In the last general social surveys in Switzerland (ESS 2010, MOSAiCH 2011 as national support to ISSP) we have a diversity of measures of non-responses biases:
1. Sampling frame indication on age, gender, nationality, civil status among others.
2. Geolocalisation of the address allowing to link to contextual data at the local and regional levels.
3. Environment description for all selected addresses.
4. Sequences of contacts, non-contacts, refusals, etc. on all individuals.
5. Non-responses surveys giving attitudinal information for more than 50% of non-respondents (as well as reference value for a sub-sample of respondents in order to control for time or mode effect).
The aim of this paper is to modelize the probability to participate (and to correct for in case of a satisfactory model) through the use of propensity scores, by mobilizing the information available through these different steps.