Causal Inference: Applications in Education, Violence, and the Scientific Process

Abstract: This thesis consists of five self-contained chapters. In chapter 1, I analyze how replications affect how underlying studies are cited. My results provide no evidence that citations changed in response to either successful or failed replications, with potentially worrying implications for the self-corrective ability of the scientific process. In chapter 2, we study whether Black Lives Matter protests had an effect on voter registrations in the run-up to the 2020 US presidential election. We find no robust evidence for such an effect and provide some anecdotal evidence that increased news coverage might have reduced the importance of experiencing local protests first-hand. In chapter 3, we investigate how online education during the Covid-19 pandemic affected student test scores in higher education. Comparing student performance across exam questions associated with either online or in-person instruction, we find no evidence for differences in learning outcomes across modes of teaching. In chapter 4, we examine the extent to which mass shooting events induce community-wide changes in sales of alcohol and tobacco, highly accessible substances frequently linked to coping with traumatic events. Leveraging a specification curve design, we find no robust support for this hypothesis. Finally, in chapter 5, we study the effect of seeding procedures in European club football on a team’s probability of advancing. Despite a marked drop in average opponent strength, we find no evidence that teams marginally placed in a higher pot performed better in the tournament.