Revisions and restatements
Provisional versus final data
Short answer
- US payroll jobs
- Latest 2 months based on incomplete returns, final after two successive revisions 2
- US GDP
- Advance estimate near the end of the first month after the quarter 1
- India annual GDP
- Six stages from first advance to third revised estimates, per the 2016 statement 3
- US provisional deaths
- About 65 percent complete at 2 weeks and 85 percent at 4 weeks, 2021 data 4
- Australia national accounts
- Basic quarterly survey data usually firm the quarter after the first estimate 5
What the labels mean
Agencies use different words for early figures. The Bureau of Economic Analysis (BEA) calls its first quarterly GDP figure the advance estimate 1. The Bureau of Labor Statistics (BLS) says payroll estimates for the most recent 2 months are based on incomplete returns 2. India's Ministry of Statistics uses advance estimates before the financial year ends and provisional estimates two months after it ends 3.
A final figure is the one an agency stops revising in its routine cycle. Even then, benchmark or base-year changes can reopen it. BEA's benchmark updates come about every five years 1. India's 2016 statement says estimates are revised if the base year of an index used in GDP changes 3.
Stages in three statistical systems
| System | Early stage | Later stages |
|---|---|---|
| US GDP (BEA) | Advance estimate, end of first month after the quarter 1 | Second and third estimates, then annual updates and benchmark updates 1 |
| US payroll jobs (BLS) | Preliminary estimate from incomplete returns 2 | Two monthly revisions, then a yearly benchmark to universe counts 2 |
| India annual GDP (MoSPI, 2016) | First advance estimates on 7 January, second advance on 28 February 3 | Provisional on 31 May, then first, second and third revised estimates each 31 January 3 |
Why early counts differ from final ones
The main reason is late reporting. The National Center for Health Statistics (NCHS) said it can take several weeks for death records to be submitted 4. It said provisional death counts are underestimated relative to final counts, given the steps of certification, state processing and coding 4. Based on 2021 data, provisional counts were about 65 percent complete within 2 weeks and 85 percent within 4 weeks 4.
NCHS said completeness cannot be fully determined until final data are ready, typically around December of the following year 4. Its technical notes page was last updated on 20 September 2023 and archived on 27 September 2023, so these figures are historical 4.
Survey data follow the same pattern. BLS routinely revises the prior 2 months of payroll data as more returns come in 2. In Australia, the ABS says basic quarterly survey data are as a rule firm the quarter after the initial estimate 5.
Revisions to quarters after the year closes
Annual results feed back into quarterly data. India's 2016 statement says quarterly estimates of earlier years and the first quarter of the current year are revised to keep consistency 3. The ABS publishes annual benchmark results first in the September quarter release, which normally allows up to sixteen quarters of revision 5.
Using provisional figures safely
Common questions
Is a provisional figure an error?
No. It is an early estimate from incomplete data. BLS says the latest 2 months of payroll estimates rest on incomplete returns 2.
When is a US monthly jobs figure final?
BLS says only after two successive revisions to a monthly estimate, with a yearly benchmark adjustment on a lagged basis 2.
When does India release provisional GDP estimates?
The 2016 statement set 31 May, two months after the financial year ends 3. Check the current release calendar for later changes.
Why do recent death counts look low?
Records take weeks to arrive. NCHS said provisional counts were about 85 percent complete within 4 weeks, based on 2021 data 4.
Rules and charges change. Check the official source linked on this page before you pay or travel. Spotted an error? Email hello@recordsdisagree.com.
Sources
- Bureau of Economic Analysis: Comparisons of Revisions to Real GDP (October 2026)
- U.S. Bureau of Labor Statistics: Employment Situation Technical Note (2 October 2026)
- Ministry of Statistics and Programme Implementation: Statement on Release Calendar of Annual and Quarterly Estimates of GDP and policy of revision (16 November 2016)
- CDC National Center for Health Statistics: Technical Notes, Provisional Death Counts for COVID-19 (20 September 2023)
- Australian Bureau of Statistics: Revisions, Australian System of National Accounts Concepts, Sources and Methods 2020-21 (9 July 2021)
Related
- Statistical revisions explained
Why GDP, jobs and other official figures change after release, how agencies schedule revisions, and how a revision differs from a correction of an error.
- Why official records disagree
The main reasons two official records show different figures for the same thing: timing, revisions, corrections, scope, definitions and restatements.
- Scope and period mismatches between official records
How differences in coverage, counting units, reference periods, fiscal years and geography make two correct official records show different numbers.
- How to compare two conflicting records
A step by step method for comparing two official figures that disagree, from vintage and scope to corrections, restatements and documentation.