A revision is defined as the change to statistical data previously disseminated.
The “Revision policy” refers to the set of rules that defines in which way data are subject to revision.
According to the European Statistics Code of Practice, the Quality Assurance Framework of the European Statistical System (QAF) and the ESS Guidelines on Revision Policy for PEEIs, Istat is committed to guarantee that principles on which revisions are based are respected. In particular:
- the revisions are pre-announced in the release calendar (see Revision Cards);
- in the moment of publication, the statistics are classified as preliminary, provisional or final;
- the causes of revisions are explained and possible breaks in time series are pointed out;
- both qualitative (revision analysis) and quantitative (revision measures) information on revisions are disseminated;
- revisions due to unexpected errors are disseminated out of the release calendar.
Dissemination of information on revisions
The dissemination of information on revisions reflects the fundamental principles of statistical process and product quality.
In order to ensure the standardization of disseminated information, for each short-term survey, Istat publishes a Revision Card. This Card reports information on the revision policy adopted for raw and seasonally adjusted (if produced) time series and a list of the reasons for the ordinary and extraordinary revisions. The calendar of the complete cycle of ordinary revisions is included.
Consistently with the principle of Clarity, for the main short-term indicators, Istat publishes revision triangles (real-time database) in which different versions of data released over time are collected in tabular form. Each row contains the time series released on a certain date; this permits to read by column the story of the released estimates of a given indicator, from the first to the last available release.
In addition, for each indicator, the results of revision analysis are made available. The revision analysis is carried out through the main quality indicators that provide measures of the average size, direction and variability of the revisions, with the aim of improving statistical processes.
Short-term indicator press releases contain both information on the revision measures and on the revision policy. The latter is included in a section the “Methodological note” that is released jointly with the press release in case of ordinary revisions or it is described in an ad-hoc “Information note” accompanying the press release in case of extraordinary revisions.
Classification of revisions
ORDINARY REVISIONS
Frequency: one or more infra-annual and/or annual revisions
Causes: availability of new data, updating of adjustment and processing procedures
Length of time series subject to revision: only the last part of the time series is revised (the number of revised observations for each release can vary with the revision policy applied to the specific statistics)
Channel of communication: section included in the methodological note that is released jointly with the press release
Type of revision: updates
EXTRAORDINARY REVISIONS
Frequency: regular or non-regular intervals greater than one year (usually 5 years)
Causes: changes in the methodology used to process collected data, extraordinary changes in the classification and in the definition of variables
Length of time series subject to revision: all the time series is revised
Channel of communication: production of an “Information note”, where the reasons for revisions are specified and information on the impact of revisions on data are provided
Type of revision: improvements
OTHER UNPLANNED REVISIONS
Frequency: random
Causes: errors in collected data, computation errors
Length of time series subject to revision: revisions of the single erroneous datum
Channel of communication: note on the website or on the document with which the corrected data are released
Type of revision: corrections
Short-term indicators subject to revision analysis, the revision cards and the cycles of ordinary revisions are avalilable in Italian language.