All time references are in CEST
Why are respondents less likely to consent to data linkage in web than in-person interviews, and what can we do to increase informed consent in web? |
|
| Session Organisers | Mr Jim Vine (ISER, University of Essex) Dr Sandra Walzenbach (University of Konstanz) Dr Jonathan Burton (ISER, University of Essex) Professor Annette Jäckle (ISER, University of Essex) |
| Time | Tuesday 18 July, 09:00 - 10:30 |
| Room | |
Respondents’ propensity to consent to data linkage is generally much lower in self-completed modes than in interviewer-administered surveys. Previous studies have found differences of up to 40 percentage points in consent rates for the same survey conducted in different modes. This appears to be a causal effect of the mode on willingness to consent, not due to selection of less willing respondents into web. As more surveys transition to the web, and data linkage increases in importance, there is more to be done to understand why respondents are less willing to consent in web than interviewer administered surveys, and what can be done to close this mode gap.
This session will focus on the processes by which respondents make the decision whether to consent to data linkage, how these processes differ between survey modes and why, and what can be done in terms of questionnaire design and survey protocols to increase informed consent amongst web respondents.
We welcome papers that focus on:
• understanding the differences between modes in consent decision processes,
• how consent question format, presentation, and wording influence the decision, and
• other features of questionnaire and survey design that can increase consent in web surveys.
Keywords: consent to data linkage, decision processes, questionnaire design, survey design
Professor Lisa Calderwood (Centre for Longitudinal Studies, University College London) - Presenting Author
Professor Alissa Goodman (Centre for Longitudinal Studies, University College London)
Professor Pasco Fearon (University College London and University of Cambridge)
Dr Erica Wong (Centre for Longitudinal Studies, University College London)
Dr Alyce Raybould (Centre for Longitudinal Studies, University College London)
Ms Karen Dennison (Centre for Longitudinal Studies, University College London)
Many studies ask participants for consent to link data from administrative records to survey data, and data linkage is becoming increasingly important as a way to supplement survey response and to mitigate the effects of attrition in longitudinal studies. Obtaining high consent rates can be challenging, however. In this paper we present findings from public engagement work conducted prior to fieldwork around the acceptability and implementation of these approaches, and findings from the experiment which trialled two approaches to obtaining consent – Opt in vs Opt Out. In the Opt-in approach participants are asked for permission to link each type of record; in the Opt-out approach participants are informed of the study’s intention to link their survey data to administrative records and provided an avenue to select records to link or refuse altogether. Information provided to respondents about data linkage was similar across the two approaches.
The experiment was conducted in the Early Life Cohort Feasibility Study (ELC-FS) which collected information about several thousand babies aged between 8-12 months old, and their families in 2023-2024, in order to test the feasibility of conducting a new UK-wide birth cohort study. It is funded by Economic and Social Research Council and led by the Centre for Longitudinal Studies at University College London with interviewing carried out by Ipsos. Interviews were carried out using web, phone and video-interviewing as well as face-to-face.
The experiment found different rates of consent between the two approaches: opt-out consent rates were similar across modes, ethnic groups and socioeconomic status, whereas the opt-in consent rates varied by mode and socio-demographic groups. We discuss the implications of these differences, including participant feedback as well as ethical considerations in choosing an approach for a possible future study.
Mr Sebastian Hülle (Institute for Employment Research (IAB)) - Presenting Author
Mrs Marieke Volkert (Institute for Employment Research (IAB))
Mr Benjamin Baisch (Institute for Employment Research (IAB))
Professor Joseph Sakshaug (Institute for Employment Research (IAB))
Linkage consent rates can be increased by re-asking linkage non-consenters for their consent. The typical approach to optimizing this cumulative consent is to ask once in each panel wave, i.e. by conducting repeated requests between waves. However, repeated requests can be also be conducted within a wave to improve linkage consent rates. Recent studies provide evidence that substantial proportions of linkage non-consenters change their minds when confronted with a second linkage request at the end of the questionnaire, ranging from 21% (web) to 44% (CATI).
Building on this previous research, this paper covers two experiments to optimize linkage consent in a recruitment sample of the panel study "Online Panel for Labour Market Research" (OPAL) hosted by the Institute for Employment Research (IAB). Experiment 1 randomly tests how one (placed at beginning) vs. two (placed at beginning and end) requests for linkage consent affect cumulative consent. However, it is unknown whether repeated requests within a panel wave harm panel participation in subsequent waves. This potential trade-off could occur if some respondents become upset or unwilling to continue participating in the panel study because of repeated requests.
A completely informed linkage consent requires not only that a respondent is informed about the linkage of the current wave with auxiliary data (small request) but also about the linkage with upcoming survey data due to future panel participation (full request). Therefore, a linkage request within a panel is only fully informed if the prospective nature of linkage is explicated to the respondent. Accordingly, experiment 2 randomly tests how linkage consent and next waves` response rates are affected by asking for linkage regarding the current wave (small request) in comparison to additionally requesting linkage consent for upcoming waves (full request).
Professor Thomas Crossley (University of Michigan) - Presenting Author
Dr Jonathan Burton (University of Essex)
Professor Mick Couper (University of Michigan)
Professor Annette Jäckle (University of Essex)
Dr Lewis Mitchell (University of Essex)
Dr Sandra Walzenbach (University of Konstanz)
Past research on how respondents decide whether to consent to data linkage has shown that some respondents use a reflective decision process, considering the consequences of consenting. A majority of respondents, however, make decisions based on gut feeling or habit. Those making more reflective decisions are more likely to read additional information, have a greater understanding of the request, and are also more likely to consent. Which decision process respondents use is influenced by the survey mode: web respondents are less likely to make reflective decisions, and less likely to consent, than respondents interviewed in-person.
We report on an experiment aimed at increasing informed consent in web surveys, implemented in the 2024 wave of the Understanding Society Innovation Panel. We test two treatments designed to induce more reflective processing. In both treatments, respondents are told that we would like to link data about social security benefits and State pensions to their survey data. In treatment 1 we then ask them to list different reasons why they might want to consent or not consent. In treatment 2 we ask them a series of true/false questions about the data linkage, to test their understanding of the request. Both groups are only asked for their consent decision once they have completed these tasks. The control group is asked a usual consent question. We will test the effects of these treatments on self-reported consent decision processes, indicators of cognitive effort, understanding of the consent request, and consent rates.
At the time of writing, fieldwork is nearing completion, with currently 2,110 respondents in households across Great Britain.
Dr Sandra Walzenbach (University of Konstanz ) - Presenting Author
Dr Jonathan Burton (ISER, University of Essex)
Professor Thomas F. Crossley (ISR, University of Michigan)
Professor Mick P. Couper (ISR, University of Michigan)
Professor Annette Jäckle (ISER, University of Essex)
Mr Lewis Mitchell (ISER, University of Essex)
Aiming at increasing informed consent when asking for the permission to link survey data to other sources, researchers are trying to find the most effective strategies to design consent requests. For this purpose, we have combined web survey experiments on consent wording and layout with the collection of eye-tracking data. We recorded respondents’ pupil movements and dwell times while they were asked for permission to link their survey data to administrative records. These data allow us to evaluate to what extent different wording and layout conditions affect how participants process consent requests, that is, where they look and for how long they do so.
We asked participants for permission to link their data to education records in a lab study conducted in May and June 2024 with Essex Lab at the Centre for Behavioural Science (Essex University). The 298 participants were randomly assigned to one of six different versions of the consent request that differed in how the introductory information on the linkage process was presented. While keeping content and text length constant, it varied in its layout and complexity of wording. That is, we crossed two different wording conditions (easy / difficult) with three layout conditions (block of text / breaking up the text into bullet points / splitting the text over several pages).
The presentation provides practical insights into the effect of consent design on (1) depth of processing (measured by eye-tracked dwell times and areas of interest looked at) and (2) several key consent outcomes: willingness to consent, comprehension of the request and confidence in the decision made. Initial analyses suggest that respondents found the easy text in bullet point format easiest to process: dwell times were shortest, understanding of the request was similar as in other treatments, and consent rates were highest compared to other formats tested.
Mr Jim Vine (ISER, University of Essex) - Presenting Author
Professor Annette Jäckle (ISER, University of Essex)
Dr Jonathan Burton (ISER, University of Essex)
Professor Mick P. Couper (University of Michigan)
Professor Thomas F. Crossley (University of Michigan)
Many surveys link to administrative records, with respondent consent, and many surveys now use web as a primary mode of data collection. However, respondents are substantially less likely to consent when asked online than in person. In panel surveys, respondents who decline a request for data linkage once often do consent when asked again in a subsequent wave. Many longitudinal surveys therefore routinely re-ask consent questions of those who do not consent initially.
In this paper we examine the following research questions: Is this re-asking of non-consenters as effective in the web mode as it is in face-to-face interviews? Are web consenters just a smaller set, or are they also a different set of respondents than face-to-face consenters?
We analyse data from repeated consent requests that made in different modes in a major UK household panel survey (Understanding Society: The UK Household Longitudinal Study). As well as raw differences by mode of completion — the ‘as-treated’ analysis — we can use experimental allocation to web-first or face-to-face-first in the Understanding Society Innovation Panel to calculate intention to treat (ITT) differences based on mode of allocation, and to estimate causal effects of mode. As consent to link to a variety of domains of administrative data have been (re)asked in Understanding Society, we replicate aspects of our analysis with consents relating to different domains.
Our initial findings suggest that while some non-consenters do provide consent when asked a second time in the web mode, the proportion who do so is lower than the proportion of non-consenters who consent when re-asked face-to-face. These findings mirror typical mode differences in initial data linkage consent requests. We will present further results related to our research questions at the conference.