Medicaid Expansion

What the evidence says

The strongest and most consistent finding across sources is that Medicaid expansion decreases mortality, and several independent studies converge on similar magnitudes. Miller (2021) found ACA-era expansion decreased annual mortality among near-elderly adults by 0.132 percentage points (a 9.4% reduction versus the sample mean), with the effect growing over time — from a 0.089 percentage-point decline in year 1 to 0.208 percentage points by year 4 — as internal, disease-related and health-care-amenable causes of death fell while external-cause mortality did not. Sommers (2012, NEJM), studying earlier state-level Medicaid expansions, found a 19.6-per-100,000 (6.1% relative) reduction in all-cause mortality, concentrated among nonwhite adults (−41.0/100,000 vs. −14.0/100,000 for whites), adults 35–64 (−30.4/100,000, versus no measurable effect under 35), and high-poverty counties (−22.2/100,000 vs. −11.3/100,000 in low-poverty counties). Khatana (2019) isolated cardiovascular mortality specifically and found expansion counties held steady (146.5 to 146.4 deaths per 100,000) while nonexpansion counties rose (176.3 to 180.9), a difference of 4.3 fewer deaths per 100,000 per year, translating to roughly 2,039 fewer cardiovascular deaths annually in expansion states, with the association strongest where insurance gains were largest. Maternal mortality shows the same pattern: Eliason (2020) found 7.01 fewer maternal deaths per 100,000 live births in expansion states, concentrated among non-Hispanic Black mothers (16.27 fewer) and Hispanic mothers (6.01 fewer), with no statistically significant effect for non-Hispanic White mothers (about 3.74 fewer, p=.062) — and Guernsey (2026) found expansion decreased severe maternal morbidity by roughly 50 fewer instances per 10,000 deliveries (including transfusion), an effect that appears at least 21 months after a state's expansion takes effect.

The mechanism behind these mortality effects runs through coverage and utilization, and that chain is well documented on its own. Coverage effects are large and repeatedly replicated: Medicaid enrollment or coverage rose by 12.8 percentage points (Miller 2021), 15.6 percentage points by year 2 (Miller 2017), 2.2 percentage points/24.7% relative (Sommers 2012), and self-reported uninsurance fell by 4.4 to 22.7 percentage points across studies (Miller 2021; Miller 2017; the JAMA Internal Medicine cohort). Frean (2017) further decomposed this: expansion cut the uninsured rate among newly-eligible adults by 8.9 points in 2014 and 13.7 points in 2015, while also producing a "woodwork effect" — a 2.6 to 4.6 percentage-point coverage increase among adults who were already eligible before the ACA, observed in expansion and nonexpansion states alike, and adding to roughly 60% of ACA-attributable coverage gains versus 40% from exchange subsidies. On utilization, the Oregon Health Insurance Experiment (Finkelstein 2012) found coverage increased hospital admission probability by 2.1 percentage points (30% relative), outpatient visits by 21 percentage points (35% relative), prescription drug use by 8.8 percentage points (15% relative), and preventive-care compliance by three-tenths of a standard deviation. Post-ACA cohort studies found similar increases in checkups (+16.1 points), office visits (+0.69/year), diabetes screening (+6.3 points), and regular care for chronic conditions (+12.0 points).

Financial protection is one of the most uniformly positive outcomes in these sources. Coverage reduced the probability of any out-of-pocket medical spending by 20 percentage points (35% relative) and cut medical bills sent to collections by 6.4 points (25% relative) in the Oregon experiment; a separate cohort found a 29.5% reduction in annual out-of-pocket spending from a $434 baseline, an 18.2-point drop in cost-related barriers to care, an 11.6-point drop in skipped medications due to cost, and a 14.0-point drop in difficulty paying medical bills. Miller (2017) found similar but smaller year-2 effects: a 3.4-point decrease in inability to afford follow-up care and a 7.9-point decrease in worry about paying medical bills.

Where the evidence is thinner or genuinely absent: no source here quantifies a long-run (beyond 4 years) mortality effect, no source measures Medicaid expansion's effect on chronic disease prevalence with a significant result (Chan 2020 found no measured change in diabetes, cardiovascular disease, or asthma prevalence), and depression rates showed no statistically significant change (−6.9 points, p=.08). Employment effects are reported only in direction — "no negative effects" — with no magnitude given (Obama 2016). State-level fiscal cost data is limited to a single qualitative report that one state's per-member capitation rate was underestimated by more than 12%, persistently rather than as a one-time surge (Sommers 2013).

Where the evidence splits

On self-reported general health status, the sources disagree. Sommers (2012) and a JAMA Internal Medicine cohort both found expansion increased reports of excellent or very good health (2.2 points and 4.8 points respectively) and decreased reports of fair/poor quality of care (7.1 points). Miller (2017), examining year 2 of ACA-era expansion specifically, found no significant change in health status as reported by respondents. Chan (2020) went further and found an adverse association — expansion increasing reports of fair or poor health — but flagged this result itself as low confidence, counterintuitive, and possibly reflecting omitted-variable bias or reverse causality rather than a true effect, and called for further study. The claims do not state a design difference that fully reconciles this beyond the differing time windows and specifications each study used.

On external-cause mortality, Miller (2021) found no measured decline ("mortality from external causes... does not appear to decrease after the expansions"), while Sommers (2012) found a significant 3.8-per-100,000 (7.6% relative) decrease in external-cause mortality. These are two different studies of two different waves of Medicaid expansion (Sommers 2012 examines earlier state-level expansions; Miller 2021 examines the ACA-era expansions), but the claims themselves do not explain the discrepancy — it is unexplained by the sources on this page.

On the "woodwork effect" — enrollment among adults already eligible before an expansion — Sommers (2013), based on interviews with state Medicaid officials, reported no observed evidence of a woodwork effect, attributing any enrollment increase among previously-eligible groups to economic conditions rather than expansion spillover. Frean (2017), using quantitative triple-difference analysis of survey data, found a clear woodwork effect of 2.6 points in 2014 and 4.6 points in 2015, present in expansion and nonexpansion states alike. The split tracks a method difference stated in the sources: qualitative official interviews versus a quantitative econometric estimate.

On hospital and emergency-department utilization, Finkelstein (2012), using the Oregon lottery, could not reject the null of no change in emergency room utilization and found no significant effect on the likelihood of a doctor visit or overnight hospital stay was separately confirmed by Miller (2017) for ACA-era expansion in year 2. Against this, a JAMA Internal Medicine cohort found expansion decreased any ED visit likelihood by 6.0 points and decreased reliance on the ED as a usual source of care by 6.1 points. The sources differ in design — an individual-level randomized lottery in one state versus a multi-state difference-in-differences design — but do not themselves state this as the reconciling explanation.

Thresholds and dose-response

Miller (2021) shows a temporal dose-response: the mortality-reduction effect of expansion grows with time since implementation, from a 0.089 percentage-point decline in year 1 to 0.208 percentage points by year 4. Moreno-Serra (2014) reports a spending dose-response for government health spending broadly: each $100 increase in per-capita government health spending is associated with a reduction of 13.2 per 1,000 in under-5 mortality and roughly 2.2 to 2.6 per 1,000 in adult mortality, with this marginal mortality-reducing effect reported as larger in low- and middle-income countries than in richer ones. Khatana (2019) reports a related dose-response for cardiovascular mortality: the association between expansion and mortality reduction is stronger in counties with a larger increase in the number of low-income residents gaining insurance coverage.

Open empirical questions

• No source measures mortality, cardiovascular, or maternal-mortality effects beyond four years post-expansion — whether effects continue to grow, plateau, or reverse over a longer horizon is unmeasured here.
• Chan (2020) flags but does not resolve whether its adverse self-reported-health finding reflects a true effect, omitted-variable bias, or reverse causality.
• Why Medicaid acceptance rose among other specialists but not among primary care physicians or psychiatrists is not explained by the sources.
• State-level fiscal cost data is limited to a single qualitative report from one state; no source here quantifies aggregate or long-run state budget effects of expansion.