Centrality in Corporate Interlock Networks: Reliability and Stability

I There is a terminological difficulty to be noted here. Centrality has both a generic and a particular meaning in this paper. Generically, refers to any of three intuitive conceptions: degree (connectedness), closeness, and betweenness (Freeman, 1979). In its particular sense, refers to any variation of a measurement technique introduced by Phillip Bonacich (1972a, 1972b) and further elaborated in Bearden et al. (1975), Mariolis (1975), and Mariolis, Schwartz, and Mintz (1979). When using it in this latter sense, we generally refer to scores, or to and centrality. In other cases, the context should make clear which of the two meanings we intend. This paper addresses a series of empirical, methodological, and theoretical questions raised by examining the reliability and stability of centrality in corporate interlock networks. Data on the interlocking directorates of 1094 large U.S. corporations in 1962, 1964, and 1966 are analyzed with a test-retest simultaneous equation model. The results confirm the common, but little tested, assumptions that centrality measures are highly reliable and stable. Further, we find that, of three measures examined (number of interlocks, nondirectional centrality, and directional centrality), number of interlocks is slightly more reliable or stable than the other two. Finally, the results show that the centrality of banks is more stable than the centrality of nonbanks. We conclude with a discussion of the implications of these findings.*