Conferences

Paradata—the data captured by computerized systems during the process of data collection—are increasingly being used in Web surveys. Web paradata includes characteristics of a respondent’s browser captured from the server logs (server-side paradata) and respondents’ behaviour on each Web page as they answer the survey (client-side paradata). The latter are typically collected by embedding JavaScript code in survey instruments, and include such measures as response latencies (e.g., time to first click), order of responding, changing of answers, keystrokes, mouse movements, and the like. Initially researchers collected paradata to improve the design or performance of the instruments; increasingly they are being used to understand or even control respondent behaviour. This raises questions about whether and how survey respondents should be informed about the collection of paradata.

We conducted a vignette-based experiment in the probability-based online LISS panel developed by the University of Tilburg. Respondents were exposed to a description of a hypothetical survey in which we varied 1) whether we mentioned the collection of paradata and how much information was provided, 2) the topic of the survey (sensitive or not) and the sponsor of the survey (government versus commercial). A total of 5,198 responded to the vignette and answered a set of background measures in a prior wave.

Respondents were significantly more willing to participate in the hypothetical survey in the condition with no mention of paradata than in any of the three conditions in which paradata was mentioned. Willingness to participate (WTP) also varied by survey topic and sponsor. Among those who expressed WTP in the survey, there was also variation in willingness to permit capture of the paradata. In this paper we examine the sociodemographic, behavioural, and attitudinal correlates of WTP under different paradata disclosure scenarios, and discuss the implications for informing respondents about the capture and use of paradata in Web surveys.