An individual company earns its profit by selling for more than its costs. But all companies together cannot do that against each other — one firm's cost-cutting is another firm's lost revenue. For the corporate sector as a whole, profits have to come from somewhere outside the circle of firms selling to each other. National accounting says exactly where: business and government investment, government deficits, and dividends recycled back into spending pour in; household saving and money flowing abroad drain out. That is the Kalecki–Levy profits identity — not a theory, an accounting fact. Think of profits as the water level in a bathtub: the taps and drains account for it exactly, every quarter.
The trap this page exists to avoid: reasoning about all firms the way you would reason about one firm. A company can save its way to higher profit; the corporate sector cannot, because the identity must balance.
Business investment, net. When companies build plants, buy equipment, or write software beyond what wears out, that spending is revenue to other companies without being a cost of current production for anyone. It pours straight into the profit pool. Net of depreciation, because replacing worn capital is not a profit source. Historical middle band: [+4.80, +7.92]% of GDP; the mid-century engine — its era-average contribution was 7.99% of GDP in 1947–1979 and reads 4.15% today, below its band for 5 quarters.
Government investment, net. The same mechanism with a public buyer. Band [+0.91, +1.93]; era-average 2.15% in 1947–1979 versus 0.78% in the current era — below its band for 22 straight quarters.
Dividends recycled. A dividend paid is money out of the sector — but when households spend or reinvest it, it comes back as someone's revenue with no matching production cost. The bigger the payout economy, the bigger this tap. Band [+2.58, +4.52]; it has run above that band for 63 consecutive quarters — the longest abnormal period on the panel — and contributes +6.75% of GDP today, the largest single source.
Household saving. Every dollar households DON'T spend is revenue firms never receive — the classic drain. Band [−7.83, −4.43] (a drain, so the band is negative); the drain today (−3.24%) is unusually SMALL — thinner household saving than in almost all of history (p92), which supports profits.
The federal deficit, and the state & local balance. A government that spends more than it taxes is adding net revenue to the private sector — deficits pour into the profit pool. Federal band [+0.99, +4.40]; its contribution is +5.22% of GDP today, above the band for 15 quarters, and its era-average went from 1.55% (1947–1979) to 7.87% (2020–). State & local: band [−0.31, +0.89], above it today.
Money flowing abroad. When the country buys more from the world than it sells (the full current account — goods, services, income and transfers), that spending leaks out of the domestic profit pool. Band [−2.79, +0.22]; the leg has sat below its band for 24 straight quarters (−3.37% today).
The statistical discrepancy. The national accounts measure the economy twice — once from spending, once from income — and the two never agree to the dollar. The gap is a real, small, published series (band [−0.97, +0.11]), and it is a leg of the identity, not a plug: with it included, the columns sum to reported corporate profits exactly, every one of the 318 quarters since 1947.
2000–02 — the capex unwind. The era mix shifted from investment-led 1980–2000 (net investment averaging +6.35% of GDP) into a 2001–2008 era where the investment tap averaged +5.76% and the external drain deepened to −4.91%. The drill chart carries the mechanism: info-processing equipment peaked at 2.91% of GDP in 2000Q4 and structures at 3.28% in 2001Q3. One corporation's capex is another's revenue, so the unwind pulled the investment leg down. It did not pull profits down: the federal budget's swing from surplus to deficit more than made up for it, and the profit share rose from late 2000 to the end of 2002. History, not forecast.
2008–09 — the collapse the first prints understated. The identity's era rows bracket the break: margins averaged 7.83% of GDP in 2001–2008 and 9.88% in 2009–2019 — but the real-time record was rewritten under readers' feet. The main page's tested claim (VALIDATED, n=3) documents it: the 2009 recovery's profit share was first reported far weaker than it proved, with episode revisions of +1.57 points of GDP share at the median — the deepest under-report of the three qualifying recoveries.
2020–21 — the deficit era, arithmetically. The 2020– era's computed mix: the federal deficit contributing 7.87% of GDP on average — the largest single source, and the only era where it dominates — beside dividends at 7.34%, with household saving draining just −6.10% on average. The margin averaged 10.67% of GDP, the highest of any era. And a correction the arithmetic forces on the popular story: for 2009–2019 the dominant source was dividends (5.73%), NOT the deficit (5.33%) — the "QE deficit era" narrative fails its own arithmetic there.
The page's equation, in the national accounts' own ids:
CPATAX = A557 + A889 + DIVIDEND − PSAVE − FGDEF − SLDEF + NETFI − (GDP − GDI)
It was reached by a documented ladder, not assembled to fit: the first mapping used gross nonresidential investment and the goods-only trade balance, and missed reported profits by a median 74% of their level. Each term was then corrected strictly per the identity's own documentation. Net total private investment replaced gross, and the full current account replaced the goods balance. The statistical discrepancy came in with its sign derived, which temporarily made the gap WORSE. Last came government net investment, which the investment term always included. Together these corrections closed the residual to zero at every one of the 318 quarters. No term was added or dropped for its effect on the number; the full ladder is in the public record.
Before this page was built, a rules-based "acceleration flag" on the capex components was designed and frozen in writing. It was tested on as-known-at-the-time data against two pre-named episodes — the 1990s telecom boom and the post-2022 buildout. It failed its own bar, and the failure is published rather than tuned away. The 1990s episode cannot be honestly tested at all, because the archived vintages for these series do not reach it. The ALFRED archives for Y034RC1Q027SBEA, B985RC1Q027SBEA and B009RC1Q027SBEA are the published record of what was knowable when. On today's revised data the frozen rule did not earn its keep — a share that compounds persistently stays ahead of an expanding percentile of its own growth, structurally. Reviving any successor rule requires a fresh, forward-tested pre-registration; nothing on this page flags accelerations.
All series are BEA NIPA data via FRED, quarterly, seasonally adjusted annual rates. They are net private domestic investment (A557), government net investment (A889), net dividends (DIVIDEND), personal saving (PSAVE), and federal and state & local net saving (FGDEF, SLDEF). The rest are the current-account balance (NETFI) and the capex components (Y034, B985, B009). The statistical discrepancy is derived here as GDP minus GDI from the two published levels (GDP, GDI) — BEA's own definition, exact on every stored quarter. FRED's A030 is held beside it as a check and never enters the identity. As HYG-34 found, FRED posted the annual update's revision of the discrepancy days after the other legs, leaving the identity open. Corporate net interest (FRED W255RC1Q027SBEA) is not shown on the site. Its ALFRED vintage archive is too shallow for this page's standards; that archive is the published record to check. Everything here is latest-basis — today's revised history, labeled as such, never presented as what was knowable at the time. The September annual update revises these levels, and this page's history moves with it, disclosed. Era boundaries are frozen in the feature's charter; a new era row enters only by recorded amendment.