Online Program

A Theoretical Framework for Adaptive Collection Designs
*David Haziza, Université de Montréal 

Keywords: Nonresponse; responsive design; nonresponse bias; nonresponse variance; calls prioritization

We present a theoretical framework for adaptive collection designs in the context of computer-assisted telephone interview surveys. By adaptive collection designs, we mean any procedure of calls prioritization and resources allocation that is dynamic as data collection progresses; i.e., the procedure uses paradata or other information to adapt itself to what is observed during data collection. We focus on calls prioritization. The goal of an adaptive collection design is to increase quality for a given cost or alternatively to reduce cost for a given quality. The literature has essentially focused on finding collection designs that lead to a reduction of nonresponse bias of an estimator that is not adjusted for nonresponse. Thus, improvement of quality is associated with nonresponse bias reduction. We argue that it is not the best criterion to use as the bias that can be removed at the data collection stage of a survey through an adaptive collection design can also be removed at the estimation stage through an appropriate nonresponse weight adjustment procedure. Instead, we minimize the nonresponse variance of an estimator that is adjusted for nonresponse. We develop a procedure of calls prioritization that attempts to achieve this goal. The results of a simulation study using data from the Workplace and Employee Survey will be shown.