Evaluating one's life: A judgment model of subjective well-being.

Research on rhe use o f ‘no opinion'-filters suggests that respondents are the less likely to offer a substantive response the more strongly the filter question is worded. A series of experiments is reported that demonstrates that filter questions influence respondents’ perception of their task: the more strongly the filter question is worded, the more re­ spondents assume that they will have to answer difficult questions, and that they may not ha\e the required knowledge. Accordingly, filter questions discourage respondents from offering global opinions that they may hold. In line with this assumption, all respondents who reported not having an opinion in response to a filter question, subsequently pro­ vided substantive responses on a global opinion question— presumably because the global question asked was less demanding than expected on the basis of the filter. Analyses of these substantive responses indicated that respondents who initially reported not having an opinion differed from respondents who reported having one. M ethodo­ logical implications of these findings for the use of filter questions and for research on the nature o f ‘floating’ are discussed. C O N S E Q U E N C E S O F ‘ N O O P I N I O N ’ F I L T E R S The major goal of public opinion research is the description o f opinions held by a population. Accordingly, public opinion researchers frequently attempt to screen out respondents who do not hold an opinion on the issue under study because they assume that these respondents may provide meaningless responses. T o accomplish this screening task, they developed a variety o f filter questions that allow the identification of respondents who do not hold an opinion. This methodological research resulted in some of the most reliable findings in the area of question wording (cf. Schuman and Presser, 1981; Sudman and Brad burn, 1974 for reviews). In general, respondents are more likely to report not having an opinion on an issue when this alternative is explicitly offered as * A previous version o f (his paper w u presented al the m ating o f (he A m m an Association for Public Opinion Research, Toronto, Canada, May 1988. The reported research was supported by ZU M A's program on Cognition and Survey Methodology and by a Feodor-Lynen Fellowship from the Alexander von Humboldt Sri/rung 10 Norbert Schnarz (C. World Allocution fo r Public Opinion Reuorch ig8g part of a ‘filtered’ question, than when it has to be volunteered in response ro the ‘standard form’ of the question that does not explicitly offer a ‘don’t know’ re­ sponse alternative. I f a ‘don’t know’ option is offered, the increase in 'no opinion’ responses depends on the specific form of the filter used. Generally, the use of a so called ‘quasi-filter’ results in smaller increases than the use of a ‘full-filter’ . In the former case, a ‘no opinion’ option is offered as part of a precoded set of re­ sponse alternatives, whereas in the latter ease respondents are explicitly asked whether they have an opinion on the issue before the interviewer proceeds to ask the question proper. Finally, the increase in ‘don’t know’ responses to a full filter depends on the strength of the wording of the filter question, with stronger wordings resulting in higher rates of no opinion responses. For example, more respondents provide a substantive answer when the filter question is worded, 'D o you have an opinion on this?’, than when it is worded, ‘Have you thought enough about this to have an opinion?’. Several processes are likely to contribute