What is the ROC's national accounts revision mechanism?
In order to achieve timeliness, stability, and correctness, Taiwan's national accounts revision mechanism was determined by the 197th Committee of National Accounts in May 2007 and the 211th Committee in November 2010 as follows:
1. Advance estimates: The advance estimates for each quarter is announced by the end of the month after the end of that quarter. Although the chief goal of issuing advance estimates is to promptly communicate the state of the economy, many sources of data are not available in time to the detriment of correctness. As a result, revision must be performed after obtaining data subsequently from other sources.
2. Preliminary estimates: Preliminary estimates for the quarter are announced seven or eight weeks after the end of the quarter. Although preliminary estimates can be used to further strengthen the accuracy of the advance estimates, because some sources of data are still not fully available, other correlated indicators must be used via correlation analysis or the moving average method, or reverse inference of economy activities and other statistical procedures used to provide estimates.
3. Annual revision: The statistics for the previous two years are revised each November. The chief goal of this annual revision is to revise preliminary estimates results on the basis of updated relevant departmental final accounting, balance of payments, sampling survey results, and factory operation census, but not to improve compilation concepts, methods, or announcement content.
4. Quinquennial revision: A comprehensive revision is performed in accordance with international convention once every five years, and includes tracing of all time series. The content of this revision includes:
(1) Revision of statistical classifications, compilation concepts, and announcement content:
Revision chiefly seeks to achieve harmony with international norms. For instance, in the quinquennial revision in 2014, research and development (R&D) is recognized as part of capital formation, social security schemes are compiled as general government sector and the chain-linking method for computing aggregate growth is adapted from the fixed-base-year method.
(2) Benchmark revision year level correction and time series interpolation:
In accordance with the Industry, Commerce and Service Census, Agriculture, Forestry, Fishery and Husbandry Census and other special survey data not used in annual revisions are employed to perform data correction. (the benchmark year is 2011). In
addition, statistical methods are reviewed and revised. Following the benchmark revision, other time series interpolation tasks were performed.