國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
87
build cooperative network. Firms with a history of innovating new products tend to continue performing well in innovation. Innovative firms are attractive to potential cooperative partners because they can serve as a springhead of creative ideas. Hence, I included past innovative performance measure, the number of new products or new services introduced in a firm in 2006, in my statistical analyses.
Firm Size
Large firms have more managerial resources to develop new knowledge and attract potential partners to build cooperative relationships with them. To control for a possible size effect, the number of employees was used as an indicator of firm size.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
88
CHAPTER 6. EMPIRICAL RESULTS
Table 6.1 presents the descriptive statistics and correlations for all measured variables. To check for multicollinearity, I compute the VIFs (variance inflation factors) and they are all below 3.4, which is well below the cut-off point of tolerance and suggests that multicollinearity is not a serious problem (Wooldridge, 2002). I use polynomial regression (Edwards, 1993, 1994) which uses unconstrained
regression equations to examine the relationship among identity differentiation, identity integration, and cooperation. The equation includes separate measures of identity differentiation and identity integration, their squared terms, and their interaction term. To reduce collinearity problems, the measure of identity
differentiation and the measure of identity integration are scale-centered before they are squared. The equation is:
Cooperation = β0 +β1 D +β2 I +β3 D2 +β4 D*I +β5 I2 + e
D: the degree of identity differentiation of a group member I: the degree of identity integration of a group member
Table 6.2 offers the result of polynomial regression analyses about the
association between identity differentiation and identity integration and cooperation.
I argue that firms with high identity differentiation and high identity integration are more likely to cooperate with other group members. After I control past innovative performance, ownership, Bonacich power of transaction network, and firm size, table 6.2 shows thatΔR2 for this set of identity-related predictors is statistically significant.
The result suggests that some kind of association between identity differentiation and identity integration affects cooperation.
In figure 6.3, the congruence line (D = I), along which identity differentiation
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
89
and identity integration are equal, runs from the near corner to the far corner of the contour plot. The incongruence line (D = -I), along which identity differentiation and identity integration are different, runs from the left corner to the right corner. If the curvature of the surface moves upward along the D = I line, it means that
cooperation would be more when values of identity differentiation and values of identity integration were both high than when both were low (see figure 6.1 and figure 6.2); then, H1 is supported. The three-dimensional surface plot created by
unstandardized regression coefficients also supports my theoretical argument (see figure 6.1). Furthermore, following the method used by Edwards and Rothbard (1999), I calculated high score and low score to examine whether cooperation is higher when both values of identity differentiation and values of identity integration are high. High score is 1.42 and low score is (-0.54). High score is greater than low score. Therefore, Table 6.2, figure 6.1, figure 6.2, figure 6.3, high score and low score all show that Hypothesis 1 is supported. The implication is that imitating ingroup members’ behavior and ignoring uniqueness development cannot help a firm have more cooperation with ingroup members.
In figure 6.3, I further investigate the distribution of cooperation on the
horizontal plot. Intuitively, if a firm would like to gain more cooperation, it should develop more congruent values or become more and more similar with pot ential cooperative partners to gain their recognition and willingness to cooperate with this focal firm. However, in figure 6.3, I, surprisingly, find that the value of cooperation is greater in the below right-hand corner (identity differentiation is high) than that in the top left-hand corner (identity integration is high). Counter-intuitively, figure 6.3 shows that identity differentiation contributes more to high cooperation.
In Hypothesis 2, I argue that peripheral brokers are more likely to have high
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
90
identity integration and high identity differentiation. I use binary logistic analysis to test H2 and results in table 6.3 support my argument (regression coefficients = 0.04; p
< 0.05). Regression coefficients = 0.04 means that, compared with the probability of occurrence of non-event (DV=0), the probability of occurrence of an event (DV=1) increases 0.04 times for every additional value of independent variable. The implication of this result is that, peripheral brokers, who are not constrained by a certain group and not too integrated into the business group, are more likely to accept partners’ good ways of behaving and have more flexibility to try something new or develop uniqueness.
To know how much additional variance is explained by brokerage effect and identity effect, I conduct hierarchical regression analyses to test H3. In Hypothesis 3, I propose that high identity integration and high identity differentiation plays a
mediating role in the relationship between brokerage and cooperation. Following Baron and Kenny (1986), I estimate the following regression equation to examine whether the mediating effect is supported. First, the mediator (high identity
integration and high identity differentiation) is regressed on the independent variable (peripheral broker). Table 6.3 shows that peripheral broker positively affects high identity integration and high identity differentiation (regression coefficients = 0.04; p
< 0.05). Second, the dependent variable (cooperation) is regressed on the
independent variable (peripheral broker). Table 6.4 shows that peripheral brokers are more likely to have more cooperation with other group members (regression coefficients = 0.32; p < 0.05). Third, the dependent variable (cooperation) is
regressed both on the independent variable (peripheral broker) and the mediator (high identity integration and high identity differentiation). Table 6.4 shows that high identity integration and high identity differentiation exerts a significant positive
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
91
impact (regression coefficients = 0.4; p < 0.001) on cooperation and the effect of peripheral broker on cooperation is less when the mediator (high identity integration and high identity differentiation) is controlled. Hence, hypothesis 3 is supported.
While the association between brokerage effect and cooperation is widely accepted, I open the black box, high identity integration and high identity differentiation, in this association.