Studying Social Change ― Methodological and Empirical Analysis of Retrospective and Longitudinal Methods

Abstract

Social change can be studied by cross-sectional, longitudinal, and retrospective general population surveys. Along with important advantages of retrospective surveys which includes the lack of sample attritions, time flexibility, and lower costs, the quality of retrospective data remains questionable due to possible measurement errors. To determine the extent of data quality issues, the author of this article compares data of retrospective questions obtained in RUSSET longitudinal surveys on 5-years interval (in 1998 versus 1993) and on 30-years interval (new round “RUSSET after 30 years” in 2024) with the concurrent data of the same study (collected in 1993). The retrospective questions were asked about two factual indicators (family and employment statuses) and five subjective indicators referring to general satisfaction with life and four specific life domains. At the aggregate level, retrospective methods produce results quite close to the concurrent ones over a five-year interval and less accurate estimates of average values over a 30-year interval. The observed differences indicate an exaggeration of retrospective estimates. The greater the time distance, the greater the overestimates; however, the order of categories or life domains among themselves and the distances between them are reflected accurately. At the individual level, the agreement between concurrent and retrospective factual indicators is 80—90%, while for subjective indicators it is 60% over a 5-year period and 50% over a 30-year period. Among the hypotheses put forward to explain the causes of bias in retrospective data, i.e. idealization of the past and accounting for changes and the influence of social context, the most likely is idealization of the past. No significant differences in the agreement between retrospective and concurrent data were found across gender, age, and educational groups. Measurement error is lower when there were no changes in the situation during the observation period and higher when changes occurred, as well as when the question concerns uncertain or not confirmed officially marital or employment statuses. Taking this into account, the author concludes that retrospective methods can be a good addition to longitudinal studies, used to reconstruct life events, unfold a life calendar, fill in the gaps in information between waves of a longitudinal study, but it cannot fully fulfill the task of assessing social changes.

Acknowledgments. The research was carried out with the support of the Russian Science Foundation, project No. 23-18-00635 “Life course, values, expectations of the generation entering the adult life in 1990x, — longitudinal survey after 30 years”.