Assessment of individual preferences is of interest to many disciplines, includ ing economics, marketing, business, and health.
Discrete choice methods involve situations wherein individuals evaluate several policy alternatives and then choose one alternative over others, expressing their willingness to make tradeoffs among alternative attributes. A discrete choice model is then used to recover individuals' preference functions from these responses. In spite of the popularity of discrete choice methods, the reliability and accuracy of these methods in providing correct assessment of individual preferences are often questioned. This dissertation investigates ways to improve the discrete choice method in recovering this information with higher precision. To achieve this, I focus on two essential components of discrete choice experiment implementation: the design process and the econometric analysis process.
Manuscript 1 and Manuscript 2 of this dissertation research focus on the discrete choice experimental design process. Here I specifically address designing stated choice methods, a specific form of discrete choice methods, for valuing public goods.1 Stated choice methods have become popular among researchers for their ability to value a range of public goods and services and for their ability to include a range of attributes in the stated choice questions. Stated choice typically involve surveys, which may either take a hypothetical form, wherein respondents are not actually expected to pay for their choices, or a real-money form, wherein respondents make payments for their choices. The incentive properties in both hypothetical surveys and real-money surveys, continue to be a subject of debate. Hypothetical surveys may be subject to hypothetical bias, which could overestimate values of public goods because respondents might not treat monetary costs in hypothetical surveys the same way they treat such costs in their actual daily transactions. On the other hand, real-money surveys may be subject to free-rider bias, which could underestimate the values of public goods because respondents might recognize their opportunity to benefit from the financial contributions of others. These measurement biases may lead to incorrect welfare measures and, therefore, suboptimal policy decisions.
Researchers have developed various theoretical and econometric methods to reduce or correct for these measurement biases. I develop a dominant strategy incentive compatible mechanism for designing stated choice surveys in order to elicit individuals' true preferences, eliminating the incentive to free ride in realmoney choice questions. I adapt Clarke's (1) pivotal mechanism to stated choice surveys in order to motivate truth telling. I present theoretical proofs of the incentive compatibility of this mechanism for a binary choice case and a multiple alternative choice case. I design and conduct induced-value experiments to verify if respondents indeed adopt their dominant strategies while faced with the incentive compatible mechanism. I also compare the dominant strategy equilibrium property of my proposed mechanism with that of alternative value revelation mechanisms.
From this analysis, I find that this proposed mechanism performs quite well for the binary choice case but fails to perform as well in the multiple alternative case. However the study provides a better understanding of individuals' incentives behind incorrect revelation of demand for public goods.
Manuscript 3 deals with the discrete choice modeling process. For over a decade, the research took a direction of accounting for the heterogeneity of the populations, thus providing estimates of willingness to pay (WTP) distributions. One such method, the random parameters logit model have became quite popular in obtaining WTP distributions from the marginal utility distributions for heterogeneous populations. However, most random parameters logit model applications suffer from one limitation because, they keep the marginal utility of income constant to obtain the distributions of WTP. I suggest that shifting the distributional assumption from marginal utilities to the welfare measures themselves directly yields the distributions of WTP. An empirical application reveals that this proposed model and the random parameters model yield similar mean WTP, but welfare measure are more readily interpretable in the proposed model.
Finally, Manuscript 4 sums up the findings of this dissertation and discusses the contribution, the limitation and future direction of this research. (Abstract shortened by UMI.)
1Stated choice methods are also useful for valuing goods and services that are new in the market or when there is insufficient variability in actual choices to allow analysis of the attributes of interest. The focus here is only on public good valuation.