New discrete choice methods for valuing environmental amenities: Theory and evaluation

Categorical data analysis has been an important tool used in recovering individual preferences because of the increased availability of detailed micro-data sets that describe people's choices about different types of environmental commodities. Despite its popularity, there is little empirical evidence revealing the reliability of welfare measure derived from discrete choice models. The central issue is the limited information regarding individual preferences that can be conveyed through discrete response data.;The purpose of this research is to develop alternative discrete choice valuation methods and evaluate the existing techniques for valuing changes in environmental quality. Issues related to welfare measurement are also addressed. Nonparametric functions, in specific, the cubic smoothing splines, are utilized to develop the new discrete choice methods. The new nonparametric discrete choice methods provide innovations on both model specification and the welfare estimator. Two groups of sampling experiments, in the cases of binary and multiple choices, are conducted to compare the new methods with the conventional parametric models based on the welfare measures derived from them. In addition, an empirical study of recreational demand at the Monongahela River basin area applying the new estimation techniques is also reported.;The simulation results suggest that in the dichotomous choice setting there is a jointness in which changes in nonmarketed resource influence individuals' behavior giving rise to some observable responses in their consumption of market goods. It appears that these differences across individuals lead to sufficient diversity in the measures of the true willingness to pay individuals would have for the nonmarketed good that a constant estimate derived from the nonparametric methods cannot perform as well as the parametric estimators. Equally important, it appears that the performances of both parametric and nonparametric methods are quite sensitive to the specific features of the assumed preferences. Although the nonparametric estimator does allow greater flexibility in that dimension, it does not necessarily assure an improved estimate. In addition, both estimators perform better when the preferences hypothesized to underlie individuals' choices for nonmarketed commodities are likely to be dominated by nonuse values.;In the multiple choice setting, it appears that the parametric and nonparametric estimators are quite comparable for all true underlying models. As in the binary choice case, the nonparametric estimator appears to have more variability across different hypothetical individuals that comprise the sample. Again, both estimators are sensitive to assumed preference structures. The precision of estimation decreases with the complexity of true preference structure. In addition, it appears that choice specifications significantly affect both estimators and omission of choices has contributed to substantial estimation bias.;Finally, nonparametric methods are applied to the Monongahela River Basin study to measure recreation benefits. Resampling experiments indicate that nonparametric estimates differ from the known results and are more efficient than the parametric estimates. In addition, models with different considerations of substitute sites lead to significantly different welfare measures.