The Expenditure Side Slowed to 1.5 Percent. The Income Side Sped to 2.2.

On August 26, 2026, the Bureau of Economic Analysis published one document covering one quarter, and that document carried two growth rates for the same three months. In Bureau of Economic Analysis, GDP (Second Estimate) and Corporate Profits, 2nd Quarter 2026, the headline sentence reads: “Real gross domestic product (GDP) increased at an annual rate of 1.5 percent in the second quarter of 2026”. Further down the same release: “Real gross domestic income (GDI) increased 2.2 percent in the second quarter, compared with an increase of 1.2 percent in the first quarter.” Neither figure is a forecast, a different vintage, or a different country. They are two measurements of the same quarter, printed on the same page, and they differ by 0.7 percentage point.

The size of the gap is not the interesting part. The sign is. The same release states: “In the first quarter, real GDP increased 2.1 percent.” So the expenditure side moved from 2.1 to 1.5, a deceleration of 0.6 percentage point. The income side moved from 1.2 to 2.2, an acceleration of 1.0 percentage point. A reader watching GDP saw the economy lose most of a point between the first and second quarters. A reader watching GDI saw it gain a full point. Both readers are quoting the same release.

There is also a third answer in the same release: “The average of real GDP and real GDI increased 1.8 percent, compared with an increase of 1.7 percent.” On that reading the quarter barely moved at all, 0.1 percentage point. Three published answers, one quarter, one document. What follows tracks the 0.7 point down to the specific line of the account where it is recorded, and then asks what the September 30 release can and cannot do about it.

One Quarter, Two Measurements, One Document

The two measures are not rival estimates from rival agencies. They are the two sides of a single account, and BEA builds both. The income side is defined component by component in the agency's methodology volume: “Thus, GDI is equal to compensation of employees, plus taxes on production and imports less subsidies, plus net operating surplus, plus consumption of fixed capital.” The expenditure side is the familiar consumption, investment, government, and net exports sum. In concept the two totals are the same number. In practice they are assembled from different survey and administrative inputs, and the totals land apart.

BEA does not leave the ranking of the two implicit. Chapter 2 of Bureau of Economic Analysis, Concepts and Methods of the U.S. National Income and Product Accounts, Chapter 2: Fundamental Concepts states it directly: “In general, the source data for the expenditures components are considered more reliable than those for the income components, and the difference between the two measures is called the 'statistical discrepancy.'” That sentence does two things at once. It names the gap, and it tells a reader which side the agency trusts more when the two disagree.

Holding both of those in mind produces the tension that makes the second quarter of 2026 awkward. The side BEA calls more reliable says growth slowed. The side BEA calls less reliable says growth accelerated. And BEA publishes the average of the two as a third series, which treats them as equally informative. All of that sits inside one release.

Where the Difference Is Booked

A gap between two totals has to live somewhere, and in the national accounts it lives on one specific side. The same chapter describes the layout of the domestic income and product account: “The left (income) side of the account shows GDP as measured by the incomes earned in production—GDI—plus the 'statistical discrepancy'”. Read that as an accounting instruction rather than as commentary. The product side is presented at its own measured total. The income side is presented as GDI plus a residual that makes the two sides balance. The residual is not spread across the components; it is a single line.

That placement matters for interpretation, and this is where a writer is tempted to say that the methodology stops short of explaining the choice. It does not stop short. Footnote 23 hangs on the reliability sentence quoted above, pointing to an article in the agency's own journal, and the article gives the reason in full: “Given the important role that corporate profits and proprietors' and partnership income play as components of GDI and given the challenges of adjusting source data to accord with economic accounting measures of profits and income, BEA records the statistical discrepancy with GDI in order to reflect the relative reliability of the source data and required adjustments underlying GDI relative to GDP.”

So the placement is two claims at once. Profits and proprietors' and partnership income are named as the components carrying the measurement problem, and booking the residual there is said to reflect GDI's source data being the less reliable of the two — Chapter 2's ranking restated as a purpose. Readers who have followed this blog's earlier work on measures that turn out to rest on an estimated denominator will recognise the pattern from Capacity Utilization Is a Ratio: The Denominator Is an Estimate Revised Every Year: the published figure is exact, and one of its inputs is not.

The Deflator Closes the Easy Exit

The first reflex on seeing 1.5 percent against 2.2 percent is to reach for prices. Two real series, two different deflators, and the gap is an artifact of the price indexes. That exit is closed, and it is closed in the same chapter: “Real GDI is calculated as current-dollar GDI deflated by the implicit price deflator (IPD) for GDP”.

One deflator, applied to both. Whatever separates real GDP from real GDI in the second quarter of 2026 was already there in current dollars before any price adjustment. The gap is a source-data gap, full stop. That single sentence removes a whole family of comfortable explanations and forces the question back onto the survey and administrative inputs feeding each side.

It also means the 0.7 percentage point and the 0.9 percentage point of the prior quarter are directly comparable quantities. The gap was positive 0.9 in the first quarter, with GDP above GDI, and negative 0.7 in the second, with GDI above GDP. The swing across a single quarter boundary is 1.6 percentage points of measured growth, with no price mechanism available to absorb any of it.

The Footnote Names the Suspects

Following the footnote pays off a second time. The article that Chapter 2 cites is Bureau of Economic Analysis, Survey of Current Business: The Role of Profits and Income in the Statistical Discrepancy, and it states what the gap is taken to represent: “The difference between GDP and GDI is known as the statistical discrepancy. Both GDP and GDI provide a complete picture of economic output, so the statistical discrepancy provides an indication of net measurement error.”

Net measurement error, not a second economic fact. The same article then names two mechanisms that move the profits and income lines after the fact. First: “Because tax-based source data are used to estimate profits and income for annual and benchmark revisions, adjustments are required for misreporting by taxpayers.” Second: “To the extent that BEA cannot identify capital gains and losses attributable to corporate partners or mark-to-market accounting, measured profits and income in the NIPAs may be affected.”

Now put the second quarter of 2026 next to that. The same August release reports: “Profits from current production (corporate profits with inventory valuation and capital consumption adjustments) increased $400.9 billion in the second quarter, compared with an increase of $74.4 billion in the first quarter.” The profits line, the line the article identifies as the hard one, moved more than five times as much in the second quarter as in the first. The income side accelerated in the same quarter that the profits input jumped, and the discrepancy flipped sign.

That is a correlation inside one release, not a demonstration of causation, and it is worth saying plainly. But it does point at where the tax-based revisions will land later. The article's own long-run note is the counterweight: “However, recent work shows that GDP and GDI levels follow the same trend and rarely drift far from each other.” Levels, and trend. A quarter of growth rates is a different object.

When the Inputs Actually Arrive

The revision timetable is the part a reader can check without any modelling. Compensation of employees is the first component Chapter 2 lists in its definition of GDI. No document retrieved here states which wage series BEA benchmarks it to, so the calendar below is evidence about when wage data exist, not about BEA's source pipeline. According to Bureau of Labor Statistics, Schedule of News Releases and Full Data Availability for County Employment and Wages, first quarter 2026 data were released on Friday, August 28, 2026, and second quarter 2026 data are scheduled for Wednesday, December 2, 2026. The page also notes that “Final quarterly and annual averages data for each year will be available with the release of first quarter data (preliminary) for the subsequent year.”

That places the second quarter's own wage census more than two months after the September 30 estimate of that quarter. The profits side is slower still. The most recent tax year on the page for Internal Revenue Service, SOI Tax Stats: Corporation Income Tax Returns Complete Report (Publication 16) is 2022. The tax-return anchor for corporate profits, the anchor the misreporting adjustments are built on, is several tax years behind the quarter being estimated.

Concrete Framework

A reader who wants to use both numbers rather than choose between them can work through four checks in order.

Check one: read the sign of the gap, not its size. Compute real GDP growth minus real GDI growth for the quarter and for the one before it. Positive 0.9 followed by negative 0.7 is a different situation from two quarters of positive 0.9. A stable gap is a level offset that cancels out of growth comparisons. A gap that flips sign puts the quarter-to-quarter change itself in dispute.

Check two: ask which side the agency ranks higher, and then ask why it publishes the average anyway. Chapter 2 ranks the expenditure source data above the income source data. The release nonetheless prints the average of the two as its own series. Those are two defensible positions, and quoting one while ignoring the other is where most secondhand summaries go wrong.

Check three: locate the residual before attributing it. The discrepancy is booked on the income side as a single line, with profits and proprietors' and partnership income named as the components that carry the measurement problem. That means a surprise on the income side is a claim about those lines specifically, not about consumer spending or trade.

Check four: date the inputs against the estimate. If the wage census for the quarter has not been published and the tax-return anchor is several years old, then an early income-side figure is an extrapolation wearing a decimal point. The same discipline applied to payroll benchmarking in Benchmark Revisions Land on One Month applies here, and the same discipline applied to a trade figure that printed at two values in July's Goods Deficit Printed Twice.

What to Watch Next Week

The next observation point is fixed and close. The August release names its own successor: the third estimate for the second quarter of 2026, scheduled for September 30, 2026, bundled with industry accounts, corporate profits, state GDP, and state personal income. Three things on that page settle whether the reading above holds.

First, whether real GDI stays above real GDP. If GDI is revised down toward 1.5 percent while GDP holds, the second quarter's sign flip shrinks and the prior quarter's positive gap looks like the normal state. If GDI holds near 2.2 percent, the flip is durable through a full revision cycle.

Second, the profits line. The third estimate carries a fresh corporate profits figure. Whether the $400.9 billion increase survives, and by how much it moves, is the single most direct read on the component that the methodology names as the carrier of the discrepancy.

Third, the average series. If the average stays near 1.8 percent while its two inputs move in opposite directions, that is a reminder that the average is arithmetic rather than reconciliation.

What will not arrive next week is new wage data for the quarter. The wage census for the second quarter of 2026 is calendared for December 2, and the corporate tax anchor on the IRS page still stops at tax year 2022. The September 30 estimate will be built from the same class of inputs as the August one.

What Would Invalidate This

Several things would, and two of them are limits on what was verified here rather than disagreements about interpretation.

First, the dollar levels. No dollar figure above comes from a NIPA table; the only dollar amounts quoted are the two profits changes printed in the release text. The dollar level of the statistical discrepancy, and the dollar levels of GDP and GDI for the two quarters, sit in tables the release cites by number without reproducing, and repeated reads returned the percent changes consistently and those table cells not at all. If the discrepancy's dollar level turns out to be small relative to its own history, the sign flip matters less than the framing here implies.

Second, the share attribution. The article names profits and proprietors' and partnership income and gives BEA's reason for booking the residual with GDI, but the retrieved text assigns no numeric share of the discrepancy to any single line. Anyone claiming that profits explain a specific fraction of the second quarter's gap is going beyond that document, and so would this piece if it made the claim.

Third, the long-run counterweight. The same article notes that the levels of the two measures follow the same trend and rarely drift far apart. If a longer run of quarters shows sign flips of this size arriving routinely, then the second quarter of 2026 is unremarkable and the framing above overweights one boundary. A base rate over a decade of quarters would settle that, and building it requires the level tables that were not available here.

Fourth, the timing argument. It rests on two published calendars. If BLS moves the December 2 wage release forward, or if BEA incorporates a wage source for the quarter that does not depend on that calendar, the claim that the September 30 estimate uses the same class of inputs as the August one weakens. The calendars are public and dated, so that is checkable rather than arguable.