Lessons Learned from Implementing Response Propensity Models in the 2013 Census Test

• Activity Number: 158
• Type: Topic Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Gina Walejko*+
• Companies: U.S. Census Bureau
• Keywords: adaptive design ; response propensity models ; census

Abstract

The 2013 Census Test modified existing Census Bureau systems to study an adaptive approach to case assignment for a sample of housing units in Philadelphia between October and December of 2013. A test goal was to uncover operational "lessons learned" related to modeling daily response propensities for sample cases and automating case assignments. We used frame information and contact history paradata to model daily response propensities for sample cases. Case management systems prioritized seven cases with the highest response propensities per CAPI interviewer, and interviewers were trained to work these "high priority" cases each day. This paper discusses lessons learned. Topics include: systems testing (e.g. how do we construct systems tests that test response propensity models and automated case assignment?); systems in operation (e.g. what interviewer behavior and systems anomalies need to be anticipated?); case dispositions (e.g. how should we treat problematic cases such as demolished units?); and case reassignments (e.g. how should incomplete reassigned cases be handled?). The paper concludes with a brief discussion of propensity model revision for a 2014 Census Test.