Attrition

Unit nonresponse is a problem for any type of survey; however, unit nonresponse in panel studies can be a more severe problem than in cross-sectional studies. [Page 43]Like cross-sectional studies, panel studies are subject to nonresponse at the initial wave. In addition, attrition—which is unit nonresponse after the initial wave of data collection—can occur at each subsequent wave.

A framework for understanding attrition in panel studies divides the participation process into three conditional steps: (1) location, (2) contact given location, and (3) cooperation given contact; this process cycle is repeated at each wave. Attrition thus occurs because of a failure to relocate or recontact an eligible sample unit after the initial wave of data collection, and because of noncooperation (i.e. a refusal to participate again in the survey) or the inability to participate again.

The accumulation of attrition over several waves can substantially reduce the number of sample units, thereby reducing statistical power for any type of analysis, both cross-sectional and longitudinal. However, attrition may also introduce nonresponse bias in the survey estimates. Differential or selective attrition occurs when the characteristics of the sample units who drop out of the panel because of attrition differ systematically from the characteristics of sample units who are retained in the panel study.

Distinguishing between initial wave nonresponse and attrition is important because the reasons for attrition may be different from the reasons for nonresponse in the initial wave of a panel study or in cross-sectional studies, in general. Contrary to cross-sectional studies where sample units' judgments about participating in the survey are largely made during the brief interactions they have with survey interviewers when the request is formulated, sample units in panel studies with repeated survey requests and contacts in between data collection points have more information about the nature of the request being made and will be influenced by their personal survey experience in the initial wave or other previous waves. In addition, in the case of a panel study, and once the initial wave has been conducted, the interviewers are better informed than in the initial wave to select the best approach to successfully locate, contact, and convince sample units to participate in additional waves of the panel study.

There are two main strategies that survey researchers use to address attrition. The first is to reduce attrition rates by maximizing sample retention; the second is to develop post-survey adjustments to correct for the biasing effects of attrition. These two strategies are not mutually exclusive, and they often are used together.

The main goal of panel management or panel maintenance is to maintain participation of all sample members in the panel study after the initial wave. The specific techniques to reduce attrition in panel studies are focused on locating the sample unit and establishing sufficient rapport with the sample units to secure their continued participation. Panel studies can keep contact with the sample units and keep them interested in participating in the panel study by adopting a good panel maintenance plan and employing techniques of tracking and tracing. Acquiring detailed contact information, the organization of contact efforts, hiring skilled interviewers, and retaining staff over time are important components of a good panel maintenance plan. Tracking procedures aim to maintain contact with sample units in the period between waves in order to update addresses between interviews so that a current or more recent address is obtained for each sample unit prior to conducting the interview. Tracking procedures are adopted in an attempt to find the missing sample units and are used at the point of data collection when the interviewer makes his or her first call, discovers the sample member has moved, and tries to find a new address or telephone number.

...