A Model Study Based on Social Network Relational Dimensions and Structural Dimensions

I argue that greater complementarity increases the breadth of the knowledge and information pool of the entire group, which facilitates correct decision-making and improves the effectiveness of alliance management [11,12,13].

Group tie density refers to the number of duplicate connections in a group divided by all possible binary connections in the group [14]. The junction density aggregates the degree of association duplication in all bilateral relationships. Therefore, the group tie density improves the communication and coordination among members, which facilitates the flow and utilisation of complementary resources, as well as reduces the opportunistic behaviour.

Status represents an organisation’s position relative to other organisations throughout the industry network [15, 16]. In a multiparty alliance, there are two different types of status-based differences, which are status heterogeneity and status-based fault. Status heterogeneity refers to the different levels of network centrality among multiple members. Status-based fault refers to the degree of subgroups formed by multiple members based on their network centrality.

Network status heterogeneity leads to decreases in motivation and ability to share complementary resources among members, thereby reducing the advantages of the complementarity effect. Therefore, when there is network status heterogeneity, it weakens the advantage of complementarity.

The concept of faults is mostly seen in studies of the individual team level, which focus on the similarity within the team and the differences between groups. Network status-based faults lead to a lack of trust between subgroups and can weaken the interactions among subgroup members [6]. Therefore, due to the positional fault, the degree of sharing and utilisation of complementary resources between subgroups is reduced.

This paper uses China’s venture capital industry as an empirical scenario, using data from two major venture capital data providers, CVSource and Zero2IPO [17, 18]. Both CVSource and Zero2IPO are China’s leading research institutions, which contain the entire range of data from China’s venture capital industry. This paper combines the information from the two databases to obtain the most comprehensive data coverage and industry data for the period 1999–2018. Although the first international venture capital firm entered China in the early 1980s, China’s venture capital field did not develop until the late 1990s. Therefore, this paper uses 1 January 1999 as the starting point of the sample and 31 December 2014 as the end point, so that I can leave the 4-year window period up to 31 December 2018, to better evaluate performance. Since most Chinese venture capital firms were set up after 2000, left censoring is not a serious problem.

Performance. The investment performance is a binary variable. If the investment is finally exited by initial public offerings, it takes the value “1”; otherwise, it takes “0”.

Complementarity. Venture capital institutions invest in different stages, and different levels of expertise are required at the different stages to complement each other. Therefore, complementarity is an average of two–two correlations between the investments of the alliance partners at different stages of investment in the industry: Complementarity=1−ΣKcor(Sx;Sy)Nk{\rm Complementarity} = 1 - {{{\Sigma _K}cor({S_x};{S_y})} \over {{N_k}}} , where S_x and S_y are, respectively, the investment phase vectors of a pair of institutions x and y 5 years before the alliance; N_k is the number of all possible bilateral associations. Complementarity is a continuous variable between the values “0” and “1”; higher values indicate greater complementarity [19].

Group tie density. Group tie density is the ratio of all pre-existing bilateral connections in the multiparty alliance to all the possible bilateral connections between members. In the 5 years prior to the cooperation, when two organisations jointly invested in the same venture at least once in the same round, those two organisations had a tie. For example, suppose a group contains three parties A, B and C, where A and B have a network tie in the first 5 years. Then, the group tie density is 1/3, i.e., one of the three pairs of bilateral relations has a network interaction foundation over the period of 5 years [5].

Network status heterogeneity. This calculation is based on a joint investment event in the whole industry, which better reflects the position of venture capital institutions in the venture capital industry. Since the enterprise’s network location is relatively stable, I use a 5-year window (t − 5) to build the current t-year network matrix. This paper uses Bonacich’s (1987) [20] centrality to measure the member’s network location: C_i,t = ∑_j(α + βC_ij)R_ijt − 1, where C_i,t is the centrality of member i in t years, R_ijt − 1 refers to the common investment matrix, α is a proprietary coefficient and β represents a weight parameter that indicates the degree to which the centrality of the member i is a function of others’ centrality in the matrix R_ijt − 1 [16,20]. Referring to Zhang et al. (2017) [5], I first calculate the similarity of each pair of bilateral positions, i.e., I divide the smaller party by the larger one and then calculate the group level by subtracting the mean of all bilateral similarities.

Network status-based fault. I calculate the degree of status-based faults at the group level using the standard deviation of all possible bilateral similarities in the group [21].

I control the group-level variables that affect group performance. First, I control the variables associated with the group’s activity; as the study sample is from the venture capital industry, the variables include factors such as investment activity uncertainty and whether there are multiple rounds of investment. Among them, the factor Stage refers to investment uncertainty, which is measured by the degree of development of an investment deal [22, 23]. Early investments have high uncertainty, while late investments have relatively low risk. These take values 1–4; so, a high value represents high activity uncertainty. Unsingle measures whether the investment is a multistage one; if the investment has had two or more rounds, its value is “1”; otherwise, it takes value “0”. Second, I control the overall resources and capabilities of the group. I use the total initial public offerings record, which is the number of successful investments in the group level, to represent the Group Ability. I also control the Group Resources in each group, i.e., the total number of previous investment deals. Third, I control the composition of the group, such as size, geography and age. The Size refers to the number of members in the group. I control the average Age of each group. I also control the Proportion of Foreign Members and Average Geographical Proximity in a joint investment group. At the same time, some industries (such as high-tech industries) are fertile grounds for the successful exit form of initial public offerings. To address this heterogeneity and other factors related to industry that affect the investment performance, I added a fixed effect of industry segmentation.

Given the binary classification characteristics of the dependent variable, in this paper, I use the probit model to estimate the investment exit tendency, which refers to multilateral interorganisational performance: Pr(I = 1|X) = F(X,β), where F is the standard normal cumulative distribution function. Based on variable selection and the factors affecting performance in this paper, a probit regression model was constructed [24]. ‘Model 1 is a model that only adds control variables; Model 2 is the main effect model of complementarity to performance; Models 3, 4 and 5 are models with moderator variables; and Model 6 is a full model with all variables. Among them, X_i includes all control variables I mentioned above. (1)Performance i=β0+β1∑Xi+∑Industry+εiPerformance\,i = {\beta _0} + {\beta _1}\sum {{X_i}} + \sum {Industry + {\varepsilon _i}}(2)Performance i=β0+β1Complementarity+β2∑Xi+∑Industry+εiPerformance\,i = {\beta _0} + {\beta _1}Complementarity + {\beta _2}\sum {{X_i}} + \sum {Industry + {\varepsilon _i}}(3)Performance i=β0+β1Complementarity+β2Complementarity × Group tie density+β3∑Xi+∑Industry+εiPerformance\,i = {\beta _0} + {\beta _1}Complementarity + {\beta _2}Complementarity\, \times \,Group\,tie\,density + {\beta _3}\sum {{X_i}} + \sum {Industry + {\varepsilon _i}}(4)Performance i=β0+β1Complementarity+β2Complementarity × Network status heterogeneity +β3∑Xi+∑Industry+εi\eqalign{ & Performance\,i = {\beta _0} + {\beta _1}Complementarity + {\beta _2}Complementarity\, \times \,Network\,status\,heterogeneity \cr & {\kern 50pt} {\kern 50pt}\,\, + {\beta _3}\sum {{X_i}} + \sum {Industry + {\varepsilon _i}}}(5)Performance i=β0+β1Complementarity+β2Complementarity × Network status − based fault +β3∑Xi+∑Industry+εi\eqalign{ & Performance\,i = {\beta _0} + {\beta _1}Complementarity + {\beta _2}Complementarity\, \times \,Network\,status\, - \,based\,fault \cr & {\kern 50pt} {\kern 50pt}\,\, + {\beta _3}\sum {{X_i}} + \sum {Industry + {\varepsilon _i}}}(6)Performance i=β0+β1Complementarity+β2Complementarity × Grouptiedensity +β3Complementarity × Network status heterogenity +β4Complementarity × Network status heterogenity +β4Complementarity × Network status − based fault+β5∑Xi+∑Industry+εi\eqalign{& Performance\,i = {\beta _0} + {\beta _1}Complementarity + {\beta _2}Complementarity\, \times \,Grouptiedensity \cr &{\kern 50pt}{\kern 50pt}\,\, + {\beta _3}Complementarity\, \times \,\,Network\,status\,heterogenity \cr &{\kern 50pt}{\kern 50pt}\,\, + {\beta _4}Complementarity\, \times \,\,Network\,status\,heterogenity \cr & {\kern 50pt}{\kern 50pt}\,\,+ {\beta _4}Complementarity\, \times \,\,Network\,status\, - \,based\,fault + {\beta _5}\sum {{X_i}} + \sum {Industry + {\varepsilon _i}}}