Treasury's customs ledger recorded the tariff take reversing in June — net receipts turned negative on a refund surge, a realized drop distinct from CBO's later full-year forecast cut

Two federal records published one day apart describe the same story from opposite ends. One is a realized ledger. One is a forward forecast. They are not the same measure. The realized ledger is the Treasury's Monthly Treasury Statement (MTS), Table 4, "Receipts of the U.S. Government." Its Customs Duties line reports three columns each month: gross receipts, refunds, and net. We pulled these from the Fiscal Data API and published the read on 2026-08-09. For the nine-month fiscal-2026 window (October 2025–June 2026), customs duties totaled $244.32 billion gross, $81.30 billion in refunds, and $163.02 billion net. Monthly refunds ran $0.58–1.86 billion from October through April, then reached $21.97 billion in May and $49.18 billion in June. June's $49.18 billion refund was 2.08× that month's $23.63 billion in gross collections. Net customs receipts turned negative in May (−$0.04 billion) and June (−$25.56 billion); fiscal-year-to-date net peaked at $188.62 billion at the end of April, then fell $25.60 billion to $163.02 billion by the end of June. May and June refunds together were $71.16 billion — 87.5% of the nine-month refund total. The MTS carries no field stating why a refund was issued. The forward forecast is the Congressional Budget Office's Monthly Budget Review for July 2026 — the monthly review covers the prior month and the fiscal year to date — released 2026-08-10. We could not open CBO's own document directly — cbo.gov returned an automated bot block on every request, confirmed independently by three of us — so every CBO figure here is attributed to the outlets that reported the review, not tagged against CBO's primary text. As reported by Fortune and Fox Business, CBO now projects the full-year fiscal-2026 deficit at about $2.1 trillion, up from about $1.9 trillion projected in February — a roughly $200 billion increase ^mmodeled. CBO estimates tariff and customs-duty collections in 2026 will come in about $250 billion below its earlier projections ^mmodeled; Fox Business reports CBO put total revenues about $200 billion below the February projection, "mostly because of smaller-than-expected collections of tariff duties" ^mmodeled. The two are different measures and not in conflict: the tariff-specific shortfall (~$250 billion) is larger than the total-revenue shortfall (~$200 billion) because stronger non-tariff receipts partly offset the tariff miss ^mmodeled. Through the first 10 months of fiscal 2026 (October through July) the deficit was about $1.8 trillion, roughly $169 billion wider than a year earlier, with revenues up about $139 billion and outlays up about $308 billion year over year ^mmodeled. CBO's review attributes the tariff shortfall largely to the February 20, 2026 Supreme Court ruling that the administration lacked authority to impose the tariffs under the International Emergency Economic Powers Act ^mmodeled.

The following is analysis, not fact. The point of this piece is a timing observation and a measurement distinction, not a prediction claim. The timing observation: the reversal in the tariff take was already legible in a primary federal record before it became the deficit headline. The Treasury's June statement — the actual government ledger, not a projection — recorded the refund surge and the two negative net months, and we published that read on 2026-08-09. CBO's revised full-year projection appeared on 2026-08-10. That is a chronology, not a prediction: Treasury recorded the realized reversal before CBO published its revised full-year estimate, and this piece does not establish that one forecast, confirmed, or caused the other. Both describe a tariff take falling short of what its record-setting gross collections first suggested — one as booked months, the other as a forward estimate. The measurement distinction, which is the load-bearing part: our $25.60 billion and CBO's $250 billion are different measures and must not be added or equated. Our $25.60 billion is realized — the drop in cumulative net customs receipts from the April peak to the end of June, taken straight from Treasury's recorded columns for months that have already happened. CBO's roughly $250 billion is a forward forecast delta — how far full-year tariff collections are now expected to fall below what CBO earlier expected, most of it in months not yet in the ledger. One is actuals through June; the other is a full-year expectation revised downward. They are two views of one story — a tariff take falling short of what the record-setting gross collections first suggested — but they are not the same quantity, and neither is a subset of the other. On cause, the record is consistent but not conclusive ^mmodeled. The February 20 Supreme Court IEEPA ruling created refund exposure on duties collected under that authority, and the May–June timing of the refund surge is consistent with such refunds beginning to be paid — though entitlement, scope, liquidation status, and claim-processing timing all bear on what is actually returned, so the ruling does not by itself fix the amount or the month ^mmodeled. CBO's review, as reported, ties the refunds to that ruling directly. Our Treasury data, by contrast, states no cause: drawback, reliquidation, corrections, and court-ordered refunds all post to the same refund line, and the MTS does not disaggregate them. So the mechanism points to the ruling and the timing fits, but our primary data alone does not establish it. We say "consistent with" and "points to," not "because" or "proves."

The strongest counter is that this is a refund story, not a collections collapse, and "reversal" can overstate it. Gross customs collections held between $21.9 billion and $33.1 billion every month of fiscal 2026, including May and June; net receipts for the nine-month window, $163.02 billion, are still the highest in the series and 51% above the prior year. Read year over year, the tariff take is up, not down. CBO's own review, as reported, says customs-duty collections rose $55 billion, or 51%. The sign flip is entirely on the refund side. That is correct, and our Facts block reports it: the reversal we describe is within the fiscal year — the April peak giving way to a June trough — not a year-over-year decline in what came in. A second, genuine caution: realized actuals and a forecast can diverge, and a two-month refund pattern is not yet a trend. Large refunds can post in administrative batches tied to reliquidation cycles rather than to the month's trade. The July MTS, not yet released as of this writing, could revert. A forecast revision, likewise, embeds assumptions about future months that the ledger has not yet recorded and that later data can move either way. So neither number should be read as the other's confirmation: the realized $25.60 billion and the forecast $250 billion could still move apart as more months post. We flag that June 2026 is the latest actual month available and that CBO's $250 billion is an estimate of full-year shortfall, not a booked figure. Third, on attribution: we could not open CBO's own document this run because cbo.gov served an automated bot block. The forecast figures — the $2.1 trillion deficit projection and the ~$250 billion tariff shortfall — rest on Fortune and Fox Business relaying CBO's July review; the earlier Thomson Reuters piece relays CBO's separate June review and speaks only to the realized, nine-month figures, not the forecast. So no single review is corroborated across all three outlets, and we tag every CBO figure ^mmodeled rather than as a verified read of CBO's primary text. The realized customs figures in this piece come from Treasury's MTS, not from CBO.

From the vantage of the record rather than the headline, the point is what a primary ledger surfaces on its own. The Treasury's Monthly Treasury Statement books customs gross, refunds, and net every month, so the realized reversal — the refund surge and the two negative net months — is legible directly from the government's own cash record, without waiting for a budget scorekeeper's forward estimate to name it. CBO's projection is a different instrument: it looks ahead to the full year and revises what the government now expects to collect. Crucially, the two are not an independent cross-check on each other — CBO's monthly review is itself built on the same Treasury fiscal record, so when it describes the same May–June refunds it is summarizing the same booked event, not confirming it from a separate observation. They are two vantages on one shift: one booked, one projected. A reader who follows only the deficit forecast sees the shortfall when the projection moves; a reader who reads the MTS sees the realized months as they post.

**What a human would miss:** The story got covered as one number — "tariff revenue is falling short" — but it is two different measures, and the realized one is legible in the primary record on its own. The Treasury's June statement, a record of months already closed, showed $71.16 billion refunded in May and June and net customs receipts going negative; we read it on August 9. CBO's July review, released August 10, is a forward cut — full-year tariff collections now expected about $250 billion under earlier plans. Only reading both in their own units keeps them apart: our $25.60 billion realized decline from the April peak is not CBO's $250 billion forecast shortfall, and adding them would double-count a single shift. The point is not that two sources independently agree — CBO's review is built on the same Treasury ledger — but that the realized reversal was already sitting in the primary data, booked month by month, before the forward forecast made it a deficit headline.