Respondent experience (i.e., a respondent's information set) has long been suspected to influence contingent valuation estimates of environmental values. We assess the influence of experience by explicitly modeling the relationship between respondent experience and both fitted individual resource values and the conditional variance of these estimated values.
A set of observed and contingent behavior results for each survey respondent allows the researcher to control for individual heterogeneity by taking advantage of panel data methods. The contingent scenarios also provide opportunities to (a) test for differences between observed and contingent preferences and/or (b) assess likely demands under conditions beyond the domain of observed variation in costs or resource attributes.
We generalize upon previous analyses of dichotomous-choice with follow-up data by relaxing the assumption that the identical unobserved resource value motivates both responses. While values implied by the first and second responses are highly correlated and may be drawn from the same distribution, they are definitely not identical. Furthermore, assuming that they are identical can severely distort the estimated valuation distribution.
Contingent valuation survey responses are combined with travel cost data on actual market behavior to estimate jointly both the parameters of the underlying utility function and its corresponding ordinary demand function. This is a prototypical empirical example of a new modeling strategy, variants of which should prove useful in many applications, especially where reliance on a single valuation method is undesirable.
The luck of the draw in assigning the referendum thresholds on individual referendum contingent valuation questionnaires can produce a surprisingly wide variety of value estimates. We control for the behavioral biases that confound other comparison studies by using one sample of payment card CV data and simulating 200 samples of consistent referendum responses. Where referendum questions have produced different value estimates than other formats, elaborate explanations for the apparent discrepancies may not be necessary.
Most researchers working with referendum contingent valuation data have used ordinary logit models and have done without confidence intervals for their fitted willingness-to-pay (WTP) values derived using referendum contingent valuation data. We then how an alternative approach, censored logistic regression, makes it straightforward to produce confidence intervals (either for individual observations, at the means of the data, or for an arbitrarily forecasted set of explanatory variables).
Contingent valuation methods (CVM) have been shown to be potentially very useful for eliciting information about demands for non-market goods. We assess the sensitivity of “payment card” CVM results to the researcher's choice of estimation method. Empirical payment card data are used in both (a) a naive ordinary least squares (OLS) procedure employing interval midpoints as proxies for the true dependent variable, and (b) an efficient maximum likelihood (ML) procedure which explicitly accommodates the intervals. Depending upon the design of the payment card, OLS can yield biased parameter estimates, misleading inferences regarding the effects of different variables on resource values, and biased estimates of the overall resource value.
Closed-ended contingent valuation surveys are used to assess demands in hypothetical markets and recently have been applied widely to the valuation of (non-market) environmental resources. This interviewing strategy holds considerable promise for more general market research applications. The authors describe a new maximum likelihood estimation technique for use with these special data. Unlike previously used methods, the estimated models are as easy to interpret as ordinary least squares regression results and the results can be approximated accurately by packaged probit estimation routines.