Millions of Americans who buy health insurance through Affordable Care Act exchanges face a sharp jump in what they pay each month starting in 2026, after enhanced premium subsidies expire at the end of 2025. The Congressional Budget Office estimates that at least 2.2 million people will become uninsured as a result, while broader analyses place the potential coverage loss as high as 7.3 million. With 23.1 million people enrolled in ACA exchanges for 2026 and federal officials simultaneously flagging 2.6 million improper or phantom enrollments, the actual scale of disruption is difficult to pin down but almost certain to be significant.
Why the subsidy cliff hits hardest in high-premium states
The enhanced premium tax credits, first introduced under the American Rescue Plan Act and extended through the Inflation Reduction Act, capped what most enrollees paid for a benchmark silver plan at a fixed percentage of household income. Once those credits expire after 2025, the older, less generous subsidy formula returns. For enrollees in states where benchmark premiums already sit well above the national median, the gap between what they owe and what federal assistance covers will widen the most. That gap is the mechanism that drives coverage losses: people who cannot afford the difference simply stop paying.
The mechanics of this shift are detailed in a technical analysis from the Congressional Research Service. Under the enhanced credits, a household earning 150% of the federal poverty level paid roughly 0% of income toward a benchmark plan. When the old formula kicks back in, that same household could owe 4% or more of income, a meaningful amount for families living near the poverty line. CBO projects this change will add 2.2 million people to the ranks of the uninsured in 2026 alone, a figure that reflects only the direct effect of subsidy reduction and does not account for behavioral ripple effects like delayed care or shifts to short-term plans that carry fewer protections.
States with older, sicker risk pools tend to carry higher benchmark premiums. When enhanced credits disappear, enrollees in those states absorb the largest dollar-amount increases. The hypothesis that these high-premium states will record the steepest share of subsidy-driven coverage losses is consistent with CBO’s modeling, which ties coverage decisions to net premium cost. But no publicly available federal dataset breaks down projected losses state by state or by income bracket, leaving a critical gap in the evidence and limiting policymakers’ ability to target any future relief.
Federal enrollment data and the phantom-enrollment problem
The Centers for Medicare and Medicaid Services reported that 23.1 million people enrolled in ACA exchange coverage for 2026, a figure the agency described as reflecting continued strength and stability. That number, however, carries an asterisk. A separate analysis from the Office of the Assistant Secretary for Planning and Evaluation at HHS estimated that 2.6 million questionable accounts remain on the rolls, with enrollment having peaked in 2025 before program integrity reviews began to catch up.
These two data points create an analytical tension. If 2.6 million of the 23.1 million reported enrollees are not legitimate, the effective enrollment base is closer to 20.5 million. A loss of 2.2 million from that adjusted base would represent roughly one in ten remaining enrollees, a far steeper proportional hit than the raw numbers suggest. And the higher estimate of 7.3 million potential coverage losses, drawn from broader analyses that factor in behavioral responses and shifts between employer and individual markets, would cut the adjusted base nearly in half.
CBO’s projections and ASPE’s improper-enrollment estimates were produced on separate tracks and have not been formally reconciled in any published federal document. CBO’s 2.2 million figure appears to use total reported enrollment as its baseline, meaning it may overstate the number of real people losing real coverage if a significant share of the decline comes from cleaning up phantom accounts rather than from genuine premium-driven exits. Conversely, if improper enrollments are concentrated among people who never paid premiums or used services, then the raw CBO estimate may still be the best guide to how many households will actually feel the subsidy cliff.
Another complication is timing. Improper enrollments may be removed from the rolls gradually as verification systems improve, while the subsidy rollback hits all at once in 2026. That mismatch could cause coverage statistics to swing sharply from year to year, making it harder to distinguish between the effects of policy change and routine data cleanup. Analysts trying to measure the impact of the subsidy cliff will need to disentangle these overlapping forces rather than treating enrollment as a single, cleanly measured number.
Unresolved questions about who drops coverage and why
Several pieces of the puzzle are still missing. No primary federal source distinguishes between enrollees who will actively choose not to renew because premiums are too high and those who will simply fail to complete verification steps during the transition. The difference matters for policy: the first group needs financial relief, while the second needs administrative support. Treating both as a single number obscures where intervention would be most effective and risks misdirecting limited outreach resources.
CMS has published national average net premiums after tax credits for the lowest-cost silver plan, but the agency has not released enrollee-level survey data on price sensitivity. Without that information, projections about how many people will drop coverage at various premium thresholds rest on historical patterns that may not hold in a post-pandemic insurance market. Enrollment surged in recent years partly because of simplified sign-up rules and aggressive broker outreach, some of which contributed to the improper-enrollment problem. People who came into the market under those unusual conditions may behave differently when faced with higher premiums and stricter verification.
Income distribution is another blind spot. The enhanced credits flattened premium contributions across much of the income spectrum, making coverage more affordable for middle-income households that previously received little or no help. When the old formula returns, these households will again face a hard cutoff where subsidies phase out. Yet public summaries of CBO’s work do not provide a detailed breakdown of projected coverage losses by income band. That leaves open questions about whether the coming uninsured increase will be concentrated among the lowest-income enrollees, who still qualify for some assistance, or among those just above key thresholds, who may see the sharpest jumps in out-of-pocket premiums.
The CBO’s budget window for these coverage shifts stretches from 2026 through 2035, meaning the 2.2 million figure is a single-year snapshot within a longer trajectory. If Congress extends the enhanced credits, even partially, the coverage loss shrinks. If it does not, losses compound as premiums rise and more enrollees age into higher-cost brackets. Over time, a smaller, sicker individual market could push premiums even higher, potentially triggering a feedback loop in which each year’s premium increases drive additional healthy enrollees out of the pool.
For now, the best available data point to a sharp, subsidy-driven shock layered on top of an enrollment base that is itself less solid than headline figures suggest. How many people ultimately lose coverage in 2026 will depend not only on the statutory end of enhanced tax credits, but also on how aggressively federal and state officials police improper enrollments, how insurers price plans in anticipation of a leaner risk pool, and how households weigh rising premiums against competing financial pressures. Until more granular data are released, policymakers will be forced to navigate that uncertainty while the clock on the subsidy cliff continues to run down.
