edu/kotchen/pubs/environ.pdf

Environmental attitudes, motivations, and contingent valuation of nonuse values: a case study involving endangered species

A test is performed of the recommendation by the NOAA panel on contingent valuation that respondents be reminded about other substitute resources and their budget constraint prior to answering the willingness-to-pay (WTP) question The context for this experiment is a CVM study of the benefits from reducing fire hazards to old-growth forests in Oregon The test was performed using two mail surveys that were otherwise identical except for the inclusion of reminder statements prior to the WTP question The dichotomous choice logit equations were not statistically different between versions and the mean WTP results were identical

https://www.jstor.org/stable/pdfplus/3146643

A. Gonzalez-Caban

Papers

Do Reminders of substitutes and budget constraints influence contingent valuation estimates