CSO General Revisions Policy

This section details the CSO’s General Revisions Policy. Domain specific revisions practices must be documented by each area in their survey background notes or methodological documents which will contain more detailed information taking account of this policy together with the requirement of any domain specific revision requirements.

4.1 Notification of Revisions to Published Statistical Series

1. Statistical results that are subject to revisions must be indicated at dissemination.

1. Statistical results that have been revised must be indicated in the release and/or on PxStat. (See Section 5 for PxStat revision codes and labels.)
2. Advance notice must be given of future revision plans arising from major changes to statistical methodology and data sources.

2. Where possible advance notice must be given to users of discontinuities / series breaks.
3. Substantial revisions must be accompanied by a detailed explanation of their nature and extent.
4. Users who detect errors will be contacted directly and informed of revised data.

4.2 Revisions Calendar

The General Revisions Policy is supplemented by a Revisions Calendar, a copy of which will be made available on the CSO website. This calendar provides an overview of which sets of statistics are subject to revision and describes the respective revisions cycle.

4.3 Consultation with Users

Consultation, where appropriate, with key users on revision practices will be documented by each business area.

4.4 Revisions to Research Micro-Data

All users of research micro-data will be informed of revisions to the micro-data file.

4.5 Revision Practices

Detailed revision practices must be documented by statistical areas in their survey background notes, quality reports and methodological documents and are also kept up to date on their Methods page on the CSO website. 

4.6 Discontinuities in Statistical Series

Discontinuities in statistical series will be documented in detail when they occur, and users will be informed of the reasons and potential usage of the discontinued series, where applicable.

4.7 Timeliness of Release

Timeliness of releases will be balanced with the need to avoid frequent revisions.

4.8 Data storage and keeping records of Revisions 

All data prior and post revision (i.e. provisional, final and other revised published data) must be stored in the Disseminate location based on the agreed corporate data standard. This data should be clearly named and labelled so that each dataset associated with revised data is easily identifiable and should be marked as having a permanent retention period.

4.9 Quality of Provisional data

Revision analyses must be carried out, where appropriate, and the results will be used in the assessment of data quality and to continually monitor and improve statistical processes. The scale of revisions must be published for certain statistical series where it is the norm to issue at least one set of estimates prior to final figures. Users will thus be informed of the reliability and the utility of preliminary to penultimate estimates of the data.

The publication of provisional data in many cases is a statutory requirement. In this context it is important that the CSO evaluate, on an ongoing basis, the quality of provisional data being produced and where appropriate provide the relevant information necessary to ensure that users are made aware of the limitations of such data. However there are some statistical series (e.g. National Accounts) where it is recognised internationally that significant revisions to early estimates may be necessary and thus form an integral part of the dissemination process.

4.10 Official Series

When a statistical series is revised retrospectively (for example, due to rebasing or a change in methodology or seasonal adjustment), there is more than one published statistical value at each time point over the period of revision. Users will be informed which values for previous periods represent the official series.

4.11 Analysis of Revisions

Revisions analyses should be carried out to ensure and enhance data quality, revisions procedures and methods.

Revisions analyses study the impact of revisions by comparing provisional and final results (or preliminary and later estimates). The difference between provisional and final results (or preliminary and later estimates) indicates the scope of the revision. Systematic trends may also become apparent from revisions analyses, for instance, if preliminary values are usually revised upwards or downwards. In cases where significant systematic trends are identified, future preliminary estimates could be adjusted by the average bias established.

The results of revisions analysis should be presented in the background notes of releases and in the domain specific quality report.