Country experience: Australia: data revision policy

20.48.     The Australian Economic Statistics Revisions Policy (ASERP), published by the Australian Bureau of Statistics (ABS), states, in summary, that economic statistics are typically high-frequency data that require a trade-off between accuracy and timeliness, and it aims to maximize both factors. The policy  states that the following principles should be observed: (a) as far as possible, users should be informed in advance of significant methodological changes that will result in revisions; (b) information on revisions, including their reasons, should be kept and published as appropriate; and (c) revisions should be analysed to determine whether improvements in sources or methods could lead to a future reduction in the extent of revisions. 

20.49.    The current ASERP has been presented with challenges. Among those are: (a) the complexity and risk imposed on business statistical centres (BSCs) from having to maintain parallel data for six to nine months on systems that are not designed for that use; (b) clients agreeing to the introduction of concurrent seasonal adjustment and regular revisions based to a desire to know the true result sooner (the current policy of Australia delays the release of that true, or improved quality, data); (c) increasing reliance on administrative data and not being able to impose the ASERP revisions policy on the administrative sources; and (d) maintaining internal as well as external coherence for statistical data. 

20.50.    Trade in services statistics have developed their own revisions policy that is consistent with AESRP on a broad level (that is, maximizing accuracy and timeliness while taking both into account) but also guides decisions BSC use. In summary, it states that survey revisions may be recorded at any stage but should be applied only after the current survey data has been finalized to ensure consistency across publication cycles, particularly during periods of monthly, quarterly and annual releases. Historical revisions should be recorded as soon as they are identified, as they only occur once a year in statistics on the international supply of services and may not cover periods in which revisions are identified. Any revisions that do not fit the other two criteria may be applied at any stage. For instance, if the survey data are revised from the previous quarter, the following steps are taken: (a) revisions are identified and confirmed; (b) revisions are documented; (c) revisions are applied in the processing system for trade in services through to working spreadsheets; and (d) those spreadsheets are updated with only the revisions mentioned above and are sent to other areas for further processing.


Back to C.5. Revision policy