Home Computer Science OECD guidelines on measuring subjective well-being.
Key messages on scale labelling
A number of complex trade-offs need to be managed when making decisions about how to label response options - including interactions with the overall number of response options, the survey mode and the translatability of items - as well as, of course, the underlying construct of interest and the manner in which the question is phrased.
Scale anchors matter because they set the response frame. In any subjective measure, it remains a challenge to ensure that all respondents understand scale anchors in the same way. It appears to be advisable to adopt absolute scale anchors that encompass the full spectrum of possible experiences (i.e. “never/always/completely” rather than “very”, for example) so that there is at least conceptual clarity about where the scales end, even if respondents still define these end states subjectively. Although the difference that this approach is likely to make has not been quantified, it is less ambiguous than the alternatives, and therefore preferable. It may also be advisable to avoid agree/disagree, true/false and yes/no response formats in the measurement of subjective well-being due to the heightened risk of acquiescence and socially desirable responding - although more concrete evidence on the difference this makes would be welcome.
In terms of labelling the various points along a scale, consistency of approach is essential, and there are evidently benefits and drawbacks to both numerical and verbal labels. Numerical labelling enables a greater number of response options to be used, which may be particularly advantageous when single-item questions are used to assess complex constructs such as life evaluations, without over-burdening respondents (especially via telephone interviewing). Numerical scales are also likely to pose fewer translation problems, which is important to international comparability. For these reasons, in the context of short subjective well-being measures for inclusion in national surveys, numerically-labelled scales are likely to be preferable.
|< Prev||CONTENTS||Next >|