GDP Advance vs. Second vs. Third Estimate: Why the Number Changes

A GDP estimate describes economic activity for a period, such as a quarter. A vintage is the version of that estimate available in a particular release. The quarter answers “what period is being measured?” The vintage answers “which release's estimate am I looking at?” A chart can show the same quarter in several vintages and therefore show different values without changing the period being discussed.

BEA publishes three current quarterly estimates for each quarter: advance, second, and third. Those names describe the sequence of current estimates, not three different definitions of GDP. The BEA's [NIPA Primer]({source:beaNipaPrimer}) explains why estimates arrive in vintages as source data become available; the current [GDP release information]({source:beaGDPReleaseRevisions}) gives BEA's present schedule and definitions.

This distinction matters when comparing a number reported in an earlier release with the latest historical series. The earlier value records what BEA estimated with the information then available. The latest value reflects a later information set and can still be subject to future updates.

The estimates follow a recurring quarterly sequence. BEA describes the advance estimate as arriving near the end of the first month after a quarter ends, the second near the end of the next month, and the third near the end of the following month. Actual calendar dates are listed on BEA's current release calendar and can vary; the useful distinction is the order and information available at each release.

All three releases estimate the same quarter. The second estimate does not measure the next quarter, and the third estimate is not a separate “final GDP” series. It is the third of BEA's current quarterly estimates. BEA's [GDP learning page]({source:beaGdpLearningCenter}) describes this release sequence and the additional source data incorporated between releases.

BEA assembles GDP from many source series rather than observing every transaction in one complete survey. Those source series have different publication schedules, coverage, definitions, and revision practices. When an early GDP release is due before all relevant source information is available, BEA uses the information then available, including estimates for items that have not yet been reported in full. Later releases can replace or update parts of that information.

A change in an underlying source series can therefore flow into a GDP component and the aggregate. New information can affect current-dollar measures, price adjustments used for real measures, or the measured contribution of a component. BEA's Information Quality Guidelines describe how source information arrives over time and how BEA documents methods, assumptions, and planned revisions. The [NIPA Primer]({source:beaNipaPrimer}) also explains how estimates may rely on available data and indicators when direct source data have not yet arrived.

This does not mean every component changes in every release, or that one late data point alone determines the whole revision. The updated estimate combines the source information available for the account. To understand a particular release, read its technical notes and source-data assumptions alongside the headline.

No. “Advance,” “second,” and “third” are release labels, not grades of certainty. Later estimates usually incorporate more detailed and comprehensive source information, which can improve measurement. But a later publication date does not by itself prove that the estimate is closer to an unknowable exact value, and the third estimate is not permanently closed to revision.

A revision can raise the reported growth rate, lower it, leave it unchanged, or alter its components without changing the rounded headline. There is no rule that successive estimates must move in one direction. A downward revision is not by itself evidence of a weak economy, manipulation, or a defective first release; an upward revision is not proof that the first estimate was deliberately understated.

Treat every vintage as an estimate constructed under the data and methods available at that time. When your question concerns what policymakers, businesses, or readers knew then, preserve the contemporaneous vintage. When your question concerns BEA's current account of that period, use the latest available vintage and identify it as such.

The three current quarterly estimates are followed by separate update cycles. They should not be treated as a fourth, fifth, and sixth monthly estimate of one quarter. Annual and comprehensive updates can revise a wider span of historical quarters and years.

BEA says annual updates generally cover at least the five most recent calendar years and their associated quarters. They incorporate newly available major annual source data and may include changes in methods or definitions. Comprehensive, also called benchmark, updates occur at about five-year intervals and bring in major periodic source data as well as major conceptual improvements. The current [BEA release information]({source:beaGDPReleaseRevisions}) describes how these update scopes differ.

A comprehensive update can therefore change the historical account over a much longer period than the advance-to-third sequence. The revisions may reflect a combination of newly available observations, improved source coverage, or methodological and conceptual changes. These are distinct reasons to label the data vintage when comparing a long-run chart.

<!-- learn:illustration -->

Start by naming the period and statistic: for example, real GDP growth for a particular quarter, rather than simply “GDP.” Then decide which version answers your question. Use a contemporaneous release vintage to reconstruct what was known at the time; use the latest published history for a current description of the period.

When comparing two estimates, record the quarter, estimate label or update vintage, release date, and measure. Keep real GDP separate from current-dollar GDP, and check whether the growth rate is the seasonally adjusted annual rate (SAAR) or another comparison. Otherwise a difference attributed to a revision might actually come from comparing different measures or conventions.

BEA's [GDP data page]({source:beaGrossDomesticProductOverview}) links to current releases, previously published estimates, and a vintage history. Its [revision information page]({source:beaGdpRevisionInformation}) provides tables comparing the current estimate with the estimate released the prior month. Those are useful starting points for a like-for-like comparison.

For the difference between prices and production volume, see nominal versus real GDP. To distinguish annualized quarterly growth from year-over-year growth, see real GDP growth rates. For why the expenditure and income sides can differ, see [GDP vs. GDI and the statistical discrepancy](/learn/gdp-vs-gdi-statistical-discrepancy-explained). Each guide answers a different question about the release.

Suppose an imaginary quarter's real GDP growth is reported at a seasonally adjusted annual rate of 1.2% in the advance estimate, 0.8% in the second estimate, and 1.0% in the third estimate. In a later hypothetical annual update, assume the same quarter's estimate is recalculated to 0.9%.

These invented values show the sequence, not actual BEA data. The second estimate is 0.4 percentage points below the advance estimate; the third is 0.2 points above the second and 0.2 points below the advance. The later annual update changes the figure again. The path is not monotonic, and the example does not tell us which vintage is “the truth.” It only illustrates that successive versions for one period can differ as source information and update scope change.

The example also keeps the release cycle separate from the measured period: all four figures refer to the same hypothetical quarter, while the later annual update belongs to a broader historical review. In a real comparison, check the release notes to learn which data and methods were incorporated rather than inferring a cause from the direction of the number alone.

A revision table can show how a current estimate differs from a previous release and which reported components changed in the comparison. BEA's [GDP revision information]({source:beaGdpRevisionInformation}) currently presents comparisons between the current estimate and the estimate released last month. The related release may also provide technical notes and source-data assumptions that help explain the update.

A single comparison cannot establish that one estimate was accurate and the other was wrong, that revisions are systematically biased, or that the economy changed between the two publication dates. The underlying quarter is being re-estimated; the table does not describe two successive periods of economic activity. Nor does the sign of one revision predict the next one.

If you are evaluating the revision process as a whole, use BEA's revision studies and state the sample period and vintage pair those studies compare. Do not use a single quarter's change as a substitute for that evidence. Most routine reading needs a simpler habit: label the vintage, match the measure and period, and state whether you are describing what was known then or what BEA estimates now.