Work in Progress
Long-term Effects of Growing Up with a Disabled Sibling (with Kevin Kloiber and Felicia Stokke)
We study the long-term consequences of growing up with a disabled sibling using comprehensive Swedish administrative data covering over 900,000 children. Employing two complementary identification strategies, an across-family comparison using Double Machine Learning and a within-family sibling design, we find that having a disabled sibling reduces 9th-grade educational outcomes by 0.05-0.07 standard deviations, lowers adult earnings by approximately 3%, and shifts occupational choice toward caring professions. Within-family comparisons reveal no differential effects among siblings, indicating that these costs are shared equally rather than disproportionately falling on the more-exposed child. This pattern points to household-level mechanisms, such as the reallocation of parental time and financial strain, as the primary channels. Sweden's transfer system buffers the earnings penalty at the bottom of the income distribution but leaves significant effects at the middle and upper portions of the income distribution. Mental and cognitive disabilities generate effects two to three times larger than physical disabilities.
Draft available upon request
Improving Risk Assessment and Treatment Choice in Medical Decision-Making using Risk Assessment Tools (with Corinna Hartung, Charles F. Manski, Joachim Winter and Amelie Wuppermann).
Physicians routinely predict patient outcomes and choose treatments under uncertainty. Evidence-based risk tools can support these decisions, yet they rely on a limited set of predictors and leave clinicians to adjust predictions using additional patient information. We study how physicians make such adjustments in cardiovascular risk assessment using the Systematic Coronary Risk Evaluation (SCORE) tool, which predicts 10-year cardiovascular mortality from five risk factors. Focusing on obesity as a risk modifier, we run an online vignette experiment with general practitioners (GPs) and cardiologists in Germany. Physicians are randomly assigned to assess risk using clinical judgment alone or with access to the SCORE chart, and we benchmark their assessments against risk estimates from longitudinal population health data.
We find that physicians without decision support substantially overestimate mortality risk given risk factors. Access to SCORE improves accuracy by about 12 percentage points, with gains concentrated among GPs, but overestimation persists. The improvements in risk accuracy translate only weakly into medication recommendations, but SCORE meaningfully reduces GPs’ referrals to cardiologists. Our elicitation design, which allows physicians to report either point estimates or intervals, reveals substantial uncertainty in their probabilistic beliefs about clinical outcomes.
Draft available upon request
The Organization of Birth Care and Health Outcomes
Hospitals are complex organizations in which clinical authority is distributed across professional groups with distinct training and clinical preferences. I study whether delegating clinical authority within a hospital obstetric team generates spillovers on patients not directly affected by the organizational change. I exploit the staggered introduction of Alongside Midwifery Units (AMUs) across 48 German hospitals between 2005 and 2023. AMUs grant midwives full autonomous responsibility for managing low-risk births within hospitals that also operate a standard consultant-led unit. Using the universe of German inpatient births and the staggered difference-in-differences estimator of Callaway and Sant'Anna (2021), I restrict the analysis to the approximately 75% of mothers who are clinically ineligible for AMU care and can therefore only be affected through spillovers on the consultant-led unit. AMU introduction increases unplanned (intrapartum) C-sections among these mothers by 2.3 percentage points, a 12.4% relative increase, while planned C-sections are unaffected. The shift is not accompanied by changes in severe maternal morbidity or fetal distress, suggesting that obstetricians intervene at an earlier threshold of clinical concern rather than in response to acute deterioration. The pattern is consistent with a resource-reallocation channel: the AMU's one-to-one midwife staffing requirement draws midwife capacity from the consultant-led unit, reducing the continuous midwifery support that facilitates physiological birth progression. These findings show that organizational reforms designed to benefit one patient group can generate unintended consequences for others through within-team resource reallocation.
Draft available upon request