|Page tools: Print Page Print All RSS Search this Product|
New National Accounts Benchmarking Project
The National Accounts Branch (NAB) uses annual and quarterly data sources to compile quarterly GDP. Benchmarks (or annual data) are better quality because they come from more reliable sources (e.g. annual surveys with larger sample sizes). Indicators (or quarterly data) are often less reliable, less detailed and/or less appropriate in scope than those annual benchmarks. NAB uses a benchmarking procedure to combine indicators and benchmarks and align them to produce better quarterly economic indicators.
NAB uses the benchmarking method developed by Denton (1971)1.The estimates of the benchmarked quarterly data are derived by minimising the sum of squares of the quarterly differences of the ratio of the benchmarked quarterly values and the quarterly indicators. This is to ensure that the benchmarked series are as proportional to the indicator as possible. This method is applied over a five year moving window and subject to the constraint that the sum of the benchmarked quarterly estimates for a given year equals the annual benchmark. (Note: This is the multiplicative version, there is also an additive version of the method.)
When the annual benchmark series ends the NAB applies a simple extrapolation method to the quarterly estimates beyond the annual benchmark. For these estimates the benchmarking procedure assumes that the quarterly growth rates of the benchmarked series are the same as those of the quarterly indicator (this is done by carrying forward the quarterly Benchmark-to-Indicator (BI) ratio for the last quarter of the most recent benchmark year). This method can cause large revisions if the annual benchmark and the quarterly indicators are not highly correlated.
Many experts have proposed an ABS project to investigate more advanced extrapolation methods that can minimise quarterly revisions due to benchmarking in the Annual National Accounts (ANA). The new National Accounts Benchmarking Project will consider a number of methods including: X-12 ARIMA, Exponential Smoothing and univariate regression models.
The main outcomes of the project include (i) assessing if alternative extrapolation methods can minimise revisions and (ii) investigating the possibility of automating new extrapolation methods using the current procedure.
1 Denton, F.T. (1971) Adjustment of Monthly or Quarterly Series to Annual Totals: An Approach Based on Quadratic Minimisation. Journal of the American Statistical Association Vol 66, No. 333, 99-102
These documents will be presented in a new window.