Data Revision

Revisions as a necessary part of compilation

Revisions are an important part of the compilation of energy statistics. While the compilation and dissemination of provisional data often improve the timeliness and relevance of energy statistics, provisional data should be revised when new and more accurate information becomes available. This practice is recommended if countries can ensure consistency between provisional and final data. Although repeated revisions may generally be perceived as reflecting negatively on the reliability of official energy statistics, attempting to avoid this by producing accurate but untimely data will ultimately fail to satisfy users’ needs. Revisions affect both annual and short-term energy statistics, but are often more significant for short-term data [IRES, Ch. X, para. 10.22, PDF p. 146, 2018] — recommendations tracker row X/10.22. See Data Quality for the accuracy–timeliness trade-off that motivates the use of provisional data in the first place.

Two types of revisions

In general, two types of revisions are distinguished [IRES, Ch. X, para. 10.23, PDF p. 146, 2018]:

(a) Routine, normal or concurrent revisions — part of the regular statistical production process, aiming to incorporate new or updated data or to correct data or compilation errors; and

(b) Major or special revisions — not part of the regular revision schedule, conducted to incorporate major changes in concepts, definitions, classifications, and changes in data sources.

Revision policy for routine revisions

Recommendation (tracker row X/10.24): with respect to routine revisions, countries should develop a revision policy that is synchronized with the release calendar, and the description of that policy should be made publicly available. Agencies responsible for official energy statistics may decide to carry out a special revision, in addition to normal statistical data revisions, to reassess data or investigate in depth some new economic structures; such revisions are carried out at longer, irregular intervals and may require changes in the time series going back as far as its beginning, to retain methodological consistency. It is recommended that these special revisions be subject to prior notification to users, explaining why the revisions are necessary and providing information on their possible impact on the released outputs [IRES, Ch. X, para. 10.24, PDF p. 146, 2018].

Features of a well-run revision policy

Recommendation (tracker row X/10.25): countries are encouraged to develop a revision policy for energy statistics that is carefully managed and well coordinated with other areas of statistics, aimed at providing users with the information necessary to cope with revisions in a systematic manner — the absence of coordination and planning of revisions is considered a quality problem by users. The essential features of a well-established revision policy are: predetermined release and revision schedules; reasonable stability from year to year; openness; advance notice of reasons and effects; easy access to sufficiently long time series of revised data; and adequate documentation of revisions included in statistical publications and databases. A sound revision policy is recognized as an important aspect of good governance in statistics, helping both national users of the data and promoting international consistency. The forthcoming Energy Statistics Compilers Manual will provide detailed information on good practices in revision policy [IRES, Ch. X, para. 10.25, PDF p. 146, 2018].

Related

• Dissemination — the dissemination policy that must also cover revision policy (Section A, para 10.2(c)), and the release-timetable framework revisions interact with
• Data Quality — the accuracy–timeliness trade-off and the reliability quality dimension that revision policy manages
• IRES Chapter X — Dissemination — the Chapter X digest this page supports