Selected Working Papers
Misdated at Birth: Measurement Technology and Threshold-Based Care (Job Market Paper - JMP)
In many institutional contexts, standardized scores and classifications can determine eligibility for treatment and access to healthcare. This article analyzes how measurement technologies can affect the allocation of care or lead to inequities. More precisely, it examines the impact of changing the method of dating pregnancy: from the last menstrual period to ultrasound dating. Ultrasound dating standardized and revolutionized the calculation of fetal gestational age and, consequently, the reclassification of pregnancies as preterm, term, or post-term. However, such reclassification can be prone to measurement error and generate inequities in hospital treatment, as it determines the type of care newborns receive after birth. Using integrated Swedish registry data from 1973 to 2017 and an instrumental variable design based on the lagged intensity of ultrasound dating in hospitals, I demonstrate that ultrasound dating alters recorded gestational age and reclassifies pregnancies across clinically relevant thresholds for determining hospital care. The effects are concentrated on neonatal health indicators and postnatal healthcare utilization, whereas long-term effects—such as academic performance—appear only in selected specifications, such as in prolonged pregnancies. However, in municipalities where routine ultrasound dating was performed earlier in pregnancy, the adverse reclassification pattern is mitigated, and the models favor ultrasound. The results demonstrate that ultrasound can alter the allocation of care, making measurement error relevant to public policy formulation.
Accurate Due Date Prediction without Ultrasound: A Machine Learning Approach Using Maternal Characteristics (Working Paper)
With: Mikael Elinder and Oscar Erixson. Draft available upon request!
Accurate prediction of gestational length is vital for prenatal care and obstetric planning. Traditional methods, such as Naegele’s Rule (NR) and ultrasound (US), assume fixed gestation lengths and overlook variation due to maternal characteristics. We develop machine learning (ML) models to improve estimated due dates (EDD) in contexts without access to ultrasound. Using data from over 600,000 pregnancies in the Swedish Medical Birth Register, our models incorporate readily available maternal factors, including last menstrual period, parity, age, height, weight, and health behaviors. We compare the predictive accuracy of our ML models with NR and US. While US remains the most accurate method, the ML models significantly outperform NR and narrow the accuracy gap between NR and US. For second-time mothers, a model leveraging first-pregnancy data even surpasses both NR and US. These findings suggest that ML can provide a low-cost, scalable alternative for EDD prediction, particularly in settings where ultrasound is inaccessible or impractical.
Do Pregnancy Complications Shape Sisters’ Fertility Decisions? Evidence from Within-Family Exposure (Working Paper)
With: Mikael Elinder and Oscar Erixson.
This paper studies whether pregnancy complications affect not only the fertility decisions of the woman who directly experiences the complication, but also the subsequent fertility behavior of her sisters. The empirical strategy uses difference-in-differences models to separate the direct effect of experiencing a complication from the indirect effect of being exposed to a sister’s complicated pregnancy. Preliminary results suggest that complications reduce subsequent fertility mainly among the women who directly experience them, while the evidence for spillover effects on sisters is weaker. This points to the importance of distinguishing between direct health shocks and family-information spillovers when studying fertility responses to adverse pregnancy events.
Shifting Access to Care: Evidence from Hospital Closures and Openings in Brazil (Working Paper)
With: Valdemar Neto and Soraya Roman.
This paper studies how changes in hospital access affect newborn health outcomes in Brazil. Using administrative data on births and hospital availability, I identify municipality-level access shocks based on the loss or gain of important hospital routes and estimate their effects using a staggered difference-in-differences design with event-study specifications. The results suggest that hospital closures worsen newborn health, while hospital openings partly offset these effects. The project contributes to the literature on service delivery, spatial inequality, and human capital in developing countries.
Publications and Policy Reports (Pre-Doc)
Inequalities in the geographic access to delivery services in Brazil (BMC Health Services Research)
With: Valdemar Pinho Neto, Cecilia Machado, Felipe Lima, and Soraya Roman. (Link to the article!)
Evaluation of data on deaths due to COVID-19 from the databases of the Civil Registry (RC-Arpen), SIVEP-Gripe, and SIM in Brazil in 2020 (Cadernos de Saúde Pública)
With: Ricardo Guedes, Cecilia Machado, and Marina Aguiar Palma (Link to the article!)
The Relationship Between Disarmament and Crime Rates in the State of São Paulo (In Portuguese)
With: Lucas Adriano Silva, Pedro Rodrigues Oliveira, and Viviani Silva Lírio (Link to the report!)
The Occurrence of Crimes During the COVID-19 Pandemic: Initial Investigations for the State of Rio Grande do Sul (In Portuguese)
With: Ana Cecília de Almeida, Felipe Nathan Ferreira dos Santos, and Bruno Truzzi (Link to the report!)
Gender and Racial Inequalities in the Health Sector: An Intersectional Analysis (In Portuguese)
With: Marcos Duarte, Alexandre Almeida, Davi Costa, Soraya Roman, and Valdemar Pinho Neto (Brazilian Public Call No. 22/2023)