Backcasting Time Series During 2008 SNA ANZSIC 06

Backcasting Time Series During 2008 SNA / ANZSIC 06 Implementation Michael Davies, Division Head, Macroeconomic Statistics Division, Australian Bureau of Statistics September 2014 1

2008 SNA / ANZSIC 06 backcasting • • Implementation of 2008 SNA and ANZSIC 06 – a fundamental reworking of the Australian National Accounts Annual benchmark series backcast from 1994 -95 to 2007 -08 Annual series backcast from 1959 -60 to 199394 Quarterly series backcast from September quarter 1959 to June quarter 2009 2

Challenges (1) • The new industry classification system is fundamentally different to the previous industry classifications Starting periods for national accounts series across individual components and underlying data sources vary significantly Historical series are divided into two periods: • • ‘Live’ compilation period – in which time series can be recalculated in response to changes in standards & classifications; Maintained compilation period – in which no detailed data available making impossible to ensure changes in standards & classifications are represented 3

Challenges (2) • • • Most input data series were not available for the full time period (only capital expenditure data went back to 1959 -60) There was often little information available for constructing historic estimates for new series such as R&D There were significant breaks and inconsistencies between historical series and new series 4

Backcasting & Splicing Principles • Backcasting involves removing breaks in series between ‘live’ and ‘maintained’ periods in national accounts Converting growth levels of historical series into those comparable with that for new estimates while keeping growth rates of the former intact The following splicing principles were adopted • • • Undertaking splicing at lowest feasible level The results should be economically plausible No unusual seasonal variation at the splicing point Pragmatic approach for extrapolating time series in ‘maintained’ periods in case of no historical data 5

Bridging • • • Bridging involves converting historical data in old industry classifications into series in new industry classifications This is done using known relationships between series in a common period (“bridging factors”) to extrapolate time series in new industry standards in ‘maintained’ periods Bridging factors are calculated over several time periods to remove influence of short-term volatility 6

Managing the backcasting process • • Closely managing the incorporation of backcast data into the national accounts Using common techniques as far as possible to help ensure the coherence of the results Maintaining flexibility and finding the right balance between backcasting approaches Ensuring that the results were plausible and that economic history, particularly quarterly growth rates, did not change unless affected by 2008 SNA implementation (e. g. capitalisation of R&D) 7