Research

Abstract: This paper investigates how sibling gender affects substance use during adolescence. I utilize a sample of dizygotic twins from the National Longitudinal Study of Adolescent to Adult Health (Add Health), exploiting the random assignment of their sex at birth. This design helps me address two of the key methodological concerns in prior studies on sibling gender: (1) parental selection on fertility based on the gender composition of their children, and (2) the complex interaction between birth order and sibling gender. I find that among male adolescents, having a brother increases the likelihood of using cigarettes, alcohol, and marijuana by 9.9% points, 15.5% points, and 5.5% points, respectively. Regarding potential mechanisms, I find that the results are consistent with the channel of direct sibling interactions, but not with the channels of differential parental investment or family structure.

Abstract: This paper provides the first causal evidence on whether the adoption of the Diagnosis-Related Group (DRG) system harms high-risk patients, a question central to provider opposition against prospective payment systems worldwide. I exploit the mandatory DRG transition for Cesarean deliveries at advanced hospitals in Korea in July 2013, employing a difference-in-discontinuity (DiDC) design using restricted-access individual-level daily insurance claims data. The sample covers the universe of singleton births at advanced hospitals between 2011 and 2013, with health outcomes tracked for both mothers and children for up to five years. I find that within the analysis window, the mandatory DRG adoption did not change Cesarean-section rates or patient and hospital composition. Hospital stays for Cesarean deliveries fell significantly by 11.8 percent. Despite this reduction, health outcomes show no adverse changes over five years, including mortality, hospital utilization, Cesarean-related complications, and subsequent fertility. Notably, these null effects hold not only among average patients at advanced hospitals, who are already substantially higher-risk than the general obstetric population, but also among the higher-risk patients even within this group. These findings suggest that excessively long hospital stays prior to DRG adoption represented overtreatment without added health benefits even among high-risk patients.

Abstract: We use data on siblings near the minimum drinking age to provide causal estimates of peer effects in alcohol consumption, exploiting the increase in consumption of the older sibling in a regression discontinuity design. Preferred point estimates imply that younger sibling binge drinking decreases 27% of the mean at the cutoff. These negative reduced form spillover effects are concentrated in subgroups where the first stage discontinuity is largest, among siblings who are likely to spend more time together, and for measures of excessive alcohol consumption. While our results are somewhat imprecise, we argue that these patterns of heterogeneity are consistent with younger siblings learning from the costs of their older siblings' drinking behavior. Our results are directly interpretable as the effect of peer alcohol consumption, whereas most prior work identifies the effect of exposure to the peer. We explain how this distinction matters for policy.

Abstract: This study investigates whether changes in risk perception play a critical role in improving of preventive behaviors and health outcomes by examining the 2009 H1N1 influenza (or swine flu) pandemic in Korea. We employ a difference-in-differences estimation strategy by comparing the differential effects of the H1N1 outbreak on the confirmed cases of diseases which can be prevented by preventive behaviors (e.g., intestinal infections) and the cases of diseases which cannot (e.g., injuries). Using unique administrative data from South Korea’s National Health Insurance Service (NHIS), we find that the exogenous increase in health risk reduced the incidence of intestinal infections compared to the injuries during the H1N1 influenza outbreak. The reduction was the most substantial among children under five years of age, with a 25.4% decline in cases of intestinal infections relative to injuries. Our findings are robust across various alternative specifications. We provide suggestive evidence that active adoption of preventive behaviors is one of the channels underlying the unexpected decrease in diarrhea cases. The effects, however, faded away shortly after the end of the pandemic and did not last in the long run.

Abstract: In this paper, we introduce minimum cost spanning tree problems with multiple sources. This new setting is an extension of the classical model where there is a single source. We extend several definitions of the folk rule, the most prominent rule in the classical model, to this new context: first as the Shapley value of the irreducible game; second as an obligation rule; third as a partition rule and finally through a cone-wise decomposition. We prove that all the definitions provide the same cost allocation and present two axiomatic characterizations.