Comparison of the Contingent Valuation Method and the Stated Choice Model for Measuring Benefits of Ecosystem Management: A Case Study of the Clinch River Valley, Tennessee

Several recent studies have compared the stated preference contingent valuation method (CVM) and discrete choice analysis for non-market value. The studies suggest that values derived from the two different methods differ because of the information presented in the contingent market. One explanation is that in the CVM consideration of substitutes typically amounts to a statement reminding the respondent of a budget constraint. In the choice analysis, consideration of substitutes is part of the survey design and the decision process. An alternate explanation is that information on the suite of complementary changes is explicitly recognized in the choice models and is assumed to be constant in the choice analysis. Another difference between the two analyses is experimental aspects; the choice model has an iteration format in questions, but not the CVM.

The subject of this dissertation pertains to issues of substitutes and experiment aspects between the CVM and discrete choice model, comparing the values of environmental quality changes in the upper Clinch River, Tennessee. In the first test, three sample surveys are created: a choice model survey; a standard CVM survey where complements to the policy change are not considered; and a “rational expectations” version of contingent valuation where the complements to the policy change are explicitly stated in the survey. In the second test, an additional survey is created for examining the experimental differences between the two models. It is a CVM survey containing multinomial questions like the choice model.

These two tests provide evidence that the welfare estimates derived by the choice model are much higher than the corresponding CVM, even though questioners for both models are conditioned to provide either the same information on substitutes for a policy or the same number of multinomial questions between the two models. Another finding is that people may not be sensitive to embedding, but depend upon attributes in questions. These findings suggest that when individuals face the two different formats, such as referendum and choice formats, the psychological aspects of individual decision behavior are carefully concerned.