Research

Trust and Information Demand With Biased Preferences: A Prediction Game Experiment

Concerns about bias distorting the quality of information from even the most informed experts are growing with digital media, so understanding the mechanics of how people demand and use information in a biased setting can yield valuable insights. I study exactly such with a laboratory experiment where participants bid to observe either a private signal or a prediction made by an ‘expert’ to inform predictions about a dichotomous state. In the control, payment schemes are symmetric between the two states. Treatments introduce small additional transfers to either the participants or experts if they predict a certain state. In the control, participants do not heavily discount the experts’ predictions, but under-utilize information that is inconsistent with their initial information set. When the experts are treated with the additional transfer, participants’ value of experts’ predictions is consistently reduced while their utilization is unaltered. However, the participants who are treated with the asymmetric transfer display a significant difficulty in evaluating the information’s usefulness. They overvalue information when their existing information set already indicates the higher-payoff state, then under-utilize realized information inconsistent with the state. These combined results suggest participants are significantly better at parameterizing external bias and its effects, as opposed to internal bias. (LINK)

Valuing Accounts on Social Media With Biased Beliefs

Social media has revolutionized the way people consume and interpret news. Anyone can follow an account whose posts are deemed valuable. But doing so may segregate people into ‘echo chambers’ of isolated thinking. To better understand what account characteristics matter to people, I administer a survey asking subjects to evaluate different versions of an X account profile while varying the account’s political affiliation, credentials, and number of followers. I find four results. First, subjects are wary of political bias - even if it is consistent with their beliefs - and credentials are insufficient to overcome bias concerns. Second, followers are a valuable resource for all accounts, regardless of account credentials and party affiliation. Third, subjects who appreciate echo chambers do not need additional verification to value congruent political accounts. Finally, subjects who are concerned they have believed fake news are relatively critical of politically opposite accounts. These results contribute to two discussions. When explicitly asked about them, people are aware of political bias and use a variety of account characteristics when assessing value. However, there are still learning asymmetries on social media: people may lose the need to validate same-typed accounts or become mistrusting of different-typed accounts, leading to the filtering of dissenting information. (LINK)

Responding to Expertise in Information Cascades

Information cascades form when people value the information derived from the decisions made by others more than their own private news, causing them to ignore it and follow the masses. This phenomenon is of policy and research interest because social media influencers have the ability to initialize multiple cascades across networks, a concern because it results in an equilibrium with imperfect information. I study the role that signaling expertise plays in information cascade formation and propagation by exogenously sorting subjects on ability and historical performance in a laboratory experiment. Signaling ability results in a significant increase in cascading behavior and aggregate welfare. Signaling historical performance does not affect behavior. In an era where expertise is questioned at a heightened rate, the results provide insights into how people form beliefs about others and the mechanics behind evaluating the quality of information. (LINK)

Free Riding Toward Personal Protection: Relating Parental Cooperative Behavior With Vaccine Hesitancy (with Aaron Enriquez and Mariah Ehmke)

Small urban clusters and rural communities, which have historically had low vaccination rates, are especially vulnerable to healthcare system overloads. We conducted a study in which parents from such locations played a Voluntary Contribution Mechanism experiment and then answered survey questions about influenza vaccinations. We observed parents’ cooperative actions in the experiment (i.e., contributions to a shared group account) and their relationship with flu vaccination decisions for themselves and their child. This article classifies different player-types based on parents’ propensities to cooperate and react to their partners’ actions, including “free riders” (keep the majority of tokens), “contributors” (contribute the majority of tokens), and “conditional cooperators” (adjust contributions based on their partner’s actions). We also control for the intensity of reciprocation among all players. We find that free riders and parents who tend to reciprocate are the most likely to vaccinate. Our result about free riders is a departure from previous literature. The findings shed light on behavioral motives behind people’s vaccination decisions. Policies that amplify free riding and reciprocation may increase vaccination rates, which would be critical for mitigating the damaging effects of COVID-19 and other preventable diseases. (LINK)

A Gamble of Life and Death (with Gregory Marchal and Mariah Ehmke)

Vaccine hesitancy in rural communities relates to outbreaks of vaccine preventable diseases like influenza, burdening healthcare systems with hospitalizations and deaths. Research on largely urban populations has shown that parents who mistrust the healthcare system perceive uncertainty on information related to vaccine efficacy and risks (i.e. side effects), leading to a decline in vaccine uptake. Our research objective is to test the role of parents’ economic risk preferences and vaccine information ambiguity in their decision to forego influenza vaccinations for their children and themselves. We collected data using a lab-in-field economic experiment to measure parents’ constant relative risk aversion coefficient (CRRA) and a survey to obtain data on their vaccine beliefs, practices, and information sources. The data were then analyzed using a logit model regression to test the role of economic risk preferences and vaccine information ambiguity in their influenza vaccination decisions. We control for trust in the healthcare system, community characteristics, and personal demographic information in the model estimation. We find parents’ influenza vaccination decisions are significantly dependent on their ambiguity aversion, but not their risk aversion CRRA measurements. Parents who perceive greater uncertainty in the risks of vaccines relative to the risks of diseases tend to vaccinate their children for the flu at lower rates. This relationship exists after controlling for trust in the healthcare system, suggesting that policies addressing the perceived ambiguity in the vaccination decision independent of healthcare trust may be most effective to reduce hesitancy. (LINK)