Premium
Construct validity of an objective (entropy) categorical measure of diversification strategy
Author(s) -
Hoskisson Robert E.,
Hitt Michael A.,
Johnson Richard A.,
Moesel Douglas D.
Publication year - 1993
Publication title -
strategic management journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 11.035
H-Index - 286
eISSN - 1097-0266
pISSN - 0143-2095
DOI - 10.1002/smj.4250140305
Subject(s) - discriminant validity , econometrics , diversification (marketing strategy) , entropy (arrow of time) , convergent validity , categorical variable , statistics , mathematics , construct validity , economics , psychometrics , business , marketing , thermodynamics , physics , internal consistency
This study measures the construct validity of an objective (entropy) approach to measurement of diversification strategy. Results indicate strong convergent, discriminant and criterion‐related validity for the entropy measure of diversification. In particular, support for the entropy measure of diversification strategy was demonstrated through associations with the Rumelt subjective measure of diversification (convergent validity); size, debt and R&D intensity (discriminant validity); and accounting and market‐based performance (criterion‐related validity). Using structural equations modeling, the study reports strong standardized validity coefficients with a diversification factor (0.87 for the entropy and 0.94 for Rumelt's measures). The objective (SIC count) measure exhibits a low standardized validity coefficient (0.44) with the diversification factor. In a discriminant validity test, 70 percent of the variance in the entropy measure is unique to diversification while only 2.8 percent and 7.6 percent are unique to leverage and size, respectively. However, only 6.3 percent of the variance in the SIC count measure is unique to diversification. The study suggests that it may be more appropriate to use the diversification factor with both the entropy and Rumelt subjective measures for maximum accuracy (however, using either alone would be acceptable). Also, the results suggest that the SIC measure may be appropriate in more limited circumstances.