A discrete choice conjoint analysis and a discrete contingent valuation assessment of Southern Appalachian resources: a comparison and contrast

In response to increased public concern for environment resources, economists have relied upon the contingent valuation methodology (CVM) to measure an environmental resource's existence value. However, researchers have identified biases in CVM that distort the estimated value of environmental resources. Conjoint analysis (CA) is a valuation methodology that is used by marketing professionals to predict the public’s acceptance of a new market good. This dissertation investigated whether discrete choice CA can be considered a credible alternative to discrete choice CVM.

A single CA survey and three CVM surveys that provided the same information about Southern Appalachian environmental resources were given to a convenience sample of students, staff, and faculty at The University of Tennessee, Knoxville. Comparisons between the CVM surveys' estimated regressions and mean willingness-to-pay (WTP) amounts detected a sequential question and part-whole bias, but not an embedding bias.

Four different CA models of the respondent choice decision were estimated. The first CA model (the traditional CA model) assumed the independence of irrelevant alternatives (HA). The second CA model revealed violations of DA. The third and fourth CA models measured if resource attribute quality level interactions across choice alternatives or within choice alternatives in the CA survey influenced respondent choice. The four CA models estimated different resource attribute parameter and consumer surplus estimates. Considerable multicollinearity between the regressors was detected in the latter two CA models of respondent choice.

There were several noteable findings that are of policy interest. First, violation of HA reveals that CVM's (which preserves the HA assumption) consumer surplus estimates are a function of the status quo level of resource quality given in the WTP question. Consequently, actual quality changes in the status quo level of resource quality given in each CV survey would invalidate consumer surplus estimates. Second, although the part- whole bias was detected, the survey methodologies did not possess significantly different consumer surplus estimates. Third, the sequential question bias resulted in the CVM whole bias was detected, the survey methodologies did not possess significantly different consumer surplus estimates. Third, the sequential question bias resulted in the CVM survey inflating consumer surplus estimates. Fourth, CA's multiple attribute evaluation framework compared to CVM's single attribute evaluation framework did not result in significantly different consumer surplus estimates.