Cross-Classification Analysis in the Field of Management and Organization: Comments on the DEL-Technique

The DEL-technique, a proportionate reduction in error measure, developed by Hildebrand, Laing and Rosenthal, has been applied and portrayed as a promising prediction analysis technique to evaluate theory on the basis of cross-classification data, though it was controversial at its birth in the early 70s. According to the opponents, Goodman and Kruskal, the interpretation of DEL as a proportionate reduction in error measure of knowing a prediction rule over not knowing the prediction rule, cannot be held, because it is benchmarked against independence instead of ignorance. However, even when neglecting this criticism the DEL-measure can be easily misinterpreted as a measure of acceptance of the specified customized hypothesis as the only and best relationship between two categorical variables, when the context for the interpretation is not carefully stated in terms of the adhered research paradigm: theory-testing versus prediction logic. When taking into account this criticism, the researchers need to be acknowledged for clearly addressing some of the methodological problems in prediction research, however, an alternative proportionate reduction in error measure may generate unequivocally interpretable results and outperforms the DEL-technique. We would like to thank José Meppelink for her assistance in conducting experiments and thoughtful suggestions in carrying out the research.