Institution

Eurostat

Year

2015

Level

Intermediate to advanced

Abstract

This second issue of 2015 is dedicated to research that was presented at the conference ‘New Techniques and Technologies in Statistics’ (NTTS) in March 2015. Producing QNA estimates typically involves benchmarking previous years’ estimates, where an annual benchmark is available. In the absence of a current annual benchmark, extrapolation is required based on a modelled relationship with an indicator series. As additional benchmarks become available, published QNAs are revised. The authors of the paper seek methods to improve these revisions. They explore different benchmarking and forecasting combinations applied to both the ‘forward’ and ‘back’ series on Australian economic series and on synthesised data. They introduce metrics to assess different methods based on magnitude and timeliness of revisions, apparent bias, and preservation of movements from the indicator.