Research

Abstract: This study explores the relationship between women’s representation in political power and school closures during the COVID-19 pandemic. Using a cross-country dataset in Europe, we document a striking negative relationship between the share of female members in national governments and school closures. We show that a one standard deviation increase in female members of national governments is associated with a significant reduction in the likelihood of school lockdowns by 24% relative to the average share of school closures. This result is robust to an extensive set of sensitivity checks. We attribute this pattern to a higher awareness of female politicians about the potential costs that school closures imply for families, in particular working mothers with young children.

Abstract: This study explores the enduring influence of Black peers in school on their classmates’ future political preferences. Using comprehensive panel data from a representative sample of US school students followed over time, I leverage quasi-random variation in the share of Black peers across cohorts within a school. The results suggest that a higher proportion of same-gender Black peers is associated with a lower likelihood of holding conservative political preferences in the future. I focus on political preferences in a period in the United States, when race was a highly decisive issue, i.e., the election year of Barack Obama, thus highlighting the importance of topic salience on peer effects. I provide suggestive evidence for the argument that the re-categorization of different racial groups into a shared school-belonging identity may be one of the forces behind these results.

Abstract: This paper investigates how peer risk preferences influence individual engagement in risky behaviors, particularly crime. Using the Add Health dataset, we construct peer shares of risk attitudes but face missing data due to survey structure. To address this, we employ Synthetic Minority Over-sampling Technique (SMOTE), a machine-learning method recently applied in crime prediction (Campedelli et al., 2024). We predict risk attitudes for individuals missing direct responses and construct peer risk shares from these estimates. Our empirical strategy leverages within- school, across-cohort variation to estimate causal peer effects. Our results contribute to the peer-effects literature and provide novel evidence linking risk tolerance to criminal activity (Chevalier and Marie, 2024). The study also highlights the use of ML for causal inference in economics.