General revision policy for Statistics Iceland

Policy and objectives for revisions

Results are revised when Statistics Iceland changes published results using standard procedures and the policy covers all official statistics published by the institution. The policy aims at making it clear for users when, how and why results that have been published are changed. Users include both producers of statistic and other users of official statistics. This policy applies generally for revisions of results published by Statistics Iceland but domain specific revision policy is described in the metadata and the relevant operating procedures. 

Rules for revision policy

The general rules and methods used in the European Statistical System:

I. European Statistics Code of Practice (CoP) stipulates:

• Revisions follow standard, well-established and transparent procedures (item 8.6).
• Errors discovered in published statistics are corrected at the earliest possible date and publicised (item 6.3).
• Advance notice is given on major revisions or changes in methodology (item 6.6).
• Revisions are regularly analysed in order to improve statistical processes (item 12.3).

II. Quality Assurance Framework of the European Statistical System (QAF)including more detailed guidelines of revisions.

III. ESS guidelines on revision policy for PEEI’spublication describing thoroughly the main guidelines for revisions.

Reasons for revisions

Results are mainly revised when new and better data sets become available. Traditionally regular revisions are conducted for many domains such as national accounts. Preliminary results are often released before final results are available according to advance release calendar to meet user demand for timeliness. Results will be more reliable when new data become available and revised figures are published. The main disadvantage of a timely publication is that preliminary results sometimes change drastically when final figures are published having effect on the trustworthiness of the producers and the results.

Major changes in methods or definitions occur and lead to revisions. In such cases the changes are organised and announced in advance in news releases or in another way.

Need may also arise for revisions of results that are not preannounced. In such cases the reason for this is explained in news releases or explanatory notes on case by case basis. 

Preliminary results and analyses of revisions

It should be stated clearly in table and text in a publication if the results are preliminary. It is necessary to analyse the results regularly to ensure the quality of the data by measuring the effect of the changes on results and compare the results to other statistical indicators. When new results are published instead of older one there could arise need to preserve and publish the original findings to make it easier for users to analyse the results. Decisions concerning which data set to preserve should be taken based on analysis of the results and by consultation with users. Extensive changes in results should be explained in the news releases. 

Statistical revisions are conducted in three forms

I. Regular revision (planned in advance)

Two quality dimensions are important for official statistics, timeliness and accuracy of results. As the aim of the publication of Statistics Iceland is to publish results as soon as possible it is common that results are based on preliminary figures that are later revised. Accuracy increases usually with time, both for short term statistics as well as longer time series usually as new and better data sets become available, e.g. from public registers. Statistics Iceland states if results are preliminary or final and users should treat preliminary results with care as final results are more reliable than preliminary figures.

Statistics Iceland frequently has some projects under development, for example preparation of new statistics that has not been published before or the use of new data sources in the production of statistics. In such cases results can change more often and cover longer time periods than usually is the case for regular revisions. The existence of development projects should be explained in the relevant operating procedures.

When seasonal adjustment is applied previously published results can change without any change in underlying data sets and the methods used for revisions of seasonal data should be explained in the relevant operating procedures.

II. Major revision (planned in advance)

When new international methods or classifications are implemented there is usually a need for changes in methods and the relevant time series. Examples of such changes are new classification of economic activities or changes in the methods in the compilation of national accounts and therefore it may be necessary to revise classifications and adopt new definitions. New data sources are sometimes adopted and older results revised. All major revisions should be preannounced in news releases or in another way

III. Corrections (not planned in advance)

In some cases there may be need for Statistics Iceland to correct results because of minor or major errors that are discovered in the base data or under the data production. The effect of the corrections on the total results both for preliminary or final results and the importance of the data for users are important when decision is taken to correct errors. Corrections should be explained as soon as needed and such corrections or incidents can also be a part of Statistics Iceland information security policy.  

Domain specific revision policy

Domain specific revision policy should be described in relevant work policy

Statistics Iceland aims if possible to keep consistency in revision policy between domains, especially in the case of related statistics.

In the metadata and relevant operating procedures for each domain there is a need for following reporting:

• Frequency and timing.
• Reasons, coverage and time interval.
• Analysis of changes when relevant.

General revision policy for Statistics Iceland

General revision policy available

General guidelines and rules for revisions are available and the decision to store older results is taken in cooperation with users.  

Revisions are preannounced

At Statistics Iceland website revisions are preannounced (in advance release calendar, in a special news release or in the Statistical Bulletin). Timing of revisions and reasons are explained to users so they are aware of what is to be expected. It should be clearly stated if the results are preliminary and information given about if, when or how previous data sets will be changed.

Domain specific revision policy available

The revision policy for each domain is described in Statistics Iceland metadata and in the relevant operating procedure Major corrections are described to facilitate revision analysis for data users.

Revisions coordinated between domains

Revisions should be coordinated between domains as far as possible to increase their usefulness.

Methodological changes introduced as soon as possible

When methodological changes are planned users should be informed as soon as possible about the content and the need for these. Revised results should be published available to all users with explanations about the impact of the methodological changes introduced.

Analysis of revisions

Revisions analysis should be carried out regularly to improve the statistics.

Impact of corrections estimated

When results have to be corrected the impact on the results should be analysed as soon as needed, taking into account the importance of the data for users. Necessary corrections should be introduced in a clear way explaining the reason for them and at the same time securing the data confidentiality.

 Reykjavík 27 April 2016