When Welfare Goes Digital: Administrative Automation and Birth Outcomes for Low-Income Mothers

• In Event: New Research on Administrative Burdens

Abstract

In 2006, Indiana contracted with IBM to automate eligibility determination across Medicaid, SNAP, and TANF, replacing county-based caseworkers with centralized call centers and standardized digital workflows. Within one year, participation fell significantly across safety net programs, with losses concentrated among recipients who had the lowest incomes, the least education, and the longest histories of program reliance. Although statutory eligibility rules were unchanged, the reform fundamentally altered how low-income households accessed the safety net during a period when income support, nutrition assistance, and insurance coverage are known to affect prenatal health. This paper asks whether administrative reorganization of welfare delivery, holding formal eligibility constant, produced measurable downstream effects on birth outcomes, and whether effects depended on the trimester of prenatal exposure. I use restricted-use county-level natality detail files from the National Center for Health Statistics for all Indiana live births between 2006 and 2011, totaling approximately 377,000 singleton births after sample restrictions. These records are linked to county-level economic and demographic data from the American Community Survey, the Small Area Income and Poverty Estimates, and the Bureau of Labor Statistics Local Area Unemployment Statistics.  I exploit the staggered county-level rollout of the automated system across three waves (October 2007, March 2008, May 2008) using a difference-in-differences design with county and year-month fixed effects. Outcomes include birth weight, low and very low birth weight, gestational age, preterm birth, and first-trimester prenatal care initiation. A trimester-specific specification estimates separate treatment effects for exposure during the first, second, and third trimesters to identify whether timing of administrative disruption during pregnancy matters for fetal development. Robustness checks include heterogeneity-robust estimators and event-study specifications. Baseline intent-to-treat estimates pooling all post-treatment exposure yield small and statistically insignificant effects on birth outcomes. Trimester-specific estimatesreveal significant estimated effects. Third-trimester exposure reduced birth weight by approximately 21 grams and gestational age by 0.17 weeks, while increasing preterm birth by 1.6 percentage points and low birth weight by roughly 0.8 percentage points. Estimated effects attenuate but persist for earlier trimesters. Effects on prenatal care utilization emerge primarily for full-pregnancy exposure. These findings suggest that administrative design functions as an intermediary determinant of health: decisions about how benefits are delivered, not only who qualifies, can produce biological consequences for the most vulnerable populations. As states increasingly adopt digital eligibility systems, these results indicate that efficiency-oriented reforms can generate unintended health costs concentrated among low-income mothers and infants, with implications for equity-conscious safety-net administration.

Author

• Jack Baldwin, Syracuse University
  Presenting Author