Bibliography

Title Questionnaire length and breakoffs in web surveys: a meta study
Workshop Workshop 2014
Year 2014
Introduction Purpose of the study: We first separately define introduction breakoff and questionnaire breakoff. This is then different from the prevailing reporting of all breakoffs together. Next, we present a case study of 5,752 surveys and analyze questionnaire length and its impact on both types of breakoffs. Design/methodology/approach: In an aggregated file of 5,752 surveys with around 1,250,000 questionnaire responses we recorded the following paradata for each survey: share of respondent statuses (i.e. complete, partial, breakoff, empty, etc.), number of pages, number of variables and questions, and eventual usage of email invitations. We then explored breakoff rates (introduction and questionnaire breakoffs at survey level) with linear regression, where predictor variables were above-mentioned paradata. Findings: The mean total breakoff rate is 40%, where the introduction breakoff highly dominates among all breakoffs (75%). The analysis shows significant correlation between questionnaire breakoffs and questionnaire length, while there is almost no correlation between questionnaire length and introduction breakoffs. Linear regression shows that on average we can expect questionnaire breakoff rate to increase by 0.06 percentage points for each additional variable included in the questionnaire and by 0.14 percentage points for each additional page. With our model we explain 10% of the variability. Originality/value: Studies usually confirm that lengthy questionnaires contribute to lower data quality. However, the relation between questionnaire breakoff rate and survey length often remains unclear, which is exactly the added value of our research. By separating introduction and questionnaire breakoffs, we then in the empirical study demonstrated that questionnaire length only impacts questionnaire breakoffs. Research limitations/implications: Limitation: 5,752 surveys were from mostly related to DIY research in Slovenia. Implication: we further extend AAPOR final disposition code 2.12 “Break-off or partial with insufficient information”. Practical implications: Measures for introduction and questionnaire breakoffs need to be defined separately, as well as their prevention. Reporting only total breakoff rate has little practical value. 
DomovBibliography
Home2014
Print