Association for Public Policy Analysis & Management

In longitudinal surveys, the term “seam bias” is used to describe a phenomenon in which transitions for key variables, such as when someone starts a job or loses health insurance, are more likely to be reported as having occurred between two successive interviews.  This bias has been present in the Survey of Income and Program (SIPP) since its inception.  After the 2008 Panel, the SIPP underwent a major redesign when interviews became annual rather than occurring every four months.   To mitigate any detrimental effects on data quality, the event history calendar was introduced to help interviewers collect retrospective data more accurately.  In this paper, I investigate how the nature and magnitude of seam bias changed in the SIPP after this revision.   Using the 2008 SIPP Panel and the 2011 SIPP Test Panel, my preliminary results show mixed effects, with seam bias increasing for employment and some social assistance variables but decreasing for school enrollment and health insurance variables.