Social structure and alliance formation patterns: a longitudinal analysis - includes appendix

Administrative Science Quarterly,  Dec, 1995  by Ranjay Gulati

• E-mail
• Print
• Link

The positive and significant coefficient of common ties and the negative coefficient of distance in columns 6 and 7 support hypotheses 4 and 5 and suggest that having common third partners or even indirect connections enhances the probability that two firms will enter an alliance. The positive coefficient of common ties indicates that the larger the number of common third partners shared by two unconnected firms, the more likely they are to enter an alliance. The negative coefficient of distance in column 7 indicates that the more distant two firms are in an alliance network, the less likely they are to seek mutual alliances. In column 7, common ties becomes insignificant when distance is introduced. Because common ties refers to a path distance of two between firms, while distance captures all distances of two or greater, this result is not surprising.

The effects of the interaction of interdependence with common ties and distance are introduced in columns 8 and 9. Hypotheses 6a and 6b predicted that connected firms would enter alliances more frequently if the firms were interdependent to begin with and that there would thus be interactions between interdependence and common ties and between interdependence and distance. The effects of the interaction terms are strong and significant. The coefficient of the interaction in column 8 supports hypothesis 6a, indicating that there is a higher likelihood of alliance formation between interdependent firms if they have numerous third-party connections. The negative and significant coefficient of the interaction in column 9 supports hypothesis 6b and indicates a lower likelihood of alliance formation between interdependent firms that are further apart in an alliance network. Taken together, the results for the interaction effects suggest that prior indirect connections are more influential in alliance formation among interdependent dyads than among other dyads.

The random-effects models applied here generate a parameter rho, an indicator of the presence or absence of heterogeneity. An estimate close to zero implies little heterogeneity, and all the time dependence in the alliance formation rate can be ascribed to the independent variables included in the models. A significant rho implies heterogeneity. Economists typically treat such heterogeneity as an exogenous factor indicating the personal propensity of a unit (here, the dyad) to engage in the activity characterized by the dependent variable (cf. Black, Moffitt, and Warner, 1990). Thus a significant coefficient here would suggest that observationally identical dyads display different alliance propensities that remain fixed because of permanent differences in their alliance preferences and other unobserved factors. These could include pairs of firms developing specific managerial routines for entering into and managing alliances with each other. Or, in the case of firms not forming alliances, it could indicate that some pairs of firms are separated by insurmountable barriers that preclude their ever entering into an alliance. Given the possibility of multiple and confounding interpretations of heterogeneity (cf. Granovetter, 1988), I am cautious in drawing direct inferences from this coefficient. For my purposes, rho primarily indicates that heterogeneity exists and is accounted for in the panel model, and its significant coefficients do indicate that unobserved factors not completely captured by the independent variables are accommodated by the statistical model.