NBER took 12 months to date the 2007 peak and 4 to date 2020

On December 1, 2008, a committee of economists announced that the United States had entered a recession in December 2007. By the time they said so, the recession was twelve months old. Payroll employment, by the committee's own account, had peaked in December 2007 and declined in every month since.

That is not a failure of the committee. It is how the system is designed to work, and the design is worth understanding, because markets use the word “recession” as if it were a switch someone flips. It is a label applied to history, by a body with no formula, no deadline and no obligation to say anything at all.

Who decides, and what they look at

The recession dates everyone uses come from the Business Cycle Dating Committee of the National Bureau of Economic Research, a research organisation. The committee does not define a recession by real GDP alone. Its own words:

A recession involves a significant decline in economic activity that is spread across the economy and lasts more than a few months.NBER, Business Cycle Dating Procedure: Frequently Asked Questions, retrieved October 12, 2026

Two of those six series come from the monthly jobs report. The August 2026 release put payrolls at +162,000 and unemployment at 4.1%. Payroll employment is one of the committee's two heaviest-weighted inputs; the unemployment rate is not on the list at all.

Why two negative quarters is not the definition

The rule of thumb that a recession is two consecutive quarters of falling GDP appears nowhere in the committee's procedure. The FAQ addresses it directly and gives four reasons for not using it.

1. The committee does not treat economic activity as the same thing as real GDP. It looks at a range of measures.
2. The definition requires a significant decline. A rule based on the sign of GDP growth ignores how deep the fall is.
3. The committee's main product is a monthly chronology. GDP is quarterly, so the monthly indicators do most of the work.
4. It gives equal weight to real gross domestic income, the income-side measure. The FAQ notes that the gap between GDP and GDI, which statisticians call the statistical discrepancy, was important in both 2001 and 2007 to 2009.

The 2007 announcement makes the same point from the other direction. The committee reported that real GDP fell slightly in the fourth quarter of 2007, rose slightly in the first quarter of 2008, rose again in the second quarter and fell slightly in the third. On a two-quarter rule, no recession existed on December 1, 2008. The committee dated one to December 2007 anyway, because payroll employment had peaked that month and declined every month since, and real personal income less transfers had peaked the same month.

The lag, episode by episode

The committee publishes every announcement it has made since 1979, and the FAQ tabulates the elapsed time for each. Peaks are called in 4 to 12 months; troughs take longer.

The 2010 announcement explains the delay in its own terms. The committee waited for the national-accounts revisions of July 30 and August 27, 2010 to settle the path of real GDP and GDI through 2009. It also noted that payroll employment reached its low 21 months after the official trough in the 2001 to 2003 episode, and 6 months after it in 2009. Employment turns later than the economy does.

What the chronology does not do

The trough announcement of July 19, 2021 contains a sentence that is easy to skip: the chronology “does not identify the precise moment that the economy entered a recession or expansion”. By convention the first month of a recession is the month after the peak and the last is the month of the trough. March and April 2020 were the recession; February was the peak.

Nor is the chronology revised. The FAQ states that the committee has not changed any turning-point date it has announced since 1978. That choice, to wait rather than call early and correct later, is the source of the lag. The underlying data are revised constantly, as the site's piece on data revisions shows.

Finally, the committee does not predict. It has no model and issues no outlook. Indicators that markets watch for early warning, such as the slope of the Treasury curve on this site's yield curve page, are a separate exercise. The 10-year Treasury yield was 5.00% on September 15, 2026; yields are not among the committee's inputs and it has never commented on them.

What it means for your money

• A household that waits for the official word “recession” before adjusting is, on the committee's own record, acting 4 to 12 months after the peak. In 2008 the announcement arrived after twelve months of falling payroll employment.
• The lag runs in reverse at the bottom. Fifteen months of expansion passed after June 2009 and after April 2020 before either was official. A rule of “stay out until the recession is declared over” would have kept a saver out for over a year of recovery in each case.
• The two series the committee weights most heavily, payroll employment and real income less transfers, are published monthly by the BLS and the BEA. They are the closest thing to an up-to-date version of the committee's inputs.
• An emergency fund is sized to the possibility of job loss, not to the announcement of a recession. Employment bottoming 21 months after the 2001 trough is why such a fund is measured in months of expenses, not in recessions.