Abstract
This paper is concerned with the question of finding sets of similar micro-level industries that can be grouped to obtain a smaller set of macro-level industries. We argue that factor analysis based on the Leontief inverse is a promising approach when multiple objectives of classification must be taken into account, provided that the cosine, rather than the correlation coefficient, is adopted as the measure of similarity. We apply this approach to the 1967 US input–output table because that table has been used by researchers experimenting with alternative classification methods.