Tid: 15 mars 2017, kl 13-14
Plats: B705


One year ago I gave a seminar with the title ‘Does nonresponse in surveys matter?’. The background was the rapid increase in nonresponse levels in surveys in Sweden and many other countries. Common wisdom before the turn of the millennium was that a high response rate is essential. However, a number of studies have shown that increasing the response rate by just working harder to obtain more responses is in general not very constructive. Focus among practitioners (at least many of them) and researchers has now changed to the composition of the nonresponse set rather than the size of it. I shall show some theoretical expressions for nonresponse bias and discuss how well they correspond to actual biases obtained in a simulation study and attempt to answer the question in the title. I’ll also point out some implications for our teaching in the first-term course Survey Methodology (undersökningsmetodik).