Roundtable 2
Presentations

Some reflections concerning EU harmonisation related to the ECHP and EU-SILC projects
Eric Marlier, Senior Advisor, CEPS/INSTEAD Research Institute, L

For those who are not familiar with the 2 key acronyms I will be using in my talk, “ECHP” means European Community Household Panel and, probably a bit less known, “EU-SILC” means European Union Statistics on Income and Living Conditions.

Rather than discussing the details of the ECHP and EU-SILC, I suggest providing you with a general introduction to these projects and concentrating more specifically on what is our main focus of today, that is the issue of harmonisation –in this case: harmonisation at EU level.

The European Community Household Panel
The ECHP is an EU harmonised cross-national longitudinal survey focusing on household income and living conditions in their multi-dimensionality.

This survey ran from 1994 to 2001. In the first wave (1994) a sample of some 60,500 households i.e. approximately 130,000 adults aged 16 years and over were interviewed across the then 12 Member States. In 1995 Austria, and in 1996 Finland joined the ECHP.  As from 1997, Sweden has provided the cross-sectional data it could derive from its National Survey on Living conditions so as to ensure a complete coverage of EU-15.

For most of the countries the surveys were carried out using the harmonised ECHP questionnaire (also referred to as the blueprint questionnaire), with the sole exception being Belgium and the Netherlands, where ECHP data were coming from a modification of existing national panels to satisfactorily meet the ECHP requirements.From 1997 onwards, the situation became slightly different; in Germany, Luxembourg and the UK the institutes in charge of the production of the ECHP converted national data surveys into ECHP format to replace the ECHP, creating thereby a significant loss in the initial input harmonisation of the ECHP project. Care is needed in analysing the converted data for these countries, as some information might not have been collected in the national surveys so that they will appear as missing in the ECHP. In other cases, variables that were not collected in the national surveys were imputed based on similar though not identical variables.

Within each country covered by the ECHP the surveys were carried out by the “National Data Collection Units” (often referred to by the acronym “NDUs”), which are either the National Statistical Institutes (7 out of the 14 NDUs) or research centres (7 also). The results of the interviews were then transmitted to Eurostat using a format very close to the questionnaire.

These datasets were checked and formatted by Eurostat as the ‘Production Data Base’. This Production Data Base, which is only available to NDUs, is then used by Eurostat for weighting, imputation and construction of the Users’ Data Base. The Users’ Data Base is the standardised, anonymised and more user-friendly user version of the ECHP data made available to researchers under ECHP research contracts signed with Eurostat.

Three central features of the ECHP make this dataset a valuable source of information for researchers.

1. The first feature is the multidimensional nature of the topics covered. The ECHP provides micro-data (i.e. data on individual persons and households) on a wide range of socio-economic topics, which include: income, demographics, education, employment, housing conditions, health, social life, and so on.
2. The second feature is its longitudinal nature. Individuals who were members of a household in the first wave (the “sample persons”) are followed over time allowing data analysts to examine how their circumstances change over time. The ECHP provides therefore information on relationships and transitions over time at the micro level.
3. The third and final main feature of the ECHP, which is key to us today, is the cross-national comparability of the data. The ECHP (apart from those countries using data derived from national sources, as I mentioned earlier) is a harmonised and comparable dataset across countries. This has been achieved through the implementation of common procedures at all stages from the design of a harmonised questionnaire, harmonised definitions and sampling requirements.

The European Union Statistics on Income and Living Conditions
As mentioned earlier, the ECHP ran from 1994 to 2001. From 2004 onwards, it is to be replaced by a rather different approach to producing data across all the Member States, known as the EU Statistics on Income and Living Conditions (EU-SILC), which should become the EU reference source of micro-data on household incomes and social exclusion in the EU.

The priority with EU-SILC will be the provision of quality, timely cross-sectional information (as opposed to longitudinal information) on household incomes and social exclusion. For this reason, whereas the ECHP was a full panel (with all sample persons from the first wave followed for the entire life of the panel), EU-SILC will be more flexible: it will allow for a rotational design in which an individual is followed only for four years (at most). Countries seeking to launch a full panel (like the ECHP, for instance) will, however, be allowed to do so. Luxembourg, for instance, has decided to conduct a full panel, which has started a few weeks ago.

EU-SILC will be organised under a framework Regulation adopted in 2003 by both the EU Council of Ministers (the ECOFIN Council) and the European Parliament. Unlike the ECHP, it will therefore be compulsory for all Member States. Among the EU-15 Member States, 6 have already launched EU-SILC this year or will be soon launching it this year, on the basis of gentlemen’s agreements signed with the Commission; those 6 countries are Belgium, Denmark, Greece, Ireland, Luxembourg and Austria. Apart from Germany, the Netherlands and the UK, the other Member States will start in 2004 as foreseen in the Regulation; Germany, the Netherlands and the UK will only start in 2005. Depending on the country, accessing and candidate countries will launch EU-SILC between 2004 and 2007.

The emphasis in EU-SILC is more on output harmonisation rather than, as was the case for the ECHP, on input harmonisation. Member States will be allowed –and even encouraged- to use both one or several survey(s) and, for those countries that can do it, administrative registers. The sole condition for using multiple sources is that all the data are to be “linkable” at the micro-level, i.e. at the level of both individual persons and households. Member States will, however, be allowed to separate the cross-sectional element from the longitudinal, panel element if they so wish. (We may come back on the use of administrative data if you so wish, given that in my opinion the future of social statistics is directly linked to a full use of administrative data, even though this still poses huge technical and/or legal problems in most Member States… though we are making progress each day in this fairly complex area).

Precisely because, contrary to the ECHP (at least in the first three waves), EU-SILC data will come from different sources in different EU Member States, EU-SILC will thus not be based on the use of harmonised questionnaires in all the participating Member States. What is “harmonised” is the so-called list of target variables that all Member States will have to provide in the context of EU-SILC. The detailed list of topics (but not the harmonised target variables themselves) is included in the EU-SILC framework Regulation that I have just mentioned; it has thus been adopted at highest political level and was clearly a key issue during the negotiations. This was even more so that paragraph 4 of Article 15 of the framework Regulation explicitly foresees that: “In each Member State the total duration of the interview relating to the target primary and target secondary variables of the cross?]sectional component, including household and individual interviews, shall not exceed one hour on average”.

The objective of this major shift in data collection, which encourages full use of existing national data sources (and should thus also improve the cost/efficiency ratio), is to “anchor” EU-SILC in the different national statistical systems. This is definitely a valid aim. And, in the light of the ECHP experience, I think that the methodological choice that was made for EU-SILC was in fact the best one to ensure data quality, with “quality” referring to both timeliness and reliability of the figures; this is crucial given that it is a sine qua non for further progress. This methodological choice was also the most appropriate one to ensure acceptance of this data at national level, which is probably the major failure of the ECHP, which has created important data quality problems. It is obvious that data quality, on the one hand, and acceptance and actual use of the data by Member States, on the other hand, are closely inter-linked.

This being said, we have to be realistic and recognise that the approach now being adopted also has its risks. It is not difficult to see problems relating to harmonisation and non-comparability arising – since even the way a household is conventionally defined varies from one country to the next (we can come back to this in our discussion if you so wish).

Eurostat and Member States are working together on the development of common guidelines and procedures aimed at maximising comparability through a series of implementing Commission’s Regulations to be adopted in the context of the framework regulation recently adopted in 2003, by the so-called committology procedure (under this procedure, these Commission’s Regulations do not need to be presented to the Council and/or the European Parliament: they are adopted by the Commission once approved by the Director-generals of national statistical institutes; one of these Commission’s Regulation will, for instance, describe the details of the agreed target variables mentioned earlier). This work on common guidelines and procedures is certainly crucial, but it certainly does not mean that utmost care will not be required and detailed examination needed as EU-SILC data actually become available.

Another worrying problem with EU-SILC at this stage is timing: as I said it earlier, EU-SILC will only get under way in D, the NL and the UK in 2005 (whereas it will be launched in 2003 or 2004 in the other Member States). First data for all the (current) Member States will then not be available before end-2006 at the earliest, which means that there will be a significant data gap at EU level (as well as, for several Member States, at national level). Methodological work, directly linked to the issue of harmonisation, will also inevitably be required before a link can (possibly) be made, for the agreed social indicators to be used in the monitoring of national socio-economic policies, between the existing ECHP time series and the new EU-SILC ones.

Thank you

Annex 1

Table 1: Nature of ECHP data by country and year (Harmonised original ECHP data and data derived from existing national sources)

Countries	Full ECHP Data Content	ECHP Data Format derived from National Surveys
Belgium*, Denmark, France, Greece, Ireland, Italy, the Netherlands*, Spain, Portugal	1994-2001	-
Austria	1995-2001	-
Finland	1996-2001	-
Germany	1994-1996	1994-2001 (SOEP)
Luxembourg	1994-1996	1997-2001 (PSELL)
United-Kingdom	1994-1996	1994-2001 (BHPS)
Sweden	-	1997-2001 (SLCS) (Cross-sectional data only)

* The ECHP data for Belgium and the Netherlands come from a modification of existing national panels to meet the ECHP requirements.  These are listed in the first column above because this system was in place from the beginning of the ECHP and national questionnaires were substantially modified to meet the ECHP requirements.

Presentation for the Nessie Round Table II on Harmonisation of Data and Technical Access (Cologne, 13-14 June 2003).

Cross-sectional data are data pertaining to a given time or a certain time period, whereas longitudinal data are data pertaining to individual-level changes over time, observed periodically over a certain duration.

See Annex 1 for the details of these variations across countries and national surveys.

Eurostat has prepared a number of detailed documents on the ECHP, ranging from the ‘blueprint’ ECHP questionnaires, through documents dealing with methodological issues, to the agenda and minutes of ECHP meetings.  These are available on the Eurostat CIRCA library web site page dedicated to ECHP.