What is an error?
All publications of statistics and analyses are quality assured, however, errors do sometimes occur in figures, statistics, graphs and related text. Even where figures or text do not contain errors per se, they may include ambiguous or incomplete text, giving rise to misinterpretation by the users.
Errors in Statistics Norway’s releases are unplanned occurrences. Such occurrences require the relevant product to be corrected on ssb.no, and users to be informed as necessary. Planned changes leading to the revision of previously published figures are not covered by this document.
Errors can be identified at various points in time, from just after release to several years later. They can occur in time series and releases dating back a number of years, and in statistics or publications other than those where the error was initially identified.
In some cases, errors can be corrected immediately. For other errors, it may take time to ascertain the reason for the error and when the correct figures or text will be available. It may therefore be necessary to first make users aware that an error has been found, and then publish the corrected products.
Which products should be corrected?
The principles for correcting errors apply to the following products: statistics releases, web articles, publications and tables in StatBank.
Types of error
The main distinction in error correction is between errors in content (serious and less serious) and proofing errors. In rare cases, very serious content-related errors can occur. In practice, a good deal of discretion is used in the classification of errors. Different user groups may have different perceptions of the seriousness of an error. Single errors may not have a large bearing on a release but when combined with other errors may have a major impact, and hence give a negative impression to users. Errors that are less central to the statistics can also attract media attention and lead to criticism of Statistics Norway, which would make the error serious for the institutions. Meanwhile, errors in key variables can also go undetected, making them less serious for Statistics Norway. An error that is considered to be of little significance may ultimately prove to be more extensive, and vice versa. The following discretionary criteria are used as a basis for managing corrections and reporting:
- Errors in key variables and text, e.g. key figures, aggregate quantities, news titles and introductions
- Errors that occur on a large scale, e.g. where a large number of tables and large pieces of text need to be changed
- Errors in particularly sensitive material relating to the economy and society
- Errors that lead to incorrect use of the results, e.g. widespread dissemination of misleading figures in the media
- Errors that lead to Statistics Norway being subjected to extensive criticism
Less serious errors:
- Where the figure is less central to the statistics
- Where this is minimal risk of incorrect use of the results, e.g. via the media
- Where this is minimal risk of criticism of Statistics Norway
Proofing errors: Where an error relates to form and does not have a bearing on the users’ understanding or application of the material. For example, typographic errors, linguistic errors and other errors in the text that do not affect the statistical content, such as errors in the number of decimal places used, and in dates and links etc.
Principles for correcting errors and correction communication
The main principles are:
- It is important that Statistics Norway corrects errors since published material is of significance to society and should thus be as accurate as possible.
- In order to maintain the public’s confidence in Statistics Norway, it is crucial that errors are acknowledged and explained in a professional and transparent manner, and that they are documented for the record.
- Errors in content that are identified in published statistics and analyses shall be corrected, and the corrected product shall be published on ssb.no as soon as possible.
- Statistics Norway’s quality assurance shall ensure that as few errors as possible occur in the statistics and analyses. Statistics Norway shall learn from its mistakes in order to prevent new errors occurring.
- Routines for correcting errors and information to users shall be adapted to the degree of seriousness of the error.
- Correcting and marking errors:
o Corrections of content errors shall be clearly visible to the users.
o The reason for an error in a product shall be given where necessary.
o Proofing errors shall be corrected when necessary.
o Errors in older paper publications shall not be corrected.
- Active information to the users:
o If some time is needed to correct a serious error, the users shall be informed of this, and also, where possible, when the corrected product will be available. If the error is extensive, it may be necessary to temporarily withdraw the release.
o Active information to the users in addition to releasing the corrected product shall be considered.
- Errors identified in older web releases:
o The figures shall be corrected in StatBank where possible.
o Because error correction entails a great deal of work, reference shall be made in older releases, where relevant, to corrections in StatBank and to the corrections in more recent releases.
- Errors identified in older, printed publications shall not be corrected.
- The corrected versions of products shall be published, and the original versions with errors shall not be visible to the users.