meta-analytic review of the diversification-performance relationship: Aggregating findings in strategic management

This study discusses the theoretical and methodological issues in the diversification-performance literature and presents a meta-analysis of 50 studies with 169 separate effects of the diversification-performance relationship. The overall objective of this study was to answer the primary question: What is the relationship between diversification strategies and corporate performance? In addition, this study aimed to determine if the inconsistencies found in the diversification-performance relationship across studies is attributable to moderator variables such as: (1) measurement of diversification, (2) measurement of performance, (3) source of data, (4) time period, and (5) industry effects. A final objective of this study was to illustrate the usefulness of meta-analytic techniques for reviewing and synthesizing empirical literature in strategic management. The results suggest that related diversified firms have higher accounting- and market-based returns and lower levels of total risk than unrelated diversified firms. These results are consistent with the predictions of agency theories of diversification. The results further indicate that related diversifiers have lower levels of systematic risk than unrelated diversifiers, a result consistent with the predictions of the traditional strategic management perspective of diversification. The results of the subgroup and combined meta-analyses of moderator effects indicate that the use of different measures of diversification (e.g., Rumelt's categories vs. SIC-based continuous), different measures of performance (accounting- vs. market-based), and source of data (e.g., primary vs. secondary) significantly influenced the relationship between diversification strategies and performance. The relationship was not affected by the time period of studies or industry effects. The findings are discussed in terms of their implications for practice and for future research.