Conferences

The techniques used for nonresponse bias adjustments in household surveys typically vary from country to country. This is largely due to differences in the availability of auxiliary data for nonresponse adjustments. Whilst some countries have detailed frame data, others have to resort to aggregate information of varying quality. In cross-national surveys these differences in availability of auxiliary information pose a problem for a coherent nonresponse adjustment strategy.

Recently, researchers have started exploring the virtues of contact and neighbourhood data in nonresponse analyses. For these data to be suitable for nonresponse bias adjustments they need to be related to both response propensity and key survey estimates. The European Social Survey (ESS) collects standardised contact data in each participating country, thereby documenting the processes leading to nonresponse across countries. The advantage of such data is that they can be collected in almost every country for both respondents and nonrespondents. Their suitability for nonresponse adjustments, however, is yet to be proven.

This paper looks at whether the ESS contact data can be useful when adjusting for nonresponse bias in key survey estimates. Specifically, it examines whether contact data can contribute to nonresponse adjustments over and above the effect of traditional post-stratification weights for gender, age and education. As a baseline the paper uses the age, gender and education distributions available from frame data in a selection of countries. To this baseline it compares the contribution of weights derived from the ESS contact data with the contribution of preliminary ESS post-stratification weights.

Keywords: nonresponse bias, nonresponse adjustment strategies, auxiliary data, cross-national surveys