Nonparticipation Issues Related to Passive Data Collection

Abstract

New passive data collection techniques on smartphones allow for the direct observation of a participant’s behavior and environment in place of self-reported information. However, such studies are not appealing to all people, especially those with higher security concerns. The current work explores the mechanisms that impact a sample member’s decision to participate in a passive data collection using three different online panels. The first study explores nonparticipation bias in a financial tracking study and finds evidence of bias in the self-reported measures of financial behaviors, and that prior experience with the research organization positively impacts a sample member’s decision to participate. Studies two and three employ deception studies (i.e., the passive data collections were presented as real rather than hypothetical, but no data was passively collected) in which respondents received experimentally varied invitations to participate in a smartphone-based passive data collection. The second study varies the type of data requested and the study topic to understand better how these study components interact. The findings suggest that the type of data requested impacts participation while the study topic does not. The second study utilized video messages presented to all sample members who chose not to participate. These videos asked the sample member to reconsider, varying whether or not they reiterated the data and security measures of the study from the initial invitation. The results suggest that offering a follow-up video increased participation. Finally, the third study experimentally varied the level of control the sample member would have over what data is shared with researchers during a passive data collection. The findings suggest that an offer of control may not increase participation in app-based passive data collection. The three studies suggest that sample members are more likely to participate in a survey when they have prior experience with such a request and may be converted to participate with a video message, but that the type of data requested greatly impacts the decision to participate. Future work should include replicating these studies with different requested data types and shifting to samples not drawn from online panels.