Credential and Class Distributions
Another very important remaining question can only be touched on here. This is the problem of defining the appropriate measures of educational attainment and occupational level of the first job. Some studies (e.g., Allmendinger, 1989a) have used ordinal scales for both: years of schooling and occupational prestige. These are not the most usual measures, however. The studies reported in the Shavit and Miiller (1998) volume all conducted analyses using catego-ries of both educational attainment and level of the first job. They adopted a revised version of a set of credential categories devised by Miiller, Liittinger, Konig, & Karle (1989), and the occupational class schema used in that volume and in many other comparative studies is one proposed by Erikson, Goldthorpe, & Portocarero (1979).
There are major societal differences in the distributions of both of those measures. For instance, a very large proportion of young French men (41.6%) have only the lowest level of educational credentials, whereas only a very small proportion (6.0%) of young German men have them. The societies also vary greatly in their distributions of occupational classes. In the
United States, 40.7% of the young men are in the lowest class, and that is four times as many
as in Germany.
These differences raise doubts about whether the credential and class categories are equally meaningful in all of these societies. They also raise questions about how well the categories lend themselves to intersocietal comparisons because of the effects the different distributions
have on measures of association. As Miiller and associates (1989) pointed out regarding the distributions of credentials, "The less concentrated the population is on the different certificates available in the educational system, the more easily can certificates be used for [job] selection purposes" (p. 9). Of course, the same can be said for the distributions of occupational classes. So, when we start comparing societies in terms of the association between those two highly varied distributions, important methodological questions need to be raised. The
comparative analyses conducted by Miiller and associates have shed some light on these questions, but there is much more to be done.