Conferences

Cross-nationally comparable measurement of education is vital for comparative social stratification and mobility research. A number of international classifications of education have been designed throughout the years. Despite the wide utilisation of measures of education in cross-national comparative research, validation studies of such harmonised measures are scarce (but see e.g. Braun & Müller 1997, Kerckhoff, Ezell & Brown 2002). This paper intends to show the effects of harmonising country-specific measures of educational attainment in to different cross-national measures, using national labour force surveys from Germany, the Netherlands, Estonia, France, Italy, the Czech Republic and the UK.

The proposed paper firstly discusses the challenges of cross-national comparabiltiy and loss of predictive power that results from harmonisation. It then introduces the ‘candidate measures’ of educational attainment. Among those, the complete ISCED 97 (UNESCO Institute for Statistics 2006 [1997]) is very difficult to implement in large-scale cross-national surveys, and too complex to be used in any statistical analyses. Therefore, different simplified versions of the ISCED-97 will be covered by this study, one of which was recently proposed (Schneider 2008). Here, different simplifications of ISCED 97, CASMIN (König, Lüttinger & Müller 1988) and years of education will be compared in terms of their predictive power with respect to core social stratification outcomes.

A previous study by the first author (Schneider 2008) which used cross-national survey data (the European Social Survey) showed that most ways of harmonising education variables basically lead to non-comparability of regression outcomes. This study will use fewer countries but larger national data sets with more detailed country-specific education information. This will allow us to check the previous results for countries that provided only crude education variables in the ESS (like the UK and Italy), perform more detailed checks for all countries, and give an independent test of the alternative simplification of ISCED 97 proposed in that study.

The actual validation uses linear and (multinomial) logistic regression models, predicting social status (measured using ISEI scores), social class attainment (EGP or ESeC) and unemployment (ILO definition) by educational attainment. For the single countries, models including the national measure of educational attainment are compared in terms of model fit/explanatory power with models using the different simplified versions of ISCED 97. Regression coefficients will also be compared. It is expected that the use of the simplified ISCED 97 (EULFS and ESS versions) or years of education strongly attenuates the association between education and labour market outcomes in many countries, making cross-national comparisons highly problematic. It is furthermore expected that CASMIN and the alternative simplification of ISCED 97 performs much better.