Modeling the Likelihood of Interviews and Refusals: Using Call History Data to Improve Efficiency of Effort in a National RDD Survey

• Activity Number: 41
• Type: Topic Contributed
• Date/Time: Sunday, August 8, 2004 : 4:00 PM to 5:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Roberta L. Sangster*+ and Brian J. Meekins
• Companies: Bureau of Labor Statistics and Bureau of Labor Statistics
• Keywords: nonresponse ; call history ; RDD survey ; call attempts ; calling rules

Abstract

Proportional hazard models for competing risks are used to develop improved calling strategies. The premise is that completed interview and refusals share similar risks for noncontact. However, the underlying structure may be different. For example, a sampling unit with five answering machine outcomes early in the survey period may be more likely to refuse to participate, while a sampling unit with no answering machine outcomes may be more likely to participate. The number of times an outcome occurs or the proximity between the occurrences may affect the likelihood of certain final outcomes. By modeling the likelihood for refusals and interviews, we hope to find ways to develop calling rules that improves efficiency and optimizes the chance of completion. This study uses call history records for two years of a large national RDD survey (2001-2002). The call records include the outcome of every call attempted for 125,932 cases (671,482 call attempts). Each survey was conducted over an eight-week period with a maximum number of 30 attempts for each implementation. Ultimately, the survey researchers plan on testing the calling strategies developed in this study.