English title: Unit Non-Response due to Refusal

Author(s): Ineke A.L. Stoop -

Language: English

Type: Book chapter

Year: 2012


High response rates are often considered to be the outstanding quality feature of surveys. According to Biemer and Lyberg (2003), nonresponse is a quality feature that many survey users and sponsors have heard of, and the response rate is often seen as indicative of the competence of a survey organization. Survey researchers know by now, however, that nonresponse rates are not linearly related to nonresponse bias (Groves, 2006). They are also aware of the fact that devoting limited resources to increasing response rates with little or no impact on total survey error is not money well spent (Merkle & Edelman, 2002). High response rates after excluding ‘difficult’ groups (minority language speakers, mobile only, far away regions) will not result in a good representation of the target population. High response rates on a poorly drafted questionnaire will not contribute to valid conclusions. As nonresponse rates are increasing (De Leeuw & De Heer, 2002; Couper & De Leeuw, 2003) survey costs and fieldwork duration increase too, and surveys are increasingly hard to plan. Also, nonresponse is a key threat to survey quality when survey participation is related to the outcome variables of a survey. The higher the nonresponse, and the more respondents differ from the nonrespondents, the larger the nonresponse bias and the lower the validity of survey outcomes. This is both true for noncontact and for refusal. This chapter focuses on refusal, both because the major part of nonresponse is usually due to refusal and because refusal is possibly related to the topic of the survey. If this is so, and if surveys on, say, political involvement are completed mainly by those who are interested in politics, the validity of survey results is indeed threatened . The chapter starts with a short introduction on nonresponse bias and on what constitutes a refusal. This is not as clear-cut as one might think. Refusal in mail, e-mail and Internet surveys can usually not be measured, and not opening the door to an interviewer may be an implicit refusal.

From page no: 121

To page no: 147

Anthology: Handbook of Survey Methodology for the Social Sciences