Understanding the patterns of mode switching in longitudinal studies

Alexandru Cernat, Joseph Sakshaug

Research output: Contribution to journalArticlepeer-review

20 Downloads (Pure)

Abstract

Using mixed-mode data collection is becoming a mainstream way of conducting longitudinal surveys. However, interviewing the same units in a mixed-mode longitudinal design can lead to respondents switching between modes over time. As a result, mode switching behaviors can be correlated with non-response and potentially influence survey responses and estimates of change in longitudinal analyses. This paper investigates the patterns by which people transition from one mode of interview to another in a nationally-representative, sequential mixed-mode (Web and face-to-face) longitudinal study. Using mixed-mode waves 5-10 of the Understanding Society Innovation Panel, we perform a latent class analysis on respondents and their mode switching behaviors. We identify five distinct classes of respondents: slow switchers, fast switchers, switch and non-response, face-to-face respondents, and Web respondents. Furthermore, we show that these classes differ with respect to respondent characteristics and significantly contribute to the prediction of future wave participation and mode of response, even after controlling for socio-demographic characteristics, interview mode at the previous wave, and previous non-response behavior. Practical implications of these results are discussed and possible strategies to use this information for targeting and correcting for non-response in longitudinal studies are proposed.
Original languageEnglish
Pages (from-to)281-298
Number of pages18
JournalSurvey Research Methods
Volume15
Issue number3
DOIs
Publication statusPublished - 10 Dec 2021

Keywords

  • mixed modes
  • longitudinal survey
  • latent class analysis
  • non-response
  • face-to-face survey
  • web survey

Fingerprint

Dive into the research topics of 'Understanding the patterns of mode switching in longitudinal studies'. Together they form a unique fingerprint.

Cite this