Essays in Education, Peer Effects and Decision Making

Abstract

Peers play an important role in shaping behavior in many contexts. In this dissertation, I study the role of peer interactions on key educational outcomes. In Chapter One, I implement a field experiment to examine the role of social interactions in creating spillover effects. Spillover effects happen when individuals are indirectly affected by an intervention through exposure to other treated individuals. In the experiment, I randomly assign college students in an introductory statistics course to a low-cost behavioral intervention. Treated students receive advice and prompts to make exam study plans. I measure a naturally formed peer network and exploit the exogenous variation in exposure to the intervention in order to causally estimate spillover effects on study behaviors that are transmitted through study partners. I construct a simple social learning model. Additional behavioral evidence further supports the model and I show that the positive spillover effects on untreated students are mostly driven by treated partners who have high beliefs about the return to the applet usage. Surprisingly, I find circumstances under which social interactions reduce the treatment effect. Taken together, this paper provides causal evidence of spillover effects on behavior due to peer interactions and unpacks the complexities behind spillover effects. My results highlight that in networked environments, policy makers should take peer effects into consideration not only to correctly evaluate, but also to leverage social learning to maximize policy impacts. In Chapter Two, I study a natural experiment that randomly assigns students into study groups and estimate the effect of studying with peers of certain characteristics. I find little evidence that peers’ background academic performances have significant effects on the course final grade using the traditional linear in the mean model. I find that the group gender mix has an economically and statistically significant impact. In particular, being in groups with more female peers leads to an increase in the course grade for both female and male students. I exploit the course website’s log data, and find that one is more likely to download course materials when in more female groups. This is a plausible mechanism through which the gender mix affects the grades. I also find that studying with peers from another lecture section marginally improves one’s course grade. My paper therefore provides practical suggestions for assigning students into study groups. In Chapter Three, we use a longitudinal survey design and follow college freshman, in order to provide evidence for two separate mechanisms (homophily and influence) behind similarity in peers’ behaviors. This paper demonstrates these effects for the subtle (but broadly important) underlying economic preferences, rather than the observable but potentially domain-specific behaviors previously studied. Subjects participate in three waves of an online experiment where we elicit their social network using an incentive compatible mechanism and then measure participants’ levels of altruism, willingness to take risks, and willingness to delay rewards using diagnostic tasks. We find that subjects’ risk and time preferences are significantly positively correlated with the preferences of their friends, consistent with peer influence on preferences. Additionally, we find that changes in subject’s social networks are significantly influenced by social preferences. Subjects are more likely to add someone as a friend, and less likely to drop as a friend, the more similar their social preferences are.