Research contexts conducive to the bias in the estimates with VAA-generated data

Voting Advice Applications have become particularly popular tools used in democracies to ask users about their political stances and match them with political parties accordingly. However, VAA users are usually an unrepresentative sample of the general citizenry and this may lead to biased estimates of public attitudes coming from VAA samples. In this paper, we capitalise on the 2009 European Election Study to assess the extent of such bias and identify factors that can explain cross-national differences in VAA bias. Our findings indicate that these differences are due to both country-level factors like poor penetration of VAAs and to question-level factors such as the extent to which participating in a VAA is associated with particular response patterns. The latter element represent an additional source of cross-national variation as we show that the association varies considerably across countries.