The Belgian project. As a cycling fan, it was a great experience to work with some Belgians on some very Belgian data. In six large-scale COVID-19 surveys among the Belgian population, we test the classic question-order bias hypothesis: do prior questions affect the answers to later questions in surveys? It turns out that they do!
The strength of this particular test of the question-order hypothesis is the repetition across 6 large-scale survey experiments. Respondents were either first asked about the effectiveness of the COVID-19 handling in Belgium and then about their trust in the following actors: federal, regional, local, EU, and experts. Or the reverse. Furthermore, in the first three experiments, respondents were asked about the actors in the mentioned order. In the last three experiments, this order was reversed.
The results show that the question-order bias is alive and kicking, though the effects on trust differ among the actors. Also, reversing the order of the actors (that is, within the trust outcome battery of questions such that trust in experts is asked first and EU last), the question-order effects seem to get negative (see the figure below).
The paper is written with Wouter van Dooren, Steven F. De Vadder and Koen Verhoest and is published in Public Administration.
