Key messages on question wording

Using different wording in subjective well-being questions can change the pattern of responses, although finding a difference between response patterns does not in itself indicate which wording should be preferred. The evidence regarding the “best” question wording to use is limited and mixed; indeed, as discussed in the section that follows, there are a wide variety of question and survey context features that can influence responses in a way that cannot be easily disentangled from that of wording effects alone. There are also some grounds for concern about the translation of certain subjective well-being constructs between languages, although the size of this problem and the extent to which it limits cross-national comparability is not yet clear.

One way to reduce the impact of potential variation in how respondents understand questions is to use multiple-item scales, which approach the construct of interest from several different angles in the hope that they ultimately converge. Current measures of affect and eudaimonia typically contain multiple items - which also enables conceptual multi-dimensionality to be examined. In a national survey context, lengthy multiple item scales may not be practical, but current evidence suggests that particular care needs to be taken when developing shorter or single-item affect and eudaimonia measures, as there is strong evidence for multi-dimensionality among these measures. Although multiple-item life evaluation measures show better reliability than their single-item counterparts, there is at present evidence to suggest that single-item measures can be successfully used to capture life evaluations, which are usually assumed to be unidimensional in nature. It would be useful, however, to have further evidence regarding the relative accuracy and validity of single- versus multiple-item evaluative questions to help ensure optimal management of the trade-off between survey length and data quality.

Psychometric analysis, including examination of factor structure and scale reliabilities, can go some way towards identifying questions that seem to function well among a set of scale items, but consideration of the construct of interest, validation studies and the purpose of measurement will also determine which question should be selected. For now, this discussion clearly highlights the need for a standardised approach to question wording, particularly if comparisons over time or between groups are of interest.

< Prev   CONTENTS   Source   Next >