Discrete/Continuous Contingent Valuation of Private Hunting Access in Kansas

Estimating the Local Cost of Protecting Koshi Tappu Wildlife Reserve, Nepal: A Contingent Valuation Approach

“…However, open-ended CVM often encounters high rates of non-response and/or reporting zero value of the resource, which are generally interpreted as protest bids (Halstead et al, 1991;Goodwin et al, 1993). Reasons for protests or zero bids are generally due to disagreement with the payment mechanism, distrust of the contingent market (Halstead et al, 1991;Bowker et al, 2003), or if compensation was paid previously during the establishment of KTWR and respondents know that they will not be considered again for compensation.…”

“…Therefore, economic values estimated using open-ended CVM can vary depending upon how the zero bids are treated in the analysis. Following Goodwin et al (1993) we use censored Tobit model and compare with 'double hurdle' models that allow the censoring rule to follow a distinct process from continuous WTA elicitations. We use these analytical techniques in econometric methods to illustrate the implications of the various potential treatments of protest bids in open-ended CVM on research results.…”

“…This expression represents the situation where zero responses are generated from the same process as non-zero responses that represent compensating surplus (Halstead et al, 1991;Goodwin et al, 1993). If the process generating the censoring is different from the process generating the WTA response (Goodwin et al, 1993), the appropriate censoring rule will be expressed as:…”

“…However, the empirical model that distinguishes the censoring rule from determining the continuous process is modeled using ''double-hurdle'' procedures (Goodwin et al, 1993). The double-hurdle procedure involves two models: a Probit model to address the yes/no censoring rule and a truncated model to account for WTA bids equation (2).…”

“…The doublehurdle model offers an advantage by allowing us to evaluate positive WTA responses separately from zeros. While maximum-likelihood estimation of a probit equation is used to evaluate the censoring rule (Z i h), the maximum-likelihood estimation that accounts for truncated normal distribution is used to account for the sub-sample of uncensored observations (Goodwin et al, 1993;Greene, 1997). Assuming the same variables are used across the models, a specification test that evaluates the restriction imposed by the Tobit model is performed by Cragg's formulation:…”