(Current Version: Jul. 23, 2025)
We investigate how information demand shapes belief formation using a randomized experiment on COVID-19 vaccines. We isolate the causal effect of information choice by orthogonalizing exposure from demand. We find that belief updating is substantially larger when subjects receive information they previously chose to read. This pattern shows that endogenous information demand plays a central role in how individuals process evidence, even when information exposure is exogenously assigned.
We utilized the 2016 Taiwanese General Election to see how partisan news can change voters' behavior. We find that when the voters read the information in favor of their own political ideology, they were more likely to be influenced. We also find a strong "new party" effect that the information about the new parties largely persuaded the subjects to vote for the new parties. We suggest this is due to the exposure of unfamiliar information that conveyed closer political ideology to the subjects.
(Preliminary version available upon request)
We conduct the canonical "balls-and-boxes" ("balls-in-urns") experiment, where we provided different types of reports about the signals. We found that people are further away from the Bayesian posterior when they receive the theoretically more informative report: the counts or the difference of the different colored balls, while receiving less the informative report of the ratio of different colored balls yields better guesses. We also elicit the subjects' willingness to pay for the reports prior to the updating tasks. We find the subjects are willing to pay more for the reports with which they perform better in the belief updating tasks.
(Preliminary version available upon request)
We study whether casting a first vote makes voters less responsive to new political information and fix their beliefs. Leveraging the 2021 Taiwanese Referendum, we randomly exposed young individuals—some eligible to vote and others just underage—to partisan information about the referendum’s propositions. This design allows us to isolate the effect of voting experience from self-selection. We find that eligible voters respond more to information on pro-environmental propositions: they reduce support for reopening a nuclear power plant and increase support for protecting algal reefs after reading the corresponding information. These treatments also make eligible voters more likely to stick with their original pro-environmental votes, suggesting a confirming effect of having voted. The heterogeneity in treatment effects is strongest among those who care most about environmental issues, indicating that the first act of voting can serve as an ideological assertion. Our findings show how early political participation can entrench beliefs—especially among individuals with strong identities—and thus can contribute to polarization.