1. Summary 

The compilation and dissemination of statistical information is constantly balancing opposing goals of timeliness and accuracy.  Generally, accuracy increases as time passes as more complete and comprehensive data and more robust statistical methods can be employed to produce statistical information. The General Authority for Statistics (GASTAT) must produce information for stakeholders that balances these two somewhat opposing goals.  
It is for this reason that GASTAT revises its National Accounts estimates, including those from the Quarterly National Accounts (QNA) program, on a regular basis; to provide both timely data and an opportunity to update estimates as accuracy increases with the passing of time and the availability of additional inputs.  The official introduction of this policy will provide users with a clear understanding of both the expected update frequency and period(s) to be revised and the reasons behind them in a similar way to providing users with a release calendar which programs follow pre-determined release dates for statistical products.  Before the introduction of this policy, National Accounts estimates  were revised in a largely systematic way but this policy will officially formalize the way forward. 
Such policies are standard practice in many countries and follow international recommendations associated with both compilation techniques and stakeholder communication.  Differences across countries in the type of revision and number of periods subject to revision result largely from the timing and availability of necessary data inputs.

2. National Accounts Revision Policy

2.1. Definitions

1.    Revision – A revision is defined as the numerical difference between two vintages of the same data point.
2.    Routine Revision – Revisions carried out with each new release of data, usually for the most recent prior periods.  For QNA estimates these are generally sub-annual revisions for periods within the same or the previous year.
3.    Annual / SUT Revisions – Annual revisions refer to those revisions conducted once a calendar year. These revisions cover the current year and previous year(s) to incorporate updated source data and information from annual programs including updates resulting from the compilation of Supply and Use Tables (SUT). Annual revisions include updates to seasonal adjustment models and may also include other select methodological updates.
4.    Comprehensive or Comprehensive Benchmark Revisions – Comprehensive or Comprehensive Benchmark revisions are revisions to macroeconomic accounts where not only final vintages of data are incorporated but concepts, accounting treatments (e.g. System of National Accounts updates), classification(s) or large-scale methodological changes are introduced.

2.2. Flash Estimates of GDP

Flash GDP estimates provide users with the timeliest estimates of GDP and its main economic activities on a real basis using partial data and simplified assumptions to extrapolate some monthly or quarterly indicators. As such, Flash GDP compilation does not introduce revisions to any previous period estimates and only focuses on and estimate of the most recent period. Revisions between Flash GDP and the Full QNA are therefore only applicable to total real GDP and the main economic activities which are estimated as part of flash estimates of real GDP compilation.

 2.3. Quarterly National Accounts

The revision policy for the full QNA sets forth a repeated yearly cycle to outline the reference period(s) open to revisions with each new release and identifies the first value produced by the full QNA program as a first estimate. Sub-annual QNA results are presented in both unadjusted and seasonally adjusted forms and as such, each has a distinct revision cycle.
1.    Unadjusted Data
o    A first quarter release includes routine revisions to all quarters in the previous year.
o    A second quarter release includes routine revisions to the previous quarter.
o    A third quarter release includes revisions to the first two quarters of the same year and all quarters in the previous two years (Annual Benchmark and SUT revisions)
o    A fourth quarter release includes routine revisions to the previous quarter.
2.    Seasonally Adjusted Data
o    A first quarter release includes revisions to all quarters in the previous two years.
o    A second quarter release includes revisions to the previous quarter of the same year and all quarters in the previous year.
o    A third quarter release includes revisions to the first two quarters of the same year as well as revisions to all quarters in the previous three years.
o    A fourth quarter release includes revisions to all quarters of the same year and all quarters in the previous year.
*Note: Seasonally adjusted data are based on unadjusted data and as such when unadjusted data are revised (routine, annual, SUT or otherwise), those revisions are included in seasonally adjusted estimates.

2.4. Annual National Accounts

Annual National Accounts (ANA) estimates are prepared and published once a year, with the initial estimates of these accounts for the previous calendar year (Y-1) published at the end of October (e.g. First estimates of ANA disseminated in 2025 cover the results for 2024). At the same time, ANAs are revised for the year prior in accordance with the availability of the Supply and Use Tables.
As noted above, following the dissemination of the ANAs, results are subsequently incorporated into the QNAs with their next release.

2.5. Supply and Use Tables

SUTs are prepared and published once a year in conjunction with the ANAs. SUTs cover the period (Y-2)  (e.g. SUTs disseminated in 2025 cover the year 2023) and are only revised at the time of a Comprehensive Revision.
As noted above, following the dissemination of the SUTs, the results are subsequently incorporated into the QNAs with their next release.

Table1:

3. Why are revisions necessary?

Revisions are an essential part of statistical compilation and dissemination as they allow statistical agencies to balance the primary dichotomy of time vs. accuracy that is an incumbent part of statistical compilation.  To manage this, GASTAT like other statistical agencies compiles and disseminates preliminary data that are subsequently revised as new or more accurate data become available.
Revisions will typically include new or updated data or methodologies used to compile statistics but may also be associated with corrections. Within National Accounts compilation and specifically sub-annual estimation such as the QNA there are a myriad of data sources used in the compilation of estimates; each of those data sources has their own distinct properties including their availability and the possibility of subsequent update/revision.  Data used to compile QNA estimates are sourced from a variety of locations including internal statistical products, administrative data, and other publicly available data. Each of these different data types may or may not meet the ultimate requirements of statistical compilation and the use of one data source may supplant another as time passes.

4. What are the sources of revisions?

Revisions can occur as a result of many different reasons including but not limited to:
•    Additional reported values from survey respondents.
•    More accurate/revised reporting by survey respondents.
•    Availability of late or more comprehensive data.
•    Difference in methods used in statistical data compilation between different vintages.
•    Incorporation of updated seasonal factors.
•    Benchmarking.
•    Changes in concepts, classifications, or definitions.
•    Updates to base period/weight period.
•    Error correction.