303 – Data Collection Strategies
Case Reassignment: When Making Contact Is a Two-Person Job
Rachael Walsh
U.S. Census Bureau
Julia Coombs
U.S. Census Bureau
As data collection agencies balance response rates, cost, and data quality, the use of paradata in the form of contact histories benefit response propensity models and have applications with respect to adaptive design. However, when interviewers record information regarding each contact attempt with a sample unit, these data violate most modeling assumptions because they are clustered and not randomly assigned. Focusing on the role of the interviewer, particularly with respect to case reassignment, we use data from seven demographic surveys collected over an eight-month period to examine the effects of interviewers on the response propensity of both responding and nonresponding sample units. Multilevel modeling compensates for both the nesting of the sample units within interviewers and nonrandom case assignment. This research includes case reassignments from one interviewer to another, a phenomenon largely ignored in previous research, and further explores the utility of a new indicator we developed, the scaled evenness of finding attempts, or SEFA. When modeling case reassignment, the characteristics of the interviewer who finished the case should be used. Whenever possible, difficult to contact cases should be reassigned to interviewers with a higher than average SEFA score and with fewer cases in difficult strata.