ACA Medicaid Expansion and Insurance Coverage Among New Mothers Living in Poverty

We use 2010–2017 American Community Survey (ACS) data from IPUMS.³⁷  We incorporate edits to the insurance coverage measures to account for apparent misreporting and we generate income relative to the FPL for each woman’s health insurance unit (HIU) using an approach developed by the State Health Access Data Assistance Center.^38,39  The Urban Institute Institutional Review Board determined that this analysis did not require institutional review board review because ACS data are publicly available and deidentified.

We define new mothers as women aged 19 to 44 years who report giving birth to any children born alive in the past 12 months. We limit our sample to include only US citizens because Medicaid eligibility is limited for noncitizens, and we exclude women who receive Supplemental Security Income (SSI) or Medicare because they have other pathways to Medicaid. We define the target population for the Medicaid expansion as women with HIU incomes <100% of the FPL. We exclude women with incomes 100% to 138% of the FPL because they may have access to Marketplace subsidies in nonexpansion states.

Health insurance coverage is self-reported at the time of the survey. When women report multiple types of coverage, we assign a single coverage type on the basis of the hierarchy: employer-sponsored insurance (ESI), Medicaid, private nongroup or Department of Veterans Affairs (VA) coverage, and uninsured. We focus on 3 categories of coverage: uninsurance, Medicaid, and private or other (VA, ESI, and nongroup coverage). We also use ACS measures of women’s sociodemographic characteristics, including income, age, race and/or ethnicity, educational attainment, employment, marital status, and metropolitan status (Table 2). Finally, we draw annual state unemployment rates from the IPUMS Current Population Survey database.⁴⁰ 

We use a binary measure of Medicaid-expansion status that classifies states that expanded Medicaid under the ACA by January 2017 as expansion states and all others as nonexpansion states. We also use a continuous measure of Medicaid eligibility that captures the income eligibility thresholds for each year from 2010 to 2017, drawn from Kaiser Family Foundation reports.^2,41  Specifically, we use the eligibility thresholds for nonworking parents in 2010–2013 and for all parents in 2014–2017. Eligibility is no longer differentiated by work status after changes under the ACA.⁴² 

We first describe patterns of insurance coverage among new mothers living in poverty by year and expansion status and estimate unadjusted differences. We then estimate the effect of Medicaid expansions on insurance coverage using a difference-in-differences approach with a continuous policy variable that reflects the Medicaid-eligibility threshold for parents in a given state and year. By including each year’s threshold, we capture both the presence and the magnitude of eligibility expansions across states over time. This is important because there was considerable variation in the pre-ACA Medicaid-eligibility thresholds for parents (Table 1).

We estimate linear probability models for ease of interpretation and top-code eligibility thresholds at 138% of the FPL to avoid capturing changes in eligibility for higher-income populations that do not affect the women in our sample. All models control for women’s sociodemographic characteristics, state unemployment rate, state fixed effects, and year fixed effects. SEs are clustered at the state level, and models use survey weights.

For simplicity, we report the estimated effect of a 100-percentage-point increase in the Medicaid income eligibility threshold for parents. Estimates can be scaled to inform the expected effects of expansions of other sizes (eg, 50-percentage-points) because the model assumes a linear relationship between eligibility and our outcomes of interest.

Because Medicaid expansions may have disproportionately affected different groups of poor new mothers, we also stratify our sample by race and/or ethnicity (non-Hispanic white, non-Hispanic African American, Hispanic, and other race or multiple races) and income (<50% of the FPL and 50% to <100% of the FPL).

We consider alternate definitions of new mothers living in poverty: new mothers with incomes <138% of the FPL and women with incomes <100% of the FPL who report giving birth in the past 12 months and have a child aged <1 in their household. This second sample is smaller because the ACS undercounts young children in the United States.⁴³  Next, we consider alternate measures of Medicaid-eligibility thresholds: a nonworking parental Medicaid threshold not capped at 138% of the FPL and a capped Medicaid threshold for working parents. We also consider an unweighted model, a model with state-specific linear time trends, and a model using a Medicaid-topped insurance hierarchy. Because evidence suggests that the ACA dependent coverage mandate increased coverage for young women with a recent birth,⁴⁴  we estimate separate models for young adult women eligible to remain on their parent’s insurance (ages 19–25) and for women not eligible for this provision (ages 26–44). We also test the sensitivity of our estimates to the inclusion of noncitizen women because some are eligible for Medicaid depending on their legal status, time in the United States, and state of residence. Finally, we test the sensitivity of our results to policy-relevant state and year exclusions.

Before ACA Medicaid expansion, poor new mothers in nonexpansion states were nearly twice as likely to be uninsured as those in expansion states (35.7% compared with 18.9%; P < .001; Fig 1). After implementation of the major coverage provisions of the ACA, the uninsurance rate for poor new mothers fell 41%, from 25.9% in 2013 to 15.2% in 2017 (P < .001). The percentage-point decrease in uninsurance was similar in expansion and nonexpansion states, but the increase in Medicaid coverage was larger in expansion states (63.7% to 75.4%; P < .001) than in nonexpansion states (45.7% to 51.6%; P < .001). Although private or other coverage increased in nonexpansion states, we do not observe a significant change in expansion states. By 2017, the uninsurance rate for poor new mothers in nonexpansion states was 3.2 times higher than in expansion states, and the Medicaid coverage rate was 1.5 times higher in expansion states than in nonexpansion states. As shown in Table 2, the characteristics of poor new mothers living in poverty change over our study period and vary between those living in expansion and nonexpansion states, highlighting the importance of controlling for these characteristics in our impact analysis.

After accounting for observed characteristics of new mothers, state fixed effects, and secular trends, we find that a 100-percentage-point increase in the parental Medicaid-eligibility threshold is associated with an 8.8-percentage-point decrease (P < .001; Table 3) in uninsurance among poor new mothers. The same increase in eligibility is associated with a 13.2-percentage-point increase (P < .001) in Medicaid coverage among poor new mothers. We also find a negative association between Medicaid eligibility and private or other coverage (4.4 percentage points; P = .001), suggesting that approximately one-third of the gain in Medicaid coverage reflects substitution of other coverage sources. These findings suggest that the average increase in Medicaid eligibility across expansion states of 64 percentage points is associated with a 5.6-percentage-point decrease in uninsurance, a 28% decrease from the pre-ACA mean in expansion states. The same average increase is associated with an 8.5-percentage-point increase in Medicaid coverage and a 2.8-percentage-point decline in private or other coverage among poor new mothers, all else equal.

When stratified by race and/or ethnicity (Table 4), we find that a 100-percetage point increase in Medicaid eligibility is associated with a 10-percentage-point reduction in uninsurance among Hispanic and white new mothers (P = .002 and P < .001), a 7-percentage-point reduction for new mothers of other or multiple races (P = .02), and a nearly 5-percentage-point reduction for African American new mothers (P = .09). Eligibility increases were associated with gains in Medicaid coverage for all groups. Coverage gains in the other race or multiple races category appear to be driven by gains for American Indian and Alaskan native new mothers (Supplemental Table 6).

When stratified by income (Table 5), a 100-percentage-point increase in eligibility is associated with a 9.0-percentage-point increase in coverage (P < .001) for both groups. For new mothers with incomes <50% of the FPL, expansion is associated with an 11.6-percentage-point increase in Medicaid coverage (P < .001), compared with a 16.7-percentage-point increase for new mothers with incomes 50% to <100% of the FPL (P < .001).

Our results are robust to alternate definitions of new mothers (Supplemental Table 7), alternate eligibility thresholds (Supplemental Table 8), and alternate model specifications (Supplemental Table 9), varying modestly in magnitude but maintaining the same direction and general patterns. Effects appear stronger for women not affected by the dependent coverage provision than for the full sample or young adult women (Supplemental Table 10). Including noncitizen women produces slightly weaker estimates than our main model. We find no statistically significant effects for the small sample of noncitizen women; estimates are consistent in direction with those for citizen women but are smaller in magnitude (Supplemental Table 11). Results are somewhat stronger when we omit 2014 or late Medicaid-expansion states (Supplemental Table 12).

Our model suggests that if the nonexpansion states had expanded, the uninsurance rate among our sample of new mothers in these states would have fallen to 17.6% in 2017 (Fig 2), suggesting that ∼29 000 poor new mothers in nonexpansion states could have gained coverage. In 2017, there were ∼451 000 uninsured new mothers overall, 31% of whom were citizen women with incomes below the poverty level (the target population for the ACA Medicaid expansion) (Fig 3). Of these ∼142 000 uninsured citizen new mothers living in poverty, 29% lived in expansion states and the other 71% lived in nonexpansion states. An additional 161 000 noncitizen new mothers were uninsured in 2017, some of whom may be eligible for Medicaid coverage, along with ∼146 000 new mothers with incomes above the FPL, including 21 000 with incomes 100% to 138% of the FPL in nonexpansion states.

Consistent with the growing literature on the effects of ACA Medicaid expansions on insurance coverage, we find that ACA Medicaid expansions are associated with decreased uninsurance and increased Medicaid coverage for an important population: new mothers living in poverty. Our findings suggest that the average increase in the Medicaid-eligibility threshold under the ACA is associated with a 28% decrease in uninsurance, a 13% increase in Medicaid coverage, and an 18% decline in private or other insurance among poor new mothers in expansion states.

The effects of the average increase in Medicaid eligibility on uninsurance in our sample of poor new mothers were pronounced for Hispanic and white women, with Hispanic women experiencing a 32% decline and white women a 30% decline, in contrast to a 19% decline for African American new mothers. This difference may be due, in part, to differences in preexpansion coverage by race and/or ethnicity. In expansion states in 2010–2013, African American new mothers were less likely to be uninsured than white or Hispanic new mothers and more likely to be covered by Medicaid. The racial and ethnic distribution of new mothers in expansion versus nonexpansion states may also contribute to this pattern because African American women disproportionately reside in nonexpansion states.^45,46 

The average increase in Medicaid eligibility is associated with an 18% increase in Medicaid coverage for new mothers with incomes 50% to <100% of the FPL, compared with an 11% increase for women with the lowest incomes. Again, this may be due to differences in preexpansion coverage because new mothers residing in expansion states in 2010–2013 with incomes <50% of the FPL were more likely to be eligible for and enrolled in Medicaid than those with higher incomes.

Although we find that the Medicaid expansions are associated with increased Medicaid coverage and reduced uninsurance, we also find reductions in private or other coverage. We do not actually observe reductions in private or other coverage among poor new mothers in Medicaid-expansion states. Rather, new mothers living in poverty in nonexpansion states were more likely to be covered by private or other insurance in 2017 compared with 2013, whereas we observe no change among new mothers in expansion states. Some of this may reflect the availability of subsidized Marketplace coverage at lower income levels in nonexpansion states.

In this analysis, we cannot assess the quality of the available private coverage. Private or other insurance is generally more costly than Medicaid, given typical cost-sharing and benefit structures, but may offer a broader choice of providers.⁴⁷  Maintaining Medicaid coverage in the postpartum period for women already covered by pregnancy-related Medicaid may reduce disruptions in a woman’s usual source of care.^4,36,48  Thus, there are important trade-offs between Medicaid and private coverage to consider when evaluating the importance of shifting between private and public coverage.

As policy makers at the state and federal levels consider legislation to extend pregnancy-related Medicaid coverage from 60 days to 1 year post partum,³⁵  the results of this study suggest that such policies would reduce uninsurance for women at a time when they and their families have important health care needs. In 2017, we estimate that >122 000 new mothers with incomes <138% of the FPL were uninsured and residing in nonexpansion states. These women could gain coverage if their state implemented ACA Medicaid expansion or expanded pregnancy-related Medicaid eligibility post partum. In addition, we estimate that ∼51 000 citizen new mothers with incomes <138% of the FPL were uninsured and residing in Medicaid-expansion states. These women are likely eligible for Medicaid and could benefit from investments to increase take-up of Medicaid coverage, such as outreach, education, and policies that support continuous eligibility as women transition from pregnancy-related Medicaid eligibility to eligibility as low-income adults. Additional policy efforts would likely be required to reduce uninsurance among the other nearly 277 000 uninsured new mothers, including those with higher incomes and noncitizens.

This study has important limitations. Identification of new mothers is based on ACS respondents’ answer to the question of whether a woman in their household gave birth to a living child in the past 12 months. To address concerns about measurement error, we conduct a sensitivity analysis limiting the sample to women with a birth in the past 12 months with a child <1 year of age in their household and find similar results. However, this alternative sample is likely an undercount of new mothers, given the well-documented problem of young children being undercounted in the ACS.⁴³  Because we do not know when in the past 12 months a woman gave birth, some women in our sample may still be within 60 days post partum and may still be eligible for pregnancy-related Medicaid, resulting in an underestimate of uninsurance for new mothers outside of the 60-day window. Our measures of insurance coverage are self-reported and subject to recall and social desirability biases. Although we incorporate edits to the insurance coverage measures to account for apparent misreporting, measurement error likely remains.³⁸  Finally, there is an additional measurement error in the measure of HIU income relative to poverty that we use to define our sample. There are likely reporting errors as well as family units that do not line up precisely with Medicaid-eligibility rules. Furthermore, the income measure reflects annual income, and Medicaid eligibility is determined by monthly income.⁴⁹  Finally, this analysis highlights important areas for future research, such as the relationship between state Medicaid eligibility for noncitizen women and coverage among noncitizen new mothers, and the effects of ACA Medicaid expansion on new mothers’ employment status.

Approximately 451 000 new mothers were uninsured in 2017, ∼142 000 of whom were citizen women living in poverty. Nonexpansion states could reduce uninsurance among poor new mothers by expanding Medicaid. Alternatively, they could seek waivers to extend pregnancy-related Medicaid eligibility for >60 days post partum, which could help more new mothers attend to their health needs but would not address coverage gaps among new fathers or poor parents outside of the postpartum period. Expansion states could increase coverage through education, outreach, and enrollment efforts such as waivers providing 12-month continuous Medicaid eligibility to parents. Ensuring that new mothers living in poverty have insurance coverage has the potential to help women access and afford necessary physical and mental health care, reduce stress and financial hardship, and improve the health and well-being of their entire family.

ACA

Affordable Care Act

ACS

American Community Survey

ESI

employer-sponsored insurance

FPL

federal poverty level

HIU

health insurance unit

SSI

Supplemental Security Income

VA

Department of Veterans Affairs