Deep dive 1 of 3 · Veterans' compensation

Veterans' Disability Compensation: Splitting the Stance Increase Between Legislation and Administration

Spending Decomposition 2000→2025, a deep dive on one line. Written 2026-09-03 after the split was built on Jason's decision of 2026-09-02. All figures are percentage points of CBO potential GDP unless stated (the research baseline; the website shows actual spending over actual GDP, a hundredth apart on this line); dollar figures are as published by the source named. Years are federal fiscal years, October to September.

1. Why the split matters, and what the rule is

The exhibit sorts the rise in spending into seven buckets. Three are stance: things Congress enacted inside the window, things that happened inside programs whose statute did not change (more eligible people signing up, agencies interpreting rules more generously, administrative expansions), and pre-window laws whose terms kept binding. The distinction between the first two carries weight in the argument the exhibit is built for: "the government is doing new things" reads differently when the new things were voted on than when they accreted through rulemaking and claims practice. Elsewhere the exhibit applies the rule strictly. SNAP's broad-based categorical eligibility is a state administrative option, so it is administrative. Medicaid's ex parte renewals and the ACA welcome-mat effect on people who were already eligible are administrative. The Medicare Advantage coding wedge, which no law created, is administrative. DI adjudication tightening after 2010 is administrative.

Until 2026-09-02 the veterans' line was the exception: its whole stance increase, +0.51, was labeled "ratings/presumptives/PACT (stacked)" and booked as legislation. The review flagged that most of the presumptions in the window were not statutes. The Agent Orange Act of 1991 gave the Secretary authority to add conditions by regulation, and VA used it: Type 2 diabetes in 2001, chronic lymphocytic leukemia in 2003, ischemic heart disease, Parkinson's disease and hairy-cell leukemia in 2010. The Gulf War presumptions were extended by regulation. The rating schedule was revised by regulation. Predischarge claims programs and the fully-developed-claims process are administration. Under the rule the exhibit uses for SNAP and Medicaid, those are administrative. Jason accepted the review's recommendation to split the line once the statutory pieces could be measured, and the measurement was done the same evening.

2. The line and how its stance number is built

The line is OMB's "income security for veterans" (budget subfunction 701): disability compensation, dependency and indemnity compensation for survivors, and pensions. Its counterfactual is built from disability compensation alone, which is where the growth is: the 2000 rate of compensation recipients per veteran (8.7%) applied to the actual veteran population each year, times the 2000 average payment held constant in real terms. That counterfactual falls, because the veteran population fell by a third (26.7 million to 17.6 million on VetPop2023), and the fall is split half to demographics and half to income/economy by the exhibit's rule for ambiguous residuals. Stance is actual spending minus the counterfactual.

Two things in that table are worth holding onto. Recipients per veteran quadrupled, and the average payment grew faster than the economy: both are stance under a constant-real-services benchmark, whether the cause was a statute, a regulation, or a rating decision. The recipient and payment figures are veterans receiving disability compensation (VA Annual Benefits Reports 2000 and 2025); survivors' compensation and pensions are inside the dollar line but not inside the counterfactual, a simplification the review flagged and left.

3. The events in the window, sorted by who acted

The approach follows from the table. The three statutes are sized from the scores and estimates attached to them, and everything else in the stance increase is the administrative remainder. That is the same construction the exhibit uses on the Medicaid line, where the ACA expansion group is sized directly and the traditional-group takeup is what is left.

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

4.1 CBO's cost estimate of the PACT Act, via House Report 117-249

What it is. CBO's December 2021 cost estimate of H.R. 3967, the House version of the Honoring our PACT Act, reproduced in full in the House Veterans' Affairs Committee report (H. Rept. 117-249, Part 1). CBO's own site refuses scripted requests, so the report on govinfo is where the tables were read. The estimate has a summary table for the whole bill and one table per title; the compensation rows are "Increases in Direct Spending Outlays."

What it says. The whole Act adds $17.3 billion to compensation outlays in 2025 as scored, on a path that rises to $51.5 billion by 2031 and $279.6 billion over the decade. The dip from $19.6 billion in 2024 to $17.3 billion in 2025 is the retroactive lump from re-opening previously denied claims, which CBO front-loads. The Title II process (the Secretary's new presumption-making procedure) is the largest single piece, $126.2 billion over ten years, of which $36.7 billion comes from re-adjudicating denied claims; the rest of the compensation cost is in the specific presumptions the bill writes into law (burn-pit conditions, Agent Orange locations such as Thailand, Guam and Laos, radiation sites) and the survivor benefits that follow.

How it was used. The 2025 compensation figure for the whole Act, $17.3 billion, divided by CBO's 2025 potential GDP ($30,025 billion), is the PACT piece: +0.058 percent of potential GDP. The 2023 and 2024 figures give the annual path for the chart series; the 2022 figure is dropped because the Act was signed on August 10, 2022, seven weeks before the fiscal year ended. Why the whole-bill row rather than the Title II row: the exhibit's new-legislation bucket is "what the statute added," and the burn-pit and location presumptions are as much the statute as the new process is. Why a score rather than a spending figure: no published spending series separates PACT-related compensation (section 4.3), and CBO's score is the one number that is defined the way the exhibit needs, spending caused by the Act in the year, including increased ratings for people already on the rolls.

Vintage caveat. This is the House-bill estimate of December 2021. The enacted law (the Senate amendment of June 2022) was scored at $278.5 billion of direct spending over 2022–2031, essentially the same total, but its year-by-year table could not be reached. The 2025 figure should be read as "about $17 billion," not "$17.3 billion."

4.2 VA's PACT Act performance dashboard

What it is. A monthly (now quarterly) VA publication with the operating numbers since the Act took effect. Two issues were read: Issue 53, data through September 6, 2025, and the third-anniversary issue, data through August 2, 2025. Archived in Data/veterans/pact/.

What it says, and what it does not. The dashboard counts claims and people, not the annual benefit dollars the Act generates. Its one dollar figure, $12.0 billion, is defined in the issue's glossary as the total of retroactive payments made when a claim is granted (the lump covering the months between the claim and the decision, and for claims filed within the Act's first year, back to August 2022). That is not an annual flow and cannot be used as one. The 1.61 million veterans with an approved claim include both people newly added to the rolls and people already on the rolls whose rating rose; the dashboard does not separate the two.

How it was used. As a check on the score's pace. CBO's memorandum expected 774,000 new ratings and 645,000 increased ratings by the end of 2025, about 1.42 million people. VA had approved 1.61 million veterans by early September 2025, about 15 percent more. Take-up ran ahead of the score, so the scored $17.3 billion is a floor for what the Act actually added to 2025 compensation; scaling it by that 15 percent would put the PACT piece nearer 0.067 than 0.058. The report keeps the scored figure and states the floor, because the overshoot in headcount need not translate one for one into dollars (the extra approvals could skew toward rating increases, which are smaller).

4.3 VA Annual Benefits Reports 2022–2025

What they are. VBA's annual statistical reports (246 pages each for 2023–25; benefits.va.gov/REPORTS/abr/). They are the source of the exhibit's recipient counts (2,308,186 veterans on the compensation rolls in 2000; 6,338,253 in 2025) and of the new-award series that first flagged the surge: 280,694 veterans newly awarded compensation in 2021, 322,904 in 2022, 366,971 in 2023, 457,919 in 2024 and 476,802 in 2025.

What they do not say. The reports do not separate PACT recipients or PACT dollars. The 2025 compensation chapter says only: "The annual benefit report includes PACT claims. Consolidated PACT specific reporting data is available at: https://department.va.gov/pactdata/" — that is, it defers to the dashboard. So the "PACT share of the award surge," the measurement the decision of 2026-09-02 named, cannot be read from the ABR directly; the score plus the dashboard is the closest one can get. (A first download of these reports returned a 14-page VA web page rather than the report, because the file names use a hyphen; the real files were then read in full.)

4.4 CRS report R48968 on the 2026 VA appropriation: where PACT compensation is paid from

The Act created the Cost of War Toxic Exposures Fund, a mandatory account, and one could reasonably wonder whether PACT compensation is paid from it, in which case it would sit in the exhibit's separate "VHA Toxic Exposures Fund (mandatory)" line rather than in compensation. CRS is explicit: "TEF funding does not pay the disability compensation or survivor benefit payments related to the Honoring our PACT Act of 2022 presumptions." The fund pays medical care ($55.8 billion of the $58.7 billion for 2025–26 combined) and administration. So all of the PACT compensation is inside subfunction 701, the line being split, and none of the split leaks into the TEF line. CRS also confirms the direction of the story: the rise in compensation recipients "is also a result of the expansion of eligible service-connected conditions through the Blue Water Navy Vietnam Veterans Act and the Honoring our PACT Act of 2022."

4.5 The two smaller statutes

Blue Water Navy Vietnam Veterans Act of 2019. CBO's estimate of H.R. 299: direct spending up by "almost $1.2 billion over the 2019–2028 period" (VA's own estimate at the time ran as high as $5.5 billion). Taken as roughly $0.15 billion a year.

2021 National Defense Authorization Act, section 9109. Added bladder cancer, hypothyroidism and parkinsonism to the Agent Orange list by statute. CRS Insight IN11718 reports VA's estimate of a $3 billion increase in compensation and pension obligations, including about $2.2 billion of retroactive payments, with 171,000 claims expected over five years. Taken as roughly $0.25 billion a year of ongoing benefits after the retroactive lump.

How they were used. Together about $0.4 billion a year in 2025, +0.001 percent of potential GDP: a rounding-level amount, carried as a named component so the record shows the statutes were considered, as the exhibit does for BBA-2015 on Social Security.

4.6 Sources reached but not used, and sources not reachable

• Secretary McDonough's April 2024 House testimony (docs.house.gov): PACT claim counts as of March 2024 and the TEF request; no compensation-dollar figure.
• VA's 2026 Budget Submission, Volume 3 (benefits programs): the Compensation and Pension budget model would give VA's own PACT obligations in the account. The file did not download from this machine. It is the one document that would replace the CBO score with a VA number and should be read when a machine that can reach it is used.
• cbo.gov (the enacted-law estimate of June 2022 and CBO's January 2025 discussion of compensation growth) and the Wayback Machine were blocked or rate-limited, as they were throughout the review.
• The VA PACT dashboard state tables give approved claims by state, not dollars; not needed.

5. The arithmetic

The annual path used in the chart series follows the same rule year by year: the PACT piece is CBO's 2023 and 2024 figures ($14.1 billion and $19.6 billion) over each year's potential GDP and zero before; the other-statute piece is a small ramp from 2019; the administrative piece is the line's total stance path minus those two. Stance in every year is unchanged; only its labeling changed.

Effect on the exhibit

The administrative bucket for the whole exhibit is six pieces: DI adjudication, the Medicare Advantage coding wedge, Medicaid traditional-group enrollment, SNAP takeup and BBCE, veterans' compensation growth from ratings and VA-rule presumptives, and VHA patients per veteran. The veterans piece is the largest of the six. On the website's default basis, actual spending over actual GDP, the line reads +0.41 in total, stance +0.50 (new legislation +0.06, administrative +0.44), conditions −0.09; the figures in this document are on the research baseline, cyclically adjusted spending over potential GDP, and differ by a hundredth.

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

1. "The admin bucket is a remainder, so it inherits every measurement error on the line." True, and the same is true of Medicaid's takeup piece. The line's stance total is measured well (recipients and payments are audited counts), so the remainder is only as uncertain as the statutory pieces, which are small relative to it: even scaling the PACT piece up by the 15 percent take-up overshoot moves the split by 0.01.
2. "Ratings growth is not administration; it is sicker, older veterans and a Vietnam cohort aging into disease." Partly. A constant-real-services benchmark counts more disabilities per veteran as stance because the 2000 rules, applied to the 2025 population, would not have produced them; the exhibit says so on the line and offers the "treat ratings growth as need" alternative in the program-by-program document. Within stance, though, the question here is only statute versus rulemaking, and VA's 2001 and 2010 presumptions were rulemaking. A referee who wants the aging-cohort piece out of stance altogether is arguing with the benchmark, not with this split.
3. "CBO's score is a forecast; use the actual PACT spending." There is no published actual. The dashboard's $12.0 billion is retroactive lumps; the Annual Benefits Report folds PACT into its totals; VA's budget model, the one document with VA's own PACT obligations, could not be reached. The score is the only number defined as "spending the Act caused in 2025," and the dashboard shows it is, if anything, low.
4. "The House-bill vintage is the wrong estimate." The enacted law's ten-year total ($278.5 billion) is within half a percent of the House bill's ($279.6 billion for compensation, $281.5 billion in total), and the year-by-year shape is driven by the same phase-in and re-adjudication rules. A difference of a billion dollars in 2025 is 0.003 of GDP.
5. "Survivors' compensation and pensions are in the dollar line but not in the counterfactual." Also true, and flagged in the review. Pensions fell (364,000 recipients in 2000 to 123,000 in 2025) as the World War II and Korea cohorts died, and that decline is inside the stance total as an offset of a hundredth or two; DIC rose with the PACT Act's survivor presumptions and CBO's compensation row includes survivors. Fixing this needs the ABR's pension and DIC dollar series and is a follow-up, not part of this split.
6. "The 50/50 split of the counterfactual decline is arbitrary." It is the exhibit's stated rule for residuals that have no measured cause, adopted for OASI and DI, and now for this line by Jason's decision. It affects only the demographics/income columns, not stance.

7. What would sharpen the split

• VA's 2026 Budget Submission, Volume 3, Compensation and Pension budget model: VA's own 2024–26 PACT obligations, which would replace the CBO score with an agency actual.
• VBA's compensation dollars by diagnostic code, which would let the 2001 and 2010 Agent Orange presumptions (diabetes, ischemic heart disease) be sized directly rather than left in the remainder; the 2010 rule alone was estimated at the time at roughly $13 billion over its first two years including retroactive Nehmer payments (VA's 2010 impact analysis, as recalled; not re-verified here).
• The ABR pension and DIC series, to put survivors and pensions into the counterfactual on their own terms.

Files: engine Program-Accounting/engine/engine2.py (constants PACT_COMP, OTHER_STAT; the admin component is in ADM_COMPONENTS); sources and notes Data/veterans/pact/; the veterans section of the program-by-program document; Combined/policy_components_2000.csv for the three named components.

← Back to the methodology page