Lai, Yufeng2023-09-192023-09-192021-04https://hdl.handle.net/11299/257114University of Minnesota Ph.D. dissertation.April 2021. Major: Agricultural and Applied Economics. Advisor: Chengyan Yue. 1 computer file (PDF); viii, 122 pages.This disseration investigates potential sources of bias in demand estimation and proposes bias reduction practices. The first chapter of this dissertation examines the external validity of consumer experiments by comparing consumer willingness to pay (WTP) derived from store scanner data to existing experiment-based findings. The finings are the WTP premium estimated for organic eggs is consistent with experimental results, while estimated WTP premiums for animal welfare attributes are significantly lower than experimental findings. The results highlight the importance of considering biases when estimating the price premium for animal welfare attributes in experiments.The second chapter combines List Experiment (also known as the Item Count Technique) and Choice Experiment to propose an alternative approach, i.e., List Choice Experiment, to alleviate Social Desirability Bias. Such bias is a prevalent concern when eliciting preferences for products associated with social norms, such as eco-friendly, fair-trade, animal welfare, common resources, and public goods. List Choice Experiment allows respondents to conceal their actual responses and avoid social norm judgments, while researchers can recover consumers’ preference parameters via a maximum likelihood procedure. Chapter 2 demonstrates the usefulness of List Choice Experiment using onsite experiment data and shows that the estimations with List Choice Experiment fit market data better than direct Choice Experiment and Inferred Valuation Method. The third chapter of this dissertation examines the numerical error in the random coefficient differentiated product demand model. In this chapter, the numerical error is charaterized in terms of the estimated market share error, which proposes a criterion to compare the exponentially transformed fixed-point iteration and the iteration in linear form. This study numerically demonstrates that the exponential iteration introduces larger error and gains no convergence speed than the linear form. In addition, the error characterization suggests that the mathematical program with equality constraint (MPEC) approach by Dubé, Fox and Su (2012) is equivalent in precision compared to the nested fixed-point (NFP) approach, only if the equality constraint bound estimated market share error as tight as the fixed-point iteration. However, under the recommendation of tight fixed-point tolerance, the precision level of the NFP is unattainable by the MPEC.enBias ReductionConsumer ExperimentNumerical ErrorThesis or Dissertation