Web Survey Bibliography

Title Investigating Respondent Multitasking in Web Surveys with Paradata
Year 2015
Access date 04.07.2015

Multitasking refers to concurrent or sequential combinations of activities. While cognitive sciences have been elaborating on multitasking since the 1930s and media multitasking (multitasking involving media activities) has been extensively researched across different disciplines in recent decades, this behavior is rarely discussed in the context of survey methodology. Respondent multitasking (RM) happens when a respondent does any secondary activity while responding to a questionnaire. RM has been theoretically associated with worse response quality (e.g. Holbrook, Green, & Krosnick, 2003; Lynn & Kaminska, 2013) and scarce empirical evidences from telephone surveys have partly supported this assumption (Kennedy, 2010; Lavrakas et al., 2010). Studies on telephone and web surveys (Zwarun & Hall, 2014) predominantly use self-reports to measure RM. In such studies, reported shares of respondents who had engaged in RM range between 30 (for web surveys) and 50% (for telephone surveys). However, people tend to underestimate the amount of time spent on multitasking (e.g. Iqbal & Horvitz, 2007). While using paradata to measure RM (e.g. Stieger & Reips, 2010) could circumvent this issue, more investigation is needed to refine this approach. We bring new insights to under-researched topic of RM in web surveys and present a new way of measuring it via advanced paradata (Vehovar et al., 2012). We have developed two different sets of RM indicators: (a) focus-out events (indicating whether a respondent has switched to another window or tab); and (b) time-out events (indicating whether a respondent has spent a substantially longer time to complete a questionnaire subpage). Preliminary results from a case study (n=267) show that 56% of respondents had done some form of RM. While focus-out indicators and time-out indicators are significantly correlated, there are also some considerable differences. Further, we used a negative binomial regression and found a relationship between RM, questionnaire complexity, and item nonresponse.

Year of publication2015
Bibliographic typeConferences, workshops, tutorials, presentations

Web survey bibliography (8390)