The aim of Eurofound's survey work is to provide high quality information on the quality of living and working conditions in Europe. In order to realise this goal each of the surveys targets a different population to get specific information.
- The European Working Conditions Survey questions workers to gain insight on the quality of work and employment.
- The European Company Survey targets managers and employee representatives in companies to gather information on workplace practices.
- The European Quality of Life Survey interviews European citizens to get a perspective on living conditions and perceptions of quality of life.
Despite these differences in terms of target population, Eurofound aims to harmonise survey methodology whenever possible, to ensure that lessons learned in one survey are implemented in the others.
As with all research projects both Eurofound stakeholders and experts in the relevant field are involved in the development, implementation and evaluation of every wave of each survey, ensuring their relevance for European and national level policy makers and social partners.
Through expert involvement in questionnaire development, as well as the involvement of our national level experts in the translation process, Eurofound aims to ensure that all survey questions capture the real-world phenomena they are intended to capture (validity).
Through carefully selecting representative samples, choosing the most suitable mode of administering the questionnaires using state-of-the-art technology, recruiting experienced interviewers and providing comprehensive training for interviewing and coding, as well as through minute checking of the collected data, sophisticated weighting, and appropriate analysis Eurofound aims to ensure the consistency of survey measurements (reliability).
Eurofound’s commitment to producing high quality information is further shown in its quality assurance strategy for surveys.
Sampling
Regardless of whether the target population of a survey are European citizens, workers or establishments, it is not feasible to collect information from all members of that population. A sample of respondents is chosen to be as representative as possible of the total population.
Eurofound aims to use sampling frames of the highest possible quality. In each country, it tries to find a register which covers at least 95% of the target population. For the ECS these registers preferably contain contact information of establishments and otherwise of companies. For the EWCS and the EQLS the registers usually contain address information of households and sometimes of individuals. In the EWCS and EQLS, when no such register is available, a list of possible respondents is generated through a so-called ‘random route’ procedure.
Eurofound aims to draw samples that are large enough to give valid results on the level of individual countries and that reflect the distribution of the European population well enough to enable generalisations to Europe as a whole. The first aim requires that the sample size on a country level is at least 1000 respondents, reducing the margin of error to an acceptable level. The second aim is the reason for selecting bigger samples in bigger countries, as done in the most recent EWCS and EQLS.
Coding
Sometimes it is necessary to ask open-ended questions i.e. questions that are not directly answerable by a pre-determined set of responses. This may be the case for sector, for example. The interviewer will first note the given answer in all detail and later on assign it to the appropriate category in the Europe-wide statistical classification of economic activities (NACE).
Coding will typically have to be applied for open-ended questions related to income, the respondents’ educational level (via ISCED), as well as the region they are living in (via NUTS).
Weighting
Once the fieldwork phase is completed, the datasets have to be weighted to compensate for various possible causes of imbalance in the sample. For example, different people have a different probability of being selected for the survey has to be taken into account statistically. The bigger the unit (household/company) people are living or working in, the smaller their chances of being interviewed. Also, variations in the level of willingness to participate in a survey may result in the under-representation of certain respondent groups. The differences between countries in the size of their workforce, are not (fully) reflected in the sample size in each country so weighting is used to ensure that larger countries weigh heavier in the EU level results.
The European Labour Force Survey (LFS) is frequently used as source of reference data for judging the extent to which the data is representative of the population.
The methodology is adapted to the specific requirements of each survey - more information is available on the individual survey pages.
Exploring the potential of big data for Eurofound research
Surveys have never been the only source of large-scale, quantitative data, and discussions about better integrating data from other sources in social research have been ongoing for quite some time. However, in recent decades, with the increasing connectedness of individuals and organisations to the internet, the amount of data available from sources such as web applications, mobile devices, sensors, video streams and social media has exploded.
In this context, Eurofound is exploring the potential use of big data to complement and contextualise the data from its existing research efforts, its surveys in particular. The project currently focuses on three strands, and this work is discussed in three working papers.
- Working paper: Using big data to improve survey sampling
Reflects on the current and possible future opportunities for using big data to enhance sampling frames or even to generate sampling frames for online surveys, and examines the extent to which a full probability approach is achievable or can at least be approximated. Also discusses the possible gains in efficiency.
- Working paper: Using big data analytics to support decision-making
Offers insight into the most appropriate analytical techniques for collecting and analysing ‘true’ big data (such as the number of people visiting a place or doing an action online, movements of people based on mobile data, and geolocation). Indicates where approaches differ from ‘regular’ statistical analysis, highlights techniques that Eurofound would need to familiarise itself with, and describes the software and hardware requirements.
Explores the potential and limitations of using social media content as a data source and of the various social media platforms to which this approach could be applied (for example, Facebook, LinkedIn and Twitter), using one or more practical examples.