The headline slowed. The domestic economy did not.

Every rate below is real, meaning adjusted for inflation, and annualized, meaning the change over three months multiplied out to a yearly pace: a 12.9% annualized fall is roughly a 3.4% fall over the quarter itself. Contribution means percentage points of GDP growth — a component's growth rate multiplied by its share of the economy. That is the number worth ranking, because a violent percentage move in a small component barely shifts GDP.

What moved

Consumer spending rose 3.2% against 0.5% in the first quarter, with goods up 5.2% and services up 2.2%. Because consumption is roughly two-thirds of the economy, that was worth about 2.2 points. Business fixed investment grew 8.4%, led by equipment and intellectual property, adding roughly 1.2 points. Together they produced private domestic final demand of 3.9%, or about 3.4 points, more than double the headline.

Everything below the headline was taken back by two components: a drawdown in business inventories and a wider trade gap, together removing something close to 1.8 points. These are the noisiest lines in the national accounts, the most heavily revised, and the ones an advance estimate knows least about.

Inflation does not explain any of that — real GDP is already adjusted for it — but it does explain the size of the nominal economy. Current-dollar GDP rose 7.9% against real growth of 1.5%, a wedge of more than six points. The price index for gross domestic purchases jumped to 5.7% from 3.6%, and headline PCE inflation to 5.1% from 4.6%. Core PCE, which strips out food and energy, fell to 3.4% from 4.4%. That split points at energy, food and tariff-affected import prices rather than at broad domestic inflation.

Contribution figures are estimates, derived by multiplying each component's published growth rate by its approximate share of GDP and solving for the residual. Two decimals are shown so the arithmetic can be checked, not because the estimates are that precise: each carries roughly ±0.1 to ±0.2 points of error. The inventories-and-trade line is a combined residual that cannot be split between the two without the detailed source table. The underlying growth rates are as published.

Government: a large number attached to a small line

Federal nondefense spending fell 12.9%, the most violent number in the release. It cost the economy about a third of a point, because federal nondefense purchases are under 3% of GDP. Defense rose 2.4% and state and local rose 1.1%, offsetting most of it. Net government drag: roughly 0.14 points, the smallest of the four drivers.

^† The Q1 federal figure comes from the first-quarter advance vintage and may have shifted in later revisions. Q1 sub-components are not shown because no equally reliable published figure exists for them.

The 0.9-point swing from Q1 comes almost entirely from the fact that Q1 was itself distorted. Federal spending surged around 9% then, clearing a backlog that the fourth-quarter 2025 shutdown had suppressed, when federal spending fell at a 16.6% rate. The likeliest reading — an interpretation, not a measurement — is that the three quarters form a single timing distortion, and that federal purchases across them run closer to flat than any individual print suggests.

Spending levels, and the question of cash

The level of government spending is not higher than in the first quarter. A 0.8% annualized decline is a 0.2% decline over the quarter itself, so real government purchases sit marginally below Q1 — essentially flat, not a cut. In nominal dollars the level almost certainly rose, since the relevant price index ran near 5%. Real terms and nominal terms give opposite signs here, and the phrase "spending fell" holds only in the real, annualized one.

Nor is the government accumulating cash. The Treasury General Account, the government's operating balance at the Federal Reserve, averaged $869.3bn in Q1 and $865.3bn in Q2 — a decline of four billion, which is noise. The quarter-end reading rose from $893.0bn to $919.1bn, but quarter-end is the least informative moment to look: corporate tax receipts and settlement timing spike it every quarter, and the intraquarter range in Q2 ran from $687.0bn to $1,038.0bn. More fundamentally, that balance is funded by issuing debt. It measures cash management, not thrift.

The distinction that matters: the government line in GDP is not the federal budget. It counts only what government buys — goods, services, investment. It excludes Social Security, Medicare, Medicaid and interest, which are transfers and make up most of the budget. Purchases can therefore fall while the fiscal position deteriorates, and that is what happened. The first nine months of fiscal 2026 ran a $1.4tn deficit, $29bn worse than the same period a year earlier, on about $4.2tn of receipts against about $5.5tn of outlays, with spending up $172bn year over year on entitlements and interest. June alone ran a $120bn deficit, $147bn worse than June 2025. The Committee for a Responsible Federal Budget projects borrowing above $2tn for the full year — a forecast, not an observation.

What it means

Government purchases and the deficit currently point in opposite directions, and only purchases enter GDP. The fiscal impulse in the growth accounts looks contractionary while the borrowing requirement grows. For anyone holding a view on rates or duration, the second drives issuance and the first does not.

On growth, roughly 1.9 points of drag — the whole distance between the 1.5% headline and the 3.4 points private demand contributed — came from components that mean-revert. If federal nondefense merely stops falling and inventories stabilise, the headline rises toward private demand without anything real improving. The risk to that reading is that this is an advance estimate resting on incomplete trade and inventory data — precisely the components carrying the argument. The second estimate lands on 26 August 2026.