Conferences

In 2006-2007, non-response survey (NRS) was implemented in the context of European Social Survey (ESS) Round 3 in 4 countries (BE, CH, NO, PL). Its study design was unique making use of the same sampling frame as in the main survey collecting additional information of non-respondents. Focusing analyses on Belgium and Poland, this paper documents measurement of non-response bias through number of socio-demographic characteristics and attitudinal variables between two different groups: cooperative and reluctant respondents on one hand, and cooperative respondents and non-respondents on the other hand. On the basis of core socio-demographic and attitudinal variables between cooperative respondent and nonrespondents, propensity scores are obtained through logistic regression model where these scores are transformed into weights.

Comparing unweighted and weighted survey responses among cooperative respondents, we evaluate the model (propensity scores) by offering different possibilities in calculating these weights.

The paper ends with some discussion and reflection on nonresponse survey design, bias and use of weights in correcting this bias by making references to other possible ways in obtaining weights (ex. observable type of information or additional response units from refusal conversion activities).

Key words: Non response survey, European Social Survey, nonrespondents, propensity scores, nonresponse bias, correction of bias.