Tracking Respondents in Longitudinal Surveys

Many of the problems social scientists are called upon to address require information best obtained through longitudinal research. Whether the emphasis of the investigation is practical or theoretical, applied or basic, data on the same people across time are important for understanding the causes or conditions out of which present situations arose and the effects of these situations on future events., However, a relatively small number of reported studies has utilized true time-series information. Instead, investigators usually infer the effects of changes over time from data which are not strictly longitudinal. Several methods are commonly used, including cross-sectional information, aggregate analysis of census cohorts, and respondent-recall data.2 Each of these methods is intrinsically weak. The first two do not follow the same people or else have no way of identifying individual sample members and their specific changes over time. The third, dependent on memory, may be inadvertently or deliberately selective. If these techniques are unsatisfactory, why are true longitudinal studies so rare? Among other reasons, maintaining contact with respondents over a period of years is difficult, costly, and time-consuming. Moreover, almost nothing has been offered in the social science literature concerning the relative effectiveness of different techniques for retrieving respondents.3 The purpose of the study reported here was to assess empirically a variety of tracking procedures and to arrive