Assessing the construct validity of strategic and SIC-based measures of corporate diversification

Several new methods of measuring corporate diversification have been developed since Rumelt's (1974) seminal work. These methods use varying criteria but they have not been backed up by construct validity The present study investigates the content, convergent, discriminant, unidimensionality, and predictive validity by comparing 54 Fortune 500 companies on Rumelt's classification (self-reported and calculated), entropy, Berry-Herfindahl index, and broad and mean narrow spectrum diversification measures, Results indicate only mild support for convergence, discriminant and predictive validity, and no evidence of unidimensionality amongst these measures, This raises problems in the measurement of diversification and subsequent relationships to performance linkages.