HSE statistics revision policy
The Code of Practice for Official Statistics requires all producers of Official Statistics to publish a policy on revisions.
HSE aims to avoid the need for large revisions to National Statistics unless absolutely necessary and has procedures in place to minimise the number and scale of any revisions made. Where any changes to previously published data come to the attention of HSE statisticians, they will form a professional view as to whether a revision to published data is in the public interest. If necessary, the HSE chief statistician will seek further advice from the National Statistician's office.
There are four key reasons why revisions may be necessary:
1. A change in data collection method or a methodological improvement
- Where major changes to data collection systems or methodology are planned, we will, where possible, inform users in advance of the likely impact on our data series. Once the change has been implemented, we will endeavour to estimate the effect of any discontinuity in our headline data series and produce a consistent back-series. The methodology used will be explained in full and published on the HSE statistics website. If it is not possible or practical to adjust the series, a discontinuity line will be presented in charts and tables and appropriate footnotes supplied.
- HSE's model for estimating the cost of workplace injuries and common work-related ill health conditions is continually evolving to ensure it reflects as accurately as possible the full range of costs incurred. With each annual update, we will inform users of any model enhancements, and we will endeavour to produce a consistent back series (in constant year prices).
2. Revisions in population data used to calculate rates or weights for survey data
We make use of employment and population data from the Office for National Statistics in the production of both our Labour Force Survey estimates and injury rate series. These external data series are routinely revised following a census, and since 2016 annually between censuses. The latter generally involves revisions to the previous three years. HSE statisticians will make an assessment of the impact on our data series. We will decide whether and when to incorporate the revisions and how far back to revise by considering both the extent of impact and potential disruption to our users.
3. Receipt of late or revised data from a supplier
All administrative systems are, by their nature, subject to constant updating. In order to minimise the scale of revision to published data, release dates are set by balancing the need for timely data release with that for accurate data. Hence, we aim for the data to be largely complete at the time of first release. Nonetheless, this balance means that some late reports have to be incorporated in subsequent years.
4. Errors in our statistical systems and processes
We do everything we can to minimise errors in our statistical processes. However, occasionally they will occur. Where errors are small, revisions will be incorporated at the next scheduled release date and tables annotated accordingly. If errors are substantial, immediate action will be taken with users notified via the HSE statistics website and the e-bulletin service. We will also be proactive in identifying key internal and external users to further highlight the changes. Where a significant error is identified in statistical information published on the HSE internet, and correcting it is likely to take more than one week, we will annotate the web page to identify the error and that it will be corrected as soon as possible.