Introduction
This is Statistics New Zealand's current policy on handling errors in published statistics. It is included here to guide statistical producers if errors occur in their statistical outputs.
Policy
Any error (or perceived error) reported in published statistics is to be referred to the relevant manager who is responsible for deciding on any further action.
The guidelines below should be used by business units to develop their own detailed process for managing errors.
Rationale
Members of staff or the public may perceive that information is in error if they do not fully understand particular subject matter statistics or the methodology used. The relevant manager should determine whether the information published contains errors and/or needs some correction or clarification.
Because managers are responsible for the original published data and information, they are also responsible for managing and resolving all reported or discovered errors, whether real or perceived.
The policy and guidelines aim to ensure that all errors are handled in a standard manner which is consistent with their significance rating (see below).
Guidelines
Each business unit should use these general guidelines to develop their own detailed guidelines and process for managing errors.
Discovering errors
Errors (or perceived errors) may be reported to the information centre or other Statistics NZ staff by the data users, or noticed directly by staff. It is important that any apparent error is acted on as soon as possible.
Resolving errors
The manager is responsible for checking the information and determining whether what is published actually contains errors and/or needs some clarification. If it does, the manager should prepare an error report (detailed below) which is used as a plan to resolve the error. The error report becomes a permanent record of the actions taken.
Each error must be given a significance rating, based on the table below. Business unit guidelines need to outline how each error rating will be handled.
Guide for rating errors
| Significance |
Characteristics |
Action |
| High |
Errors which affect a Tier 1 statistic output. Business, economic or social policy decisions may be misguided by the error.
Errors which affect a baseline, denominator, important indicator or important time series relied on by data users. |
Requires urgent action and a communications plan for key data users in order to lower the risk of negative impact.
The communications plan must also consider the impact of the error on key stakeholders.
Business processes should be reviewed. |
| Medium |
Errors which are considered to have a low risk effect on business, economic or social policy decisions.
Examples are: missing or incorrect labels on graphs and tables, or missing footnotes. |
Corrections must be made to published online information.
Correction of printed information could be handled in a future release of the same report or subject.
Business processes should be reviewed. |
| Low |
Errors which are not likely to affect the interpretation of a trend, or which are obvious in the context of surrounding data.
Errors found in data published some time previously which are not likely to cause major problems for data users. |
Corrections could be made through a routine update of online information, or when reprinting reports.
The manager may decide not to correct errors of a low significance.
Business processes should be reviewed. |
Communications
Managers should consider the impact that an error and its correction will have on internal and external stakeholders, and must manage all communications about the release of corrected information in a manner appropriate to the significance of the error.
Making corrections
The normal embargoed/first-release publishing processes should be followed when correcting errors. These processes should be 'fast-tracked' if necessary.
Corrected errors of 'high' or 'medium' significance must include an erratum (correction notice) explaining the corrections made. An erratum may be published as either a separately released statement or as a short correction notice for incorporation into previously published information.
Generally, errors of 'low' significance but relating to data would also require an erratum. However, some errors of low significance (for example minor text errors) could be corrected without a formal statement. The manager and publishing manager should determine whether an erratum is necessary in these cases.
Error reports
An error report should outline at a minimum:
- the nature of the error and how the error occurred
- the significance rating established
- suggested courses of action and the final action taken
- suggestions for avoiding that type of error in future.
A central repository of all completed error reports must be maintained to:
- ensure that all necessary information is available for staff to respond to questions about the published and/or corrected information
- provide a resource that enables staff to review processes and make improvements to prevent errors occurring
- provide a permanent record of how previous errors were resolved and how similar errors may be handled.