Dependent interviewing and seam effects in work history data

Abstract

Dependent interviewing has been introduced by a number of panel surveys as a means
of reducing measurement error, in particular the typically observed concentration of
transitions at the seam between waves, the ‘seam effect’. Little evidence exists,
however, of the effects on survey estimates. We report on a large scale randomised
experiment comparing dependent interviewing with traditional independent methods.
Proactive dependent interviewing improves the quality of work history data by
reducing seam effects, and reduces bias in estimates of monthly labour force
transitions and spell durations. Proactive interviewing does not have any effect on
measures of cumulative experience and does not appear to lead to under-reporting of
change. Seam transitions in continuous work histories are largely explained by editing
rules used to reconcile reports from repeated panel observations. Proactive methods
reduce seam effects by precluding overlapping non-corresponding reports. The
potential for eliminating seam effects is, however, limited by item non-response to
date questions.
Keywords: Seam Bias, Recall Error, Labour Force Transitions