Measuring the extent to which assortative matching differs between two economies is challenging when the marginal distributions of the characteristic along which sorting takes place (e.g. education) also changes for either or both sexes. Drawing from the statistics literature we define simple conditions that any index has to satisfy to provide a measure of change in sorting that is not distorted by changes in the marginal distributions of the characteristic. While our characterisation of indices of assortativeness is not complete, and hence cannot exclude the possibility of multiple indices providing contradictory results, in an empirical application to US data we find that all indices satisfying our conditions indicate that homogamy by education has increased over time.