A Learning Design for Reducing Hypothetical Bias in the Contingent Valuation Method

Over the last few years a great deal of research has focussed on hypothetical bias in value estimates obtained with the contingent valuation (CV) method and on means for ameliorating if not eliminating such bias. To date, efforts to eliminate hypothetical bias have relied on calibration techniques or on word-smithing of one kind or another to induce subjects to provide responses to hypothetical questions that mimic responses made by subjects facing actual payments in the valuation experiment. This paper introduces a different approach for eliminating hypothetical bias. A design for a CV survey format is presented which provides subjects with the opportunity to learn how the CV institution works. Sequential referenda are conducted where respondents gain experience in CV settings by participating in both hypothetical and real referenda. The logic of this Learning Design is a straightforward application of the trials process used in experimental economics. We demonstrate that the Learning Design is effective in eliminating hypothetical bias in surveys concerning donations to two different public goods.     

Donation Payment Mechanisms and Contingent Valuation: An Empirical Study of Hypothetical Bias

Donation payment mechanisms are well suited forsome contingent valuation studies. In aneffort to better understand the discrepancythat has been consistently found between actualand hypothetical donations, we investigate anapproach to estimating actual willingness todonate using contingent donations with afollow-up question in which respondents ratethe level of certainty about their response tothe contingent donation question. The approachallows us to estimate the magnitude of thehypothetical bias and identify the respondentsresponsible for the bias. Identification ofthe respondents responsible for thehypothetical bias is the first step towarddeveloping an understanding of the causes andpossible remedies. In this study we find thatmost of the respondents (80%) to thecontingent donation question provide a responseconsistent with how we predict they wouldrespond in an actual donation situation.     

A Meta-analysis of Hypothetical Bias in Stated Preference Valuation

Individuals are widely believed to overstate their economic valuation of a good by a factor of two or three. This paper reports the results of a meta-analysis of hypothetical bias in 28 stated preference valuation studies that report monetary willingness-to-pay and used the same mechanism for eliciting both hypothetical and actual values. The papers generated 83 observations with a median ratio of hypothetical to actual value of only 1.35, and the distribution has severe positive skewness. We find that a choice-based elicitation mechanism is important in reducing bias. We provide some evidence that the use of student subjects may be a source of bias, but since this variable is highly correlated with group experimental settings, firm conclusions cannot be drawn. There is some weak evidence that bias increases when public goods are being valued, and that some calibration methods may be effective at reducing bias. However, results are quite sensitive to model specification, which will remain a problem until a comprehensive theory of hypothetical bias is developed.JEL classifications: C9, H41, Q26, Q28     

Researching Preferences,Valuation and Hypothetical Bias

A number of recent papers in environmental economics have focused on the process of researching preferences – agents are uncertain about preferences but with effort may narrow their uncertainty. This issue has arisen in formulating bids in contingent valuation (CV) as well as the debate over the divergence between WTP and WTA. In the context of CV, it has been suggested that the hypothetical nature of the preference elicitation process biases responses. This paper provides both a theoretical model and experimental evidence to contribute to this debate. The model is a model of competitive bidding for a private good with two components that are particularly relevant to the debate. The first component is that bidders are unsure of their own value for the private good but may purchase information about their own value (researching preferences). The second component is that there is a probability that the auction is hypothetical – that the winning bidder will not get the private good and will not pay the winning bid. The experiment tests this theoretical model of bidding equilibrium and analyzes the effects of variations in the parameters (hypotheticalness, information costs and number of agents) on the endogenous variables (such as the proportion of bidders who become informed and the winning bid). Experimental results suggest that an increase in the hypotheticalness of an auction tends to decrease the likelihood that bidders pay for information on their valuation with an ambiguous effect on the winning bid.      

A Comparison of Value Elicitation Question Formats in Multiple-Good Contingent Valuation

This paper provides a convergent validity test of two types of multinomial choice questions vis-a-vis a dichotomous choice question by formally testing whether these stated preference elicitation question formats provide comparable welfare estimates. In particular, a dichotomous choice question, a traditional multinomial choice question, and a modified multinomial choice question suggested by Carson and Groves （2007） were applied in split samples to assess the influence of the informational and incentive properties on the respondents＇ annual willingness to accept compensation for adopting costly conservation practices in agriculture that benefit the environment. Our findings suggest that the two multinomial choice question formats elicit a similar mean willingness to accept distributions, but they are both different from a standarddichotomous choice question. Further, the willingness to accept distributions derived from the multinomial choice question formats are more dispersed than those from the dichotomous choice question.     

Minimising Payment Vehicle Bias in Contingent Valuation Studies

The payment vehicle is a crucial element inapplications of the contingent valuation methodbecause it provides the context for payment. However,in many countries a relative unfamiliarity with theuse of tax levies and referenda can affect theplausibility of payment vehicles and lead to paymentvehicle bias. The most commonly used approach fordetermining whether payment bias exists is to usetests of convergent validity. It is demonstrated thatsimple tests of convergent validity can be ineffectivein diagnosing the existence of payment vehicle bias.Payment vehicle bias is found to occur because ofdifferences in the coverage of payment vehicles anddoubts about payment being one-off. When respondentsare found to be protesting against a particularpayment vehicle, the current state of the art approachis to delete them from the sample. In this paper analternative approach that relies on the recoding ofprotest responses is proposed.     

Testing the Effectiveness of Certainty Scales,Cheap Talk,and Dissonance-Minimization in Reducing Hypothetical Bias in Contingent Valuation Studies

Stated preference methods such as contingent valuation and choice modeling are subject to various biases that may lead to differences between actual and hypothetical willingness to pay. Cheap talk, follow-up certainty scales, and dissonance minimization are three techniques for reducing this hypothetical bias. Cheap talk and certainty scales have received considerable attention in the literature, but dissonance minimization has not previously been experimentally tested. Using a four-way split sample design involving over 600 subjects, results from an actual referendum on provision of a quasi-public good were compared with three similar but contingent referenda employing the three bias-reducing techniques. Hypothetical bias was again present. Certainty scales, when properly calibrated, and dissonance minimization were found to be most effective in reducing the bias.     

Combining Contingent Valuation and Choice Experiments. A Forestry Application in Spain

When many good variants are to be valued, like in several forest management programs which differ in the attribute levels, the Contingent Valuation Method may not be a practical option, given the cost and time involved to value each variant separately. There are at least three alternative procedures in such situations. One (i) is to apply an attribute based valuation method like a Choice Experiment (CE); another (ii) is to conduct a CVM and a CE exercise and estimate a joint model; a third one (iii) is to use the CVM estimates to value a base scenario, and then use the CE results to adjust the CVM estimates for any new different scenario. A numerical simulation and an empirical application to an afforestation program in Spain show that the latter is the procedure that yields closer values to the CVM estimations, followed by (ii).      

Existence Value and the Contingent Valuation Method

People express concern about the possible extinction of animal species which they will never see, about damage to wilderness areas they will never visit, and so on. Such concerns are often interpreted by economists in terms of 'existence' or 'passive use' values. Three views have been proposed regarding the appropriate way of dealing with existence values in benefit-cost analysis. First, the existence values and concerns about environmental preservation that are not related to use may be disregarded. Second, concerns about environmental preservation may be taken into account, but not interpreted as existence values or evaluated in monetary terms. Third, the benefit-cost analysis may be modified to incorporate monetary estimates of existence values. Advocates of the third approach have developed a range of techniques for eliciting existence values, collectively referred to as the contingent valuation method. In this paper, the concept of existence value and the techniques of contingent valuation are surveyed, with the aim or providing a basis for comparisons of the three approaches described above.     

A Bayesian Approach to Double Bounded Contingent Valuation

In this paper we propose a Bayesian approach to model double bounded contingent valuation data. The double bounded elicitation method is interpreted as a two tier iterated process in which the subject is allowed to have a second thought about his/her valuation for the environmental good. Prior information is modelled from the answers to the first dichotomous choice question. The model is Quasi-Bayesian (Q-B) in that the prior distribution refers to mean willingness to pay while the likelihood function refers to the proportions of a multinomial distribution. This model is applied to empirical data from a contingent valuation survey involving the valuation expressed by European tourists for access to natural areas in the Canary Islands. Results show that point estimate of consumer surplus computed with the Q-B model does not differ substantially from single bounded model estimation. In addition, double bounded seems to be quite robust to the choice of the prior model of willingness to pay responses. Comparison with open ended suggests that the Q-B model might be useful to control for strategic response and starting point biases.     

Hypothetical Bias in Risk Preferences as a Driver of Hypothetical Bias in Willingness to Pay: Experimental Evidence

This study provides new insight on the established issue of elicitation bias in hypothetical choice settings. In particular, using an experiment that elicits real and hypothetical willingness to pay for the protection of otters, we provide evidence that the observed bias is driven by those who are both uncertain about their value and exhibit hypothetical bias in risk attitudes. This finding provides a reason why calibration approaches based on follow-up certainty questions have proven useful, but also suggests a degree of caution in applying the approach. Our results favour combining uncertainty calibration with CVM-X in which hypothetical bias in risk attitudes is also taken account to correct for hypothetical bias in WTP.     

The Economic Value of Air-Pollution-Related Health Risks in China: A Contingent Valuation Study

The economic value of preventing adverse health effects related to air pollution is estimated using contingent valuation in three diverse locations in China. Values are estimated for three health endpoints: cold, chronic bronchitis, and fatality. Alternative statistical models are tested to study their impact on estimated willingness to pay (WTP) and on the relationship between WTP and respondent characteristics. Using the official exchange rate, the sample-average median WTP to prevent an episode of cold ranges between US$3 and US$6, the WTP to prevent a statistical case of chronic bronchitis ranges between US$500 and US$1,000, and the value per statistical life ranges between US$4,000 and US$17,000. Estimated mean values are between two and thirteen times larger. Our estimates are between about 10 and 1,000 times smaller than estimates for the US and Taiwan using official exchange rates. Indoor air quality, measured for a subset of respondents, shows no consistent relationship with WTP.     

Implications of Bid Design and Willingness-To-Pay Distribution for Starting Point Bias in Double-Bounded Dichotomous Choice Contingent Valuation Surveys

We examine starting point bias in double-bounded dichotomous choice contingent valuation surveys. We investigate (1) the seriousness of the biases for the location and scale parameters of the willingness-to-pay (WTP) in the presence of starting point bias; (2) whether or not these biases depend on the distribution of WTP and on the bid design; and (3) how well a commonly used diagnostic for starting point bias—a test of the null that bid set dummies entered in the right-hand side of the WTP model are jointly equal to zero—performs under various circumstances. Monte Carlo simulations suggest that the effect of ignoring starting point bias depends on the bid design and on the true distribution of WTP. A well-balanced, symmetric bid design may result in very modest biases even when the anchoring mechanism is very strong. The power of bid set dummies in detecting starting point bias is low. They tend to account for misspecifications in the distribution assumed by the researcher for the latent WTP, rather than capturing the presence of starting point bias.     

A Semi-parametric Estimator of Willingness-to-pay Applied to Dichotomous Choice Contingent Valuation Data

The use of dichotomous choice questions (Are you willing to pay $x?) has been advocated as superior to open ended questions (What would you be willing to pay?) in contingent valuation studies of non‐market goods. A supplementary dichotomous choice question, raising or lowering the bid price depending on the initial response, gives increased precision. However, a problem with these methods is that identifying the mean willingness to pay (WTP) is more complex, and often conditional on making a distributional assumption. This paper will present a semi‐parametric method for estimating WTP without invoking any such assumptions, using the interval inherent in the double bounded format, and based on estimating the underlying hazard function for the distribution of WTP. An advantage of the approach over other non‐parametric methods is that it allows one to include respondent characteristics as determinants of the individual decisions, and does so within a computationally straightforward framework.     

Modelling Heterogeneity and Uncertainty in Contingent Valuation: an Application to the Valuation of Informal Care

Both the lack of market data and the need to adopt a more holistic approach in the valuation of non‐market activities within health care have pointed towards the use of contingent valuation (CV) methods. However, to date, few studies have employed such techniques to value informal care, despite its provision being an important public policy question. We propose an analytical framework that through the use of random parameters models and respondents' certainty scales can incorporate both unobserved and observed heterogeneity in the CV modelling. This is the first CV study of informal care for Scotland (UK) and a £7.68 per hour value is estimated.     

A Contingent Valuation Estimate of the Benefits of Wolves in Minnesota

In this paper we estimate the willingness topay for a wolf management plan and a wolfdamage plan in Minnesota using the contingentvaluation method. The theoretical definition ofwillingness to pay for wolf protection iscomposed of use and non-use values. Weincorporate a don't know response option in thedichotomous choice valuation questions. A largenumber of respondents answered don't know. Themultinomial logit model is used todifferentiate between don't know and noresponses. Non-use motives are importantfactors that explain willingness to pay. We usethese benefit estimates in combination with twoalternative cost estimates to consider theefficiency of the wolf management and damageplans. Both plans have estimated benefitsgreater than costs.     

Addressing Negative Willingness to Pay in Dichotomous Choice Contingent Valuation

This paper has four purposes. First, we outline the controversy surroundingthe issue of negative willingness to pay (WTP)in contingent valuation (CV) studies. Second,we use Monte Carlo simulation to examine theperformance of alternative distributionalassumptions in estimating WTP in the presenceof varying proportions of the populationholding negative WTP values. We focus ondichotomous choice CV (DC-CV), where negativeWTP values may be especially difficult todetect. Third, we extend the simulation toinvestigate the performance of the mixturemodels that have recently been proposed forhandling/identifying non-positive WTP values. Fourth, we extend the simulation to investigatethe performance of the nonparametric lowerbound Turnbull approach. Results indicate thatthe relative performance of the DC-CV modelingalternatives evaluated here, which assumepositive WTP, varies across the simulationsetting (e.g., proportion of negative WTP); butnone can be said to reasonably ``solve' theproblem ex post. This underscores theimportance of ex ante efforts to identify ifnegative WTP is likely to be prominent in agiven valuation setting. In such cases,appropriately handling negative WTP must beaddressed through ex ante survey design andmodeling choices that allow negative WTP.     

Protesting and Justifying: A Latent Class Model for Contingent Valuation with Attitudinal Data

This article develops a latent class model for estimating willingness-to-pay (WTP) for public goods using simultaneously contingent valuation (CV) and attitudinal data to identify individuals with similar characteristics, such as WTP and protest attitudes. We find evidence that the answer to the CV question influences the responses to the attitudinal questions. In our case, this influence reflects rational behavior (budget constraint issues) and justification biases related to protest reasons, such as lack of trust in institutions or fairness issues with respect to the distribution of the burden of preservation. The results from our empirical application confirm the importance of accounting for those biases.     

‘Adopt a Hypothetical Pup’: A Count Data Approach to the Valuation of Wildlife

The willingness to pay (WTP) for a coyote conservation program is estimated using a novel payment-vehicle, based on how many ‚problem’ coyotes respondents would be willing to sponsor for a year. This hypothetical scenario mimics an increasingly popular type of actual market. Data from a phone survey conducted in Prince Edward Island are analyzed using count data models that consider different processes explaining zero responses and the level of positive responses. This is particularly important in the case of coyotes, often regarded as an economic bad. Estimates of WTP per coyote per year around $18–$22 and annual WTP per contributor of about $46–$57 are obtained.      

Application of Sample Selection Model to Double-Bounded Dichotomous Choice Contingent Valuation Studies

Modeling households' behavior with the data from a contingentvaluation (CV) survey is often complicated by samplenon-response, which can cause non-response bias and sampleselection bias, leading to inconsistent parameter estimates and adistorted mean willingness-to-pay estimate. This paper reportsthe results of empirical tests for both biases using householdsurvey data in which the double-bounded dichotomous choice CVquestion involved the benefit of a tap water quality improvementpolicy in Korea. No non-response bias, but sample selection bias,is detected in the sample. To correct for sample selection bias,a sample selection model is employed. The authors also discusshow failure to correct for bias may distort aggregate benefitestimates.