Explorations of the use of random utility models in nonmarket benefit estimation

The random utility model is frequently used to examine the economic value of environmental quality changes. This dissertation extends this model by developing and applying procedures that allow: (i) examination of ex ante values of unknown attributes; (ii) incorporation and understanding of sources of preference heterogeneity, in particular, heterogeneity over tastes; and (iii) incorporation of endogenous feedback.;The first paper examines the hypothetical discovery of new attributes in recreation site choice models using joint revealed-stated preference data. The empirical application involved the discovery of aboriginal rock paintings along wilderness canoe routes in eastern Manitoba. A four year study of wilderness recreation trips included a stated preference experiment in which canoeists were asked if they would change site choices in response to the presence of two types of rock paintings: a "pristine" painting and another spoiled by human vandals. The resulting stated site preferences (with new attributes) were combined with the revealed site preferences (without the attributes) in the econometric analysis. The results suggest that preferences over the SP and RP models were not statistically different. Estimated welfare measures for the presence of "pristine" paintings range from ;The second paper develops a finite mixture approach to random utility models. In this analysis, attitudinal and other individual-specific information was used to develop probabilistic membership in latent classes. The probability of membership in these classes was estimated jointly with choice parameters using choice experiment data in a model that examined wilderness park choice in the Precambrian Shield region. Four classes were found in the data, and these classes reflected different motivations for taking wilderness trips and different behaviour in response to environmental quality changes. Two approaches for assessing the welfare of environmental changes were developed that took into account membership in each class, and the distribution of welfare impacts among the classes.;The third paper examines congestion in wilderness recreation areas. In this analysis, a random utility model was developed in which utility for one individual is a function of the utility of other recreationists. This model formalizes the notion that an attractive quality change for one recreationist is also attractive to others, thereby increasing the chances of congestion. Thus, in this framework, congestion can be considered endogenous feedback. To develop this notion in a random utility model, an instrumental variables approach was utilized. An ordered logit model of anticipated congestion was developed in which anticipated congestion levels were estimated as a function of individual characteristics and some environmental quality variables. This model was applied to canoeing in a system of wilderness parks and predictions of anticipated congestion were developed. These predictions were used as instruments in park choicc models and the results compared to other choice models in which reported congestion levels were used instead of the predictions. The comparisons revealed significant differences in welfare estimates from the two models.