This document describes the overarching principles and guidelines for the revision of published statistics in Statistics Norway. The practice followed for the individual statistics is documented on the relevant statistics’ home page under the tabs “About the statistics” and, where applicable, “About seasonal adjustment” on ssb.no (see below).
The principles follow the recommendations in the European guidelines on statistics (European Code of Practice − CoP) along with the accompanying Quality Assurance Framework – QAF and the European statistical system’s guidelines on revision policy for Principal European Economic Indicators – PEEIs. This means that the revisions follow common procedures and principles, and are documented and made public.
Published figures may be revised due to errors, but revisions can also be planned. The overarching revision policy described in this document relates solely to the handling of planned revisions. Principles for handling errors are described in Principles for correcting errors in publications in Statistics Norway.
Revisions are planned when new and improved data sources become available (often data in administrative data systems that has arrived late), when adjustments are made to figures in short-term statistics after comparison with the corresponding annual statistics, or when methodologies, classifications or definitions have been changed, for example with a new base year or new weights.
It is useful to distinguish between regular and non-regular revisions. Regular revisions comprise the publishing of final figures that are published after preliminary figures and, for example, revisions as a result of seasonal adjustment. Non-regular revisions cover the main revisions, such as the national accounts, but can also be undertaken for individual statistics.
The following main principles apply to revisions:
- The users should know why a revision is being performed.
- When publishing preliminary figures, details shall be given of when these will be revised.
- The size of the revisions shall be specified when revised figures are published. When particularly large or unusual adjustments are made, the reason shall be explained.
- Major changes to methodology shall be notified to the users in advance.
- Consideration shall be given to whether back calculations/changes shall be made to previously published figures, and if so, what these changes will be.
- Regular revision analyses shall be carried out to support the continuous improvement of the statistics.
- Revisions shall be discussed regularly with users.
More details are given below for some of the principles.
Criteria for revisions – why and how
Timeliness and accuracy are two key quality dimensions in statistics. These must be balanced against each other as the accuracy normally increases with a longer production period. This particularly applies to statistics based on registers, where updates may be subject to delays. In sample surveys, it may also be necessary to choose between concluding the data collection early where there is a low response rate and sending several reminders to respondents in order to get as high a response rate, and thus accuracy, as possible. In order to safeguard both the timeliness and accuracy of statistics, regular preliminary figures are published for many of the statistics, and these can be revised once or several times to improve the accuracy. Revision analyses provide a basis for such considerations.
In addition, seasonal adjustment in itself will lead to changes in previously published figures - even though there are no changes in the historical non-seasonally-adjusted figures.
Non-regular revisions/main revisions take place mainly in macroeconomic statistics that are based on a range of statistics. These revisions follow international guidelines, such as those for the national accounts. An important characteristic of such statistics is that they describe the development of the economy over many years. New data sources and methods (which are also reflected in new guidelines) may be incorporated into the statistics every few years in main revisions, which will ensure consistency between different statistical systems and comparability between countries. Such revisions are undertaken in line with a common plan that the users will be informed about in advance, and as early as possible.
Non-regular revisions of other statistics may also be carried out as a result of new data sources and methods. Users shall be informed of all revisions as early as possible.
When publishing figures that are preliminary, this should be stipulated in the text and tables. This also applies to the StatBank. When the preliminary figures are replaced by final figures or a later version of preliminary figures, the original figures are normally overwritten in the StatBank. However, the original figures can be stored in separate tables at the request of the user.
For some statistics, preliminary figures are revised several times (e.g. monthly figures for foreign trade of goods, which are updated in each release within a year), while some statistics are not revised even though corrections are made to the data on which the statistics are based (e.g. the population statistics, which are based on the last register status and not revised retrospectively). It is important to consider whether statistics should be revised, and if so, how many times and for what period of time. Revision analyses and discussion with users provide a basis for such considerations.
Revision for each survey
When preliminary figures are published for surveys, this should be stated in “About the statistics” for the relevant statistics. A description shall be given of regular revisions and planned major revisions. The procedure for the revision of seasonally-adjusted series should also be described in “About seasonal adjustment.” “About the statistics”/”About seasonal adjustment” should also include a description of regular revision analyses.
Discussions with users
Methods for the revision of published figures shall be discussed regularly with the users of the relevant statistics. For statistics with an advisory committee, this should be a natural point on the agenda.