Impact of mixed modes on measurement errors and estimates of change in panel data

Abstract

Mixed modes (MM) are receiving increased interest as a possible solution for saving costs in panel surveys, although the lasting effects on data quality are unknown. To better understand the effects of MM on panel data I will examine its impact on random and systematic error and on estimates of change. The SF12, a health scale, in the Understanding Society Innovation Panel is used for the analysis. Results indicate that only one variable out of 12 has systematic differences due to MM. Also, three of the SF12 items overestimate variance of change in time in the MM design. I conclude that using a MM design leads to minor measurement differences but it can result in the overestimation of individual change compared to a single mode approach.