Impact of mode design on measurement errors and estimates of individual change

Publication type

Conference Paper


Alexandru Cernat

Publication date

Series Number



Mixed modes are receiving increased interest from survey methodologists
as a possible solution to saving costs while retaining high quality
data. In recent years this interest has extended also to panel studies
which are looking to save costs by including a cheaper mode for some of
their respondents. The current presentation aims to tackle some of the
issues linked to such a design. First, I analyze if using a mixed mode
design will increase systematic and random error compared to a single
mode CAPI survey by applying equivalence testing in a Confirmatory
Factor Analysis. Secondly, I investigate if estimates of individual
change are influenced by mode design by comparing latent growth models
across the two designs. The first four waves of the Innovation Panel,
part of Understanding Society (UKHLS), are be used for the analysis. The
second wave of the data randomized respondents to either a single-mode
CAPI design or to a CATI-CAPI sequential design. The SF12 health scale
is used to investigate both measurement equivalence and estimates of
individual change.