Introduction

Many types of survey respondents are difficult to access, to locate, and (in longitudinal research) to stay in contact with throughout the course of a study. These types of respondents fall into two main categories. The first category includes people who are difficult to reach by nature, such as young adults whose lives are in transition, the mentally ill, the homeless, and drug users. These population groups are extremely mobile and, in some cases, less likely to maintain close ties with relatives who might serve as a means of locating or contacting them. The second category includes people who are consciously avoiding being located in an attempt to avoid contact with the justice system, immigration authorities, debt collectors, stalkers, or others. People falling into either of these two categories may lack fixed addresses, or be “cell phone only,” with episodic cell service and numbers that change frequently, or list residences or phones in the name of another person.

The hardest subjects to reach in a target population group might provide fundamentally different responses than members of the group who are relatively easier to find and survey (Groves, Fowler, Couper, Lepkowski, Singer, & Tourangeau, 2004). Not including certain segments of a population leads to nonresponse bias, which threatens the quality of survey statistics and the validity and generalizability of research findings (Cottler, Compton, Ben-Abdallah, Horne, & Claverie, 1996). The goal of maximizing power and minimizing potential nonresponse becomes even more difficult when the study population by definition is hard to reach. Researchers face a trifecta of challenges to data reliability when such studies are longitudinal: maximizing power, minimizing systematic nonresponse, and maintaining the respondent pool over time. Tracking efforts can minimize these threats by maximizing participation among sample members (Brown & Nederend, 1997), minimizing nonresponse among respondents with certain characteristics or reflecting sample subpopulations (Teitler, Reichman, & Sprachman, 2003), and reducing subject attrition in research requiring multiple waves of data collection (Cottler et al., 1996).