Targeted letters: effects on sample composition and item non-response

Summary

General population surveys as well as business surveys are affected by low response rates. There is a need to investigate possible ways to increase participation in surveys and to motivate interviewee in answering questionnaires. Even in official statistics the problem is relevant since lack of motivation undermines not only participation to the survey, but it is related to the timeliness in participation, the quality of the answers and hence the quality of the measurements of socio-economic phenomena. A possible approach to cope with this problem, which has received attention in recent years, is targeted design. In targeted designs, several survey features may be the object of interventions (timing of contact, number of contact attempts, mode of contact, interviewers’ assignment, communication features, etc.). In our paper, we consider the invitation letter, which is one possible targeted feature. This is related to motivation to respond to surveys and to respond accurately, completely, and timely. We analyze a case study in the UK Household Longitudinal Study Innovation Panel. At wave 6 of the panel, a randomized experiment was carried out to study the effects of targeted initial letters [1]. We investigate effects of targeted initial letters on sample composition, representativity, and item non-response.