CAUSES OF REVISIONS
Data are always based on up to date information at the time of their release. Any revisions to data will be transparent and governed by a clear policy applied consistently across all outputs, they rarely occur beyond two years.
CAUSES OF REVISIONS
Revisions to data can be generated by a number of factors which include:
- the receipt of additional information or corrections to or refinements of existing data – in practice, such data revisions rarely occur beyond two years. An example of where new information becomes available is where smaller banks only report annual data and this can alter the grossing up of the quarterly series for instance for banks’ income and expenditure);
- the routine updating of systems or methodologies (for example routine updating of estimated seasonal factors);
- the introduction of a material change to a key methodology, or the inclusion of a previously unused data source.
Revisions will be included within published outputs at the earliest possible opportunity (usually the next scheduled release date). A statistical release may be re-released outside of the scheduled release timetable when revisions are received - or an error is discovered - that have an impact on aggregate data which could be considered significant to good decision making. Any such re-release will be agreed internally with a member of the Executive Team, and where possible a pre-announcement of the intention to re-release the data will be made. The revised release will carry an explanation for the re-release respecting confidentialities as necessary.
Where revisions result from the introduction of a material change to a key methodology, they will either have been subject to user consultation prior to incorporation and/or will be announced in advance of inclusion in the data, normally via a Bankstats article.
Revisions over a certain threshold will be clearly identified. The relevant thresholds for revisions are normally indicated within the section of the release to which they relate. Substantial revisions will be accompanied by an explanation of their nature and extent, subject to preserving respondent confidentiality. In many cases, though, revisions will be minor and will be of little or no interest to the majority of users and will therefore not normally be identified.
Revisions tend to be made sooner in series published by the Bank than in comparable series published by the Office for National Statistics (ONS), because of the ONS policy of making revisions only at certain set points in the cycle of data releases.
Bank of England (June 2010 ), ‘Statistical Code of Practice for the Bank of England’.