The Mode is the Message: Using Predata as Exclusion Restrictions to Evaluate Survey Design

Changes in survey mode (e.g., online, offline) may influence the values of survey responses, and may be particularly problematic when comparing repeated cross-sectional surveys. This paper identifies mode effects by correcting for both unit non-response and sampling selection using the sample profile data (predata) — the individual’s number of previous survey invitations, the number of completed past surveys, and the reward points balance. The findings show that there is statistically significant evidence of mode effects in recall and subjective questions, but not for factual ones.