PUB - Publikationen an der Universität Bielefeld

Over the last decade, there has been a marked increase in the number of studies on attitude and decision research which use the factorial survey (FS) design. The FS integrates experimental set-ups into a survey: respondents react to hypothetical descriptions (vignettes) while the values of each attribute (dimension) of these descriptions systematically vary in order to estimate their impact on respondent judgments. As the vignettes are based on a number of dimensions and as respondents evaluate several vignettes, FSs are demanding in terms of individual cognitive and information-processing abilities. So far, there is little empirical knowledge of whether and to what extent this complexity is feasible in general population samples with heterogeneous respondents. Using data from a study on the fairness of earnings (with a mixed mode sample consisting a computer assisted personal interview [CAPI], computer assisted self interview [CASI], and paper and pencil [PAPI] mode), the complexity of FSs is analyzed in terms of: 1) design dimensions, such as the number of vignette dimensions (five, eight, or 12) and the number of vignettes for single respondents (10, 20, or 30), which were varied in a 3x3 experimental design; and 2) respondent characteristics that are associated with cognitive abilities (age and education). Two different indicators for cognitive load as well as learning and fatigue effects are analyzed: 1) latency time and 2) response consistency. The results show that raw reaction times but not latency times are longer for older respondents, suggesting that the cognitive effort needed for the evaluation of vignettes is not particularly high. Consistency measures reveal that respondents with a lower educational level show greater inconsistency in their evaluations when the number of vignettes is high. The number of dimensions has an effect on consistency only when respondents have to rate a large number of vignettes. In short, the results demonstrate that FSs are applicable in general population samples but should be used with a limited number of vignettes and dimensions per respondent.