Invariably, full response is not achieved with a single survey solicitation, and so a sequence of follow-up attempts typically ensues in an effort to mitigate the potentially detrimental effects of nonresponse. Rather than permitting the follow-up campaign to continue indefinitely or until some preset response rate is met, a potentially more efficient alternative is to track a key point estimate in real-time as data is received and alter the survey design phase (i.e., modify the recruitment protocol) once the point estimate stabilizes. The notion of point estimate stability has been referred to as phase capacity in the survey methodology literature, and several methods to detect when it has occurred have been proposed in recent years. Noticeably absent from those works, however, is statistical theory providing insight into how point estimates can change during the course of data collection in the first place. The goal of this paper is to take a first step in developing that theory. To do so, the two established perspectives of survey nonresponse – deterministic and stochastic – are extended to account for the temporal dimension of responses obtained during a survey design phase. An illustration using data from the 2014 Federal Employee Viewpoint Survey is included to provide empirical support for the new theory introduced.
Temporal perspectives of nonresponse during a survey design phase