Purpose of the study: Respondent multitasking refers to any secondary activity performed by respondents between the time they start responding to a survey and the time they finish. We introduce, compare and evaluate different indicators of respondent multitasking in web surveys, based either on self-reports or paradata.
Design/methodology/approach: We use paradata from an assessment survey among incoming (international) and outgoing (Slovenian) students in student exchange programmes to measure respondent multitasking. Web questionnaire also prompted respondents to report if they have engaged in any secondary activities while responding. We use statistical tests to compare these two data sources on respondent multitasking, as well as their relationship with item non-response and straight-lining.
Findings: Paradata indicators show statistically higher frequencies of interruptions than self-report indicators (p < 0.05). Based on paradata indicators, incoming/international students (who responded to a questionnaire in English language) multitasked more than outgoing/Slovenian students (questionnaire in Slovene language), while no such statistically significant differences were found with self-report indicators. Similarly, higher paradata indicators are also positively correlated with item non-response rates, while there are no such significant associations for self-reports. Weak associations are found for straight-lining indicators. Results also show importance of selecting and omitting particular parts of questionnaire (e.g. introduction page) and respondents (e.g. drop-outs) in the analysis.
Originality/value: Studies on telephone surveys suggest that respondent multitasking can be relatively prevalent in surveys and associated with data quality. However, there is little-to-no published research about respondent multitasking in web surveys. Besides this novelty, we also introduce a new way of utilizing paradata. Moreover, contrasting self-reports with paradata is a valuable addition to scarce research that compares and combines self-reports with electronic tracking data in time use and media use studies.
Research limitations/implications: Limitation: survey population is very specific. Implications: in context of web surveys, we confirm findings from research on multitasking in other settings: respondents might under-estimate their multitasking behavior; respondent multitasking is associated with data quality. 
Practical implications: Paradata could potentially be used for development of valid and reliable measures of respondent multitasking. In conclusion we offer several practical advices, considerations and limitations of employing such approach.