Dealing with the Seam Problem for the Survey of Labour and Income Dynamics - ARCHIVED

In recent years a considerable amount of attention has been focused on what is known as the "seam" problem in surveys having a longitudinal design. This refers to the fact that the number of transitions or changes in status observed across the seam when the data for two consecutive reference periods are juxtaposed is considerably larger (so metimes, an order of magnitude larger) than the average number observed in the data reported for a single reference period.

Response errors are the most probable cause of seam biases. For characteristics such as employment status or income recipiency, errors can be due to omissions or to misplacing events in time. However, standard explanations for response errors based on "forgetting theory" are not supported by the data. Results concerning proxy effects are mixed but generally show no clear association.

Dependent interviewing (i.e., feeding back to respondents responses provided on a previous interview) would appear the most appropriate strategy for dealing with seam effects. However, not all feedback techniques will necessarily work. A comparison of two such techniques, one which failed to eliminate seam effects (SIPP) and one which was successful (LMAS) and which has also been successfully tested by SIPP, attempts to identify the key features required.

The paper argues in closing that dependent interviewing should not be viewed as a necessary evil that is required for reducing seam biases but as an integral part of the interview process in longitudinal surveys.