BUSINESS SEGMENT PERFORMANCE REDUX: A MULTILEVEL APPROACH

摘要

Over the last 20 years, strategy scholars have explored the relative influence of industry and corporate effects on business segment performance. The present research acknowledges the inherent multilevel, cross-classified nature of the data and demonstrates that results are not robust to statistical method. Using 1995-99 data from 'operating segments,' the multilevel analysis presented in this article found that (1) business segment effects explain twice as much variance in business segment performance as do corporate effects, (2) corporate effects explain almost four times more variance in business segment performance than industry effects, and (3) segments explain almost eight times more variation than industry effects. Methodologically, multilevel analysis offers statistical advantages over variance components analysis, nested ANOVA, and two-stage least squares. More importantly, the method provides a means for modeling relationships between the industry, corporation, and business segments, thereby moving beyond the descriptive nature of explained variance.