CSO General Revisions Policy

Revisions

In this document, revisions refer to changes made to published statistical data when the information used in its production has been updated or corrected. This information includes all data used in compiling the statistic e.g. respondent data, administrative data, weights and factors, methodology, classifications, definitions, modifications to survey questionnaires, survey scope and data collection methods.  Typically, there are three types of revisions:

3.1 Planned Routine revisions

Routine revisions are changes in published data which are related to the regular data production process. The majority of revisions are planned, and include regular updates where:

1. It is the practice to publish preliminary estimates in advance of final estimates primarily due to the receipt of late responses, the production cycle or the availability of administrative sources.
2. Seasonal factors are updated.

3.2 Planned Major revisions

Major revisions are changes in published data, often substantial, due to one of the following reasons:

1. Changes to definitions, classifications and methodology are incorporated.
2. Updating / rebasing of the series takes place to reflect both the changes in the relative importance of sub-components and the insertion of new items into the statistical series (e.g. updating the basket of goods and services for the Consumer Price Index every 5 years).
3. Updating of key statistical series (e.g. series based on Census of Population) is necessary.
4. New data sources become available.
5. Cessation of existing data sources occurs.

Data producers often take the opportunity of a forthcoming major revision to introduce methodological improvements. Therefore it is common that major revisions are not determined by one single cause, but a combination of them.

3.3 Unplanned revisions

Unplanned revisions generally occur when:

1. Published statistics, that are considered final, need to be revised due to the receipt of updated data that impacts significantly on the figures.
2. Existing data sources are suddenly lost.
3. Errors are detected.
4. Improved estimated data become available from other sources.