Relevance & Research Question:
Compared to other survey modes, web-surveys are more easily conducted worldwide. Due to the absence of proper sampling frames, most of these surveys will be volunteer surveys, either inviting respondents from an Internet panel or inviting web-visitors to complete a survey. For the latter ones, there is a three-step selection process, namely through the access to Internet, the interest in the website's topic and the decision to take up the web-survey. While most studies so far analysed selection bias for one or a few countries, this paper firstly explores the nature of selection bias for 28 countries and secondly analyses in how far cross-national differences in the observed selection bias can be attributed to explanations related to Internet access and social and cultural indices.
Methods & Data:
The paper analyses data of the worldwide, continuous, volunteer, multilingual, cross-country comparable WageIndicator web-survey on work and wages. Using the 2010 web-survey data, for the 28 countries of interest, simple weights have been computed for six categories (gender * 3 age groups) using labour force projection data from the International Labour Organisation (ILO). Country-level explanatory variables were taken from Internet Access Tables and from the Indices of Social Development (ISD-database). Besides the usual description of selection biases, multi-level analysis has been applied in order to account for differences within and between countries in an appropriate way.
Results:
Not surprisingly, in almost all countries, the underrepresentation in the survey increases with age. Moreover, in all age groups it is higher for women than for men. Across countries, the ISD “intergroup_cohesion” index correlates negatively with the underrepresentation of women aged 40+ (-0.62) and 30-39 (-.58). The “civic_activism” index only correlates negatively with the underrepresentation of women aged 40+ (-0.44). Surprisingly, the “gender_equality” index does not correlate with women's underrepresentation.
Added value:
Most studies about selection bias have focused on one or a few countries, exploring within-country differences. Thanks to a large number of observations in a large number of countries, this paper adds to the body of knowledge by introducing country-level explanations for differences in response to volunteer web-surveys.