Key messages on survey mode
From a data quality perspective, face-to-face interviewing appears to have many advantages. Interviewer-led techniques do, however, appear to be at slightly greater risk of prompting respondents to make more positive self-reports, relative to self-completion methods - and the finding that this tendency may be exacerbated in telephone surveys indicates that the potential privacy benefits of the telephone method could be outweighed by the additional rapport that interviews can establish in face-to-face conditions (Holbrook, Green and Krosnick, 2003). The presence of partners at the time of interview may also influence responses to some sensitive questions. Much of the evidence with regard to social desirability is ambiguous, however, and the significance of findings varies from study to study. There is relatively little evidence to suggest that subjective well-being measures are uniquely susceptible to mode effects, and in some cases, other social survey questions appear to be more sensitive. The evidence reviewed above suggests that in some cases the magnitude of mode effects on subjective well-being can be quite small, but where they do exist they may affect subsequent analyses of the data. The mixed findings in relation to the survey mode are likely to reflect both differences between studies in terms of the specific questions being asked, and also within-mode differences (e.g. with or without the use of show cards, with or without computer-assisted designs, etc.). It is also possible that the variation in results has arisen because some cultures may show stronger social desirability and audience effects than other cultures (see Section 5). As mixed-mode methods are increasingly being used for household surveys, our knowledge about the impact of the survey mode across a variety of cultures will continue to develop.
Given the impact that different question wording and response formats can have (discussed in Sections 1 and 2 of the current chapter), a critical issue for comparability where mixed-mode methods are used will be selecting questions that do not require extensive modification for presentation in different survey formats. Graphical scales (i.e. those that rely on visual presentation or a series of images, such as those with a series of faces from “smiley” to “sad”), very long question wording, and response formats that contain more than around five verbally-labelled response categories should in particular be avoided, because these will not translate well between different modes.
Where mixed-mode methods are employed, it will be essential to record details of the survey mode for each respondent and subsequently test for and report the presence or absence of survey mode effects. This will provide a larger pool of evidence that will enable more systematic examination of the role of the survey mode in subjective well-being data - including whether or not it may be possible to “correct” data for mode effects in the future.
The evidence described here focuses almost exclusively on evaluative subjective well-being. Much less is known about how mode effects could influence affective and eudaimonic measures. There is perhaps reason to expect that measures of recent affective experience would show a positive bias as a result of interviewer presence - it may be less easy to admit to a stranger (rather than record anonymously) that you’ve been feeling miserable lately. This is supported by the ONS (2011b) finding that self-completion respondents reported higher anxiety than those participating in face-to-face interviews. The pleasant mood induced by the positive social interaction of an interview experience may also influence retrospective recall for recent affect. These issues require empirical examination.