Conferences

Measures of educational attainment in cross-national surveys tend to be post-harmonized from indigenous educational classifications, using a common denominator approach such as the International Standard Classification of Education [ISCED]. Another common approach has been to use duration as a common standard. The European Social Survey (ESS) documents how countries have recoded their detailed national education classification into ISCED-97. Besides these two measures, ESS data contain also another indicator, the duration of the entire educational career. We exploit this presence of multiple education measures to assess the relative quality of three measures: (A) the indigenous classification, optimally scored, (B) ISCED and© duration. We develop an optimal scale for indigenous classification using a MIMIC approach, that estimates how the respective national education levels are scaled relative to each other with regard to a number of inputs (in particular parental characteristics) and a number of outputs (in particular the acquisition of occupation and partner). We investigate to what extent the increase in detail achieved by optimal scaling has yielded a noticeably better measurement of educational attainment relative to the cruder ISCED representation. We also test a widely held assumption that duration measures are a particularly bad representation of level of education, in particular in countries with tracked system. We conclude that improvement of cross-national measurement of education can best be obtained by using at least two independent indicators, and not by asking more detailed or better questions.