Content area
Full text
1. Introduction
While an industry might contain an extremely large number of competing firms, typically managers only focus on a small subset of direct rivals. The question is, which firms will fall into that subset? Which firms will engage each other as rivals? More broadly, what shapes the structure of rivalry within an industry? A variety of categorizational schemes has been used to study patterns of rivalry (Cattani et al., 2017), and a strategic groups analysis is particularly useful in this regard (McGee and Thomas, 1986; Tang and Thomas, 1992).
Given the strategic positioning of the firms, a strategic groups analysis can be used to infer: which firms are likely to interact with each other; what the nature of those interactions might be; and how those interactions could affect the performance of the firms. This applies the logic of the S-C-P model at the level of the group rather than the industry. This means that the intensity of competition might not be homogeneous within an industry. One strategic group might be engulfed in perfect competition while another group is able to maintain a pocket of oligopolistic competition. These group dynamics can generate true group-effects that cannot be captured by industry-level or firm-level factors (Dranove et al., 1998; Murthi et al., 2013).
Unfortunately, progress in strategic groups research has been hampered by the poor fit between theory and methods. This has been due primarily to the lack of significance tests for cluster analysis. In lieu of such tests, researchers have looked for differences in performance as a sign that discrete strategic groups exist. Unfortunately, the empirical results have been disappointing. Frustrated by the lack of progress, Barney and Hoskisson (1990) proposed three options. Preferably, researchers should develop significance tests for cluster analysis to detect the existence of discrete groups; failing that, they should develop a theory explaining when strategic groups would and would not exist; failing that, they should abandon the entire field of research. Several decades later, the first two options have not been realized, and the level of research activity in this area has indeed declined (Cattani et al., 2017).
Fortunately, ecologists have been busy developing significance tests for cluster analysis. When Clarke et al. (2008) illustrated a permutation...