Measuring and modeling the (limited) reliability of free choice attitude questions

Brand attitude data often are collected through free choice brand attitude questions. However, during a second interview, respondents to such questions are unlikely to repeat a positive answer they gave during the first interview, which makes individual answers unreliable even if the aggregate results are constant. A simple stochastic model explains this lack of reliability well. The authors validate the model using eight separate data sets and discuss its consequences for consumer targeting and market research.