Controlling for Seam Problems in Duration Model Estimates: With Application to the Current Population Survey and the Computer Aided Telephone Interview/Computer Aided Personal Interview Overlap Survey
Inconsistencies in duration data at the "seams" between surveys make the problem of determining where one spell ends and the next begins impossible to resolve with certainty. In this paper I develop a statistical approach to partitioning data records with inconsistent duration data. The basis of this paper is a set of definitions as to what establishes consistent reporting at each seam. These definitions are derived from the questionnaire, and they enable me to quantify duration data inconsistencies as errors. The errors are then employed to generate probabilities over data record partitions. A duration model that uses these probabilities to control for seam problems is estimated for both the Current Population Survey and the Computer Aided Telephone Interview/Computer Aided Personal Interview Overlap Survey.