Quality & Quantity 33: 261–276, 1999.
© 1999 Kluwer Academic Publishers. Printed in the Netherlands.
Event History Data and Making a History
out of Cross-Sectional Data
How to Answer the Question ‘Why Cohorts Differ?’
NAN DIRK DE GRAAF
Department of Sociology, Nijmegen University, PO Box 9104, 6500 HE Nijmegen, The Netherlands
Abstract. At the turn of 20th century social scientists have built up a large stock of cross-sectional
data-sets to study social change. However, scholars more and more collect event history data contain-
ing the exact timing of events. Comparing the (dis)advantages event history data are to be preferred.
However, for research on value change the event history approach is inapplicable, since it is not
possible to ask the timing of a value change retrospectively. I will illustrate that value change (i.e.
cohort differences) can be studied adequately with cross-sectional data, if information about the
historical context is added. For this purpose I test Inglehart’s value change thesis.
Interestingly, there are also topics in which cross-sectional data-sets are unnecessarily being used.
Using research on secularization as an example, I show that the event-history approach can be used
to answer the question whether the decreasing number of religious people concerns a cohort-effect.
However, whatever data-set is being used, to study cohort differences, one should always give a
theoretical answer to the key-question: what exactly makes cohorts different?
Key words: cohort identiﬁcation, dynamic analysis, postmaterialism, religion.
1. Introduction: Changes Over the Life-Course and Choice of Data
Many questions social scientists raise concern social change and cohort-
replacement issues. Inglehart’s intergenerational value change thesis is one of the
best known examples (1977, 1990). Another example is the question whether the
decline of traditional religion encountered in many Western societies is mainly due
to cohort replacement. What kind of data can we use to tackle such questions? In
order to answer this question we should understand the methodological issue that
there are three types of changes in attitudes or behaviour of cohorts, i.e., categor-
ies of persons usually linked by year of birth (cf. Glenn, 1977): The ﬁrst type of
change – called a cohort or generation effect – is associated with year of birth, and
concerns all events that one generation experienced and other generations did not.
Changes of the second type, called life cycle effects, concern all events associated
with changes in age. Finally, period effects concern those events which affect all
A more extensive version of this paper was presented at the Workshop on Longitudinal Analysis,
University of Padova, 14–16 May 1998. I would like to thank Ariana Need and Elisabetta Ruspini
for helpful suggestions.