Gross domestic product estimates are published three times per quarter — advance, second, and third — and then re-estimated annually for years afterward, by the Bureau of Economic Analysis. The advance estimate lands about 30 days after a quarter closes, built from source data covering roughly two-thirds of eventual content, with the missing third estimated by trend and judgment. The median revision from advance to third estimate is small; the distribution is not. At turning points, revisions of a full percentage point or more on annualized quarterly growth have repeatedly flipped the narrative — the 2022 quarters initially reported as negative were later revised positive, dissolving the technical-recession reading that dominated headlines that summer.
The Daily News 24 publishes information, not investment or policy advice. This explainer covers statistical procedure.
Why can't the first number be right?
The components arrive on different clocks. Trade data, corporate profits, and many service-sector surveys come quarterly, months late; the monthly retail-sales and employment figures arrive faster but need reconciliation. The BEA fills gaps with statistical estimates and replaces them as real data land. That is why the second estimate, at day 58, incorporates the first full quarter of quarterly source data, and the third, near day 90, adds the rest — with each release restating the full accounting identity of consumption, investment, government, and net exports.
What are annual and comprehensive revisions?
Every July, annual revisions fold in revised source data and methodological updates for the past several years. Every five years or so, a comprehensive revision rebuilds the whole accounts — rebasing to a new reference year, incorporating new data sources, and sometimes redefining what counts as investment, as with the 2013 switch to treating R&D and entertainment content as capital formation, which lifted the level of GDP by roughly 3 percent. Comprehensive revisions can change growth rates for decades past, which is why historical GDP series always carry vintage footnotes.
How big do revisions get?
The BEA's own revision studies put the average absolute revision from advance to third estimate at a few tenths of a percentage point on quarterly growth — but with a wide tail. The 2008–2009 recession was substantially deeper in later vintages than first reported; the pandemic quarters saw revisions of historic size because the underlying data — business closures, stimulus payments, imports of goods — behaved unlike anything in the estimation models. The practical rule analysts learn: the advance estimate tells you the direction the data point in, not the number.
Why do markets still react to the first print?
Because policy reacts to it, or is expected to. The Fed cannot wait for revised data to set rates; traders price the reaction, not the eventual truth. The information content is genuinely there — advance estimates correlate strongly with final vintages — but the noise band around a 2 percent print spans the border of "solid growth" and "stall speed," and the market treats the border crossing as news.
How should readers handle GDP reports?
Treat the advance number as a first draft with error bars: watch the direction of revisions across the three estimates for the signal, use the quarterly pattern rather than any single print, and for history, quote the latest annual revision. The BEA publishes its methodology and revision history openly — the honest way to consume a statistic is to know which vintage of it you are reading.
For more context, read What Recession Dating Actually Requires.
For more context, read consumer confidence.
For more context, read Productivity and Unit Labor Costs, Explained.