Guidance

DSIT statistical error and revision policy

Published 14 December 2023

This guidance documents the error policy (including revision policy) for statistics produced by the Department for Science, Innovation and Technology (DSIT), in line with practices recommended in the Code of Practice for Statistics, specifically setting out the department’s commitments on T2: Independent decision making and leadership, and Q3: Assured quality. 

This guidance is designed to be used by DSIT staff responsible for the production, use and dissemination of statistics, or in the handling of statistics. It contains a set of rules to ensure that all DSIT officials use statistics effectively when making decisions in DSIT and handle the statistics in a way that minimises risk to the DSIT. This will be reviewed at least annually.

1. DSIT error and revision policy for published statistics 

This policy covers the production and publication of DSIT official statistics, as well as any statistics that are used externally. 

1.1 Scheduled revisions 

A scheduled revision is where statistics are initially published as provisional, which means that they are subject to change. At a later scheduled date, as announced on the upcoming statistical releases on GOV.UK, the statistics will be revised and reissued. 

Statistics published as provisional should be clearly marked by a [p] in the relevant column heading, denoting that the figures presented in the column include provisional estimates. If only some of the estimates in the column are provisional, then include a footnote detailing which individual figures this applies to. Information should be provided on the reasons why the data is provisional and why a provisional release is necessary in addition to the finalised output. 

Once data initially published as provisional is made final, any [p] markers should be removed and replaced with an [r], with the footnote updated accordingly. Information on the reasons why the data has been revised should be provided.

1.2 Errors 

Corrections to statistics will be made if errors are found or if figures are amended because of changes to methodology or definitions of the data collection. 

Where an error has been found, the Chief Statistician should be notified as soon possible. They should be given details of the error, its scale and proposed corrective actions. 

The Chief Statistician will decide if a major or minor error has occurred and advise the team responsible for the publication of any required corrective actions. 

Major errors are errors which have a significant impact in the use made of the data, for example where the overarching message changes. These errors should be corrected as soon as possible by reissuing the statistics with the corrected figures. In such circumstances key users and/or ministers may be alerted. 

Minor errors are all other errors which are not classified as major errors. These can be corrected in slower time as appropriate (always within 12 months) and can be timed to coincide with, but are not restricted to, the release of new editions of statistical products. 

In cases where errors are found in time series that appear in more than one edition, the errors should be corrected in the most recent edition only. Older editions need not be corrected. 

Where changes are made to methodology or to definitions in a time series, corrections to the historical data in the time series should be considered to bring it in line with the new methodology or definitions. This may not be possible and depends on the data availability and the amount of work involved. The requirements of the users should be considered when making such decisions. 

If an error is identified but corrected figures are not yet available due to a need for further investigation, a note should be added to the statistics to explain the situation, making it clear the figures are subject to change. 

Corrections should be shown by the addition of a [r] in the relevant column heading, denoting that the figures presented in the column include corrected figures. This should be accompanied by a footnote explaining which individual figures were corrected and the reason for the correction. Corrections will only be marked in the first edition of each publication that the corrected figures appear in. If the figures then appear again in future editions. they will not be marked as corrected. 

The report should state when it was first published and when it was revised. 

In cases where errors are corrected in cumulative spreadsheets, that is, where the latest data is added to a spreadsheet, which then replaces the previous edition on GOV.UK: 

• in the first edition after revision, corrections should be show by the addition of a [r] in the relevant column heading, denoting that the figures presented in the column include corrected figures. This should be accompanied by a footnote explaining which individual figures were corrected and the reason for the correction. 
• in subsequent editions, the [r] marker and original note should be removed from the spreadsheet and replaced by a note explaining what was revised, the reason for and scale of the correction 
• the old uncorrected publication will be removed but should be filed and available on request 
• steps should be taken to determine what corrective actions/training/communications are needed to prevent similar errors from occurring in the future. The emphasis is on learning from mistakes and improvement

2. DSIT error policy for Parliamentary questions 

There should be no instance in which a Parliamentary question (PQ) response is expected to be revised at a later date. If data is not currently held in a format that enables a sufficient response to the PQ, an explanation should be given for why an answer could not be provided. 

If a PQ answer is found to contain an error, a correction needs to be made. The corrected answer will appear on the Parliamentary website alongside the earlier response in a ‘correction to the official report’.