Every Federal Dollar, Mapped

Study IV ended at the account — the unit where an appropriation lands and a Treasury symbol is assigned — and showed that twenty of them do most of the work. This study starts there and asks the questions the account level makes answerable: which accounts moved, by how much, and why, over five years; what the receipt lines that feed them did over the same five years; what the money actually bought; where it landed; how the twelve bills that are supposed to set a quarter of it were marked and enacted for four fiscal years; who lobbies for which dollar; what Congress directed by name; and what is already promised beyond the horizon.

Two rules run through every plate. The first is about basis: account-level figures are gross outlays from USAspending, which do not sum to the $7,010.0B net total that the Monthly Treasury Statement publishes, because the budget nets out collections and because a general-fund payment into a trust fund is counted once leaving the general fund and again leaving the trust fund. Those internal transfers are hatched wherever they appear. The second is about time: Treasury counts in fiscal years, the Trustees and the lobbying registers in calendar years, and FY2026 is ten months old and not closed. No plate mixes them.

I. The map

Eighteen agencies account for the bulk of federal outlays. Between them they ran 1,312 accounts in FY2025, and the shape of the map is the shape of the argument: two accounts — Social Security’s Old-Age and Survivors Insurance and interest on the public debt — are more than a quarter of it, and $1,300B of what looks like spending is money the government paid to itself.

II. Five years at the account level

The fifty largest accounts, one row each, with the whole five-year path. Read the sparklines for shape and the last two columns for size. Three shapes recur: a steady climb (Social Security, Medicare, Medicaid, veterans, interest), a pandemic spike that collapses (unemployment, student loans, the child tax credit), and a policy step (the toxic-exposures fund, CHIPS, the ACA premium credit).

Concentrate the change into one picture and the five years have two stories, not one. Interest on the public debt rose $653B — more than the entire gross outlay of Veterans Affairs and Homeland Security together. At the other end, five pandemic accounts gave back $1,339B. Net those two against each other and the government’s outlays grew 2.8% in five years while its receipts grew 29.4%. Neither number means what it appears to mean without the decomposition.

III. Every receipt line

The receipt side has fifty-seven lines in Treasury’s Table 4. Sixteen of them carry the money; the rest are adjustments. Over five years the payroll taxes grew a third, the individual income tax a third, the corporate tax a fifth — and then reversed in FY2025 — and two lines moved in ways that no tax law explains: Federal Reserve remittances fell from $100B to $5B when the System’s interest expense overtook its earnings, and customs duties went from $80B to $195B, and are now being refunded at a rate that exceeded collections in July 2026.

IV. What it buys, and where it lands

An account tells you who spends. An object class tells you what for. The federal government is two governments on this cut: one that employs people and buys things (Defense, Homeland Security, Justice, NASA) and one that writes cheques (HHS, Treasury, SSA, Education, HUD). Defense spends 40% of its obligations on people and 16% on hardware; HHS spends 0.7% on people and 91% on grants and fixed charges.

Geography is where the public data most often lies, and it is worth showing the lie. USAspending’s own per-capita figure puts North Dakota first in the nation at $130,295 per resident. Nobody in Fargo is receiving that; Medicare’s administrative contractor is. Grants and contracts are located by place of performance and can be ranked. Direct payments cannot.

V. The twelve bills, four years

Study IV established that Congress enacted none of the twelve regular bills for FY2025. Here is what the process produced across four fiscal years, bill by bill: what the House marked, what the Senate marked, what was enacted, and what the President asked for where the request is on a comparable basis. Two things are visible at once. The House–Senate gap is routinely larger than the year-on-year change in the enacted number. And for FY2027 there is no Senate column at all: the committee postponed its June markups over the defense and non-defense topline and had reported no bill by 4 September 2026.

VI. Who lobbies for which dollar

Federal lobbying reported $5.13B in 2025, a record, from $3.78B in 2021. The single most-lobbied subject in Washington is the federal budget and appropriations — 5,204 clients, up a third in four years — and the two most-named targets on every report are the House and the Senate. Map that lobbying to the accounts it works and a pattern appears that is the reverse of the account map: the biggest accounts draw the least lobbying per dollar, and the largest of all, interest on the debt, draws none.

The other channel by which money is put into named accounts is the one Congress reserves for itself. Earmarks — community project funding in the House, congressionally directed spending in the Senate — returned in FY2022, reached roughly $15B a year, were zeroed for FY2025 by section 1111 of the full-year continuing resolution, and came back at roughly $17B in FY2026. What counts as an earmark depends on who is counting, and the plate shows three counts side by side rather than choosing one.

VII. The long funds and the horizon

Most of the money in the account map is not decided year to year, and a large part of it is not decided at all: it flows from trust funds whose reserves are drawn down on a schedule the Trustees publish, from balances Congress appropriated years ago and has not yet spent, and from statutes whose ten-year cost is known to the billion. The plate puts those on one page: reserves and depletion dates, who holds the $7.34 trillion of debt the government owes itself, $2.4 trillion appropriated but unobligated, and CBO’s path to 2036 and beyond.

VIII. The flow, whole

Study IV drew the flow as two columns. With the account and object-class work done, it can be drawn as three: what came in, which agency it left through on the official net basis, and what it bought. The dark inflows are the payroll taxes, which by law fund the accounts they connect to; every other band is an allocation in order, which is the honest way to say that a dollar of income tax has no destination until Congress and the permanent law give it one.

IX. The cycle

The last plate is the one Study IV said no pie chart could draw, and it is now built in seven movements. First a bridge: the year begins with $5,617B collected gross, gives back $383B in refunds, borrows $1,775B, and publishes $7,010B — but $928B more was collected from the public and netted against outlays, and $1,251B more moved between agencies, so $9,189B actually left an account. Second, who paid: the top one per cent of returns paid 38 per cent of the income tax on 21 per cent of the income and the bottom half paid 3 per cent; 4,759 firms with assets above $2.5 billion paid 80 per cent of the corporate tax; employers and employees split $1,584B of FICA to the dollar; $90.5B of the $195B in customs came from the IEEPA tariffs the Supreme Court has since struck down; and every dollar of the $1,775B borrowed was lent by the private sector — mutual funds, hedge funds and households above all — while the Federal Reserve let $168B run off. Third the channels: seventeen receipt lines enter, three legal doors let the money out — permanent law, the twelve bills, and the permanent appropriation for interest — it reaches the seven categories CBO scores, and it buys benefits, grants, interest, people, contracts and assets. Fourth the accounts: the sixty-odd Treasury accounts that carry each of those seven categories, from the OASI trust fund at $1,422B gross to the NIH institutes at a few billion each. Fifth, who received it: 77 million people on Medicaid and CHIP, 70 million on Social Security, 69 million on Medicare, 42 million on SNAP; California’s $172B of grants; Lockheed Martin’s $74B of contracts; $267B of civilian pay across eighteen agencies and $177B of military pay; and the $970B of net interest allocated to the foreigners, funds, banks and households who hold the debt. Sixth, what comes back, itemised: Medicare premiums, student-loan re-estimates, Postal fees, TVA power, foreign military sales, the interest the Treasury pays its own trust funds. Seventh, the loop closed: five arrows between the public and the government, drawn to their dollars, and they reconcile to the decimal — $8,320.4B in, $8,320.4B out. That is the cycle, and it is why the same year can be $7.0 trillion, $7.9 trillion or $9.2 trillion depending on where in it one chooses to stand.

X. The cycle, from the household and from the state

Plate XII follows the dollar through the Treasury. Plate XIII stands outside it and asks the two questions the Treasury tables cannot answer: which households the cycle takes from and gives to, and which states. The first is CBO’s work and the latest year is calendar 2022. Rank households by income and cut them into fifths, and the lowest fifth paid $400 in federal taxes per household — its income tax was minus $2,700, refundable credits exceeding liability — and received $26,200 in social insurance and means-tested transfers: $7,200 of Social Security and Medicare, $12,600 of Medicaid and CHIP at cost, $2,700 of SNAP. The highest fifth paid $106,700 and received $17,800, almost all of it Social Security and Medicare. Net, the bottom three fifths took $1.5 trillion out of the cycle and the top two put $2.4 trillion in; the top one per cent alone, 1.2 million households, paid $850,000 each, $147,000 of it corporate tax imputed to their shareholdings, and put in a trillion dollars. CBO’s incidence rules are the second finding: 29 per cent of the average household’s $29,400 in federal taxes never appears on anything it signs — the employer half of the payroll tax, borne as wages not paid; the corporate tax, borne as dividends not received and wages not paid; the excise taxes, borne in prices. The second question is the Rockefeller Institute’s, and the latest year is federal fiscal 2023. Assign 96 per cent of receipts to the residents who paid them and 91 per cent of outlays to the places where they landed, and forty-seven states got back more than they paid in. Virginia got $2.22 per dollar and Maryland $1.92, because of contracts and federal wages; New Mexico $2.91, Alaska $2.26 and Hawaii $1.91 because of grants and wages; Mississippi, West Virginia and Kentucky between $2.28 and $2.57 because direct payments met the lowest receipts in the country. New Jersey, Massachusetts and Washington were the three that paid in more than they got. The fifty balances sum to $1,303 billion, which is the allocable part of the year’s deficit: what the states got back that nobody paid for was borrowed, and Plate XII shows from whom.

What the map says

Put the thirteen plates together. The federal government’s money is concentrated in a handful of accounts whose growth is set by demography and interest rates, not by appropriation; the twelve bills that Congress fights over decide the shape of a quarter of the whole and, in three of the last four years, did not decide even that on time. The lobby follows the decidable dollar, not the large one: the most lobbying per dollar attaches to the small discretionary accounts — highways, housing, nutrition — and the least to Medicare and Social Security, and none at all to the largest single mover of the five years, interest. Congress’s own instrument for directing money by name, the earmark, is at most $17B a year against $7 trillion, and was switched off entirely for one of the five years by a formula extending the previous year.

And the horizon is already written. The reserves that pay Social Security retirement benefits are on a published path to Q4 2032; Medicare’s hospital fund to 2033; the Highway Trust Fund’s transit account to FY2027. $2.4 trillion is appropriated and waiting. CBO’s baseline carries net interest past $2 trillion in 2035 and debt held by the public past the 1946 record in FY2030. None of that appears in any pie chart of federal spending, and none of it can.

And the cycle has a shape when it is seen from outside the Treasury. It takes from the top two fifths of households and from three states, and it gives to the bottom three fifths and to forty-seven states, and the part of what it gives that nobody paid for — $1.3 trillion in FFY 2023 — it borrows, largely from the same top fifth that pays the taxes, who are then paid interest on it. The redistribution is real, it is large, and a third of what funds it is invisible to the people who fund it.