Predictive validity of ethnic identification measures: An illustration of the English/French classification dilemma in Canada

To establish the predictive validity of ethnic identification measures, respondents in the Montreal Urban Community were surveyed and measured on a variety of ethnic dimensions. The data was then fed to a clustering algorithm. Two-group clustering assignments were then compared to traditional taxonomic measures. It is shown that some measures do an excellent job in classifying respondents, while others, which are commonly used in practice, do not fare as well.