Does Paid Family Leave Prevent Infant Abusive Head Trauma? Evidence from Staggered Policy Adoption in the United States

Abstract

Background: Abusive head trauma (AHT) is the leading cause of fatal child maltreatmentamong infants under one year of age. Paid family leave (PFL) policies may reduce AHT by alleviating caregiver financial stress and extending parental time with newborns during the highest-risk postpartum period. Yet existing evidence relies on simple pre-treatment and post-treatment comparisons or two-way fixed effects (TWFE) regressions that can yield biased estimates under staggered treatment adoption. Methods: We use Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases from seven states (California, New Jersey, New York, Rhode Island, Maryland, Florida, and North Carolina) spanning 2003–2021, with four states adopting PFL at different times (California in 2004, New Jersey in 2009, Rhode Island in 2014, New York in 2018). Our primary estimator is the Borusyak–Jaravel–Spiess (BJS) imputation approach, which avoids the forbidden-comparison problem inherent in TWFE by imputing counterfactual outcomes using only never-treated and not-yet-treated observations. We implement both pooled and state-specific analyses; the latter use restricted donor pools retaining cross-treated-state observations only from pre-treatment periods. A TWFE event study serves as a sensitivity analysis. Covariates include state-level unemployment rates, poverty rates, minimum wages, and per-capita Earned Income Tax Credit. Standard errors are clustered at the state level throughout. Results: Pre-trends tests across event-time periods t = −3 through t = −1 are consistent with parallel trends. The pooled BJS estimator yields a negative static average treatment effect on the treated (ATT), indicating that PFL is associated with a reduction in infant AHT hospitalization rates per 100,000. Dynamic event-study estimates show the effect emerging in the first post-treatment period and persisting across the observed horizon (t = +1 through t = +5). State-specific analyses for California and Rhode Island show protective effects consistent with the pooled finding. New York’s available post-treatment window (2018 to 2021) overlaps with the COVID-19 pandemic and a phased program rollout, precluding clean identification of a protective effect for that state. Results are robust to population weighting and to the event study specification. Conclusions: Paid family leave policies are associated with reductions in abusive head trauma hospitalizations among infants, with effects robust to heterogeneous-treatmenteffect- consistent estimation. These findings strengthen the evidentiary basis for PFL as a child maltreatment prevention tool and underscore the importance of using modern staggered-DiD estimators in policy evaluation.