Project Description

Research Framework
CHINTEX is a shared-cost research project supported by the European Commission‘s Fifth Framework Programme for Research and Technological Development. It applies to the second key action "Information Society Technologies" (IST-Programme) and there to the cross programme action "New Indicators and Statistical Methods" (CPA.4).

Objectives
The harmonisation of European surveys run by national institutes is a task of growing importance for the European statistical system. The European Community Household Panel (ECHP) switched in three countries from input harmonisation with strict comparability of questionnaires and statistical routines to ex-post harmonisation with reduced comparability of questionnaires and statistical routines. This switch of method was caused by the stop of three national sub-samples and their replacement by ongoing household panels run by other institutes. It is the overall objective of CHINTEX by means of this unique data situation to clarify if it is necessary to have centralised, standardised survey instruments to achieve harmonisation and comparability or if this objective can also be achieved by ex-post harmonisation, by which independent national sources are satisfactorily converted to common concepts, definitions, survey questions etc. Furthermore, the project investigates important hypotheses about the data quality of panel surveys (non-response, reporting errors and panel effects) which are of general interest for survey statisticians.

Description of Work
The project starts with an assessment of the attained level of harmonisation in the three national conversion projects with respect to variable conversion and application of unique imputation rules and weighting schemes. In a second step a taxonomy of conversion problems for non-trivial cases shall be established to produce results of general interest. The project will demonstrate exemplarily how statistical tools may be used. Another branch of harmonisation is the use of common rules for imputation and for the construction of weights. The variability of population estimates according to these two sources is investigated. Furthermore we will look upon the impact of field related factors that cannot be harmonised ex-post. Their potential impact on non-response, quality of income data and estimation of statistical models will be examined to check whether long running panels have lower non-response and a better quality of income data. On the other hand panel attrition is regarded to introduce a bias for the estimation of statistical models. CHINTEX will investigate this hypothesis by comparing estimates from the old panels with those from new ones and by using register information in the case of the Finnish ECHP sample. This register information will also be used for a quality assessment of the ECHP imputation rules with respect to income. In addition, a cross-validation approach is used to exploit the information of the parallel panels. If the evaluation indicates a poor performance of the ECHP imputation rules, alternative suggestions shall be made. In this respect CHINTEX will concentrate on the longitudinal aspects of imputa-tion. The project will organise three workshops open to academic and official statistics to disseminate and discuss its results.

Expected Results
The project establishes new methodological framework for the conversion of data that can be used for the ECHP but can be generalised for other surveys. It presents empirical evidence on the effect of switching from input to ex-post harmonisation and analyses the effectiveness of the ex-post approach in producing comparable information. Insight on the impact of panel duration on the quality of the data is given and important hypothesis about possible panel effects are tested. The project presents an evaluation of the ECHP imputation rules on income and suggests alternatives whenever necessary.

Research Consortium
The project is carried out in co-operation of national institutes of official statistics and research institutes with high expertise in the field of panel surveys.

Statistisches Bundesamt
Institut für Forschung und Entwicklung
Wiesbaden – Germany
Johann Wolfgang Goethe Universität
Institut für Statistik und Mathematik
Frankfurt am Main – Germany
DIW (German Institute for Economic Research)
German Socio Economic Panel Study
Berlin – Germany
University of Essex
Institute for Social and Economic Research
Colchester – United Kingdom
CEPS/INSTEAD
Household Division
Differdange – Luxembourg
Statistics Finland
Social Statistics
Helsinki – Finland
Åbo Akademi University (Subcontractor)
Department of Economics and Statistics
Turku – Finland
Statistisches Bundesamt will act as the co-ordinator of the project.

Data Base Used
CHINTEX uses a data base which is unique in multiple aspects. For the three years 1994 to 1996, parallel panel surveys on private households had been carried out in Germany, the United Kingdom, and Luxembourg. In each country, one panel was part of the input harmonised European Community Household Panel (ECHP) used by official statistics, and the other panel was a sample for scientific use. After the stop of the three ECHP sub-samples, the national scientific panels had to be converted to become part of Eurostat’s harmonised ECHP data base.

In Finland, ECHP data are available for the waves 1996 until 1999. The 1996 wave was merged at the individual level with register information on income. CHINTEX is going to extent the register merge for the entire period.

Eurostat holds two harmonised data bases for the ECHP: the production data base (PDB) and the user’s data base (UDB). The PDB is the original ECHP data base including variables very close to the blueprint-questionnaires. The UDB includes variables at a higher level of aggregation and under strong anonymisation. It is the data base passed to scientific and official data users for analysis.

• Assessment and Evaluation (WP 8),
• Dissemination and Exploitation (WP 9) and
• Project Management (WP 10).

• WP 1 "Assessment of the Attained Level of Harmonisation"
• WP 2 "Non-trivial Conversions and their Methodology"
• WP 3 "Imputation and Weighting Homogenisation"
• WP 4 "Non-Response"
• WP 5 "Quality of Income Data"
• WP 6 "Panel Effects"
• WP 7 "Imputation"