Association for Public Policy Analysis & Management

Classical economic theories generally model that people could correctly update their beliefs as Bayesians. It is costly yet prevalent that people fail to learn from objective information on important issues. To explain such learning failure, one of the theories proposed by behavioral economists is that people have motivated beliefs (Bénabou, 2015, Bénabou and Tirole, 2011, 2016). While a series of lab experiments have demonstrated the potential of the theory (Eil and Rao, 2011; Mobius et al., 2011; Di Tella et al., 2015), there are large gaps in the evidence that whether/how people can distort objective information (vs. subjective opinions) in high-stakes decisions.

We combine a large-scale high-stakes natural experiment with a novel field experiment to provide the first piece of empirical evidence in a real setting. We show that such learning failure can be partly explained by motivated cognition. We focus on a unique empirical setting, where Muslim students were about to take the high-stakes College Entrance Exam (CEE) during the month of Ramadan, due to the occasional overlap of these two events.

Using a Difference-in-Differences estimator with administrative data on all students of high school graduation cohorts between 2011 and 2016, we find that taking the exam during Ramadan has a quantitatively important negative impact on the exam performance of Muslim students, as compared with their non-Muslim counterparts. To help students make informed fasting decisions, we invited well-respected Chinese Muslim leaders to grant explicit exemptions to delay the fast until after the exam. We then showed all Muslim students the same information on the cost of taking the CEE during Ramadan (in terms of exam score loss).

We conducted a randomized controlled experiment in a large Muslim high school in China. By randomly providing the exemptions to some of the Muslim students who were about to take the CEE (during Ramadan) in 2018, we created experimental variation in the stringency of religious practices: some students believed that they had to fast during the exam, while other students thought that the fast could be delayed until after the exam.

Using a visual-based survey module, we find that students who thought they had to fast during the exam in 2018 (control group) show patterns of motivated cognition: they distort the objective information in the graphs by underestimating the negative impacts of taking the exam during Ramadan, but as much as 40% of such cognitive bias is eliminated among the students who received an exemption (treatment group). Using a list experiment, we provide suggestive evidence that alleviating motivated cognition makes students better informed about the costs of Ramadan, and thus more willing to delay the fast until after the CEE.

With a direct and compelling test for the theory of motivated cognition bias, this paper contributes to the growing strands of literature on motivated beliefs, visual bias, religious participation, and Ramadan fasting. This paper also has essential policy implications about the importance of identifying and tackling the psychological motives in increasing the effectiveness of informational policy interventions.