Synthetic Cohorts from Panel Surveys: An Approach to Studying Rare Events

(1985) - Richard Campbell, Cathie Mayes Hudson

Journal: Research on Aging
Link::
DOI:: 10.1177/0164027585007001004
Links::
Tags:: #paper #Methods #SyntheticCohorts
Cite Key:: [@campbellSyntheticCohortsPanel1985]

Abstract

Notes

one possible means of avoiding at least some of the biases is to form sythnetic cohorts of indiviudals who have experianced events by pooling observations from panel surveys

students of the life course in general, study events and transitions that in one sense are universal or nearly so, yet in another sense are rare

whereas it is easy to get a sample of persons who have reited, been widowed, or whatever, it is much more difficult to design studies that obtain pre-transition data

yet it is clear that a more or less standard panel design is quite effective at capturing pre- and postdata on individuals going through an important life course transition

our approach is to pool cases across the various waves of a panel design.

we cosider several alterantives to the synthetic cohort approach, arguing that they will not suffice for most purposes

most important, the timing of the event is not controlled. Some subjects will have experianced the transition recently, others long in the past. Thus the amount and quality of retrospective pretransition data will vary as well.

finally, we canc onsdier what might be called the ‘‘nonspecfic panel design’’ in which a group of subejcts represtnetative of a range of birth cohorts are studied over time

nonspecific panel designs have the major advantage of providing preevent information on a broad population. However, the events themselves occur with relatively low freuqencey. In any given year, a small fraction of the sample will experiance the particular event or transition of interest.

at the same time, we will have to deal with the fact that for some cases we will have a short period of preevent data, whereas for others the string of information will be quite long. Conversely, for some cases the amount of postevent data will be small and for others large. Finally, the longer hte panel runs in time, the more variation in the historical and enivromenental circumstances surroundign the event.

Sociologists will be inclinded to focus on structural equation models and the application of LISREL to the problem