IMPROVING THE RELIABILITY OF RETROSPECTIVE SURVEY MEASURES Results of a

Data from a panel study of the impact of a low-income cooperative on its members are employed to compare the relative advantages of three different methods of measuring change. Although analysis confirmed the often reported unreliability of retrospective compared to true panel measurements, it also showed that "pseudo panel " retrospective approximations of panel measurements are greatly superior to change assessments made by respondents themselves. Finally, the three different types of change measures are compared in actual estimations of the impact of project participation on members. Results of this comparison confirm that, although true longitudinal designs arepreferable, pseudo panel measures of change may sometimes be acceptable and should definitely be preferred to the third, more subjective assessments of change. he superiority of longitudinal or panel research designs hasL long been recognized by social scientists (Campbell and Katona, 1953; Glock, 1955; Hyman, 1955; Campbell and Stanley, 1963; Cannell and Kahn, 1968). A large part of the advantage of panel over cross-sectional designs concerns the superior capacity of data gathered over time to establish causal priorities among variables, although a rich and growing literature attests to the difficulties of doing so. ’ In addition, panel designs are preferable because they make possibl