Deep dive 2 of 3 · Medicare provider payments

Medicare Provider Payments: Zero, the Full CBO Score, or Something in Between

Spending Decomposition 2000→2025, a deep dive on one line: Medicare excluding Part D, net of premiums. Written 2026-09-03. Figures are percentage points of CBO potential GDP unless stated, the research baseline; the website's default basis (actual spending over actual GDP) differs by a hundredth or two on this line. Dollar figures are as published by the source named. Years are federal fiscal years, October to September.

1. The question

For most lines the exhibit's single question, what would the 2000 rules have cost at 2025 conditions, has one reading. For the price Medicare pays providers it has two. The 2000 rules included update formulas: the sustainable growth rate for physicians, the hospital market basket less the adjustments of the 1997 Balanced Budget Act. Read literally, "the same rules" means those formulas, and Congress overrode the physician formula almost every year, so on that reading it did not cut, it rescued. Read as the exhibit reads every other line, "the same rules" means the same care at what the care costs to deliver, which is what a fee that kept pace with providers' input costs would buy. On that reading Congress cut, by a third of a point of GDP, and the cut is stance in the exhibit's sense: the government set out to pay less for the same thing.

The exhibit takes the second reading, for the same reason it takes constant real services on the discretionary side and price-indexed benefits on the income-related side. The benchmark is what the program delivers, not the formula that delivers it. A formula that pays providers less each year than their costs rise is a stance that tightens by itself, in the same way that a nominally frozen block grant is a stance that tightens by itself. The three answers below differ on how much of the shortfall is a cut and how much was slack in the formulas to begin with.

2. The line and how its number is built

The stance figure is five named components. Three are the price test by sector: physician fees −0.15, hospital inpatient updates −0.12, the sequester −0.08. One is the Medicare Advantage coding wedge, +0.25, administrative, because no law created it and the one law that touched it cut MA benchmarks. One is the income-related Part B premium, −0.05, new legislation of 2003 and 2010. Under the three-way split of stance the physician piece is prior law, because the 1997 formula that held physician updates near zero was enacted before the window and MACRA in 2015 continued it; the hospital and sequester pieces are new legislation, the 2010 Affordable Care Act's productivity adjustment and the 2011 Budget Control Act.

Conditions is the counterfactual, the 2000 rules at 2025 prices and population with the price cuts and the MA wedge removed, which is 3.49 on the gross program against 2.18 in 2000, less the receipts that grow with it. It is split by the exhibit's usual factors: the 65-and-over share (+1.13 on the gross program), the age mix within that group, which got younger as the boomers crossed 65 (−0.06), enrollees per senior (−0.13), and spending per enrollee relative to GDP per person, the closure factor, +0.37 on the gross program. Read together with the price cuts: there was underlying cost pressure of about 0.4 percent of GDP, and Congress squeezed prices by about 0.35 of it.

3. The outcome test, sector by sector

The test compares what Medicare actually paid per unit of service with what it would have paid had fees kept pace with providers' own input costs, and applies the shortfall to all of Parts A and B, including Medicare Advantage capitation, because MA benchmarks are anchored to fee-for-service rates. It is built from 26 physician-fee-schedule final rules and 26 inpatient final rules, one per year, extracted and chained, with MedPAC and the Trustees Reports supplying the weights. No CBO score enters it. Laws like MACRA, the ACA's productivity adjustment and the Budget Control Act enter only through the fee updates they actually produced.

3.1 Physician fees against the Medicare Economic Index

Statutory updates cumulated +12.0 percent from 2000 to 2025; the MEI, CMS's index of physician practice costs, rose 62.4 percent. Fees ended 31 percent below input costs. The series reproduces MedPAC's published chart (MedPAC shows 14 percent against 52 percent for 2000 to 2023; this series gives 14.8 against 50.0) and the American Medical Association's "down 33 percent since 2001" on a 2001 base. What the table shows that the headline does not: the shortfall accrued in two stretches. From 2002 to 2009 Congress overrode every formula cut but with updates of 0 to 1.5 percent against an MEI of 2 to 3, so fees fell behind by about a point a year while the "doc fix" was being celebrated as a rescue. From 2015 MACRA replaced the formula with statutory zeros, and the 2021 to 2025 inflation did the rest.

3.2 Hospital inpatient updates against the market basket

Chained from 2000, realized net operating updates cumulated +75.4 percent against a market basket of +111.8 percent, a ratio of 0.828: rates fell 17.2 percent relative to input prices. Indexing from 1999 instead, so that the BBA-97 cut of 2000 is inside the window, gives 0.814. Against a market basket less the ACA's productivity adjustment, the benchmark that says hospitals owe the economy's productivity growth, the ratio is 0.887, 11.3 percent below. No single channel dominates the shortfall: the BBA-97 leftovers of 2000 to 2003, the ACA's productivity adjustment from 2011, and the documentation-and-coding recoupments are each a third of it, give or take.

3.3 The sequester

The Budget Control Act of 2011 created a 2 percent across-the-board reduction in Medicare benefit payments, triggered when the Joint Select Committee failed, in effect on claims from April 1, 2013. It has been in force ever since except for the pandemic: suspended in full from May 2020 to March 2022 by the CARES Act and its extensions, 1 percent from April to June 2022, 2 percent from July 2022. Because it applies to payments rather than to the update factor, it stacks on top of the two ratios above. It is worth −0.08 in 2025. It is the one piece of the price cut that is unambiguously new legislation enacted inside the window and unambiguously a cut, and it is also the one a skeptic can call temporary, since it has an expiry date that Congress keeps extending, now into the early 2030s.

3.4 Weights, and strict against lenient

The sector weights are inpatient hospital and physician fee schedule as shares of fee-for-service Parts A and B: 43.9 and 20.2 percent in 2000 (2005 values held back to 2000, because no Part A by-type table exists before 2005), 33.7 and 14.8 percent in 2025 from the Trustees' calendar-2025 estimates. Each sector's spending effect is its weight times one over the ratio less one, applied to all of Parts A and B.

Then the judgment. The strict benchmark says fees should have matched input costs in full: MEI for physicians, the full market basket for hospitals. It gives about −0.43. The lenient benchmark forgives providers about a point a year of productivity, MEI less one point (MedPAC's own recommended rule for physicians) and the market basket less the productivity adjustment for hospitals. It gives about −0.27. The exhibit takes the midpoint, −0.35, sector by sector: physician −0.15, hospital −0.12, sequester −0.08. The review of 2026-09-02 flagged that the midpoint is produced in the engine by scaling the strict series by a hard-coded 0.835 and that the sector split is likewise hard-coded; the totals match the documented midpoint but there is no derivation file, which is on the research queue.

4. The case for zero

The argument that none of this is a cut runs as follows. The MEI and the market basket are input-price indexes with no allowance for productivity, so a fee that tracked them in full would have paid providers for cost growth they should have absorbed. The 2000 formulas already assumed some of that: the SGR tied physician spending to GDP growth, and the BBA-97 hospital adjustments took points off the market basket. Updates below the indexes therefore undid over-indexation rather than cutting real generosity, and providers' real payment per unit of care was flat. On this reading the government's stance toward providers did not tighten; the care simply got dearer, which is conditions.

The best evidence for it is MedPAC's efficient-provider benchmark. Hospitals' aggregate Medicare margin was −12.1 percent in 2024, but among hospitals that historically achieved low costs with high quality the median margin was −1 percent, and MedPAC reads causation from commercial revenue to costs rather than from Medicare rates to losses: "hospitals with more revenue from private patients tend to have higher costs." That is why the Commission, even at −12.1 percent, asked Congress only to update rates by the amount in current law. If efficient providers break even at current rates, the rates are adequate, and a benchmark that would have paid more is paying for inefficiency.

What the setting does to the exhibit: stance on the line goes from −0.15 to +0.20, because the price pieces go to zero while the MA wedge and IRMAA remain. The 0.35 does not disappear; it moves into excess health cost, which on the net line falls from +0.24 to about −0.06. The zero setting therefore also asserts that Medicare's cost per beneficiary barely outran the economy since 2000, which is a strong claim in its own right.

5. The case for the full CBO figure

The CBO-Baseline-Deltas sub-project read every "changes in CBO's baseline" table published since the January 2000 baseline, 89 vintages, and classified each legislative line. For Medicare the legislative total to 2025 is −0.18 percent of potential GDP; taking out the coverage items the exhibit counts elsewhere, chiefly Part D at +0.40, leaves −0.58 of non-coverage Medicare legislation as scored at enactment. It is a remainder rather than a hand-built list of price laws, dominated by two vintages, the 2003 Medicare Modernization Act and the 2010 ACA, whose −0.535 is the 2020 scored value carried forward as a share of GDP.

The argument for using it is that the scores are what Congress intended and what the Trustees' baseline assumes will happen. The CMS Office of the Actuary makes the case in its own way: because most updates are cut by economy-wide productivity growth that providers do not achieve, "Medicare's payments for health services would fall increasingly below providers' costs," Medicare inpatient rates fell from 68 percent of private rates in 2011 to about 55 percent in 2025, and by 2040 "nearly 90 percent of hospitals would experience negative Medicare margins." The Actuary's point is that the cuts are real, compounding and probably not sustainable; if they are real, they are a cut in stance, and the scored figure is their full size.

Why the exhibit does not simply use −0.58: it bundles four things and only the first is what the outcome test measures. Provider payment-rate cuts, the BBA-97 leftovers, MMA offsets, the ACA productivity adjustment, the sequester and MACRA. MA benchmark changes, raised by the MMA and cut back by the ACA, which the exhibit handles inside the ex post MA wedge. Premium and cost-sharing changes, because CBO's Medicare is net of premiums: IRMAA, the Part B deductible, the Part D clawback, which the exhibit carries as separate legislated cuts. And the 2022 drug-pricing reforms, about −0.05 as scored, with negotiation starting in 2026 and so roughly zero in 2025 outcomes. Compared like for like, the outcome equivalents of what the −0.58 contains are the price test −0.35, IRMAA −0.05 and the clawback −0.06, together −0.45. The scores are −0.13 more negative than the outcomes. About half the gap is things the outcome method books elsewhere; the rest is that a score is a forecast against the prior-law baseline of its day while the test reads what happened, plus the carry-forward convention, under which the ACA's 2020 value held flat as a share of GDP understates cuts that compound.

What the setting does to the exhibit: the whole −0.13 is added to the line as one more legislated component, so stance reads −0.27, and excess health cost rises to +0.35. The honest headline of this setting is that the scores and the outcome test land in about the same place once the same things are compared.

6. How the exhibit distinguishes, and what the default asserts

The website pins the line at the default; the three settings are in the research register and the program-by-program document. What the exhibit relies on is that the total change, +0.90, and the demographic piece, +0.94, are the same under every setting. What moves is a third of a point of GDP between "the government paid less" and "the care cost more," and the exhibit's position is that the outcome test with a half-productivity allowance is the reading a careful referee would land on, while the other two are shown because a reasonable person holds each.

Two further distinctions the exhibit draws that a score cannot. The Medicare Advantage wedge, +0.25, is administrative: MedPAC estimates Medicare paid MA plans $76 billion more in 2025 than the same enrollees would have cost in traditional Medicare, about 17 percent, from more intensive diagnosis coding and favorable selection. Congress cut MA benchmarks in the ACA and coding more than undid it, the "cut more than undone" case, so the wedge sits on the stance side as something the government let happen rather than something it enacted. And the income-related premium, −0.05, is separated from the price cuts because it is a charge on beneficiaries rather than a price paid to providers, and it is the only post-2000 law on the receipts line: the standard Part B premium has been 25 percent of Part B costs throughout, 23.51 percent of benefits in 2000 and 23.54 in 2025.

7. The sources, one by one, and how each was used

CMS physician fee schedule final rules, calendar 2000 to 2025. Twenty-six rules, one per year, for the conversion factor, the statutory update, temporary overrides and the MEI. Cross-checked against the AMA's conversion-factor history, CMS OACT's 2010 SGR memorandum and CRS R40907. Used for the physician ratio, year by year.

CMS inpatient prospective payment final rules, 2000 to 2025. Twenty-six rules for the market basket, the productivity and legislated adjustments, and the documentation-and-coding recoupments; the Federal Register's own sentence ("updates of 1.1 percent for all areas, that is, the market basket percentage increase of 2.9 percent minus 1.8 percentage points") is quoted in the data file for each year. One chaining trap: the 2001 BIPA mid-year bump was explicitly not carried into the base, so 2.3 is chained rather than 3.4. Used for the hospital ratio.

MedPAC, March 2025 and March 2026 reports to Congress. The MA payment ratio and its decomposition (120 percent and $84 billion in the 2025 vintage, revised to 117 percent and $76 billion in 2026, almost entirely a lower coding estimate; 114 percent projected for 2026), the hospital margin series including the relatively-efficient-provider median, and the published physician-fee chart used to validate the series. MedPAC's March 2001 report supplied the 2000 MA wedge, 1.6 percent, read from a text box. Used for the wedge, the validation and the case for zero.

Medicare Trustees Reports, 2005, 2015, 2025 and 2026. Service-type weights (Tables IV.A3, IV.B6, IV.B7), the income-related premium (Table V.E3, $13.8 billion paid by about 5.1 million beneficiaries), Part B premiums against benefits (Table V.H7), Part D premiums and state payments (Table V.H8), and enrollment (Table V.B3). Used for weights, the receipts decomposition and enrollment.

CMS Office of the Actuary, illustrative-alternative memoranda and the Trustees' statement of actuarial opinion. The compounding-unsustainability argument, quoted in section 5, with its caveats: the current opinion says the estimates are reasonable "with the important caveats noted below," not "implausible," and the 2040 negative-margin shares are the end of every vintage's horizon, not 2050 or 2098. Used for the case for the full figure.

CRS R48880 and R45106. The sequester's legal history, every suspension and extension, and the 2 percent Medicare cap. Used for the sequester timeline.

CBO, changes in the baseline since January 2000, 89 vintages, and the January 2004 MMA estimate. The legislative remainder of −0.58 and its composition; the MMA box (−152, +14, −115, +88, +21, −28 in CBO's own lines, reached through a DTIC archive copy because cbo.gov refuses scripted requests). Used for the CBO-scored setting and the like-for-like comparison.

Not reachable. cbo.gov and the Wayback Machine blocked or rate-limited every fetcher during the build and the review, so a handful of CBO figures were read from archived copies or page snippets and are marked as such in the review file.

8. What a skeptic would say, and what the answer is

"The MEI and the market basket are the wrong benchmark; nobody owes providers full input-cost growth." Agreed in part, which is why the lenient benchmark exists and the default is the midpoint. The default forgives about half a point a year of productivity; MedPAC's own rule for physicians forgives a full point. Someone who wants the full point gets stance of −0.07, which is the lenient row.

"The Medicare Advantage wedge is not administrative; the MMA raised MA benchmarks on purpose." It did, and the ACA cut them back on purpose, and the 2025 wedge is larger than either intended because plans code more intensively and enroll healthier people. No law authorizes a 17 percent premium over fee-for-service, so it is booked as something the government let happen. A reader who wants it as legislation moves 0.25 between two columns on the stance side and changes nothing else.

"The midpoint scaling is hard-coded." True, flagged by the review, and on the queue: the 0.835 factor and the sector split reproduce the documented midpoint but have no derivation file. The strict and lenient rows bracket the answer whichever way that is resolved.

"The 2000 sector weights are 2005 values." True; no Part A by-type table exists before 2005. The weights move the price piece by a hundredth or two.

"The sequester is temporary." It has been in effect for all but about two of the last twelve years and is scheduled into the early 2030s. The exhibit reads 2025 as it was, and the annual chart shows the pandemic suspension.

"Scores are the law's intent; outcomes mix in things Congress did not decide." That is the strongest version of the CBO-scored case, and the exhibit's answer is that intent and outcome landed within 0.13 of each other once like is compared with like, so the choice is a matter of principle more than of size. The exhibit prefers outcomes because that is what it does everywhere else: it reads what the Social Security formula paid, not what the 1983 amendments were scored to save.

9. What would sharpen the line

A derivation file for the strict-lenient midpoint and its sector split. Post-acute care and Part B drugs in the test: the SNF and home-health update series exist in the data folder but the test covers inpatient hospital and the physician fee schedule only, and average-sales-price drug pricing is untouched. A settled reading of the 2022 drug-pricing law once negotiation is in the outcomes. And, under the three-way split of stance, a defensible division of the physician piece between the 1997 formula (prior law) and MACRA's continuation of it (new legislation), which this document assigns wholly to prior law.

Files: engine Program-Accounting/engine/engine2.py (constants PRICE_POL, MA_WEDGE, P_IRMAA; dial medicare_price); evidence Data/medicare-prices/ (policy_series_detail.csv, pfs_updates_vs_mei.csv, ipps_updates.csv, sequester_timeline.csv, ma_vs_ffs.csv, positions.md, sources.md) and Data/verify-medicare/; the Medicare section of the program-by-program document of 2026-09-02; the review memo of the same day.

← Back to the methodology page