Testing for Phase Capacity in Surveys with Multiple Waves of Nonrespondent Follow-Up

• Activity Number: 598
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Social Statistics Section
• Author(s): Taylor Lewis*+
• Companies: University of Maryland
• Keywords: responsive survey design ; multiple imputation ; weighting ; nonresponse

Abstract

To combat nonresponse, many surveys repeatedly follow up with nonrespondents, often targeting a response rate or a predetermined number of completes. Under a fixed data collection protocol, however, returns tend to diminish with each subsequent wave of data collected, and (nonresponse-adjusted) estimates eventually stabilize. This is the concept of phase capacity, suggesting some sort of design change is warranted (e.g., switch modes, increase the incentive, or discontinue follow-up altogether). The only known test for phase capacity appearing in the literature is one employing imputation models to adjust for nonresponse. This paper describes a test similar in spirit but applicable to surveys that conduct weighting adjustments to compensate for nonresponse. The two methods are compared via an application using data from a Web-based satisfaction survey of United States government employees. The weighting version of the phase capacity test proves more conservative in the sense that it tends to conclude more follow-up attempts are warranted.