/ EUROPEAN COMMISSION
Directorate F: Social statistics
Unit F-4: Quality of life
Working Group meeting
“Statistics on Living Conditions”
11-13 June 2013
JMO Building, Room M4Draft minutes
The meeting was opened by Mr Eduardo Barredo, Director of Directorate F "Social Statistics", and Jean-Louis Mercy, head of unit F4 “Quality of life” in Eurostat.
Mr Barredo stressed the importance of the discussion the Working Group (WG) would have about the revision of the EU-SILC legal basis. He indicated that the legal step themselves should be considered in the context of the general forthcoming Framework Regulation on Social Statistics although more "fast track" options could be investigated. However, some gentlemen (ESS) agreement would be necessary for the most urgent improvements like on the improvement of timelinessof poverty and social inclusion data for which implementation is already starting. He also highlighted the importance of usual production activities and improvements already done for both SILC and HBS, as well as the first achievements in the area of thedevelopment of quality of life (QoL) indicators.
For his part, the chairman presented the newcomers in the Eurostat SILC team.
1.2Adoption of the agenda
The agenda was adopted without any modification.
1.3Approval of minutes of meeting 11-13 May 2011
The minutes were approved without any modification.
2. Priorities from policy DG's
2.1 DG EMPL and ISG
Presentation and discussion
DG EMPL presented the on-going policy developments in the area of poverty and social inclusion, including the European Semester and the Employment, Youth and Social Investment package. In particular it informed the WG about the December 2012 Commission blue print for a "deep and genuine EMU", aiming on the next 6-18 months to focus on the increase of competitiveness, then over the next 18 months – 5 years on reinforcing the convergence of employment and social policies and later to create a fiscal capacity for the EU (with modification of the Treaty)with an EMU level stabilisation tool to support adjustment to asymmetric shocks (EMU-wide unemployment scheme as one option). The increasing use of EU-SILC and models based on SILC, including for the 2 flagship publications "Employment and Social Developments in Europe" (Annual) and "Employment and Social Situation Quarterly Review", was also underlined.
For the time being, DG EMPL reminded that its first priority for EU-SILC related statistics was on timeliness, including enhanced micro-simulation by getting more disaggregated data on benefits income from SILC. Other priorities were on developing the analysis of dynamics of poverty (persistent AROP and further studies) and onthe role of social protection and redistributive impact of social spending (incl. in-kind benefits). The mid-term review of the Europe 2020 targets in 2015 is also high on the agenda, in particular on material de privation.
The chair emphasised that the development of the "deep and genuine EMU" would mean that poverty and social exclusion data are used for the unbalance score-board and related procedure, requiring even more high quality standards for EU-SILC data.
For its part, France stated that with their national 9 years longitudinal component they have recently hugely developed their analysis of dynamics of poverty and could make it available to DG EMPL.
2.2 DG REGIO
Presentation and discussion
In the absence of the DG REGIO delegate, the chair reminded that DG REGIO wants that the regional dimension is further developed in SILC, at least at NUTS 2 or some aggregation of NUTS 2 level (possibly NUTS 1 if detailed enough), depending on the countries. The reason is that over the period 2014-2020 DG REGIO aims at monitoring implementation of regional funds with SILC data and developing benchmarking in this area, possibly using only estimated regional data on poverty and social exclusion for countries where NUTS 2 or similar data would not be already available. For the next financing period starting in 2020 the aim would be to include a criteria on poverty and social exclusion for the allocation of funds. In practicethe 2020 funds should be decided in 2018, what means that SILC data fully regionalised should be available from the 2016 data collection.
2.3 DG SANCO
Presentation and discussion
DG SANCO presented the set of European Core Health Indicators (ECHI) and its shortlist of 88 indicators which includes 7 indicators based on EU-SILC data. In this context and for further developments inpolicy setting and monitoring, DG SANCO stated that its priority were on getting data for the ECHI indicators, in particular by keeping and further improving the current health questions in the EU SILC, as well as on timeliness, as well as the development of health modules with an adequate frequency of data.
3. Revision of the EU-SILC legal basis
3.1Outcomes of the Task Force on the Revision of the EU-SILC legal basis
Eurostat presented the proposals made by the Task Force (TF) on the Revision of the EU-SILC legal basis. The background of the need for this revision and activities of the TF was first reminded. In particular, the increasing needs for data over more areas of social policy actions at EU and national levels, the huge policy interest due to the financial and economic crisis, the requirements on timeliness and analysis of dynamics and the need for further improvement of the quality of SILC statistics, in particular the accuracy of the measurements including at regional level, were highlighted. In addition, the framework of the Wiesbaden Memorandum andof the modernisation of Social Statistics, aiming at answering both increasing data requests and rationalisation needs in times of cuttings in resources in NSIs, was reminded. A key element of the modernisation of Social Statistics is the streamlining and modularisation of social surveys and the revision of SILC should be one of the main drivers of it. Finally the increased use of administrative data is also an important strand of the modernisation.
For this purpose, the TF met 6 times from the end of 2011 and agreed at its last meeting in April 2013 on a proposal to be submitted to the Working Group and to the Directors of Social Statistics (DSS) in the autumn 2013.For the latter the presentation and discussion will take place in the broader context of the first discussion for the forthcoming framework European Parliament and Council Regulation on Social Statistics, incl. the revision of the EU-LFS legal basis and options for health, education and ICT statistics. The framework Regulation is expected to be proposed by the Commission by the end of 2015 and for SILC, following outcomes of 2014 tests, a proposal for the main elements of delegated and implementing acts could be developed in parallel and discussed at the WG meeting in 2015.
Hence, the first year of implementation of the revised EU-SILC will depend on the duration of the different adoption procedures but could tentatively be located in 2018. However, some elements will be implemented already before that date on a voluntary basis,due to needs for earlier improvements like on timeliness or the regional dimension.
Eurostat also insisted on the coherence ofthe strategy proposed by the TF, where all its 5 strands (contents, modes of data collection, timeliness, methodology and longitudinal component) should be understood as a part of the whole scenario, where drawbacks of each elements are more than compensated by gains in others and where each proposal is coordinated with others, such as the rotational schemes of both non-annual modules and longitudinal component.
Regarding the contents, the proposal is structured around a reduced annual component / Nucleus, with a maximum of 115 variables, complemented by a two-speed set of rolling non-annual modules (every 3-year and every 6-year). The Nucleus should provide data for theEU-2020 and other major policy indicators.Some variables related to demographic, intergenerational, migrants, and housing topics will be also included in a first wave module (not repeated). For their part, the topics of the every 3-year modules are as follows: children, health (to be developed further as part of the integrated system of health modules), housing conditions and labour (some users would prefer an every 2-year rotation but this would need fewer variables in the Nucleus). Finally, topics of 6-year modules should include validated variables from past-experiences of ad hoc modules and well tested new variables, including new policy modules every other or 3-year. Harmonisation of variables and of small sub-modules among social surveys should also be a target. A good planning of the modules was required by the TF.
As far as different modes of data collection were concerned, the TF recommended having an integrated approach whereby the use of registers should be part of a wider strategy, not as a substitute but as a complement.While the use of administrative data allows for a more efficient data collection, there are also obstacles like timeliness and comparability that should be taken into account. More generally, other modes of data collection such as CATI and CAWI were discussed during the TF/WG. Workshops for exchange of best practices among MS could be organized by MS with support of Eurostat, taking into account the mode effect on data collections. Eurostat also expressed the need for countries using the selected respondent design to converge to the general model to meet with the increasing intra-household data demands, in particular in view of the mid-term revision of the Europe 2020 social inclusion target where material deprivation items at individual level would be needed for all 16+ years old members of the household.
Regarding timeliness, an action Plan on timeliness was agreed between Eurostat and DG Employment and Social Affairs of the European Commission. Majordevelopments of the action plan were already on-going and were presented under other items of the agenda. The main specific actionmentioned under the general discussion on the revision of EU-SILC was the proposal forimproving by 6 months the data availability of the whole EU-SILC (cross-sectional) datasets, i.e. in June N+1. Eurostat indicated it hadlaunched a DSS consultation about national plans for improvement of timeliness.
Concerning precision requirements for EU-SILC, current ones are based on minimum effective sample sizes. However, the DIME (Directors of methodology) Task Force on Accuracy for EU household surveys didn'trecommendexpressing precision requirements as effective minimum sample size. Accordingly, for percentages/proportions the prioritised option for EU Regulations is to express precision requirements in terms of precision thresholds.A solution for EU-SILC precision requirements, compliant with the recommendations of the DIME TF while being fully equivalent to the current situation is to use the following formula based on the measured standard error:
se < √((?(?−?))/?)where se = standard error;p = proportion (value of the indicator); X a figure stemming from the desired precision.The advantage of the formula is to express the requirements on the basis of the actual estimates, with no need to estimate the design effect (deff).
It was proposed to base such precision requirements on the AROPE indicator (p), setting X to a value corresponding to the current requirement in terms of minimum effective sample size for the cross-sectional component, from 3 250 for Luxembourg to 8 250 in Germany. This would yield, for a value of the AROPE indicator at 24% to a half-length of the 95% confidence interval ranging from around 1 percentage point for larger countries to 1.5 percentage points for smallest countries.
Concerning compliance monitoring towards precision requirements, three principles should underpin this activity: transparency, tolerance, and use of standard tools (the method developed by NETSILC2 to calculate the variance could be a starting point).
In addition, as explained above, DG REGIO asks for the possibility of calculatingthe EU-SILC based target indicator at regional level (NUTS 2), as it is considering using poverty-inequality indicators in the funds allocation process for the next cycle starting in 2020 (data should be available 2018 onwards) and already ensuring monitoring before that date. The desired requirements for DG REGIO would be to have a 95% confidence interval for the AROPE indicator at 24% of ±2-2.5 pp. The fine tuning of this requirement will be bilaterally discussed with the Member States (MS), and Eurostat has launched a DSS consultation about national plans for developing the regional dimension.
It was also reminded that concerning quality reports a pilot was launched at the 2012 WG meeting in order to test the new tool NRME and improvements proposed andthat the corresponding state of play would be presented under item 5.2 of the agenda.
Moreover, concerning the longitudinal component, it was indicated that based on the work of a contractor presented at the TF meeting in October 2012, Eurostat proposed and the TF supported using a 6-year rotating longitudinal component. A 6-year panel would give more opportunities for longitudinal analysis and indicators and as well as enhance the quality and visibility of the persistent AROP indicator. While maintaining stable the total number of households interviewed every year, the number of households available for calculating the persistent AROP would double.For the transition, it was proposedto keep for waves 5 and 6 only a part of the households interviewed in wave 4 (around 60%) for 4 years before keeping them all at the same time of the first introduction of a panel with a size adapted to the new rotating length (around 75% of the previous panel size at first wave);
The contractor, prof. Peter Lynn, presented the advantages of 6 waves, namely better estimates of persistence of poverty (greater statistical precision, enhanced ability to compare subdomains) and analysis of poverty dynamics over time(deeper understanding of issues that can be studied currently, and new analyses that are not currently possible). The proposal would also involve less field work effort: fewer cases issued each year, smaller proportion of issued cases from “wave 1” and therefore 24% reduction in number of “wave 1” cases issued each year. Prof. Lynn made clear that the change is unlikely to have any significant effect on wave 1 response rates, attrition should be similar to currently at waves 2-4 and very low at waves 5-6 and in any case, various measures were proposed to improve response rates.Eurostat stated these points were coherent with some preliminary studies it conducted.
Finally, Eurostat indicated that the TF will continue its work for monitoring together with Eurostat the 2014 testing phase and defining precisely the different elements of the revised EU-SILC in 2014-2015.
General comments and contents
Countries using CATI stated that it limits the length of the questionnaire and that this should be taken into account. The pressure of the timeliness action plan and related needs for priorities was also raised; Eurostat indicated that timeliness actions plans are to be already implemented while the full SILC revision will take place in a second time. Concerning harmonisation with other surveys, in particular the LFS, Eurostat clarified that this is limited to some elements such as, e.g., relevant variables or short sub-modules and methodology for precision requirements. For their part, countries using the selected respondent model stressed that they want to stick fully to it. Eurostat stated that some suggestions could be however tested, e.g., for the collection of material deprivation at personal level.
Concerning sending data in June N+1, numerous countries indicated that already in 2014 for 2013 SILC data they would be able to start submitting data earlier than they currently do.However, some MS might still have difficulties to meet the target of June 2014 but will submit the data as soon as possible (FR, IT, CY). UK providing income of year N in SILC data N is also in a specific situation.
Precision requirements and regional data
ES expressed its concern about regional precision requirements asin order to cover all regions they would need a significant increase of sample size. For DE also, producing data at NUTS1 would be a huge challenge and discussions are on-going nationally on how to proceed.Eurostat reminded MS that the requirements of DG REGIO are flexible;ideally they would need NUTS2, but regions can also be merged. In case sample needs to be increased, DG REGIO could provide funding for the transitory period, but this is not the only possibility. Rebalancing sample, three years averages and modelling are also possible. Eurostat reminded it asked to DSS to include financing needs in the regional actions plans to be submitted (1st draft) by end of June or beginning of July.
FI, CZ, IS and SE indicated that they wouldn’t support the six years panel and would prefer to concentrate on quality and timeliness. DK, CY, DE, NL and NO expressed concerns and reservations on the proposal. In particular the impact on the response rate, already decreasing in various countries, and the increase in the complexity of the analysis were mentioned as possible drawbacks. For PT enlarging the number of waves was a good solution but it had a concern about the transition and on how to ask to people to answer two more waves broking the trust set at the first interview. This concern was shared by CY, FI and SE. DE indicated it should analyse in depth advantages and disadvantages. FR expressed support and pointed out that complexity would not increase according to their experience with longer panels, using appropriate weighting from the information in previous waves on household not answering after some waves. Moreover it confirmed that it experienced 65% response rate at the 1st wave while the attrition at the 9th wave is only around 10%. FR also reminded that a similar extension took place in ECHP and the impact was low.UK supported the proposal and pointed out the advantage in terms of analysis. AT supported the proposal but needs official agreement by its Ministry. IT also supported it but needs to consider the whole revision package.