Activities per year
Relevance & Research Question: The advantages and disadvantages of web-based surveys have been heavily discussed from their beginning. Often, it is emphasized that web surveys are less time consuming and cheaper to conduct than other data collection modes. In earlier literature on mode effects, however, it was commonly held that this come at the expense of data quality because the web negatively affects respondents’ cognitive effort to answer survey questions carefully. Even though later studies acknowledge that web mode does not per se produce lesser-quality data, some reluctance still exist. The current study addresses the question of whether an actual “web mode effect” exists when self-selection and interviewer effects are eliminated. Methods & Data: The study utilized data from four national survey experiments in Denmark, Norway, Iceland, and Finland (total n=6,219). All experiments were conducted in 2014/2015 as a part of the ISSP 2014 module on “Citizenship”. Respondents were assigned randomly to one of two mode groups: One group answered the survey on web (‘web group’) whereas the other group answered on paper (‘paper group’). The experimental design enabled the elimination of self-selection effects, and by comparing web mode with another self-administered mode, possible interviewer effects were eliminated as well. To test mode main effects on data quality, multiple linear regression analyses were conducted. In these, two measures of satisficing behaviors (low to high) were used to examine differences in data quality. Results: Comparisons of web and paper modes suggested that such a “web mode effect” does not exist, since respondents in the web group were not more prone to satisfice than respondents in the paper group. In other words, no evidence was found that web mode produces lesser-quality data than other self-administered modes of data collection. Added Value: The study contributed to state-of-the-art by providing more solid empirical knowledge of whether an actual “web mode effect” exists when self-selection and interviewer effects are eliminated. Whereas previous studies used non-probability sampling procedures, small-n samples, self-selection of modes, or samples from population subgroups, this study employed an experimental design and utilized data from large-n population samples.
|Publication status||Published - 2020|
|Event||General Online Research 2020 - University of Applied Science, Berlin, Germany|
Duration: 11 Mar 2020 → 13 Mar 2020
|Conference||General Online Research 2020|
|Location||University of Applied Science|
|Period||11/03/2020 → 13/03/2020|