Set them free. Improving data quality by broadening interviewer’s task

The paper deals with a highly controversial issue in survey data collection: the standardisation of the interviewer's behaviour during the interviewee’s selection of response alternatives. In the light of a large set of data drawn from several methodological studies published in the last 50 years, the author documents a counter-intuitive issue: 1. interviewer’s errors are of secondary importance and far smaller than respondent’s errors. 2. in order to minimise respondent’s errors, we need to broaden the interviewer’s tasks. Focusing on the unsolved problem of multiple word meanings of response alternatives as a relevant part of response bias, the author argues that data quality can be achieved by entrusting to the interviewer a more active role. Of course the aim of reducing respondent’s errors by broadening interviewer’s tasks will surely produce an increase in the interviewer’s effects on answers. However the dilemma to be faced is which kind of errors we prefer (and are more useful) to minimise?