CONCLUSIONS AND FUTURE WORK
The nature of the national accounts compilation process is that source data become available with a lag. There is generally a succession of source data beginning with incomplete preliminary monthly or quarterly indicator data. These data are eventually made consistent with preliminary annual data sources and eventually by final annual data that has been balanced within a supply and use system. This may be a number of years after the reference quarter. Even then there will be ongoing changes from improvements to methods and occasional changes to national accounting standards that are backcast to earlier periods, although it can be expected that estimates will 'settle down' within a reasonable period of time. The national accounts data are also subject to transformations to produce chain volume and seasonally adjusted and trend estimates to make them more useful for economic analysis.
Revisions are a natural consequence of this process. It is efficient if successive revisions keep moving the initial estimate towards a more accurate estimate. There is clearly a trade-off between timeliness and accuracy/reliability in that case. The processes to create chain volume and seasonally adjusted estimates also make their own contribution to revisions. Although users of the national accounts estimates should be aware of the trade-offs, they also need to be informed about the reliability of initial estimates to assist them to make their own judgements about how much weight to give to early estimates. Statistical agencies are also interested in this information as a guide to where they should put scarce resources into making quality improvements.
The study presented here is aimed at both informing users of the national accounts and the national accounts compilers themselves. It has provided intelligence that can be used to input into the setting of priorities for quality improvement. Some work is already planned to examine in more detail certain aspects of revisions. In particular, the impact of the seasonal adjustment process on revisions to seasonally adjusted and trend series will be examined. The series identified in this paper as likely key drivers to revisions in GDP will also be examined in more detail to see what lessons can be learned to improve processes in the future. The results of this and other work will be made available from time to time.