State Abortion Restrictions and Infant Mortality in the United States

State Abortion Restrictions and Infant Mortality in the United States | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article State Abortion Restrictions and Infant Mortality in the United States Hugo Douma This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8810808/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Importance Restrictive abortion policies may have unintended consequences for perinatal outcomes, yet evidence on their impact on infant mortality remains limited. Objective To assess whether complete state-level abortion bans are associated with changes in neonatal and postneonatal mortality in the United States. Design Population-based serial cross-sectional study (1999–2024) using nationwide vital statistics data. Causal effects were estimated using synthetic difference-in-differences, interaction-weighted event-study models, and Bayesian hierarchical modeling. Setting and Participants All live births in the United States across all 50 states and the District of Columbia. Twelve states implementing complete abortion bans were compared with pre-ban trends and states without bans. Main Outcomes Neonatal mortality (0–27 days) and postneonatal mortality (28–364 days) per 1,000 live births. Results Complete abortion bans were associated with significant increases in neonatal mortality but not postneonatal mortality. Synthetic difference-in-differences estimates showed an increase of 0.281 neonatal deaths per 1,000 live births (95% CI, 0.128–0.430). Bayesian hierarchical models indicated increases emerging one year post-ban (0.050, 97%–3% HDI 0.006–0.090), rising at two years (0.132, 97%–3% HDI 0.089–0.178), and remaining elevated at three years (0.127, 97%–3% HDI 0.070–0.183). Event-study analyses showed increased neonatal mortality three years post-ban (0.303, 95% CI 0.033–0.573). No consistent associations were observed for postneonatal mortality. Findings were robust across multiple sensitivity analyses. Conclusions Complete state-level abortion bans were associated with sustained increases in neonatal mortality, indicating potential adverse effects of restrictive abortion policies on early-life outcomes. Figures Figure 1 Figure 2 Figure 3 Key Points Question: Are state-level abortion bans associated with changes in neonatal or postneonatal mortality in the United States? Findings: In this population-based study of all live births in 50 U.S. states and the District of Columbia from 1999 to 2024, neonatal mortality increased in states after implementation of complete abortion bans, while postneonatal mortality was unaffected. Estimates were consistent across multiple causal inference methods. Meaning: Complete abortion bans may lead to sustained increases in neonatal mortality, highlighting important public health consequences of restrictive reproductive policies. Introduction Infant mortality remains a critical indicator of population health and health equity, with profound implications for long-term human capital and societal well-being. In the United States, rates of infant mortality have historically exceeded those of peer nations 1,2 and exhibit persistent racial and socioeconomic disparities 3–6 . Extensive evidence links unintended pregnancy to adverse maternal and infant outcomes, including preterm birth, low birth weight, and perinatal depression, underscoring the influence of reproductive health access on early life trajectories 7 . Yet the population-level impact of restrictive abortion policies on infant mortality remains incompletely understood, particularly when assessed with rigorous causal inference approaches that can account for the staggered adoption of policy interventions. The U.S. Supreme Court’s decision in Dobbs v. Jackson Women’s Health Organization , which overturned Roe v. Wade , enabled states to enact complete or near-complete abortion bans, profoundly altering reproductive health access. Early evidence suggests these restrictions may have measurable impacts on infant health. Bayesian panel analyses of states implementing complete or six-week abortion bans from 2012 to 2023 indicate higher-than-expected infant mortality following policy adoption, with disproportionate increases among non-Hispanic Black infants and deaths due to congenital anomalies 8 . Other studies examining post- Dobbs birth outcomes highlight potential increases in congenital anomalies and persistent maternal health disparities in abortion-restricted states, though direct evidence remains limited 9,10 . Prior research also demonstrates that restrictive abortion policies may exacerbate preexisting inequities in maternal and infant health, particularly in states with higher baseline mortality. Despite these insights, significant knowledge gaps remain. Most prior studies rely on single modeling approaches, short study windows, and limited adjustment for socioeconomic covariates, reducing causal interpretability and the ability to examine heterogeneity by population subgroups. Moreover, no previous studies have leveraged the full span of available U.S. vital statistics from 1999 through 2024 alongside rich state-level covariates from sources such as the American Community Survey. The United States presents a uniquely informative case: it combines recent, staggered abortion restrictions with high-quality, longitudinal public health data, allowing for robust causal inference and fine-grained subgroup analyses. Findings from the U.S. experience can inform understanding of the potential health consequences of restrictive reproductive policies in other settings, as many countries consider changes to abortion legislation. To address these gaps, we use data from all 50 U.S. states and the District of Columbia across 25 years, integrating detailed socioeconomic covariates and employing a triangulation of causal inference methods, including synthetic difference-in-differences, interaction-weighted event-study (using the Sun & Abraham method), and Bayesian hierarchical models, to estimate the association between state abortion bans and neonatal and postneonatal mortality. This approach allows us to rigorously assess the timing, magnitude, and specificity of policy effects, providing insights not only for the U.S. but for global public health policy regarding reproductive rights and infant health. Methods Study Design and Data Sources We conducted a longitudinal, state-level panel study of infant mortality in the United States from 1999 through 2024. Annual state-level counts of neonatal deaths (death within 28 days of birth), postneonatal deaths (death between 28 days and 1 year), and live births were obtained from the Centers for Disease Control and Prevention (CDC) Wide-ranging Online Data for Epidemiologic Research (WONDER) database. Maternal health and birth characteristics were derived from CDC Natality Files (1999–2024) 11,12 . State-level socioeconomic characteristics were obtained from the American Community Survey (ACS) 1-year and 5-year estimates (2005–2024) 13 . State-level abortion ban legislation and effective dates were extracted from the Guttmacher Institute 14 and the Kaiser Family Foundation (KFF) 15 . The two sources were reconciled, and no discrepancies were identified (see Appendix 3 for the full dataset). Data was accessed on December 28th and December 29th 2025, we did not have access to any information that could be used to identify patients. All datasets were harmonized to a common state–year structure. Socioeconomic covariates were unavailable for years prior to the introduction of the American Community Survey (ACS) and were therefore not included for those periods. Exposure The exposure of interest was the implementation of full state-level abortion bans. States were classified as exposed beginning in the year in which the abortion ban legislation took effect. Variation in the timing of policy adoption across states enabled causal inference using staggered difference-in-differences and event-study designs. Outcomes The primary outcomes were state-level neonatal and postneonatal mortality rates, calculated as deaths per 1,000 live births. Outcomes were analyzed separately due to their distinct etiologies and policy sensitivity. Covariates We adjusted for time-varying state-level covariates capturing maternal health, birth complexity, demographic composition, and socioeconomic context. Several candidate covariates exhibited substantial multicollinearity. To improve model stability and precision, we conducted variance inflation factor (VIF) analysis and constructed three standardized indices summarizing correlated domains. The full list of covariates and indices used can be found in Appendix 1. Statistical Analysis We estimated treatment effects using multiple complementary panel-data approaches appropriate for staggered policy adoption. Our primary analysis used Synthetic Difference-in-Differences (SDID) 16,17 , which combines outcome modeling from difference-in-differences with data-driven weighting from synthetic control methods to improve balance in pre-treatment trends. Models included all covariates described above, with state and year fixed effects. Standard errors were clustered at the state level. To assess dynamic treatment effects, we estimated interaction-weighted event-study models using the estimator proposed by Sun and Abraham 18 , which is robust to staggered and heterogeneous treatment effects across cohorts. Models included covariates, fixed effects, and clustered standard errors. We further estimated a Bayesian hierarchical event-study model 19 with partial pooling of state-specific treatment effects, allowing for cross-state heterogeneity while borrowing strength across the panel. Assumptions To support causal interpretation of our findings, we explicitly outline the identifying assumptions underlying each analytic method. While all models adjust for maternal health, birth complications, socioeconomic indices, and racial composition, each approach relies on different assumptions for valid inference. Synthetic Difference-in-Differences (SDID) Conditional Parallel Trends: In the absence of abortion bans, the weighted combination of control states would have experienced the same trend in neonatal and postneonatal mortality as treated states. No Spillovers: Policy adoption in one state does not affect infant mortality in other states. Correct Covariate Adjustment: Maternal health, birth complications, socioeconomic indices, and racial composition adequately capture confounding factors affecting mortality. Stable Outcome Measurement: Mortality reporting is accurate and consistent across states and years. Sun and Abraham Interaction-Weighted Event Study No Anticipation: States do not systematically adjust behaviors affecting infant mortality prior to the enactment of abortion bans. Cohort-Heterogeneous Treatment Effects: The estimator allows treatment effects to vary across states and adoption cohorts; identification assumes that, conditional on covariates and fixed effects, variation in adoption timing is as good as random with respect to residual pre-treatment trends. Parallel Pre-Trends Within Cohorts: Within each adoption cohort, treated states follow parallel trends to untreated states in the pre-treatment period. Bayesian Hierarchical Event-Study Model Partial Pooling Assumption: State-specific treatment effects are drawn from a common hierarchical distribution; variation across states is exchangeable and informs the posterior estimates. Correct Model Specification: The hierarchical model accurately captures temporal dynamics and covariate effects. Conditional Ignorability: Given covariates and state/year effects, policy adoption is independent of unobserved determinants of neonatal and postneonatal mortality. Reliable Data Reporting: Mortality and birth covariates are measured without systematic bias. Sensitivity and Robustness Analyses We conducted extensive robustness checks, including placebo interventions assigned to pre-treatment periods, leave-one-state-out analyses, and alternative covariate specifications. We additionally evaluated sensitivity to covariate inclusion through systematic permutations of index components. Consistency of effect direction, magnitude, and timing across methods and outcomes was emphasized. Software All analyses were performed using Python version 3.11.9, using pandas and numpy for data management, statsmodels and linearmodels for frequentist panel estimation, and PyMC for Bayesian hierarchical modeling. Full VIF diagnostics and additional methodological details are provided in the Appendix 1. Results Synthetic Difference-in-Differences Estimates We applied the Synthetic Difference-in-Differences (SDID) method, which combines features of synthetic control and traditional difference-in-differences approaches. SDID is particularly useful in settings with staggered policy adoption, as it constructs a weighted combination of control units that closely matches the pre-treatment trends of treated units, improving causal inference and mitigating biases that can arise in conventional two-way fixed effects models. All SDID estimates were initially calculated using raw counts of excess deaths rather than mortality rates per 1,000 live births to avoid numerical instability in the estimation. These raw-count estimates were then converted back into mortality rates per 1,000 live births using CDC natality data for each treated state in the year that abortion was banned, allowing for interpretable effect sizes while preserving estimation stability. SDID estimates indicated a statistically significant increase in neonatal mortality following the implementation of abortion bans, with no corresponding effect on postneonatal mortality (Table 1). Using placebo-based inference, the estimated average treatment effect on the treated (ATT) for neonatal mortality was 0.281 (95% CI: 0.128–0.430), while postneonatal mortality remained near zero and non-significant (ATT 0.037; 95% CI: –0.093 to 0.166). Bootstrap-based standard errors produced consistent results, confirming the robustness of the neonatal mortality effect (neonatal ATT 0.281; 95% CI: 0.013–0.549). Figure 1 displays the SDID event-time estimates. Neonatal mortality trends were closely aligned between treated and control states during pre-treatment periods, followed by a sustained divergence after policy adoption, supporting a causal interpretation of the observed increase. The full weights of the SDID can be found in Appendix 2. Event-Study Estimates Using the Sun and Abraham Approach Event-study estimates using the interaction-weighted estimator proposed by Sun and Abraham (2021), designed to address the limitations of traditional two-way fixed effects models under staggered adoption, are presented in Figure 2. For neonatal mortality, pre-treatment estimates were consistently negative, ranging from –0.510 to –0.033, with confidence intervals spanning zero, indicating the absence of systematic pre-trends. This pattern underscores that prior to adoption, neonatal mortality was trending slightly downward rather than upward. Following policy adoption, the estimates shifted positively, reaching 0.303 (95% CI: 0.033 to 0.573) three years post-treatment, indicating a statistically significant increase in neonatal mortality. In contrast, postneonatal mortality exhibited estimates that fluctuated around zero both before and after adoption, with no consistent upward trend. Although a few pre-treatment coefficients were individually significant, they did not follow a monotonic pattern, and post-treatment estimates remained small in magnitude (e.g., +3 year: –0.036, 95% CI: –0.164 to 0.093), suggesting no discernible effect of the policy on postneonatal mortality. The leave-one-state-out robustness analysis indicated that for neonatal mortality three years post-treatment (+3 years), the estimated effects lost statistical significance at the p = 0.05 level when excluding certain states, as the corresponding 95% confidence intervals included zero. Specifically, the results were as follows: Texas (0.321; 95% CI -0.300, 0.942), New Hampshire (0.250; 95% CI -0.015, 0.516), Maine (0.247; 95% CI -0.017, 0.512), Kansas (0.271; 95% CI -0.001, 0.543), and Alaska (0.258; 95% CI -0.010, 0.527). For all other excluded states, the interval excluded zero and was statistically significant. We performed a covariate permutation robustness check to examine the sensitivity of the results to different subsets of control variables, including maternal health, racial composition, Hispanic births, and birth complications indices. Across specifications, pre-treatment coefficients were consistently negative, indicating no systematic upward trend prior to policy adoption. Post-treatment estimates at three years (+3) were positive in all specifications; the first specification (maternal health index only) slightly crossed zero (0.310; 95% CI –0.005 to 0.624), while all other specifications remained statistically significant (0.319–0.303; 95% CI 0.033–0.573). These results support that the observed increase in neonatal mortality is robust to the inclusion of a subset of covariates. The full results and complete robustness checks can be found in Appendix 2. Bayesian Hierarchical Event-Study Estimates For neonatal mortality, the posterior means from Bayesian hierarchical event-study estimates were near zero throughout the pre-treatment period, indicating no evidence of anticipatory trends. After policy adoption, posterior increases emerged: at one year post-adoption, the posterior mean increase in neonatal mortality was 0.050 (97%–3% highest-density interval [HDI], 0.006–0.090), rising to 0.132 (97%–3% HDI, 0.089–0.178) at two years and remaining elevated at three years (mean 0.127; 97%–3% HDI, 0.070–0.183). In contrast, postneonatal mortality showed no consistent post-treatment increase. Posterior means remained close to zero across post-adoption periods, with 97%–3% HDIs consistently overlapping zero, indicating no detectable effect. Consistency of Findings Across Methods Across synthetic difference-in-differences (SDID), interaction-weighted event-study, and Bayesian hierarchical models, abortion bans were consistently associated with increases in neonatal mortality, with no corresponding effect on postneonatal mortality (Table 1). The timing, magnitude, and outcome specificity of these effects were aligned across modeling approaches and inference frameworks. These results were robust to bootstrap and placebo-based inference, leave-one-state-out analyses, and alternative covariate specifications, supporting the reliability of the observed neonatal mortality increases. Discussion In this longitudinal, state-level analysis spanning 25 years, the implementation of state abortion bans was consistently associated with increases in neonatal mortality, while postneonatal mortality remained largely unchanged. These associations were observed across multiple causal inference approaches, including synthetic difference-in-differences, interaction-weighted event studies, and Bayesian hierarchical event-study models. The convergence of findings across distinct modeling frameworks and inferential strategies strengthens confidence that the observed increases in neonatal mortality reflect policy-related effects rather than model-specific artifacts. These findings align with a growing body of literature documenting adverse perinatal outcomes following abortion restrictions 8,10,20 . Prior studies using conventional difference-in-differences designs, interrupted time series, regression discontinuity approaches, and individual-level birth records have linked restrictive abortion policies to increases in high-risk births, maternal morbidity, congenital anomalies, and infant mortality. Evidence from analyses of Medicaid abortion funding restrictions, gestational limits, and clinic closures similarly points to elevated risks of neonatal death and severe neonatal complications, particularly among socioeconomically disadvantaged populations. The consistency of results across diverse methodological approaches, including those that do not rely on synthetic controls or staggered adoption designs, suggests that the association between abortion restrictions and neonatal mortality is not dependent on a single analytic framework. The outcome specificity observed in this study further supports the plausibility of the findings. Increases were concentrated in neonatal mortality, with no corresponding rise in postneonatal mortality. This pattern is clinically and biologically consistent with mechanisms operating during pregnancy, delivery, and the immediate postnatal period. Neonatal mortality is highly sensitive to prenatal conditions, maternal health, fetal anomalies, and perinatal care, whereas postneonatal mortality is more strongly influenced by environmental exposures, injury, and post-discharge care. The absence of detectable effects on postneonatal mortality argues against a generalized deterioration in infant care or surveillance and instead points to pathways concentrated around gestation and birth. The timing of effects further supports a causal interpretation. Across models, estimates were near zero in pre-treatment periods, with increases emerging after policy adoption and persisting over subsequent years. The absence of systematic pre-treatment trends reduces the likelihood that findings are driven by differential baseline trajectories or anticipatory changes. Together with robustness to placebo tests, leave-one-state-out analyses, and alternative covariate specifications, these patterns strengthen the inference that abortion bans contributed to the observed increases in neonatal mortality. Several strengths of this study warrant emphasis. First, the analysis draws on complete population-level birth and infant death data covering all U.S. states and the District of Columbia over a 25-year period, providing substantial statistical power and temporal depth. Second, the use of multiple modern causal inference methods specifically designed for staggered policy adoption addresses well-documented limitations of traditional difference-in-differences approaches. Third, the incorporation of rich state-level socioeconomic and demographic covariates from the American Community Survey reduces confounding from compositional changes over time. Finally, estimating effects using both frequentist and Bayesian frameworks allows triangulation across inferential paradigms, increasing confidence in the robustness of the results. These findings have important implications for reproductive health and perinatal policy. Abortion bans are often evaluated primarily through legal, ethical, or political lenses. The results presented here indicate that such policies are also associated with measurable population-level consequences for neonatal survival. The magnitude, timing, and outcome specificity of the effects suggest that abortion restrictions may increase neonatal mortality by increasing the proportion of births at elevated medical risk. Policymakers considering abortion restrictions should account for downstream impacts on neonatal health and ensure that perinatal care systems are adequately resourced to manage higher-risk pregnancies. Expanded access to comprehensive prenatal care, maternal-fetal medicine services, and neonatal intensive care may mitigate some adverse effects, although such measures are unlikely to fully offset the risks associated with forced continuation of medically complex pregnancies. This study has several limitations. First, the analysis was conducted at the state level and cannot capture individual-level mechanisms or heterogeneity across subpopulations. Second, although we adjusted for a broad set of maternal health, birth complexity, socioeconomic, and demographic factors, residual confounding by unmeasured time-varying factors remains possible. Third, abortion bans vary in scope, enforcement, and clinical exceptions, which may introduce exposure misclassification and attenuate estimated effects. Fourth, while modern staggered-adoption methods address many shortcomings of traditional difference-in-differences designs, they rely on assumptions that cannot be directly tested, including the absence of unobserved time-varying confounders that differentially affect treated states. Finally, infant mortality data may be subject to reporting delays or misclassification, although such errors are unlikely to vary systematically with the timing of policy adoption. Future research should examine individual-level birth and death records to identify specific pathways linking abortion bans to neonatal mortality, including gestational age, congenital anomalies, and maternal morbidity. Stratified analyses by race, socioeconomic status, and geographic access to perinatal care are needed to assess whether effects are concentrated among populations already experiencing elevated baseline risks. Longer follow-up will also be required to determine whether these effects persist, attenuate, or intensify as health systems and patient behaviors adapt to evolving policy environments. In conclusion, across multiple analytic approaches and extensive robustness checks, abortion bans were consistently associated with increases in neonatal mortality but not postneonatal mortality. These findings contribute robust causal evidence to an expanding literature on the public health consequences of restrictive abortion policies and underscore the importance of considering neonatal health outcomes in ongoing policy debates. Declarations Funding Statement: No funding went into this study. Data availability All data used are publicly available. Neonatal and postneonatal deaths and live births were obtained from CDC WONDER; maternal and birth characteristics from CDC Natality Files (1999–2024); and state-level socioeconomic covariates from the ACS (2005–2024). State abortion bans were compiled from the Guttmacher Institute and Kaiser Family Foundation (reconciled and included in Appendix 3). Analytical code is in https://github.com/LuluDuFeu/abortion-ban-infant-mortality-public-code and will be made public upon publication. Conflicts of interest The authors declare no conflicts of interest. References Chen A, Oster E, Williams H. Why Is Infant Mortality Higher in the United States than in Europe? Am Econ J Econ Policy . 2016;8(2):89-124. doi:10.1257/pol.20140224 Singh GK, Yu SM. Infant Mortality in the United States, 1915-2017: Large Social Inequalities have Persisted for Over a Century. Int J Matern Child Health AIDS IJMA . 2019;8(1):19-31. doi:10.21106/ijma.271 Lorenz JM, Ananth CV, Polin RA, D’Alton ME. Infant mortality in the United States. J Perinatol . 2016;36(10):797-801. doi:10.1038/jp.2016.63 Matoba N, Collins JW. Racial disparity in infant mortality. Semin Perinatol . 2017;41(6):354-359. doi:10.1053/j.semperi.2017.07.003 Singh GK, Kogan MD. Persistent Socioeconomic Disparities in Infant, Neonatal, and Postneonatal Mortality Rates in the United States, 1969–2001. Pediatrics . 2007;119(4):e928-e939. doi:10.1542/peds.2005-2181 Jang C, Lee H. A Review of Racial Disparities in Infant Mortality in the US. Children . 2022;9(2):257. doi:10.3390/children9020257 Nelson HD, Darney BG, Ahrens K, et al. Associations of Unintended Pregnancy With Maternal and Infant Health Outcomes: A Systematic Review and Meta-analysis. JAMA . 2022;328(17):1714. doi:10.1001/jama.2022.19097 Gemmill A, Franks AM, Anjur-Dietrich S, et al. US Abortion Bans and Infant Mortality. JAMA . 2025;333(15):1315-1323. doi:10.1001/jama.2024.28517 Gressler L, Lewis K. Changes in maternal morbidity and infant outcomes following state-level abortion bans post-Dobbs: a comparative interrupted time series study. BMC Public Health . 2025;25(1):2265. doi:10.1186/s12889-025-23468-8 Nuss E, Iwasaki M, Robbins L, et al. Maternal mortality according to state abortion legislative climate following the US Supreme Court’s Dobbs v. Jackson Women’s Health Organization ruling. Pregnancy . 2025;1(6):e70128. doi:10.1002/pmf2.70128 Natality Information. https://wonder.cdc.gov/natality.html Multiple Cause of Death Data on CDC WONDER. https://wonder.cdc.gov/mcd.html Bureau UC. Supplemental Estimates. Census.gov. Accessed January 16, 2026. https://www.census.gov/programs-surveys/acs/ Institute G. Interactive Map: US Abortion Policies and Access After Roe. Accessed January 16, 2026. https://states.guttmacher.org/policies/ Abortion in the United States Dashboard | KFF. Accessed January 16, 2026. https://www.kff.org/womens-health-policy/abortion-in-the-u-s-dashboard/ Arkhangelsky D, Athey S, Hirshberg DA, Imbens GW, Wager S. Synthetic Difference-in-Differences. Am Econ Rev . 2021;111(12):4088-4118. doi:10.1257/aer.20190159 Dong W, Zhang X, Liu S, et al. Effect of the national integrated demonstration area for the prevention and control of noncommunicable diseases programme on behavioural risk factors in China: a synthetic difference-in-differences study. Lancet Reg Health - West Pac . 2024;50:101167. doi:10.1016/j.lanwpc.2024.101167 Sun L, Abraham S. Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. J Econom . 2021;225(2):175-199. doi:10.1016/j.jeconom.2020.09.006 McGlothlin AE, Viele K. Bayesian Hierarchical Models. JAMA . 2018;320(22):2365. doi:10.1001/jama.2018.17977 Singh P, Gallo MF. National Trends in Infant Mortality in the US After Dobbs . JAMA Pediatr . 2024;178(12):1364. doi:10.1001/jamapediatrics.2024.4276 Table Table 1. Summary of neonatal and postneonatal mortality across all methods Method Outcome Relative Time (t) Estimate mortality per 1k births (95% CI / 3%–97% HDI) SDID Bootstrap Neonatal – 0.281 (95% CI: 0.013–0.549) Postneonatal – 0.037 (95% CI: -0.132–0.205) SDID Placebo Neonatal – 0.281 (95% CI: 0.128–0.430) Postneonatal – 0.037 (95% CI: -0.093–0.166) Sun & Abraham Neonatal –5 –0.033 (95% CI: –0.570, 0.505) Neonatal 0 –0.037 (95% CI: –0.546, 0.471) Neonatal +1 0.035 (95% CI: –0.481, 0.551) Neonatal +2 0.327 (95% CI: –0.266, 0.921) Neonatal +3 0.303 (95% CI: 0.033, 0.573) Postneonatal –5 –0.278 (95% CI: –0.574, 0.018) Postneonatal 0 –0.163 (95% CI: –0.449, 0.123) Postneonatal +1 –0.342 (95% CI: –0.608, –0.076) Postneonatal +2 –0.296 (95% CI: –0.577, –0.015) Postneonatal +3 –0.036 (95% CI: –0.164, 0.093) Bayesian Event Study Neonatal –5 –0.000 (HDI 3%–97%: –0.029, 0.031) Neonatal 0 –0.005 (HDI 3%–97%: –0.047, 0.034) Neonatal +1 0.050 (HDI 3%–97%: 0.006, 0.090) Neonatal +2 0.132 (HDI 3%–97%: 0.089, 0.178) Neonatal +3 0.127 (HDI 3%–97%: 0.070, 0.183) Postneonatal –5 0.017 (HDI 3%–97%: –0.013, 0.042) Postneonatal 0 0.016 (HDI 3%–97%: –0.025, 0.056) Postneonatal +1 0.024 (HDI 3%–97%: –0.017, 0.066) Postneonatal +2 –0.025 (HDI 3%–97%: –0.068, 0.021) Postneonatal +3 –0.027 (HDI 3%–97%: –0.084, 0.029) Additional Declarations The authors declare no competing interests. Supplementary Files Appendix1VIFCovariates.docx Appendix2Fullresultsweightsandrobustnesschecks.docx Appendix3.xlsx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8810808","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":587205630,"identity":"be7e8000-3d77-4f62-b47f-53cb198fc38c","order_by":0,"name":"Hugo Douma","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3UlEQVRIiWNgGAWjYBACAxDBAyLYG1AFidDCc4CxAVmQCC0SCURqMZdIv/jgTcXhaP6Zb8wffGyzs2dgb94mgU+L5YycYsM5Zw7nzridY9g440xyYgPPsTK8Wgxu5KRJ87Ydzm0AamnmqWBOYJDIMSNOy/ybZwyb/xjU2zPIvyGkJf0YWMuGGzyGzQwVhxkbJHgIaDnzhhnol/TcjWfSCmf2nDme2MaTVmyBV8vx9IfAELPOnXf88IYPP9uq7fnZD2+8gU8LMFLQYoENv3IQYH9AWM0oGAWjYBSMbAAAdxROi5tbRw4AAAAASUVORK5CYII=","orcid":"","institution":"The University of Texas at Austin","correspondingAuthor":true,"prefix":"","firstName":"Hugo","middleName":"","lastName":"Douma","suffix":""}],"badges":[],"createdAt":"2026-02-06 21:11:51","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8810808/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8810808/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":102314885,"identity":"5bdfe860-0a06-48f8-a3fb-e2e0b39f1ba8","added_by":"auto","created_at":"2026-02-10 12:37:41","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":114569,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSynthetic Difference-in-Differences (SDID) Estimates of the Effect of Abortion Bans on Infant Mortality\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eEstimated average treatment effects (ATT) with 95% confidence intervals, expressed as average excess deaths at state-year level. Negative values of relative time indicate pre-treatment periods; positive values indicate post-treatment periods. Neonatal and postneonatal outcomes are shown, accounting for staggered adoption across states. Bootstrap and placebo estimates are included to illustrate robustness.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8810808/v1/ac9f78dd39193a40a64d936f.png"},{"id":102314886,"identity":"bbb662b1-99ff-47ab-bcf3-978ea8fe46c2","added_by":"auto","created_at":"2026-02-10 12:37:41","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":125173,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSun–Abraham Interaction-Weighted Event Study Estimates of the Effect of Abortion Bans on Infant Mortality\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eEvent-time coefficients (relative to the year of abortion ban adoption, t=0) with 95% robust confidence intervals. Negative t indicates pre-treatment periods; positive t indicates post-treatment periods. Estimates are cohort- and treatment-interaction weighted following Sun and Abraham (2021), accounting for staggered adoption across states. Separate lines for neonatal and postneonatal mortality are shown.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8810808/v1/05562259cd7ab7f34a4409c8.png"},{"id":102314893,"identity":"8943ee97-7642-4a56-ae6c-eebd6d9920e2","added_by":"auto","created_at":"2026-02-10 12:37:41","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":278888,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEvent-Time Effects from Bayesian Event Study on Neonatal and Postneonatal Mortality\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ePosterior means with 95% highest-density intervals (HDI) are shown. Relative time t=0 indicates the year of full abortion ban adoption; negative values are pre-treatment, positive values are post-treatment. Estimates account for staggered adoption across states.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8810808/v1/c50c60f5b560fa2cc9c01b32.png"},{"id":102399115,"identity":"539c1fe6-76c2-4a6e-988b-7eea701eefc0","added_by":"auto","created_at":"2026-02-11 10:33:05","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1478633,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8810808/v1/63f43db8-ca8e-4932-a379-667ead04f691.pdf"},{"id":102397297,"identity":"dbf6c362-d9a1-4675-82c3-4d792ba6803d","added_by":"auto","created_at":"2026-02-11 10:15:06","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":8999,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix1VIFCovariates.docx","url":"https://assets-eu.researchsquare.com/files/rs-8810808/v1/1e50b0baf1444d3316466f30.docx"},{"id":102314888,"identity":"53890a66-0de3-49aa-8641-648f5993d919","added_by":"auto","created_at":"2026-02-10 12:37:41","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":35337,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix2Fullresultsweightsandrobustnesschecks.docx","url":"https://assets-eu.researchsquare.com/files/rs-8810808/v1/1765ffd365b72c69960aa032.docx"},{"id":102397908,"identity":"8feb2c54-2634-4f93-bd59-de7283d436cd","added_by":"auto","created_at":"2026-02-11 10:20:07","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":6752,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix3.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8810808/v1/0c3af3fcdc68b785e7614a9c.xlsx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eState Abortion Restrictions and Infant Mortality in the United States\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"Key Points","content":"\u003cp\u003e\u003cstrong\u003eQuestion:\u003c/strong\u003e Are state-level abortion bans associated with changes in neonatal or postneonatal mortality in the United States?\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFindings:\u003c/strong\u003e In this population-based study of all live births in 50 U.S. states and the District of Columbia from 1999 to 2024, neonatal mortality increased in states after implementation of complete abortion bans, while postneonatal mortality was unaffected. Estimates were consistent across multiple causal inference methods.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMeaning:\u003c/strong\u003e Complete abortion bans may lead to sustained increases in neonatal mortality, highlighting important public health consequences of restrictive reproductive policies.\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003eInfant mortality remains a critical indicator of population health and health equity, with profound implications for long-term human capital and societal well-being. In the United States, rates of infant mortality have historically exceeded those of peer nations\u003csup\u003e1,2\u003c/sup\u003e and exhibit persistent racial and socioeconomic disparities\u003csup\u003e3\u0026ndash;6\u003c/sup\u003e. Extensive evidence links unintended pregnancy to adverse maternal and infant outcomes, including preterm birth, low birth weight, and perinatal depression, underscoring the influence of reproductive health access on early life trajectories\u003csup\u003e7\u003c/sup\u003e. Yet the population-level impact of restrictive abortion policies on infant mortality remains incompletely understood, particularly when assessed with rigorous causal inference approaches that can account for the staggered adoption of policy interventions.\u003c/p\u003e\n\u003cp\u003eThe U.S. Supreme Court\u0026rsquo;s decision in \u003cem\u003eDobbs v. Jackson Women\u0026rsquo;s Health Organization\u003c/em\u003e, which overturned \u003cem\u003eRoe v. Wade\u003c/em\u003e, enabled states to enact complete or near-complete abortion bans, profoundly altering reproductive health access. Early evidence suggests these restrictions may have measurable impacts on infant health. Bayesian panel analyses of states implementing complete or six-week abortion bans from 2012 to 2023 indicate higher-than-expected infant mortality following policy adoption, with disproportionate increases among non-Hispanic Black infants and deaths due to congenital anomalies\u003csup\u003e8\u003c/sup\u003e. Other studies examining post-\u003cem\u003eDobbs\u003c/em\u003e birth outcomes highlight potential increases in congenital anomalies and persistent maternal health disparities in abortion-restricted states, though direct evidence remains limited\u003csup\u003e9,10\u003c/sup\u003e. Prior research also demonstrates that restrictive abortion policies may exacerbate preexisting inequities in maternal and infant health, particularly in states with higher baseline mortality.\u003c/p\u003e\n\u003cp\u003eDespite these insights, significant knowledge gaps remain. Most prior studies rely on single modeling approaches, short study windows, and limited adjustment for socioeconomic covariates, reducing causal interpretability and the ability to examine heterogeneity by population subgroups. Moreover, no previous studies have leveraged the full span of available U.S. vital statistics from 1999 through 2024 alongside rich state-level covariates from sources such as the American Community Survey.\u003c/p\u003e\n\u003cp\u003eThe United States presents a uniquely informative case: it combines recent, staggered abortion restrictions with high-quality, longitudinal public health data, allowing for robust causal inference and fine-grained subgroup analyses. Findings from the U.S. experience can inform understanding of the potential health consequences of restrictive reproductive policies in other settings, as many countries consider changes to abortion legislation.\u003c/p\u003e\n\u003cp\u003eTo address these gaps, we use data from all 50 U.S. states and the District of Columbia across 25 years, integrating detailed socioeconomic covariates and employing a triangulation of causal inference methods, including synthetic difference-in-differences, interaction-weighted event-study (using the Sun \u0026amp; Abraham method), and Bayesian hierarchical models, to estimate the association between state abortion bans and neonatal and postneonatal mortality. This approach allows us to rigorously assess the timing, magnitude, and specificity of policy effects, providing insights not only for the U.S. but for global public health policy regarding reproductive rights and infant health.\u003c/p\u003e"},{"header":"Methods","content":"\u003ch3\u003e\u003cstrong\u003eStudy Design and Data Sources\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eWe conducted a longitudinal, state-level panel study of infant mortality in the United States from 1999 through 2024. Annual state-level counts of neonatal deaths (death within 28 days of birth), postneonatal deaths (death between 28 days and 1 year), and live births were obtained from the Centers for Disease Control and Prevention (CDC) Wide-ranging Online Data for Epidemiologic Research (WONDER) database. Maternal health and birth characteristics were derived from CDC Natality Files (1999\u0026ndash;2024)\u003csup\u003e11,12\u003c/sup\u003e. State-level socioeconomic characteristics were obtained from the American Community Survey (ACS) 1-year and 5-year estimates (2005\u0026ndash;2024)\u003csup\u003e13\u003c/sup\u003e. State-level abortion ban legislation and effective dates were extracted from the Guttmacher Institute\u003csup\u003e14\u003c/sup\u003e and the Kaiser Family Foundation (KFF)\u003csup\u003e15\u003c/sup\u003e. The two sources were reconciled, and no discrepancies were identified (see Appendix 3 for the full dataset). Data was accessed on December 28th and December 29th 2025, we did not have access to any information that could be used to identify patients.\u003c/p\u003e\n\u003cp\u003eAll datasets were harmonized to a common state\u0026ndash;year structure. Socioeconomic covariates were unavailable for years prior to the introduction of the American Community Survey (ACS) and were therefore not included for those periods.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eExposure\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eThe exposure of interest was the implementation of full state-level abortion bans. States were classified as exposed beginning in the year in which the abortion ban legislation took effect. Variation in the timing of policy adoption across states enabled causal inference using staggered difference-in-differences and event-study designs.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eOutcomes\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eThe primary outcomes were state-level neonatal and postneonatal mortality rates, calculated as deaths per 1,000 live births. Outcomes were analyzed separately due to their distinct etiologies and policy sensitivity.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eCovariates\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eWe adjusted for time-varying state-level covariates capturing maternal health, birth complexity, demographic composition, and socioeconomic context. Several candidate covariates exhibited substantial multicollinearity. To improve model stability and precision, we conducted variance inflation factor (VIF) analysis and constructed three standardized indices summarizing correlated domains. The full list of covariates and indices used can be found in Appendix 1.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eStatistical Analysis\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eWe estimated treatment effects using multiple complementary panel-data approaches appropriate for staggered policy adoption.\u003c/p\u003e\n\u003cp\u003eOur primary analysis used Synthetic Difference-in-Differences (SDID)\u003csup\u003e16,17\u003c/sup\u003e, which combines outcome modeling from difference-in-differences with data-driven weighting from synthetic control methods to improve balance in pre-treatment trends. Models included all covariates described above, with state and year fixed effects. Standard errors were clustered at the state level.\u003c/p\u003e\n\u003cp\u003eTo assess dynamic treatment effects, we estimated interaction-weighted event-study models using the estimator proposed by Sun and Abraham\u003csup\u003e18\u003c/sup\u003e, which is robust to staggered and heterogeneous treatment effects across cohorts. Models included covariates, fixed effects, and clustered standard errors.\u003c/p\u003e\n\u003cp\u003eWe further estimated a Bayesian hierarchical event-study model\u003csup\u003e19\u003c/sup\u003e with partial pooling of state-specific treatment effects, allowing for cross-state heterogeneity while borrowing strength across the panel.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eAssumptions\u003c/strong\u003e\u003c/h3\u003e\n\u003ch4\u003eTo support causal interpretation of our findings, we explicitly outline the identifying assumptions underlying each analytic method. While all models adjust for maternal health, birth complications, socioeconomic indices, and racial composition, each approach relies on different assumptions for valid inference.\u003c/h4\u003e\n\u003ch4\u003e\u003cstrong\u003eSynthetic Difference-in-Differences (SDID)\u003c/strong\u003e\u003c/h4\u003e\n\u003col start=\"1\" type=\"1\"\u003e\n\u003cli\u003e\u003cstrong\u003eConditional Parallel Trends:\u003c/strong\u003e In the absence of abortion bans, the weighted combination of control states would have experienced the same trend in neonatal and postneonatal mortality as treated states.\u003cbr\u003e \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eNo Spillovers:\u003c/strong\u003e Policy adoption in one state does not affect infant mortality in other states.\u003cbr\u003e \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eCorrect Covariate Adjustment:\u003c/strong\u003e Maternal health, birth complications, socioeconomic indices, and racial composition adequately capture confounding factors affecting mortality.\u003cbr\u003e \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eStable Outcome Measurement:\u003c/strong\u003e Mortality reporting is accurate and consistent across states and years.\u003cbr\u003e \u003c/li\u003e\n\u003c/ol\u003e\n\u003ch4\u003e\u003cstrong\u003eSun and Abraham Interaction-Weighted Event Study\u003c/strong\u003e\u003c/h4\u003e\n\u003col start=\"1\" type=\"1\"\u003e\n\u003cli\u003e\u003cstrong\u003eNo Anticipation:\u003c/strong\u003e States do not systematically adjust behaviors affecting infant mortality prior to the enactment of abortion bans.\u003cbr\u003e \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eCohort-Heterogeneous Treatment Effects:\u003c/strong\u003e The estimator allows treatment effects to vary across states and adoption cohorts; identification assumes that, conditional on covariates and fixed effects, variation in adoption timing is as good as random with respect to residual pre-treatment trends.\u003cbr\u003e \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eParallel Pre-Trends Within Cohorts:\u003c/strong\u003e Within each adoption cohort, treated states follow parallel trends to untreated states in the pre-treatment period.\u003cbr\u003e \u003c/li\u003e\n\u003c/ol\u003e\n\u003ch4\u003e\u003cstrong\u003eBayesian Hierarchical Event-Study Model\u003c/strong\u003e\u003c/h4\u003e\n\u003col start=\"1\" type=\"1\"\u003e\n\u003cli\u003e\u003cstrong\u003ePartial Pooling Assumption:\u003c/strong\u003e State-specific treatment effects are drawn from a common hierarchical distribution; variation across states is exchangeable and informs the posterior estimates.\u003cbr\u003e \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eCorrect Model Specification:\u003c/strong\u003e The hierarchical model accurately captures temporal dynamics and covariate effects.\u003cbr\u003e \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eConditional Ignorability:\u003c/strong\u003e Given covariates and state/year effects, policy adoption is independent of unobserved determinants of neonatal and postneonatal mortality.\u003cbr\u003e \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eReliable Data Reporting:\u003c/strong\u003e Mortality and birth covariates are measured without systematic bias.\u003c/li\u003e\n\u003c/ol\u003e\n\u003ch3\u003e\u003cstrong\u003eSensitivity and Robustness Analyses\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eWe conducted extensive robustness checks, including placebo interventions assigned to pre-treatment periods, leave-one-state-out analyses, and alternative covariate specifications. We additionally evaluated sensitivity to covariate inclusion through systematic permutations of index components. Consistency of effect direction, magnitude, and timing across methods and outcomes was emphasized.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eSoftware\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eAll analyses were performed using Python version 3.11.9, using pandas and numpy for data management, statsmodels and linearmodels for frequentist panel estimation, and PyMC for Bayesian hierarchical modeling. Full VIF diagnostics and additional methodological details are provided in the Appendix 1.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eSynthetic Difference-in-Differences Estimates\u003c/strong\u003e \u003c/p\u003e\n\u003cp\u003eWe applied the Synthetic Difference-in-Differences (SDID) method, which combines features of synthetic control and traditional difference-in-differences approaches. SDID is particularly useful in settings with staggered policy adoption, as it constructs a weighted combination of control units that closely matches the pre-treatment trends of treated units, improving causal inference and mitigating biases that can arise in conventional two-way fixed effects models. \u003c/p\u003e\n\u003cp\u003eAll SDID estimates were initially calculated using raw counts of excess deaths rather than mortality rates per 1,000 live births to avoid numerical instability in the estimation. These raw-count estimates were then converted back into mortality rates per 1,000 live births using CDC natality data for each treated state in the year that abortion was banned, allowing for interpretable effect sizes while preserving estimation stability.\u003c/p\u003e\n\u003cp\u003eSDID estimates indicated a statistically significant increase in neonatal mortality following the implementation of abortion bans, with no corresponding effect on postneonatal mortality (Table 1). Using placebo-based inference, the estimated average treatment effect on the treated (ATT) for neonatal mortality was 0.281 (95% CI: 0.128\u0026ndash;0.430), while postneonatal mortality remained near zero and non-significant (ATT 0.037; 95% CI: \u0026ndash;0.093 to 0.166). Bootstrap-based standard errors produced consistent results, confirming the robustness of the neonatal mortality effect (neonatal ATT 0.281; 95% CI: 0.013\u0026ndash;0.549).\u003c/p\u003e\n\u003cp\u003eFigure 1 displays the SDID event-time estimates. Neonatal mortality trends were closely aligned between treated and control states during pre-treatment periods, followed by a sustained divergence after policy adoption, supporting a causal interpretation of the observed increase.\u003c/p\u003e\n\u003cp\u003eThe full weights of the SDID can be found in Appendix 2. \u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEvent-Study Estimates Using the Sun and Abraham Approach\u003c/strong\u003e \u003c/p\u003e\n\u003cp\u003eEvent-study estimates using the interaction-weighted estimator proposed by Sun and Abraham (2021), designed to address the limitations of traditional two-way fixed effects models under staggered adoption, are presented in Figure 2. For neonatal mortality, pre-treatment estimates were consistently negative, ranging from \u0026ndash;0.510 to \u0026ndash;0.033, with confidence intervals spanning zero, indicating the absence of systematic pre-trends. This pattern underscores that prior to adoption, neonatal mortality was trending slightly downward rather than upward. Following policy adoption, the estimates shifted positively, reaching 0.303 (95% CI: 0.033 to 0.573) three years post-treatment, indicating a statistically significant increase in neonatal mortality.\u003c/p\u003e\n\u003cp\u003eIn contrast, postneonatal mortality exhibited estimates that fluctuated around zero both before and after adoption, with no consistent upward trend. Although a few pre-treatment coefficients were individually significant, they did not follow a monotonic pattern, and post-treatment estimates remained small in magnitude (e.g., +3 year: \u0026ndash;0.036, 95% CI: \u0026ndash;0.164 to 0.093), suggesting no discernible effect of the policy on postneonatal mortality.\u003c/p\u003e\n\u003cp\u003eThe leave-one-state-out robustness analysis indicated that for neonatal mortality three years post-treatment (+3 years), the estimated effects lost statistical significance at the p = 0.05 level when excluding certain states, as the corresponding 95% confidence intervals included zero. Specifically, the results were as follows: Texas (0.321; 95% CI -0.300, 0.942), New Hampshire (0.250; 95% CI -0.015, 0.516), Maine (0.247; 95% CI -0.017, 0.512), Kansas (0.271; 95% CI -0.001, 0.543), and Alaska (0.258; 95% CI -0.010, 0.527). For all other excluded states, the interval excluded zero and was statistically significant.\u003c/p\u003e\n\u003cp\u003eWe performed a covariate permutation robustness check to examine the sensitivity of the results to different subsets of control variables, including maternal health, racial composition, Hispanic births, and birth complications indices. Across specifications, pre-treatment coefficients were consistently negative, indicating no systematic upward trend prior to policy adoption. Post-treatment estimates at three years (+3) were positive in all specifications; the first specification (maternal health index only) slightly crossed zero (0.310; 95% CI \u0026ndash;0.005 to 0.624), while all other specifications remained statistically significant (0.319\u0026ndash;0.303; 95% CI 0.033\u0026ndash;0.573). These results support that the observed increase in neonatal mortality is robust to the inclusion of a subset of covariates. \u003c/p\u003e\n\u003cp\u003eThe full results and complete robustness checks can be found in Appendix 2. \u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBayesian Hierarchical Event-Study Estimates\u003c/strong\u003e \u003c/p\u003e\n\u003cp\u003eFor neonatal mortality, the posterior means from Bayesian hierarchical event-study estimates were near zero throughout the pre-treatment period, indicating no evidence of anticipatory trends. After policy adoption, posterior increases emerged: at one year post-adoption, the posterior mean increase in neonatal mortality was 0.050 (97%\u0026ndash;3% highest-density interval [HDI], 0.006\u0026ndash;0.090), rising to 0.132 (97%\u0026ndash;3% HDI, 0.089\u0026ndash;0.178) at two years and remaining elevated at three years (mean 0.127; 97%\u0026ndash;3% HDI, 0.070\u0026ndash;0.183).\u003c/p\u003e\n\u003cp\u003eIn contrast, postneonatal mortality showed no consistent post-treatment increase. Posterior means remained close to zero across post-adoption periods, with 97%\u0026ndash;3% HDIs consistently overlapping zero, indicating no detectable effect.\u003cem\u003e \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsistency of Findings Across Methods\u003c/strong\u003e \u003c/p\u003e\n\u003cp\u003eAcross synthetic difference-in-differences (SDID), interaction-weighted event-study, and Bayesian hierarchical models, abortion bans were consistently associated with increases in neonatal mortality, with no corresponding effect on postneonatal mortality (Table 1). The timing, magnitude, and outcome specificity of these effects were aligned across modeling approaches and inference frameworks. These results were robust to bootstrap and placebo-based inference, leave-one-state-out analyses, and alternative covariate specifications, supporting the reliability of the observed neonatal mortality increases.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this longitudinal, state-level analysis spanning 25 years, the implementation of state abortion bans was consistently associated with increases in neonatal mortality, while postneonatal mortality remained largely unchanged. These associations were observed across multiple causal inference approaches, including synthetic difference-in-differences, interaction-weighted event studies, and Bayesian hierarchical event-study models. The convergence of findings across distinct modeling frameworks and inferential strategies strengthens confidence that the observed increases in neonatal mortality reflect policy-related effects rather than model-specific artifacts.\u003c/p\u003e\n\u003cp\u003eThese findings align with a growing body of literature documenting adverse perinatal outcomes following abortion restrictions\u003csup\u003e8,10,20\u003c/sup\u003e. Prior studies using conventional difference-in-differences designs, interrupted time series, regression discontinuity approaches, and individual-level birth records have linked restrictive abortion policies to increases in high-risk births, maternal morbidity, congenital anomalies, and infant mortality. Evidence from analyses of Medicaid abortion funding restrictions, gestational limits, and clinic closures similarly points to elevated risks of neonatal death and severe neonatal complications, particularly among socioeconomically disadvantaged populations. The consistency of results across diverse methodological approaches, including those that do not rely on synthetic controls or staggered adoption designs, suggests that the association between abortion restrictions and neonatal mortality is not dependent on a single analytic framework.\u003c/p\u003e\n\u003cp\u003eThe outcome specificity observed in this study further supports the plausibility of the findings. Increases were concentrated in neonatal mortality, with no corresponding rise in postneonatal mortality. This pattern is clinically and biologically consistent with mechanisms operating during pregnancy, delivery, and the immediate postnatal period. Neonatal mortality is highly sensitive to prenatal conditions, maternal health, fetal anomalies, and perinatal care, whereas postneonatal mortality is more strongly influenced by environmental exposures, injury, and post-discharge care. The absence of detectable effects on postneonatal mortality argues against a generalized deterioration in infant care or surveillance and instead points to pathways concentrated around gestation and birth.\u003c/p\u003e\n\u003cp\u003eThe timing of effects further supports a causal interpretation. Across models, estimates were near zero in pre-treatment periods, with increases emerging after policy adoption and persisting over subsequent years. The absence of systematic pre-treatment trends reduces the likelihood that findings are driven by differential baseline trajectories or anticipatory changes. Together with robustness to placebo tests, leave-one-state-out analyses, and alternative covariate specifications, these patterns strengthen the inference that abortion bans contributed to the observed increases in neonatal mortality.\u003c/p\u003e\n\u003cp\u003eSeveral strengths of this study warrant emphasis. First, the analysis draws on complete population-level birth and infant death data covering all U.S. states and the District of Columbia over a 25-year period, providing substantial statistical power and temporal depth. Second, the use of multiple modern causal inference methods specifically designed for staggered policy adoption addresses well-documented limitations of traditional difference-in-differences approaches. Third, the incorporation of rich state-level socioeconomic and demographic covariates from the American Community Survey reduces confounding from compositional changes over time. Finally, estimating effects using both frequentist and Bayesian frameworks allows triangulation across inferential paradigms, increasing confidence in the robustness of the results.\u003c/p\u003e\n\u003cp\u003eThese findings have important implications for reproductive health and perinatal policy. Abortion bans are often evaluated primarily through legal, ethical, or political lenses. The results presented here indicate that such policies are also associated with measurable population-level consequences for neonatal survival. The magnitude, timing, and outcome specificity of the effects suggest that abortion restrictions may increase neonatal mortality by increasing the proportion of births at elevated medical risk.\u003c/p\u003e\n\u003cp\u003ePolicymakers considering abortion restrictions should account for downstream impacts on neonatal health and ensure that perinatal care systems are adequately resourced to manage higher-risk pregnancies. Expanded access to comprehensive prenatal care, maternal-fetal medicine services, and neonatal intensive care may mitigate some adverse effects, although such measures are unlikely to fully offset the risks associated with forced continuation of medically complex pregnancies.\u003c/p\u003e\n\u003cp\u003eThis study has several limitations. First, the analysis was conducted at the state level and cannot capture individual-level mechanisms or heterogeneity across subpopulations. Second, although we adjusted for a broad set of maternal health, birth complexity, socioeconomic, and demographic factors, residual confounding by unmeasured time-varying factors remains possible. Third, abortion bans vary in scope, enforcement, and clinical exceptions, which may introduce exposure misclassification and attenuate estimated effects.\u003c/p\u003e\n\u003cp\u003eFourth, while modern staggered-adoption methods address many shortcomings of traditional difference-in-differences designs, they rely on assumptions that cannot be directly tested, including the absence of unobserved time-varying confounders that differentially affect treated states. Finally, infant mortality data may be subject to reporting delays or misclassification, although such errors are unlikely to vary systematically with the timing of policy adoption.\u003c/p\u003e\n\u003cp\u003eFuture research should examine individual-level birth and death records to identify specific pathways linking abortion bans to neonatal mortality, including gestational age, congenital anomalies, and maternal morbidity. Stratified analyses by race, socioeconomic status, and geographic access to perinatal care are needed to assess whether effects are concentrated among populations already experiencing elevated baseline risks. Longer follow-up will also be required to determine whether these effects persist, attenuate, or intensify as health systems and patient behaviors adapt to evolving policy environments.\u003c/p\u003e\n\u003cp\u003eIn conclusion, across multiple analytic approaches and extensive robustness checks, abortion bans were consistently associated with increases in neonatal mortality but not postneonatal mortality. These findings contribute robust causal evidence to an expanding literature on the public health consequences of restrictive abortion policies and underscore the importance of considering neonatal health outcomes in ongoing policy debates.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding Statement:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNo funding went into this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll data used are publicly available. Neonatal and postneonatal deaths and live births were obtained from CDC WONDER; maternal and birth characteristics from CDC Natality Files (1999\u0026ndash;2024); and state-level socioeconomic covariates from the ACS (2005\u0026ndash;2024). State abortion bans were compiled from the Guttmacher Institute and Kaiser Family Foundation (reconciled and included in Appendix 3).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAnalytical code is in https://github.com/LuluDuFeu/abortion-ban-infant-mortality-public-code and will be made public upon publication.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflicts of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflicts of interest.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eChen A, Oster E, Williams H. Why Is Infant Mortality Higher in the United States than in Europe? \u003cem\u003eAm Econ J Econ Policy\u003c/em\u003e. 2016;8(2):89-124. doi:10.1257/pol.20140224 \u003c/li\u003e\n\u003cli\u003eSingh GK, Yu SM. Infant Mortality in the United States, 1915-2017: Large Social Inequalities have Persisted for Over a Century. \u003cem\u003eInt J Matern Child Health AIDS IJMA\u003c/em\u003e. 2019;8(1):19-31. doi:10.21106/ijma.271 \u003c/li\u003e\n\u003cli\u003eLorenz JM, Ananth CV, Polin RA, D\u0026rsquo;Alton ME. Infant mortality in the United States. \u003cem\u003eJ Perinatol\u003c/em\u003e. 2016;36(10):797-801. doi:10.1038/jp.2016.63 \u003c/li\u003e\n\u003cli\u003eMatoba N, Collins JW. Racial disparity in infant mortality. \u003cem\u003eSemin Perinatol\u003c/em\u003e. 2017;41(6):354-359. doi:10.1053/j.semperi.2017.07.003 \u003c/li\u003e\n\u003cli\u003eSingh GK, Kogan MD. Persistent Socioeconomic Disparities in Infant, Neonatal, and Postneonatal Mortality Rates in the United States, 1969\u0026ndash;2001. \u003cem\u003ePediatrics\u003c/em\u003e. 2007;119(4):e928-e939. doi:10.1542/peds.2005-2181 \u003c/li\u003e\n\u003cli\u003eJang C, Lee H. A Review of Racial Disparities in Infant Mortality in the US. \u003cem\u003eChildren\u003c/em\u003e. 2022;9(2):257. doi:10.3390/children9020257 \u003c/li\u003e\n\u003cli\u003eNelson HD, Darney BG, Ahrens K, et al. Associations of Unintended Pregnancy With Maternal and Infant Health Outcomes: A Systematic Review and Meta-analysis. \u003cem\u003eJAMA\u003c/em\u003e. 2022;328(17):1714. doi:10.1001/jama.2022.19097 \u003c/li\u003e\n\u003cli\u003eGemmill A, Franks AM, Anjur-Dietrich S, et al. US Abortion Bans and Infant Mortality. \u003cem\u003eJAMA\u003c/em\u003e. 2025;333(15):1315-1323. doi:10.1001/jama.2024.28517 \u003c/li\u003e\n\u003cli\u003eGressler L, Lewis K. Changes in maternal morbidity and infant outcomes following state-level abortion bans post-Dobbs: a comparative interrupted time series study. \u003cem\u003eBMC Public Health\u003c/em\u003e. 2025;25(1):2265. doi:10.1186/s12889-025-23468-8 \u003c/li\u003e\n\u003cli\u003eNuss E, Iwasaki M, Robbins L, et al. Maternal mortality according to state abortion legislative climate following the US Supreme Court\u0026rsquo;s \u003cem\u003eDobbs v. Jackson Women\u0026rsquo;s Health Organization\u003c/em\u003e ruling. \u003cem\u003ePregnancy\u003c/em\u003e. 2025;1(6):e70128. doi:10.1002/pmf2.70128 \u003c/li\u003e\n\u003cli\u003eNatality Information. https://wonder.cdc.gov/natality.html \u003c/li\u003e\n\u003cli\u003eMultiple Cause of Death Data on CDC WONDER. https://wonder.cdc.gov/mcd.html \u003c/li\u003e\n\u003cli\u003eBureau UC. Supplemental Estimates. Census.gov. Accessed January 16, 2026. https://www.census.gov/programs-surveys/acs/ \u003c/li\u003e\n\u003cli\u003eInstitute G. Interactive Map: US Abortion Policies and Access After Roe. Accessed January 16, 2026. https://states.guttmacher.org/policies/ \u003c/li\u003e\n\u003cli\u003eAbortion in the United States Dashboard | KFF. Accessed January 16, 2026. https://www.kff.org/womens-health-policy/abortion-in-the-u-s-dashboard/ \u003c/li\u003e\n\u003cli\u003eArkhangelsky D, Athey S, Hirshberg DA, Imbens GW, Wager S. Synthetic Difference-in-Differences. \u003cem\u003eAm Econ Rev\u003c/em\u003e. 2021;111(12):4088-4118. doi:10.1257/aer.20190159 \u003c/li\u003e\n\u003cli\u003eDong W, Zhang X, Liu S, et al. Effect of the national integrated demonstration area for the prevention and control of noncommunicable diseases programme on behavioural risk factors in China: a synthetic difference-in-differences study. \u003cem\u003eLancet Reg Health - West Pac\u003c/em\u003e. 2024;50:101167. doi:10.1016/j.lanwpc.2024.101167 \u003c/li\u003e\n\u003cli\u003eSun L, Abraham S. Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. \u003cem\u003eJ Econom\u003c/em\u003e. 2021;225(2):175-199. doi:10.1016/j.jeconom.2020.09.006 \u003c/li\u003e\n\u003cli\u003eMcGlothlin AE, Viele K. Bayesian Hierarchical Models. \u003cem\u003eJAMA\u003c/em\u003e. 2018;320(22):2365. doi:10.1001/jama.2018.17977 \u003c/li\u003e\n\u003cli\u003eSingh P, Gallo MF. National Trends in Infant Mortality in the US After \u003cem\u003eDobbs\u003c/em\u003e. \u003cem\u003eJAMA Pediatr\u003c/em\u003e. 2024;178(12):1364. doi:10.1001/jamapediatrics.2024.4276 \u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003e\u003cstrong\u003eTable 1. Summary of neonatal and postneonatal mortality across all methods\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"607\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMethod\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOutcome\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRelative Time (t)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEstimate mortality per 1k births (95% CI / 3%\u0026ndash;97%\u0026nbsp;HDI)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eSDID Bootstrap\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.281 (95% CI: 0.013\u0026ndash;0.549)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.037 (95% CI: -0.132\u0026ndash;0.205)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eSDID Placebo\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.281 (95% CI: 0.128\u0026ndash;0.430)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.037 (95% CI: -0.093\u0026ndash;0.166)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"10\" valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eSun \u0026amp; Abraham\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u0026ndash;5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.033 (95% CI:\u0026nbsp;\u0026ndash;0.570, 0.505)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.037 (95% CI:\u0026nbsp;\u0026ndash;0.546, 0.471)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.035 (95% CI:\u0026nbsp;\u0026ndash;0.481, 0.551)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.327 (95% CI:\u0026nbsp;\u0026ndash;0.266, 0.921)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.303 (95% CI:\u0026nbsp;0.033, 0.573)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u0026ndash;5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.278 (95% CI:\u0026nbsp;\u0026ndash;0.574, 0.018)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.163 (95% CI:\u0026nbsp;\u0026ndash;0.449, 0.123)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.342 (95% CI:\u0026nbsp;\u0026ndash;0.608, \u0026ndash;0.076)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.296 (95% CI:\u0026nbsp;\u0026ndash;0.577, \u0026ndash;0.015)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.036 (95% CI:\u0026nbsp;\u0026ndash;0.164, 0.093)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"10\" valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eBayesian Event Study\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u0026ndash;5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.000 (HDI 3%\u0026ndash;97%:\u0026nbsp;\u0026ndash;0.029, 0.031)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.005 (HDI 3%\u0026ndash;97%:\u0026nbsp;\u0026ndash;0.047, 0.034)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.050 (HDI 3%\u0026ndash;97%: 0.006, 0.090)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.132 (HDI 3%\u0026ndash;97%: 0.089, 0.178)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003eNeonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.127 (HDI 3%\u0026ndash;97%: 0.070, 0.183)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e\u0026ndash;5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.017 (HDI 3%\u0026ndash;97%: \u0026ndash;0.013, 0.042)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.016 (HDI 3%\u0026ndash;97%: \u0026ndash;0.025, 0.056)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e0.024 (HDI 3%\u0026ndash;97%: \u0026ndash;0.017, 0.066)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.025 (HDI 3%\u0026ndash;97%: \u0026ndash;0.068, 0.021)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 124px;\"\u003e\n \u003cp\u003ePostneonatal\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e+3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 281px;\"\u003e\n \u003cp\u003e\u0026ndash;0.027 (HDI 3%\u0026ndash;97%: \u0026ndash;0.084, 0.029)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"The University of Texas at Austin","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8810808/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8810808/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eImportance\u003cbr\u003e\n\u003c/strong\u003eRestrictive abortion policies may have unintended consequences for perinatal outcomes, yet evidence on their impact on infant mortality remains limited.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eObjective\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo assess whether complete state-level abortion bans are associated with changes in neonatal and postneonatal mortality in the United States.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDesign\u003cbr\u003e\n\u003c/strong\u003ePopulation-based serial cross-sectional study (1999–2024) using nationwide vital statistics data. Causal effects were estimated using synthetic difference-in-differences, interaction-weighted event-study models, and Bayesian hierarchical modeling.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSetting and Participants\u003cbr\u003e\n\u003c/strong\u003eAll live births in the United States across all 50 states and the District of Columbia. Twelve states implementing complete abortion bans were compared with pre-ban trends and states without bans.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMain Outcomes\u003cbr\u003e\n\u003c/strong\u003eNeonatal mortality (0–27 days) and postneonatal mortality (28–364 days) per 1,000 live births.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003cbr\u003e\n\u003c/strong\u003eComplete abortion bans were associated with significant increases in neonatal mortality but not postneonatal mortality. Synthetic difference-in-differences estimates showed an increase of 0.281 neonatal deaths per 1,000 live births (95% CI, 0.128–0.430). Bayesian hierarchical models indicated increases emerging one year post-ban (0.050, 97%–3% HDI 0.006–0.090), rising at two years (0.132, 97%–3% HDI 0.089–0.178), and remaining elevated at three years (0.127, 97%–3% HDI 0.070–0.183). Event-study analyses showed increased neonatal mortality three years post-ban (0.303, 95% CI 0.033–0.573). No consistent associations were observed for postneonatal mortality. Findings were robust across multiple sensitivity analyses.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions\u003cbr\u003e\n\u003c/strong\u003eComplete state-level abortion bans were associated with sustained increases in neonatal mortality, indicating potential adverse effects of restrictive abortion policies on early-life outcomes.\u003c/p\u003e","manuscriptTitle":"State Abortion Restrictions and Infant Mortality in the United States","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-10 12:37:15","doi":"10.21203/rs.3.rs-8810808/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1fd717fd-fcff-4542-9950-843bf11e3f20","owner":[],"postedDate":"February 10th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-02-10T12:37:15+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-10 12:37:15","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8810808","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8810808","identity":"rs-8810808","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}