http://www.iser.essex.ac.uk/publications/working-papers/iser/2009-08

When analysing survey data to infer to the general population researchers assume that each selected person (from a probability sample) was contacted and interviewed. However, in every survey there are persons that can either not be contacted or are contacted but not interviewed, i.e. they are nonrespondents to the survey. The extent to which the sample is contacted by the interviewers (i.e. the contact rate) can differ across surveys. In a cross-national survey this contact rate can also differ across countries. Differences in contact rates can pose a problem for cross-national comparisons if the composition of the contacted sample differs across countries. Countries in a cross-national survey can differ in (1) population composition, (2) the way fieldwork is carried out and (3) the effect of (1) and (2) on the contact rate. This paper investigates which fieldwork factors are associated with the contact probability of a sampled person in the European Social Survey (ESS). We look at factors associated with contact within seven ESS countries (Belgium, Finland, Greece, Ireland, Portugal, Spain and the UK) and at factors associated with differences in contact rates between two countries. For the cross-national analysis we aim to disentangle which part of the difference in contact rates is due to differences in population composition, which part is due to differences in the way the ESS was implemented and which part is due to differences in the effect of the population composition and fieldwork implementation on the contact rate.

In the European Social Survey (ESS) contact rates differ across countries. These differences are broadly due to (1) differences in survey implementation, (2) differences in population characteristics associated with contact propensity and (3) differences in the association between 1 or 2 and contact propensity. This paper investigates correlates of contact within and across ESS countries by decomposing cross-country differences in predicted mean contact propensities into (population and fieldwork) characteristics effects, coefficients effects and a pseudo-interaction effect. The findings shed light on the cross-national comparability of the manipulable aspects of the contacting process. In addition, we distinguish factors explaining withincountry contact propensity from factors explaining cross-country differences. Working Paper 2009-08 2009-03-13 Annelies Blom Institute for Social and Economic Research agblom@essex.ac.uk http://www.gesis.org/das-institut/mitarbeiter-adressen/mitarbeiterverzeichnis/?valpha=&selcat=B%3E%3E&order=sortname&id=145&&selres=62&pagecount=1#62