6.1. This chapter discusses the accuracy, reliability and timeliness of ABS GFS output and makes an assessment of the overall quality of the statistics included in each release of data for a given period.
6.2. In any statistical undertaking, there is usually some trade-off between accuracy, reliability and timeliness. The trade-off involves balancing users’ requirements for timely statistics against the time and cost (of the ABS and data suppliers) required to collect and compile statistics of a given degree of accuracy and reliability. Generally, any increase in timeliness comes at the expense of accuracy and reliability.
6.3. In this discussion, accuracy is defined as the closeness of an estimate to the ‘true’ value. Reliability is defined as the stability of an estimate as measured by the size and frequency of revisions made to the estimate over time. These two attributes should always be considered together, as it is possible to have a statistic that is reliable (because it is revised infrequently) but always inaccurate. In general, timeliness refers to the amount of time between the end of the period to which the statistics refer and the date of first release of the statistics to users. However, this definition is not applicable to forward estimates, which users would probably prefer to have before the reference year begins. Because GFS forward estimates are usually not available to the ABS until the end of the current year, timeliness for the forward estimates is defined as the amount of time between the beginning of the forward year and the date of release of the statistics.