Final report of the DMES Task Force
on Benchmark Revision Policy
Version 29 November 2016
List of abbreviations and acronyms
BOPBalance of Payments
CMFBCommittee on Monetary, Financial and Balance of Payments statistics
COFOGClassification of the Functions of Government
COICOPClassification of Individual Consumption by Purpose
DMESDirectors of Macro-Economic Statistics
EDPExcessive Deficit Procedure
EDPSWGEDP Statistics Working Group
EHRPEuropean Harmonised Revision Policy
ESAEuropean System of Accounts
ESCBEuropean System of Central Banks
ESSEuropean Statistical System
EUEuropean Union
FRIBSFramework Regulation Integrating Business Statistics
GFSGovernment Finance Statistics
GNI Gross National Income
GNICGNI Committee
IIPInternational Investment Position
MGDDManual on Government Deficit and Debt
NAWGNational Accounts Working Group
PEEIsPrincipal European Economic Indicators
SBSStructural Business Statistics
SDMXStatistical Data and Metadata Exchange
SIMSTATSingle Market Statistics
1.Introduction
At the 12th meeting of the Directors of Macro-Economic Statistics (DMES) on 25-26 June 2015, it was decided to:
-establish a DMES Task Force on Benchmark Revision Policy (DMES TF on BRP);
-organise a DMES Seminar on benchmark revisions in December 2015.
The seminar on benchmark revisions was organised on 11 December 2015 as part of the December meeting of the DMES.
With input from the DMES Seminar a mandate (Annex A) was drafted and a Task Force was formed (See Annex B for the composition of the Task Force).
In parallel, a CMFB Task Force on Harmonised European Revision Policy was active. This Task Force focussed on (quarterly and annual) routine revisions.
The two Task Forces worked closely together and two joint meetings were organised (see Annex C for an overview of the meetings).
The Task Force met four separate occasions and had very good discussions on the various relevant issues. Based on these discussions a number of recommendations were formulated and are listed in the different sections of this report. An overview of these recommendations can be found in the final section 10 along with the Task Force'sconclusions.
First, the report outlines the background, goal of the Task Force and the terminologies used throughout the report. Then, the substantial issues are dealt with, namely the European harmonised benchmark revision year of dissemination, the consistency in the year of dissemination of revised data, benchmark revisions and administrative use of national accounts, and national major revisions. Before coming to the conclusions a separate section deals with the communication and public dissemination of benchmark revisions.
2.Background
Consistency and comparability of national accounts and balance of payments/IIP data within countries, between domains,and across countries is hampered amongst other things by a lack of a harmonised revision policy or to be more precise, an implemented harmonised revision policy. Obviously, for users this situation is far from ideal. Moreover, for compilers an implemented harmonised revision policy would be useful. Discrepancies (asymmetries for instance) would not be blurred by vintage issues.
Co-ordination between source statistics (business statistics, population census, and many others) and national accounts and balance of payments/IIP is another issue where progress can and should be made. Implementing a European harmonised benchmark revision policy for national accounts and balance of payments statistics would be a catalyst for further improvement in that respect.
The topic of harmonising revision policies has been on the agenda for many years - the CMFB discussed this issue in January 2003 while the NAWG (at that time the Working Party on National Accounts) looked at it in February 2001.
In 2011 the ESS guidelines on revision policy for the Principle European Economic Indicators (PEEIs) were finalised. The ESSC endorsed these guidelines in February 2012. The ESS guidelines follow the relevant standards defined in the European Statistics Code of Practice.
In July 2012 the CMFB agreed on the introduction of a harmonised European revision policy for national accounts and balance of payments/IIP, to be implemented by September 2014. The part on major revisions is presentedbelow.The CMFB framework on revisions is in line with the ESS guidelines.
In February 2015 the CMFB launched a questionnaire to monitor the implementation of the revision policy. The majority of respondents (19 out of 29) were in favour of a possible harmonisation of the frequency and publication of major statistical (benchmark) revisions in the EU, though some of them expressed some doubts on feasibility issues. However, only six Member Statesplanned to conduct them on a regular basis, following the European recommendation (every five years).
CMFB European Harmonised Revision Policy (part on major revisions)[1]
Furthermore, the CMFB recommendation states that “the national need for major revisions outside the major (European) ad-hoc revisions is recognised, although not formally in the proposal”.
European Union Member States have recently executed major revisions due to the implementation of ESA 2010 (in 2014), and the EU generally followed a co-ordinated timetable agreed at European level, which also extended to balance of payments statistics. According to the terminology used in the proposed European revision policy from the CMFB, this was a 'major ad-hoc revision'. Usually, during these revisions, new or changed basic data sources and/or new estimation models are also incorporated in addition to conceptual changes.
'Pure' benchmark revisions ('major regular revisions') take place on a regular basis (five to ten years) to incorporate results of changes in basic data sources (surveys and censuses) and/or new estimation methods.
According to the CMFB proposal, benchmark revisions should take place in years ending with '0' and '5', with implementation years ending with '4' or '9'.
It appears that most Member States have applied a major revision based on data concerning reporting year 2010. The common implementation year was 2014.
The revisions due to these major regular revisions substantially differed between Member States. Especially because of the consequences for EU budgetary(Own Resources)calculations, these revisions caused substantial comment in some Member States. The different sizes of the revisions were partly caused by different intervals between major revisions in the different Member States. Thepressure to come to a harmonised European policy for major regular revisions was increased substantially by these events[2].
One of the reasons for discrepancies between national accounts/rest of the world data and balance of payments/IIP data was in the past partial inconsistency of concepts. ESA 2010 and BPM6 are now fully consistent and therefore this source of discrepancies has vanished.There is now a much greater need to address the remaining reasons for inconsistency and to include the need for such cross-domain consistency in a harmonisedrevision policy.
3.Goal of the Task Force
The main goal of the work of the DMES Task Force on Benchmark Revision Policywas to prepare a concrete proposal for the next harmonised benchmark revision for the national accounts, balance of payments and related areas. For this goal, the Task Force had to investigate the main factors to take into account when establishing the timing of a benchmark revision.
The Task Force was also asked to give guidance in the case of a national need for benchmark revisions outside the harmonised approach.
Furthermore, the Task Force was expected to address the issue of consistency between non-financial accounts and financial accounts, between quarterly and annual data, and between rest of the world accounts and balance of payments/IIP. The Task Force would also elaborate on the issue of consistency with national accounts data provided to Eurostat as part of the ESA Transmission Programme and national accounts data used for administrative purposes (for example EDP and Own Resources).
Finally, the Task Force was asked to elaborate on the communication and public dissemination of benchmark revisions.
For the mandate of the Task Force see Annex A.
4.Terminology used in the Report
Benchmark revisions
In the context of the work of this Task Force, the term 'benchmark revision' is used. In the terminology of the CMFB framework these revisions are called 'major regular revisions': major revisions that take place on a regular basis to incorporate results of changes in basic data sources (e.g. surveys and censuses) and/or new estimation methods. The term 'benchmark revision' is closer to the practice of national accounts compilationand is equivalent to 'major regular revision'.
Routine revisions vs benchmark revisions
Initially, it was decided, in good co-operation with the CMFB Task Force, that routine revisions apply to annual and infra annual revision windows, combined with a specified depth to backward revisions. In the context of the harmonised European policy, this depth is 4 years (including T-1).
Some Member States however, execute (or plan to do so) 'annual benchmark revisions' with an impact on time series (usually a set number of years back but beyond the above mentioned routine revision window of 4 years). Annual benchmark revisions with an impact on time series beyond the routine revision window would thus be classified as major revisions. Some of these annual revisions do not comply with the idea of a benchmark revision (comprehensive update of all possible sources and methods with impact on all tables of the ESA 2010 Transmission Programme).
The CMFB Task Force opened the possibility to update time series in the third quarter with an undefined length. This could include the results of 'annual benchmark revisions'. The benchmark revisions in this report refer to 'real' benchmark revisions (see section on harmonised benchmark revisions below).
Focus on the year of dissemination
In the original CMFB framework the timing of major regular revisions was as follows:
"Benchmark years fixed with years ending with '0' and '5', implementing years with years ending '4' and '9', unless this is integrated with a European-wide major occasional revision in another year."
The Task Force decided that co-ordinating a benchmark revision should concern the year that the results of a benchmark revision are disseminated, not the reference year (which some countries call the 'benchmark year').Member States should be free to decide on the most appropriate reference year[3] according to the national data sources that are available[4]. So, years mentioned in this report refer to the year of dissemination only.
Harmonised benchmark revision
Before moving to the proposal for a harmonised benchmark revision, or actually a harmonised dissemination year, it is necessary to define it. The term 'harmonised' refers to a co-ordinated European major regular revision. In case of a major ad-hoc revision (due to the implementation of a new ESA for instance) this is usually combined with such a revision. A harmonised benchmark revision refers here to a co-ordinated European major revision between major ad-hoc revisions, when they occur. There are good examples of coordinating major ad-hoc revisions across the EU with a common dissemination date such as the implementation of ESA 2010, BPM6 and the move to NACE Rev.2.
In the context of the work of this Task Force, a revision was defined as a 'benchmark revision' when, at the time of dissemination,at the same time:
All ESA tables are aligned. (Although in practice this may not be fully achieved due to various deadlines for reporting ESA tables. Those tables delivered with longer transmission delays may then incorporate some further changes since the first results of the benchmark revision were disseminated.)
Time series cover a large number of years (starting in 2000 or 1995 - depending on the requirements in the ESA Transmission Programme – or further back), and in any case more than the 4-years threshold defined for routine revisions.
All pending classification issues for example regarding units have been resolved/implemented.
Major changes in sources or in compilation methods, if required, have been implemented.
Errors identified in an earlier stage have been corrected.
Furthermore, consistency between domains should be mentioned. Although consistency between domains (non-financial accounts and financial accounts, national accounts rest of the world accounts and balance of payments/IIP) is the ultimate aim for every publication, this consistency should at least be reached during the dissemination of the results of a (harmonised) benchmark revision.
Summarising, a harmonised benchmark revision should lead to a situation where at least once every five years a maximum degree of consistency (within national accounts, across Member States and between domains) is reached, and the statistics concerned are based on the best possible data.
National major revisions
For benchmark revisions outside the European harmonised approach, the term 'national major revisions outside the European harmonised approach' or simply 'national major revisions' will be used. The Task Force members disapproved of the use of the word 'uncoordinated' ason a national level these revisions are of course coordinated. Section 8 deals with recommendations regarding national major revisions.
5.Dissemination Year for the next harmonised benchmark revision
Harmonised benchmark revision 2019
In 2014 all European Member States published the results of a major revision due to the harmonised implementation of ESA 2010. According to the terminology used in the CMFB framework, this was a 'major ad-hocrevision'. The majority of Member Statessimultaneously conducted benchmark revisions. So, data published in 2014 contained the results of both revisions. Many but not all Member States chose the year 2010 as new reference year.
Following the CMFB framework the dissemination of the next harmonised benchmark revision should be in 2014 + 5 = 2019.However, for several reasons, many Member States have indicated that they would not be able to meet this. Below, first the current situation is described. Then a two-step approach is proposed. First, how to deal with benchmark revisions in the period 2018-2021 and subsequently a harmonised benchmark revision in 2024 is suggested.
Current situation
Based on inputs from Task Force members, supplemented by information transmitted during the NAWG meeting 10-11 November 2016, Countryplans are currently as follows:
In October 2016, Denmark has implemented a major revision of the balance of payments statistics, going back to 2005.The national accounts data willbe revised accordingly, in November 2016, and other known issues will be dealt with at the same time. A decision on the next major revision has not been taken yet,but preliminary thoughtspoint to2020 as the year of dissemination.
Malta will disseminate the results of a benchmark revision in 2017 and considers another benchmark dissemination in 2019.
Austria will probably disseminate the results of a benchmark revision in 2017.
The Netherlands will disseminate the results of a benchmark revision in 2018. Actually, the reference year was moved 5 years (from 2010 to 2015) but due to an acceleration of the production process the results will be published in 2018 instead of 2019.
Norway, Swedenand Belgium will disseminate the results of a benchmark revision in 2019. Possibly Spainand Finland will choose the same year.
Slovakia will choose between 2019 and 2020 for disseminating the results of the next benchmark revision.
Poland has derogations regarding the ESA 2010 Transmission programme until 2020 and due to this Poland will disseminate the results of a benchmark revision in that year (however, 2019 is also still on the table).
Czech Republic will publish the results of a benchmark revision in 2020.
France aims at disseminating the results of a benchmark revision in 2021, mainly because of changes occurring in business statistics due to profiling. Recently it was decided that in 2018 revised French national accounts data will be published due to adjusted cross border primary income flows. However, the planned benchmark revision with dissemination in 2021 remains.
Slovenia has no special benchmark revision policy, data is revised as need arises. Decision on the next benchmark year is not yet adopted.
United Kingdom and Irelandbenchmark annually, so a choice of a specific year is not relevant for these countries. Although some Member States are considering this approach, most Member States stick to multi-year intervals. Partly to provide users with stable time series, partly because of the heavy burden of executing benchmark revisions.
The other Member States represented in the Task Force (Germany and Italy) have not yet decided on the next benchmark year. Italy is in favour of planning the dissemination of the results of the next benchmark revision in 2019 but has not yet taken a final decision on this issue.
Apart from specific country reasons to decide that an (extra) benchmark revision is needed (like substantial changes in one of the main sources or major developmental programmes delivering significant revisions) there are several European issues that impact all or almost all Member States. However, this does not lead automatically to a common year for a harmonised benchmark revision.
An important issue is the implementation of FRIBS. Actually, this issue consists of two elements. First of all, it is uncertain when the legal procedure for this new Regulation will be finalised (current planning is for 2018). So it is uncertain in which year the results of the changed business statistics will become available as inputs for the national accounts. Moreover, the possible impact on national accounts data is unknown and could be very significant.
Secondly, many Member States will have to change a part of the statistical units they use and examine more complex enterprises in more detail (profiling). The impact of this change is uncertain. Most probably for some Member States the change will lead to the need for a benchmark revision. Unfortunately, at least regarding a harmonised benchmark revision, there is not one single year where every country should be compliant with the (existing) Regulation concerned. The target is reference year 2017 as the first year to have reached 'a high level of compliance', possibly to finalise it fully for the reference year 2018. Data concerning 2018 will be available in June 2020. In case of a huge impact the Member States concerned would naturally execute a benchmark revision and disseminate the results in 2019, 2020 or 2021.
Another example is the next population census. This is planned for 2021 (reference year). The deadline for data transmission is March 2024. However, some Member States will have the data available much earlier. Especially in Member States without a population register, like IE, the possible impact on national accounts estimates is huge and therefore these Member States will need to include the new data as soon as possible. The fact that data comes available at different moment in the Member States makes it difficult to use this factor to determine a common dissemination year for the revised results of national accounts/balance of payments.