Private Demand Revised Up to 4.2% (Q2 2026 Second Estimate): The Headline Held at 1.5% Because Imports Were Revised Up Too — and What It Means for the Dollar

The Bureau of Economic Analysis left second-quarter growth at 1.5% in its second estimate on Wednesday 26 August — and changed almost everything underneath it. Consumer spending was revised up to 3.4%, imports were revised up to 12.5%, and the two cancelled. Private domestic demand now reads 4.2%, the strongest in thirteen quarters, and the release carried the quarter's first profits estimate: up $400.9bn, of which $4.9bn reached shareholders. The headline that did not move is the least informative number in the file.

This is the revision the original version of this piece flagged as the thing to watch, and it resolved in the direction the composition pointed rather than the direction the headline did. It also landed in the same minute as the July personal income and outlays report, which is why most readers never saw it: a monthly inflation print with a familiar number beats a quarterly revision with an unfamiliar one, every time. The quarterly file changed the picture more.

What actually happened

The second estimate is normally a formality. This one was not, because the two largest revisions ran in opposite directions and happened to be nearly the same size.

Two of those rows did not exist a month ago. The second estimate is the first pass that carries gross domestic income and corporate profits, which is why it deserves more attention than it gets — a third of the release is new information rather than a refinement of old information.

Why the headline did not move

BEA's technical notes are unusually specific about the sources, and both stories are worth knowing.

The upward revision to consumer spending came from services, led by health care — mainly hospitals and physician services — on newly available Census Bureau Quarterly Services Survey data. That was partly offset by a downward revision to goods, led by recreational goods and vehicles (mainly information processing equipment) on revised Census Bureau Monthly Retail Trade Survey data for May and June, and by gasoline and other energy goods on new Energy Information Administration data for May. In other words: households spent more on medical care than the first draft assumed, and less on computers and fuel.

The offsetting import revision has a stranger origin.

Why the headline and the demand data point opposite ways

GDP measures domestic production, not domestic spending. The accounting starts from everything bought inside the country and then subtracts the portion that was made elsewhere, because those goods were not produced by the US economy. That subtraction is not a statement about economic health. It is bookkeeping.

So when imports rise 12.5% in a quarter — 15.9% for goods alone — the arithmetic guarantees a weaker headline, regardless of whether the imports were bought by a struggling economy or a booming one. On BEA's own contribution table, imports took 1.64 percentage points off measured growth and net exports 1.14. Inventories took another 0.72. Government took 0.16. Against that, consumer spending added 2.31 points and fixed investment 1.20.

Olu Sonola of Fitch Ratings, quoted by the Associated Press after the advance estimate, put the point in terms of what is driving the capex: "The consumer rescued the quarter. AI investment remains a powerful growth story, but the import surge underpinning the buildout is a reminder that an AI boom does not automatically translate into an equally large boost to U.S. GDP."

The second estimate strengthened that reading. Nonresidential fixed investment now reads 8.5%, with equipment at 13.6% and intellectual property products at 8.8%, against structures at −1.8%. The build-out is hardware and software, not buildings, and the hardware is substantially imported.

The profits line the release published for the first time

Profits from current production rose $400.9bn in the second quarter, against $74.4bn in the first, taking the level to $4,827.4bn at an annual rate — 22.8% above the $3,929.7bn of a year earlier. That is the number, and it is large. What it was used for is the more interesting part.

Dividends were essentially flat while retained earnings rose $292.9bn, to $1,617.2bn from $1,001.7bn a year earlier — a 61% increase in the pool of profit that corporations keep rather than distribute. Retained earnings are the cheapest source of investment funding a company has, and this is the same quarter in which equipment investment ran at 13.6% and the import line broke the GDP headline. The accounts do not prove one funded the other — they are aggregates, not a cash-flow statement — but they do show a corporate sector generating internal funds at a rate that makes an import-intensive capex cycle sustainable without going to markets for it.

By sector, domestic nonfinancial industries accounted for $309.6bn of the increase, domestic financial industries $71.6bn, and the rest of the world $19.7bn. Industry detail below that split is withheld until the third estimate. One line inside the financial total is worth flagging: on the inventory-valuation-adjusted basis BEA publishes by industry, Federal Reserve banks recorded profits of $37.5bn at an annual rate, against $8.4bn in the first quarter and a loss of $42.1bn in the second quarter of 2025. A central bank whose interest expense on reserves has fallen relative to the yield on the portfolio it bought years earlier swings back into profit — an $80bn annualised turn, and a mechanical consequence of the rate path rather than a policy decision.

Income says 2.2%, product says 1.5%

The second estimate is also the first sight of gross domestic income, and it disagreed with GDP in the direction the demand data already suggested. Real GDI rose 2.2%, against 1.2% in the first quarter. Real GDP rose 1.5%. The average of the two — a measure BEA publishes precisely because neither side is definitive — rose 1.8%.

GDP counts what was produced and sold. GDI counts the income earned producing it. They are the same quantity measured from opposite ends, and the gap is published honestly as the statistical discrepancy, which narrowed to $240.9bn at an annual rate from $291.5bn. When the income side runs hotter than the product side, the most common explanation is that the product side is missing activity — which is the argument the private-demand figure has been making all along.

The composition of that income growth is worth stating plainly, because it is where the profits data connects back to the household data. Compensation of employees rose $141.9bn on the quarter. Corporate profits rose $400.9bn — roughly 2.8 times as much. That is a fact about one quarter of nominal income accounts, not a trend and not a judgment, but it is the arithmetic reason the income side accelerated while the wage-driven measures did not.

The inflation split, revised the wrong way

Every price measure in the release was revised up.

These are quarterly changes at annual rates, and the direction of the revision matters more than the size. The advance estimate let you tell a story in which quarterly core inflation decelerated a full point, from 4.4% to 3.4%. The second estimate makes that deceleration 0.8 points instead, and lifts the headline PCE reading to 5.3%. Set against the monthly data released the same morning — July core PCE at 3.3% year over year, with the monthly increase concentrated in portfolio management fees — the picture is an inflation process that is not accelerating but is not resolving either.

How the dollar traded it, and what changes the picture

The market response was small, and honesty requires saying why attribution is impossible: the GDP revision and the July personal income and outlays report were released in the same minute, 8:30 a.m. Eastern on 26 August. Nobody can separate the two from price data alone.

What the data shows: on the Treasury's daily par yield curve the two-year note closed at 4.19% on 26 August against 4.17% the day before, the ten-year at 4.66% against 4.64%, and the 30-year at 5.18% against 5.17% — a curve two basis points higher, which is noise. On European Central Bank daily reference rates, which are fixed at 14:15 CET and therefore fifteen minutes before the release, the reaction shows up in the following day's fix: between 26 and 27 August the dollar gained 0.21% against the euro, 0.35% against sterling, 0.20% against the yen, 0.16% against the franc, 0.12% against the New Zealand dollar and 0.05% against the Canadian dollar, and lost 0.07% against the Australian dollar. A firmer dollar against six of seven, on a scale that would not register on a weekly chart.

That is the part worth carrying into September. In July the growth factor and the interest-rate factor pulled against each other: a soft headline argued for patience while the demand data argued the opposite. The revision removed that tension. On the current vintage, underlying demand is stronger than first reported and the quarterly price indices are higher than first reported — two of the five factors the meter scores now lean the same way, which is a cleaner set-up for the dollar than the one the advance estimate produced, without anything in the headline having changed.

Three things decide whether it holds.

The third quarter, which looks nothing like the second. The Atlanta Fed's GDPNow model put Q3 growth at 4.6% in its 26 August update, with the next revision due 1 September. The mechanical point this piece made in July is now doing the work: a level of imports that stays high is not a growing subtraction, so the arithmetic that broke the second quarter's headline flatters the third. Treat a nowcast eight weeks ahead of the print as a wide distribution rather than a forecast — but the direction is a consequence of the accounting, not an opinion about the economy.

The September meeting. The Committee held at 3.50–3.75% on 29 July on a 9–3 vote, with Beth Hammack, Neel Kashkari and Lorie Logan each dissenting in favour of a quarter-point increase, and the next decision comes on 16 September. Chair Kevin Warsh delivers the Jackson Hole keynote on Friday 28 August at 10:00 a.m. Eastern. Nothing in this release settles that meeting; it does make the hawkish dissenters' description of the economy harder to argue with than the 1.5% headline alone suggested.

The 30 September restatement. The third estimate arrives with the annual update of the national, industry and regional accounts — the first time BEA has started all three on the same day. That is a broader revision than a normal third pass, and a quarter this dependent on one trade line has room to move. The fiscal side of the same story is mapped in the term-premium breakdown, and the labour-market read that the September decision has to reconcile against it sits in the 57K payrolls breakdown.

The wider lesson is the one the headline keeps obscuring. A GDP print is not one number; it is roughly a dozen, and in a quarter dominated by trade flows the headline is the least informative of them. The advance estimate said 1.5% and the market sold the dollar. The second estimate said 1.5% and reported stronger demand, higher prices, faster income growth and a $400.9bn profits jump. Nothing in the top line moved, and the read underneath it changed materially. How currency strength is built from fundamentals rather than price is set out on the about page, and the current dollar read sits on the USD currency page.

Educational macro context only — not investment advice.