Before the Supreme Court ended the constitutional right to abortion in June 2022, both sides of the debate made confident predictions about what would follow. Abortion opponents predicted that restrictions would save infant lives and that fears of a public health catastrophe were overstated. Abortion-rights advocates predicted rising infant and maternal mortality, collapsing abortion access, and an exodus of doctors from restrictive states. Four years later, the studies are in, and they have produced something neither side expected: a scoreboard that cannot be read. One prediction was confirmed. One was reversed. Two split into competing studies that reach opposite conclusions from similar data. The argument over policy has quietly migrated into an argument over measurement, and the measurement is contested at every line.
A physician-essayist recently reviewed the same record and drew a strategic lesson from it, that abortion-rights advocates should argue from freedom rather than from outcomes because outcomes are unpredictable. That is an argument about how one side should litigate. The record itself supports a different observation: the predictions failed as predictions, in both directions, and the failure says more about how policy meets data than about which side was right. This analysis takes no position on the contested policy questions it describes.
The prediction that held
The clearest result in the record concerns infant mortality. A study by Alison Gemmill and colleagues, published in JAMA in 2025, examined the fourteen states that enacted complete or six-week abortion bans and found a 5.6 percent relative increase in infant mortality, about 478 excess infant deaths over the study period, with larger increases among Black infants and for deaths tied to congenital anomalies. A separate national study by Parvati Singh and Maria Gallo in JAMA Pediatrics found infant mortality up 7 percent in the months after the decision, with deaths involving congenital anomalies up 10 percent, and attributed most of the excess to chromosomal and genetic conditions. An earlier analysis of Texas after its 2021 six-week ban had pointed the same direction.
The studies' authors attribute the pattern to pregnancies carried to term that would otherwise have ended, including pregnancies with conditions incompatible with life. That attribution is where the two sides part ways, because the same deaths read differently depending on the premise. To researchers counting infant mortality, a birth with a fatal anomaly followed by a neonatal death is an excess death. To supporters of the restrictions, that infant's brief life is the policy working as intended, not a statistic to be mourned as failure. The number is the same. The ledger it lands on is not.
The researchers themselves cautioned that the mortality figures are only the most countable surface of a larger picture. Singh and Gallo described the death counts as the visible part of a broader burden, including illness, complications, and maternal mental health consequences that do not appear in mortality tables. Those cautions cut both ways too. Skeptics of the studies point out that the post-decision window is short, that the pandemic's tail and changes in reporting complicate every trend, and that the Texas pattern has not clearly repeated elsewhere. The one point neither side disputes is that the increase is concentrated in deaths tied to congenital anomalies, which is precisely the category where the policy's effect and its critics' indictment are indistinguishable.
The prediction that reversed
The most widely repeated prediction of 2022 was that abortion would become rare in America. The national count moved the other way. According to KFF and the Society of Family Planning's #WeCount project, the United States recorded about 1.11 million abortions in 2024, up from 1.05 million in 2023, with monthly totals hitting records. The mechanism was not a mystery: medication abortion dispensed by telehealth. Roughly a quarter of all abortions now happen through telehealth, up from under 5 percent before the decision, and roughly two thirds of abortions nationally use the two-pill regimen. Shield laws in eight states allow providers to mail pills to patients in restrictive states, and courts have so far left the federal framework for mifepristone in place.
That national number also contains its own opposite. Inside the states with bans, abortion did collapse, by roughly ten thousand in-person procedures a month, and travel or mail service are not the same as a nearby clinic. Both facts are true at once: access ended where the bans applied, and national volume rose anyway, because the counter-levers, telehealth, travel, shield laws, cheaper virtual clinics, were stronger than the prediction assumed. A policy lever that large rarely produces the simple effect forecasters imagine, because the system around it adjusts.
The predictions that split
Maternal mortality and physician migration were the two predictions that dissolved into competing studies. On maternal deaths, a JAMA Network Open analysis of 22 million births found no statistically significant increase in pregnancy-associated mortality in ban states, with mortality declining in both ban and non-ban states as the pandemic wave receded. The authors emphasized wide confidence intervals, a short observation window, and persistent racial disparities, and noted that the absence of statistical significance is not proof that no effect exists. Advocates on each side read the same study as vindication.
On doctors, the literature disagrees. One study found OB-GYN supply in the twelve most restrictive states fell 4.2 percent relative to comparison states. Another, larger study of 60,000 OB-GYNs found that 94 to 96 percent stayed put in every category of state, with the workforce growing everywhere. A third body of evidence reconciles the two: clinicians who specifically provided abortions relocated at dramatically higher rates, 42 percent of them, even as OB-GYNs generally did not. The exodus happened in the narrow population where it was most expected and least visible in the headline counts.
Why the scoreboard broke
The record failed to cohere for structural reasons, not because anyone counted badly. The outcomes were never cleanly separable from everything else that happened in the same years: the pandemic's tail, changes in how deaths are recorded, and baseline disparities that long predated 2022, since the ban states had worse maternal and infant outcomes before the decision. The interventions were never clean either. A ban is not a sealed border; pills cross state lines by mail, and measurement of medication abortion depends on surveys with their own disagreements. The forecasters on both sides modeled the policy. The reality included the response to the policy, and the response was the story.
The deepest contest is not over any single number but over which number counts. The infant mortality increase is a harm to one side and an intended consequence to the other. The rising national abortion count is a failure of policy to one side and a triumph of resilience to the other. When the same figure can be entered on either side of the ledger, no additional precision settles the argument, because the disagreement was never about arithmetic. It was about the ledger.
The contest has also become institutional. Advocacy organizations on both sides now maintain their own reading of the peer-reviewed record, issuing rebuttals to each study within days of publication, funding sympathetic researchers, and publishing policy briefs that cite the studies favoring their side. This is not unusual in contested policy areas, and it is not evidence of bad faith on either side. It is what happens when a measurement becomes strategically valuable: the measurement attracts advocates, the advocates attract resources, and the public ends up choosing between interpretations rather than reading the underlying tables. None of that changes what the tables say. It changes who is trusted to say what they say.
That is worth stating plainly, because the data itself is doing real work for people regardless of politics. Whatever one believes about abortion policy, both sides should want infant and maternal mortality counted accurately, the statistics agencies that count them staffed, and the studies read in full rather than in headlines. A contested scoreboard is not merely a rhetorical problem. It degrades the shared factual floor on which the next decision, legislative, judicial, or personal, gets made.
Four years on, the one clean conclusion the data supports is that almost nobody predicted it well. The side that expected catastrophe found some of it in infant mortality and none of it in maternal mortality. The side that expected protection found fewer abortions nationwide than before, not more births. Both are left pointing at studies the other side disputes, which is what a measurement looks like when it becomes a battleground. The next prediction anyone makes about this subject should probably begin by admitting how badly the last round went.
There is a quieter finding underneath the argument, and it may be the only durable one. The families, infants, and clinicians inside these statistics are not hypothetical, and the counting that tracks them is public infrastructure, maintained by civil servants whose work both sides rely on whenever they cite a number. However the policy fight resolves, the numbers will still need to be counted, accurately and completely, by people nobody is pressuring. That is the floor the debate stands on, and the past four years show how easy it is to mistake the floor for a weapon.
Primary sources
- The Gemmill study, "US Abortion Bans and Infant Mortality," published in JAMA in April 2025, for the 5.6 percent figure, the 478 excess deaths, and the subgroup results, and the Singh and Gallo study in JAMA Pediatrics, as summarized by Healio, for the national 7 percent estimate and the congenital anomaly results.
- The Abraha and colleagues study in JAMA Network Open for the maternal mortality findings and their caveats.
- KFF's abortion trends analysis for the 2024 counts, telehealth shares, and shield law figures, NPR's reporting for the mifepristone litigation context, the workforce studies by Zhu and by Staiger and colleagues as reported by MedPage Today, and the STAT First Opinion essay by David N. Hackney for the initial survey of the four predictions.