UC Santa Barbara

This dissertation examines how people update their beliefs with different information structures under various environments. The first chapter studies how the self-selection of COVID-19 vaccine information affects people's belief updates on vaccine effectiveness and their preferences for vaccines. Rational information acquisition theory predicts people select the more informative information; thus people's beliefs will be more persuaded by the information they select. We test the prediction in a critical real-world context—information about COVID-19 vaccines. We conduct an online experiment in Taiwan where the subjects select information about COVID-19 vaccines, and then the subjects update their beliefs about the vaccine effectiveness and references of vaccines. As our design distinguishes different stages of the rational acquisition framework, it allows us to diagnose the underlying mechanism of the theory. Our empirical findings demonstrate evidence that people's information acquisition generally coheres with the rational theory framework predictions; that is, people choose information when the information is more likely to alter their decisions. We show that our subjects' beliefs change more when they see the information they select. We also find evidence of change in vaccine preferences and choices after they receive the information they select, which further suggests that the subjects follow the rational information acquisition framework. Chapter 2 studies whether the first vote changes how people's voting decisions after seeing information. The first vote can be a crucial political assertion that causes people to stick to their beliefs even after reading the information. In this study, we examine the interaction of voting experience and the persuasiveness of information. To control the potential endogeneity arising from the self-selection to vote, we use eligibility as the random cutoff, as the ineligible voters can never select to vote. We utilized the 2021 Taiwanese Referendum to see whether new information heterogeneously impacts people’s voting choices between eligible and ineligible voters. We find that the eligible subjects become less supportive when they see negative information about nuclear power plants and more supportive when they see positive information about algal reefs. The treatments make the eligible nay voters in nuclear power plants and yea voters in algal reefs stick on their votes more, suggesting the confirming effect of the action of voting. The heterogeneity between eligible and ineligible subjects is more profound among the subjects who care about environmental issues the most, which indicates that the first vote can be an active assertion to environmental voters. In the third chapter, we explore how individuals use and value different statistical features in a balls-in-boxes experiment. In contrast to the literature, people in the real world are usually exposed to summarized information (e.g., proportion) instead of the raw data they can access in the laboratory. In our belief updating experiment, we experimentally investigate how individuals use and value different statistical characteristics of realized signals, referred to as sample features. We find that people align the closest to the Bayesian updating when they see Proportion. We also see people prefer Proportion same as Count or Sequence, even though Proportion is less informative than the other two features. Furthermore, we find that the belief updates are closer to the Bayesian benchmark when the subjects use their preferred sample feature, implying that people are sophisticated about the subjective value of information.