European Central Bank – DG Statistics* Seasonal adjustment of chain-linked volume measures in Quarterly National Accounts: Findings of a European Task Force Martin Eiglsperger European Central Bank – DG Statistics* The 2008 World Congress on National Accounts and Economic Performance Measures for Nations Washington D.C., 12-17 May 2008 * The views expressed in this presentation are those of the author and do not necessarily reflect the views of the European Central Bank.
1 Background Annual chain-linking of Quarterly National Accounts (QNA) volume measures By almost all statistical offices in the EU since 2005 Implementation not fully harmonised across EU countries 2007: European Task Force, chaired by Eurostat and the ECB; 19 EU countries participated → impact of chain-linking techniques and relevance for seasonal adjustment Detailed investigations by Robert Kirchner, Ingo Kuhnert, Sven Öhlén and Marcus Scheiblecker
Chain-linking techniques (1) 2 Chain-linking techniques (1) Three approaches to the annual chain-linking of Laspeyres volume indices in average prices of the previous year: Annual overlap technique: links refer to the volumes of the previous year at average prices of the year One-quarter overlap technique: links refer to the volumes of the fourth quarter in the previous year at average prices of the year Over-the-year technique: links refer to the volumes of the same quarter in the previous year at average prices of the year
Chain-linking techniques (2) 3 Chain-linking techniques (2) Use of chain-linking techniques in EU countries: Annual overlap technique: most EU countries One-quarter overlap technique: Austria and the United Kingdom Over-the-year technique: Used in the Netherlands and Sweden; change-over to the annual overlap technique planned / (partly) realised
4 Effect on growth rates Annual overlap technique: Year-on-year growth rates do not reflect pure volume changes, quarter-on-quarter rates do within the same calendar year One-quarter overlap technique: Quarter-on-quarter growth rates reflect pure volume changes Over-the-year technique: Year-on-year growth rates reflect pure volume changes
Effect on time consistency 5 Effect on time consistency Annual overlap technique: Annually chain-linked QNA and chain-linked Annual National Accounts (ANA) are consistent over time One-quarter overlap technique: Annually chain-linked QNA and chain-linked ANA are not consistent over time → benchmarking Over-the-year technique: Annually chain-linked QNA and chain-linked ANA are not consistent over time, but deviations tend to be very small
Aggregation and disaggregation 6 Aggregation and disaggregation Aggregation/disaggregation of unchained data Annual overlap technique: Nominal values of the previous year as weights → existing data One-quarter overlap technique: Volumes of the fourth quarter of the previous year at average prices of the year as weights → additionally required Over-the-year technique: Volumes of the respective quarter of the previous year at average prices of the year as weights → additionally required
Time series properties (1) 7 Time series properties (1) Chain-linked volume measures: Trend and weak substitution effect in the source data Source: Kuhnert, I. (2007), internal presentation to the Task Force
Time series properties (2) 8 Time series properties (2) Chain-linked volume measures: Trend and strong substitution effect in the source data Source: Kuhnert, I. (2007), internal presentation to the Task Force
Seasonally adjusted series 9 Seasonally adjusted series Time consistency: not present in seasonally adjusted data Task Force recommends benchmarking Aggregation and disaggregation: One-quarter-overlap and over-the-year technique require additional weights data, annual overlap-technique does not (after benchmarking) Additivity of s.a. data via “indirect” adjustment, i.e. aggregation of unchained s.a. data? → consistency vs. quality of seasonal adjusted data might be an issue
Order of procedures Chain-linking, seasonal adjustment, benchmarking 10 Order of procedures Chain-linking, seasonal adjustment, benchmarking Non-linearity of procedures: Order matters! Benchmarking as the last step in order to fulfil user requirements for time consistency Benchmarking can have a significant impact on the seasonal pattern (ARIMA-models, X-12 output) Seasonal adjustment after chain-linking Research by the Office for National Statistics (United Kingdom): “Drift” of one-quarter-overlap approach could be reduced by conducting s.a. first
11 Summary (1) Over-the-year technique can have a substantial impact on the seasonal pattern → not recommended by the Task Force Annual overlap technique: Time-consistent data + no additional data requirements for aggregation! One-quarter-overlap technique: smooth transition, but does not produce time-consistent series → benchmarking is recommended Seasonally adjusted chain-linked series: benchmarking is recommended in order to produce time-consistent series
12 Summary (2) Consistency in aggregation of seasonally adjusted data: “Indirect” adjustment by aggregating unchained data → consistency vs. quality of seasonal adjusted data might be an issue One-quarter-overlap approach: Seasonal adjustment before chain-linking? Benchmarking!