Refining the Total Survey Error Perspective

Abstract

Total survey error (TSE) is a very valuable paradigm for describing and improving surveys, but it can be improved. First, either TSE needs to be limited to covering just instances of differences between true and measured values or TSE should be rechristened as total survey measurement variation (TSMV) if other forms of measurement-related variation are to be included. Second, the TSE/TSMV typology needs to be as detailed and comprehensive as possible. Third, TSE needs to be thought of as heavily involving the interaction of error components and the concept of comparison error should be used to extend TSE to cover multiple survey types. Fourth, the minimizing of TSE is an important goal in survey research and the TSE paradigm can be used as both an applied application and a research agenda to achieve that goal. Finally, TSE has both individual and aggregate components and an absolute and situational aspect. The role of each of these needs to be kept in mind.