Online surveys are developing rapidly for producing national statistics, in particular in emergency situations. However, this survey mode is often associated with a low response rate and self-selection behaviour related to the topic of the survey, so that endogneous selection issue can arise.

Mixed-mode is an opportunity to address this issue directly, provided that the protocol is adapted to construct instrumental variables of participation. We proposed to use Heckman’s selection model to produce a weighting which corrects for endogenous selection.

This method is illustrated with the French EpiCov survey, collected online and by phone, during the Covid-19 crisis.