Abstract
State-owned private equity funds in China currently oversee assets worth more than 12 trillion RMB. Due to the uncertainty in the private equity market and the presence of information asymmetry, these state-owned private equity firms frequently engage in coinvestments with other private equity firms. The coinvestment strategy allows them to mitigate risks and exchange valuable information and resources. Which types of partners do state-owned private equity firms typically collaborate with? The existing literature built coinvestment partner selection models based on the traditional regression models and ignored the complexity of the network structure. This research analyzes cooperative relationships using exponential random graph models, considering both structural effects and node attributes. The empirical study of 4645 private equity firms operating in the Chinese private equity market shows that state-owned private equities are more likely to collaborate with foreign private equities and domestic private-owned private equities compared to collaborating with other state-owned private equities. Furthermore, in markets characterized by high marketization indexes, state-owned private equities demonstrate greater inclinations to partner with foreign and domestic private-owned private equities. When state-owned private equities allocate their investments to high-tech industries, their likelihood of collaborating with foreign private equities increases.
1. Introduction
What types of coinvestment partners are state-owned private equity firms (SPE firms) more inclined to collaborate with? Are there any limitations or conditions that influence the selection of coinvestment partners for state-owned private equity firms? Can exponential random graph models analyze the intricacies of strategies involved in choosing coinvestment partners? Those are the questions I explored in this paper. Private equity firms primarily invest in high-growth startups [1]. Given the inherent risk associated with the private equity market, establishing a coinvestment network becomes crucial for these firms to exchange information, identify innovative young companies, mitigate uncertainty, and acquire necessary resources [2]. In order to obtain complementary resources and information that are essential, state-owned private equities need to cooperate with partners with differentiated resources [3].
Two distinct theories exist regarding the selection of coinvestment partners. The first is the homogeneity theory, which posits that building connections are facilitated by the similarity between private equity firms [4]. According to this theory, enterprises with similarities are more likely to have congruent judgments and experience fewer conflicts [5]. In contrast, other studies show that establishing connections is actually more feasible between heterogeneous enterprises. An explanation for such feasibility is that diverse private equity firms possess distinct resources that can contribute to the development of invested enterprises [6]. In the context of state-owned private equities, it is essential to explore their actual choice of coinvestment partners, specifically whether they prefer partnering with other state-owned or private-owned private equity firms. Additionally, understanding how the institutional environment and investment preferences influence their partner selection strategy is of significant importance. Addressing these questions is the primary focus of this paper.
This research employs an exponential random graph model (ERGM) or p model to analyze the joint influence of endogenous structural effects (e.g., three-star effect) and exogenous effects (related to actor attributes) on network formation. By separating these effects, the study aims to investigate how they collectively shape networks [7].
The contents of this study are as follows. First of all, this study theoretically discusses the reasons why state-owned private equity chooses heterogeneous partners. Specifically, the study explores the impact of the uncertain institutional environment and the tendency to invest in high-tech enterprises on partner selection [8]. Then, the method section introduces the models and methodology of the research to test the hypotheses using the ERGM. After that, the empirical results are represented. The final section concludes the study by discussing the research’s contributions and outlining future research directions.
2. Theory
When examining the probability of forming collaborations with the state-owned private equity, it is crucial to consider the disparities in information and resources possessed by different types of private equity firms. State-owned private equity firms can access information from traditional industries and resources associated with state-owned enterprises. In contrast, private-owned private equity firms have access to market-based information and resources related to emerging industries. It is important to recognize that the resources and information held by these two types of firms are complementary to each other (Hakanson, 1993). Furthermore, private-owned private equity firms have a strong incentive to thoroughly investigate potential enterprises due to a significant portion of the manager’s salary being tied to dividends. In contrast, the salary structure is relatively fixed in state-owned private equity firms. Collaborating with private-owned private equity firms can enhance the motivation of investors to conduct thorough due diligence. In this context, the traditional theory of homogeneous cooperation is inapplicable. State-owned private equity firms seek out partners who offer complementary resources and capabilities.
Cooperation among heterogeneous partners improves performance [9]. Many intercompany relationships, for example, in terms of research and development (R&D), might have many reasons, such as market expansion, technology development, and information gathering [10]. Research indicates that the benefits of heterogeneity are evident via cognitive stimulation and resource complementarity. Firstly, diverse backgrounds among private investors provide a range of cognitive resources and bring forth varied knowledge and experiences in problem-solving. Conflicts arising from different values and approaches can foster innovation. This process, referred to as “creative reorganization” by Stark [11], involves the amalgamation of diverse elements from different domains, inspiring novel ways of thinking. Existing resources are utilized creatively [12], resulting in innovative outcomes.
In addition, diversified partners are more likely to offer complementary resources to the startups in which they invest [13]. Startups require various resources, such as experienced professionals, marketing channels, and technical support. These resources are highly specialized and specific to particular industries or locations. When startups aim to increase their market share, the need for complementary assets becomes even more critical [14].
Moreover, the collaboration between state-owned private equity firms and heterogeneous enterprises can effectively reduce principal-agent costs and enhance the motivation for conducting due diligence. State-owned private equity firms generally have salary structures that are similar to those of state-owned enterprises, with minimal variation and a low proportion of dividends and performance-based components. Consequently, the motivation of managers in state-owned private equity firms is lower than that in private-owned ones. As a result of the principal-agent problem, state-owned private equity firms face higher governance costs compared to private-owned ones. However, by engaging in partnerships with private equity firms, state-owned ones can reduce governance costs and enhance employee motivation.
In summary, when the state-owned private equity collaborates with private-owned private equity partners, they are more likely to bring a diverse range of resources to startups. These resources can include technical support, access to market channels, and government support [6]. Furthermore, such cooperation reduces governance costs and enhances employee motivation [15].
There are two distinct categories of private-owned private equity firms: domestic private-owned private equity firms (DPE firms) and foreign private equity firms (FPE firms). Both types share certain advantages over state-owned private equity firms (SPE firms), but they exhibit notable differences.
When state-owned PE and domestic PE coinvest, the state-owned ones can gain access to a wealth of information and resources concerning private enterprises in China. Furthermore, the domestic PEs demonstrate a high degree of flexibility in their investment strategies, enabling them to adapt to various circumstances and seize opportunities effectively.
Domestic PEs exhibit a remarkable capacity to conduct extensive searches for resources and information on a very large scale. Additionally, they possess the ability to offer highly tailored value-added services to startups. During the collaborative process between state-owned and domestic PEs, the former acquires a greater abundance of information and resources. Furthermore, while domestic private equity firms tend to prioritize market logic over industrial policy, the disparities in investment logic between domestic and state-owned PEs are relatively limited when compared to the differences observed between foreign and state-owned PEs. In terms of the startups’ trajectory, they pursue initial public offerings (IPOs) within the domestic capital market, leading to fewer conflicts concerning listing plans [16]. When it comes to team incentives, the profitability demands of the DPE play a crucial role in reducing the principal-agent costs associated with the SPE. The decision-making process of coinvestment assists the SPE in filtering out business models that are unable to generate substantial profits in reality. This analysis proposes the following hypothesis.
Hypothesis 1. A state-owned private equity firm is more willing to coinvest with a domestic private-owned PE firm than with a state-owned PE firm.
When the state-owned PE coinvests with the foreign PE, the SPE can also obtain more information and resources about the startup. Foreign private equity brings valuable expertise in investing in early-stage high-tech enterprises, offering a wealth of experience and a standardized approach to team management and operational processes. State-owned private equity firms engaging in cooperation with foreign PE firms can gain access to information and resources that differ significantly from their own [17]. Foreign PE firms bring valuable expertise in evaluating startup enterprises and can introduce the SPE to their previous projects, facilitating knowledge transfer. Foreign and state-owned PEs have some differences, and the latter has the option to function as a financial investor, thereby mitigating logical conflicts. These considerations propose the following hypothesis.
Hypothesis 2. A state-owned private equity firm is more willing to coinvest with a foreign PE firm than with a state-owned PE firm.
Institutional environments affect the partner selection strategy of private equity (Beckman et al., 2004). Because the coinvestment network has the function of buffering external shocks, one of the factors influencing the partner selection strategy of the PE is institutional uncertainty. The level of uncertainty within the institutional environment significantly impacts the extent to which private equity firms rely on informal social relationships and trust networks [18]. In emerging markets, institutional uncertainty is higher due to the relative immaturity of the legal system [19].
The institutional environment encompasses a range of elements such as regulations, customs, norms, and professional intermediaries [20, 21]. The level of environmental uncertainty encountered by private equity depends on various factors, including the development of market elements, the number of market intermediaries, the degree of government intervention, and the maturity of the legal system [22]. Higher levels of uncertainty in the environment can result in increased costs for private equity firms in terms of collaboration and monitoring, as the absence of clear formal institutions leaves them with limited recourse when their interests are jeopardized. In such instances, companies should avoid potential moral hazards that may arise due to the lack of formal institutions to enforce agreements [23]. In uncertain institutional environments, state-owned private equity firms may be more willing to coinvest with state-owned enterprises because more established group norms among state-owned enterprises can prevent deviant behavior [24, 25].
In a stable institutional environment, state-owned private equity firm demonstrates a greater inclination to engage in coinvestments with private-owned PE firms. As institutional uncertainty diminishes, the propensity for heterogeneous cooperation increases. In a more open and transparent market characterized by a well-developed legal system and effective market intermediaries, private equity no longer relies solely on social relations and trust to safeguard its own rights and interests [26]. Instead, it can leverage the advantages of a developed institutional framework and market infrastructure to ensure the protection of its interests. In this context of reduced institutional uncertainty, the collaboration between the state-owned PE and the private-owned PE becomes more feasible, allowing for the combination of resources from diverse sources to achieve enhanced performance.
In China, even in the provinces with the highest marketization index (such as Jiangsu, Zhejiang, and Guangdong Province), the influences of the state-owned economy are quite important in many important industries. The state-owned banks and financial companies provide various financial resources. Many state-owned companies are active suppliers and consumers. Foreign PEs and domestic private-owned PEs can improve their legitimacy by cooperating with state-owned PEs when entering the finance market and finding suppliers and consumers. However, private-owned PEs’ dependency on state-owned PEs decreased a lot compared with private-owned PEs from districts with low marketization indexes in China. Therefore, I put forward the following hypothesis:
Hypothesis 3. In areas with high marketization index, a state-owned private equity firm is more willing to invest with a domestic private-owned PE firm.
Hypothesis 4. In areas with high marketization index, a state-owned private equity firm is more willing to invest with a foreign PE firm.
Heterogeneous alliances can help high-tech companies integrate more resources and achieve better performance [13]. In the case of high-tech enterprises, the pursuit of continuous innovation serves as the fundamental pathway for achieving rapid growth. Innovation often arises from the introduction of novel technological elements or the reconfiguration of existing resources, as highlighted by Schumpeter and Stark [11, 27]. The incorporation of new technical elements and the reorganization of resources enable companies to venture into untapped markets and reap substantial financial rewards. Because of the significant disparities in resources and information between state-owned PE and private-owned PE, their coinvestments can offer distinct and complementary resources for the advancement of high-tech enterprises.
The digital economy provides evidence for why high-tech enterprises need to combine heterogeneous resources [8]. Numerous startups concentrate on Internet services and demonstrate innovation in online operations. Through the fusion of online and traditional business models, they surpass many conventional offline competitors [28]. When startups endeavor to transition traditional businesses to an online platform, they require a substantial array of complementary resources in traditional offline domains, including brands and marketing channels. This exemplifies innovative high-tech companies typically possess elevated demands for complementary resources.
Private equity has an important impact on the resources and information that can be obtained by startups. In order to meet the high demand for innovative resources, they need to diversify their coinvestment partners. If the state-owned private equity focuses on investing in high-tech companies, in order to mobilize more heterogeneous resources to help the development of high-tech companies [29, 30], the state-owned private equity is more willing to cooperate with private-owned PE [11]. The partner selection strategy of the state-owned private equity is affected by the investment strategy of the private equity firms. I put forward the following hypotheses.
Hypothesis 5. When a state-owned private equity firm tends to invest in high-tech enterprises, a state-owned private equity firm is more willing to invest with a domestic private-owned PE firm.
Hypothesis 6. When a state-owned private equity firm tends to invest in high-tech enterprises, a state-owned private equity firm is more willing to invest with a foreign private equity firm.
3. Methods
3.1. Sample
The sample to test the hypotheses comes from the Zero to One database (from the year 1993 to the year 2017). This database is a famous database specializing in the private equity industry in China. Given the substantial changes that occur in PE coinvestment networks annually, a five-year observation window ensures network stability. The choice of a five-year window is also aligned with the typical duration of private funds, which ranges from four to seven years. The first observation window spans from 1993 to 1997, and the dataset encompasses five distinct windows from 1993 to 2017. This research utilized Fan Gang’s marketization index, as outlined by Wang et al. [22], to gauge the institutional environment. This index is widely used to measure the marketization level of different provinces in China. The dataset used in this study contains 4,645 private equity firms and 31,516 coinvestment records.
3.2. Measurements
The dependent variable is the occurrence of joint investments between the two private equity firms over the five-year observation period. If a coinvestment occurs, the variable is 1; otherwise, it is 0.
The independent variables were derived from the attributes of private equity firms. If both coinvesting private equities are state-owned, the cooperation between state-owned private equity firms is 1; otherwise, it is 0. In the case where one private equity firm is state-owned and the other is domestic, their cooperation is 1; otherwise, it is 0. If one private equity firm is state-owned and the other is foreign, their cooperation is 1; otherwise, it is 0. The first moderator is the marketization index, which serves as an indicator of the institutional environment. This moderator is defined as the average value of the marketization indexes of the provinces where the two private equities’ headquarters are located. Due to its moderating role, the marketization index is centralized. The second moderator is the propensity of private equity firms to invest in high-tech enterprises. By calculating the ratio of high-tech startups to all startups invested in by each private equity firm during the five-year observation period, I got the investment preference of each private equity firm for high-tech enterprises. Given the calculated dependent variable at the dyad level, the moderator is determined by averaging the investment preferences of the two private equity firms. This moderator is centralized. The industry classification employed in this study is based on the China National Economic Industry Classification issued by the Statistics Bureau of China, which encompasses 16 first-class industries. These industries are classified into high-tech and nonhigh-tech industries based on the national industry classification code [26]. High-tech industries have specialized technical human resources focusing on research and development activities.
I controlled the external factors that can influence coinvestment performance and internal structural factors of the networks. Firstly, this study controlled the cooperation between domestic and foreign private equity firms, as well as the cooperation between foreign private equity firms. If a coinvestment occurs between a domestic and a foreign private equity firm, the variable “cooperation between domestic private-owned PE and foreign PE” is 1; otherwise, it is 0. Similarly, if the two foreign private equity firms engage in a coinvestment, the variable “cooperation between foreign PEs” is 1; otherwise, it is 0.
Additionally, this study incorporated the geographical distance and industry distances between the two private equity firms as control variables. The geographical distance of a coinvestment is determined by calculating the distance between the cities homing the respective private equity firms. As private equity firms have investment teams distributed across different regions, this research selected the closest group as the geographical distance between the two private equity firms [31].
The construction of industry distance had the following steps. The first step determined the industries in which each private equity firm invested during the specified window period and represented this combination as a vector. The next step calculated the number of industries in which both private equity firms invested, divided by the total number of industries in which the two private equity firms individually invested. Equation (1) represents the Jaccard coefficient of the industry vector.
In equation (1), sets A and B are the industries that two PEs invest in, respectively. The industry distance is the difference between 1 and the Jaccard coefficient [32]. A higher value of industry distance indicates a greater disparity between the two private equity firms in terms of the industries they prioritize or focus on.
Thirdly, the embeddedness was controlled. The relationship embeddedness effect between the two private equity firms was measured by the number of joint investments made by both firms within the last five years. The structural embeddedness effect between the private equity firms was assessed by the number of common neighbors they shared within a five-year time window [7].
Finally, the structural factors were controlled. I included edges, which is the number of joint investments in the network in the window period divided by the maximum possible number of joint investments. I controlled whether the node in the network is an isolated node, which means during the window period, the PE firm only makes the solo investment and has no joint investment with other PE firms. I also controlled a three-star effect, which means the PE firm cooperates with the three other PE firms in the observation window.
3.3. Models
After years of advancement, the theories and methodologies for analyzing entire networks have reached a level of maturity. The scholars who have contributed to statistical models for whole network analysis primarily come from statistics and sociology, such as Garry Robins and Tom Snijders [33, 34]. Statistical models for the whole network analysis offer several advantages, including the consideration of network structures and their relative ease of operation and interpretation. Some commonly used statistical models for whole network analysis include ERGMs (p models), stochastic actor-oriented models (SAOM), quadratic assignment procedures (QAP), and block models. Table 1 provides a comparative analysis of these models.
Powell et al. [35] employed modified logit models using clustering methods to address the correlated variances. Their study identified four fundamental trends that could impact the cooperation of biotechnology enterprises: homogeneity, preferential connection, following the trend, and diversification. The use of modified logit models reduces variance correlation, paving the way for analyzing large datasets. However, this approach may ignore the influence of complex network structures.
QAP, proposed by Krackhart [36], is a well-established method for analyzing network data. QAP utilizes matrix permutation to calculate matrix correlation coefficients, thus mitigating the issue of spurious correlations. QAP is applicable to large-scale networks without incorporating complex network configurations, such as three-star effects. QAP is more effective than modified logit models but weaker than ERGMs when estimating parameters in the whole network.
ERGMs offer several advantages in the analysis of network data. They can effectively model the influences of various complex network configurations, allowing for the examination of multiple mechanisms that impact network formation simultaneously. ERGMs also provide effective estimations compared to traditional regression models. ERGMs have found wide application in the social sciences. I gave a detailed explanation of ERGM in the first part of the supplementary materials and Table S1. I used an example to illustrate the basic functions of ERGMs. ERGMs also have their limitations. They only model small to medium-sized networks and do not apply to large-scale networks.
SAOM is a social network model developed based on the characteristics and behavior of individual nodes [34]. This model utilizes longitudinal data to distinguish between the social influence process and the social choice process. It explains the emergence, dissolution, and maintenance of connections within the network through different mechanisms. The social influence process refers to the impact of peers within the social network on the behavior of individual nodes, while the social choice process emphasizes how the attributes and behaviors of nodes influence their selection of network partners. However, its limitation is its applicability only for modeling very small-scale networks.
The block models emphasize the social structure manifested within the network [37]. The block model employs hierarchical clustering to examine the distinct roles of groups within the network and assign nodes to partitions. An advantage of block models is their ability to capture and describe the roles played by the individual nodes. However, their limitation is that they do not make assumptions about individual behavior when describing the overall network structure.
In this study, the dataset is medium-sized, which is larger than the datasets that SAOM can study. Therefore, I could not use stochastic actor-oriented models. QAP model can consider the influence of structure embedding, but it cannot consider three-star effects and other complex configurations. I used QAP models in the robustness tests. The exponential random graph models can be used to deal with medium-sized datasets with complex network structures and produce more effective estimation results, so I chose to use the exponential random graph models.
When using ERGM models, in addition to controlling for independent variables and external control variables, it is essential to also control for the relevant structural factors in order to eliminate correlations among the residuals in the network data. In exponential random graph models, the Monte Carlo method simulates the fundamental characteristics of the graphs [33, 34, 38]. A network with n nodes can be represented by an nn matrix. In the matrix, xij represents the number of connections between nodes i and j. I got matrix Y from matrix X through equation (2). The matrix Y is the dependent variable of ERGM.
Among them,
The following equation represents the ERGM:
In equation (3), is the probability that Y is equal to the observed value y, is the graph statistic, is the corresponding parameter, and k is the normalized index. The selection of variables in ZA is chosen by the theory of network formation and may vary in different studies. The conventional approach to studying private equity treats each variable as independent [7]. However, this assumption neglects the interdependence of relationships within the network. For instance, when considering a scenario where node i is connected to node j, and node j engages in coinvestment with node k, the traditional approach assumes that the relationship between nodes i and k is independent. However, if the relationship between nodes i and k relies on their connection through node j, the conventional model violates the assumption of independence. The ERGM considers the interdependence of nodes is proper for analyzing the configuration of networks compared to other methods.
In this model, “A” is the network configuration, a specific structure of a small set of nodes. For example, the structural embedding effect is defined as one node connecting to the other two nodes. The three-star effect is defined as one node connecting to the three other nodes. A three-star effect is composed of two structural embedding effects, so there is a high correlation between those two configurations. When interpreting the coefficients of configurations, it is necessary to interpret the two coefficients at the same time. The strong correlations between the configurations can be processed by the simulation method, which is the difference between the exponential random graph models and the traditional regression models [39]. The configurations controlled in the models according to the theory in this study are shown in Table 2. For focal node i (which is the black node in the second column), the name of the configurations, the diagrams, and the mathematical definitions are shown in Table 2.
Edge is the most basic network structure. For node i, the mathematical definition of edge is the number of all edges associated with node i. Embeddedness can be divided into relational and structural embeddedness [40]. The relational embeddedness effect refers to the former connection between two nodes. The lowercase letter t in the definition of the relational embeddedness effect in Table 2 represents time rather than node. The structural embeddedness effect refers to the common neighbor of two nodes. It will also affect the formation of ties. Having common neighbors can prevent opportunism because common neighbors can supervise actors and punish their opportunistic behavior [40]. The three-star effect shows the cumulative advantage effect. Cumulative advantage means that more connected nodes tend to connect to more nodes [41]. The isolation effect is also commonly-controlled, indicating that some nodes prefer to invest alone rather than participate in coinvestments.
The ERGMs employ Monte Carlo Markov chain maximum likelihood estimation (MCMLE) to estimate the model parameters [34]. This study constructs and estimates the models using the “ERGM” package available in [42].
For robustness check, QAP (quadratic assignment procedure) is also used to test the hypotheses. QAP and the exponential random graph models both use the Monte Carlo simulation method to analyze the basic characteristics of the graphs [36].
4. Empirical Results
4.1. Descriptive Statistics and Regression Results
Table 3 provides the definitions of key variables, while Table 4 presents the descriptive statistics and correlation analysis of the primary variables. The correlations between variables are all below 0.5. Table 5 reports the results of the ERGM for state-owned private equity joint investments. In order to emphasize the independent variables and moderating effects, I only showed the parameters for the main variables in Table 5. The complete empirical results of ERGMs are shown in Table S2 in the second part of the supplementary materials. Model 1 includes the basic control variables, while Model 2 incorporates the main effects. Models 3 and 4 introduce one moderator each, and Model 5 represents the full model, including all variables and moderators. According to the AIC information criterion, the model fittings are getting better. Model 2 to Model 5 support Hypotheses 1 and 2. Compared to coinvest with the SPE, the SPE is more willing to invest with the DPE. The probability of cooperation between the SPEs (β = 0.14, , Model 2) is lower than the probability of cooperation between the SPE and the DPE (β = 0.32, , Model 2). The SPE is more willing to coinvest with the FPE than with the SPE. The probability of cooperation between the SPEs (β = 0.14, , Model 2) is lower than the probability of cooperation between the SPE and the FPE (β = 0.89, , Model 2). According to the estimated coefficients, the probability of cooperation between the SPE and the FPE is greater than the probability of cooperation between the SPE and the DPE. A possible reason is that foreign private equity can provide more information and complementary resources needed by state-owned enterprises.
In Table S2 of the supplementary materials, I showed the complete results from ERGMs. In Model 5 of Table S2, both enterprises are foreign PEs significantly increases the possibility of coinvestment in the window period. Compared with the situation in which both enterprises are domestic private-owned PEs, the possibility of cooperation between two foreign PEs increases by 1.6 times. The geographical distance between private equity firms and the industrial distance exert a negative influence on the likelihood of joint investment between the two PEs during the specified window period. Specifically, a 10% increase in geographical distance leads to a 23% decrease in the probability of cooperation. A 10% increase in industrial distance results in a significant 60% decrease in the probability of cooperation. Both relationship embeddedness and structural embeddedness positively affect the likelihood of joint investment. When the number of previous cooperation between the two enterprises increases from zero to one, the probability of joint investment in the window period increases by 2.8 times. Similarly, when there is a transition from having no common neighbor to having one common neighbor, the probability of cooperation increases by 63%. However, the coefficients associated with the three-star effects are insignificant. This result implies that cumulative advantage effects do not have an insignificant impact on tie formation in this particular graph. The significant and positive coefficients observed for isolation effects indicate that private equities prefer independent investments rather than coinvestments.
Figure 1 illustrates the moderating effect of the marketization index on the cooperation between the SPE and the DPE, as well as the cooperation between the SPE and the FPE. Model 5 and Figure 1 support Hypotheses 3 and 4. The result implies that the cooperation between the SPE and the DPE is moderated by the marketization index of the places where the PEs are located. The higher the marketization index is, the more willing the SPE is to coinvest with the DPE (β = 0.03, ). The cooperation between the SPE and the FPE is moderated by the marketization index. In areas with a higher marketization index, the SPE is more willing to conduct coinvestment with the FPE (β = 0.08, ). In Figure 1, the horizontal axis is the average value of the marketization index of the two PEs; the vertical axis is the possibility of cooperation between the two PEs. When the marketization index takes the average value, the possibility of cooperation between the SPE and the DPE is 1.6 times the possibility of cooperation between the two SPEs. When the marketization index is equal to the average value plus one standard deviation, the possibility of cooperation between the SPE and DPE is 2.3 times that of the two SPEs. When the marketization index takes the average value, the possibility of cooperation between the SPE and the FPE is 2.8 times the possibility of cooperation between the two SPEs. When the marketization index is the average value plus one standard deviation, the possibility of cooperation between the SPE and the FPE is 4.2 times higher than that of the two SPEs. SPEs are more inclined to cooperate with the FPEs, followed by the DPEs and, finally, the SPEs. According to the results, state-owned private equity needs to integrate complementary resources and enhance team incentives in joint investment.
In Figure 2, the horizontal axis represents the average tendency of two companies to invest in high-tech enterprises, while the vertical axis represents the probability of cooperation between private equity firms. The results do not support Hypothesis 5. Model 5 and Figure 2 support Hypothesis 6, indicating that the cooperation between the SPE and the FPE is moderated by the tendency to invest in high-tech enterprises. When the SPE tends to invest in high-tech enterprises, it is more willing to invest with the FPE (β = 0.11, ). When the tendency to invest in high-tech enterprises is equal to the average, the possibility of cooperation between the SPE and the FPE is 2.1 times the probability of cooperation between the two SPEs. When the tendency to invest in high-tech enterprises is equal to the average value plus one standard deviation, the possibility of cooperation between the SPE and the FPE is 2.3 times that of the two SPEs.
Comparing the results of H3a and H3b, the empirical results show that the SPEs are more likely to coinvest with FPEs than DPEs. One of the possible explanations is that foreign PEs can provide more useful resources to high-tech firms. In fact, many successful Chinese startups succeed by imitating the players in other mature markets. For instance, Tencent started a social media company three years after the Mirabilis company in Israel, and Baidu began as a search engine company two years after Google. Most of the investors of Tencent and Baidu in the earlier stages are foreign PEs. The investing experiences and the asymmetry information that FPEs accumulate from foreign markets can benefit high-tech companies. As for DPEs, their experiences are often limited to the Chinese market, just like SPEs.
4.2. Robustness Check
In order to verify the robustness of the conclusions, QAP (quadratic assignment procedure) is also used to test the hypotheses. QAP employs a simultaneous permutation of both rows and columns in the matrix to examine the self-organizational structure within the network. If two columns are independent of each other, the outcome will remain the same after the permutation. However, if the two columns are not independent, there will be different results. The empirical results presented in Table 6 indicate that the SPE firms exhibit a higher propensity to invest with DPE or FPE firms compared to coinvesting with other SPEs. These findings provide further support for Hypotheses 1 and 2. Furthermore, the likelihood of cooperation between the SPE and the FPE is higher compared to the possibility of cooperation between the SPE and the DPE. The marketization indexes of the cities where the PEs locate positively moderate the cooperation between the SPE and the DPE. As the marketization index increases, the likelihood of cooperation between the SPE and the DPE also increases. This result accepts Hypotheses 3 and 4. When the marketization index is higher, the probability of cooperation between the SPE and the FPE increases faster. This phenomenon can be attributed to the significant differences between the SPE and the FPE. An environment with well-established intermediaries can resolve conflicts between these entities. In Model 9, as the tendency to invest in high-tech enterprises increases, the probability of cooperation increases between the SPE and the FPE. Those findings confirm Hypothesis 6 but do not support Hypothesis 5.
Regarding the control variables, when both PEs involved in the joint investment are FPE firms, there is a significant increase in the likelihood of coinvestment during the specified window period. Furthermore, the geographical and industry distances between the two PEs have a negative influence on the probability of joint investment. The relational embeddedness has a positive impact on the likelihood of joint investment. These findings align with the conclusions drawn from the ERGM analyses.
5. Summary
This research examines the coinvesting behavior of state-owned private equity firms. The findings reveal that state-owned private equity firms have greater inclinations to engage in joint investments with both domestic and foreign private equity firms, as heterogeneous alliances can provide access to additional resources and information for the involved enterprises. Furthermore, compared to cooperating with the DPF firms, the SPE firms demonstrate higher propensities to engage in coinvestment activities with the FPE firms. Although FPE firms differ significantly from SPE firms, they offer valuable information and resources, particularly in high-tech industries where the state-owned private equity may be lacking. Moreover, in regions characterized by a transparent institutional environment and a high marketization index, the probability of cooperation increases between the SPE and the DPE, as well as between the SPE and the FPE. The reason is that regions with a high level of marketization tend to have lower costs associated with addressing cooperative conflicts. Marketization exerts a stronger influence on the cooperation between the FPE and the SPE compared to the cooperation between the DPE and the SPE. Furthermore, the higher the tendency of private equity firms to invest in high-tech enterprises, the higher the likelihood of cooperation between the state-owned private equity and the foreign private equity. The state-owned private equity firms that invest in high-tech industries require greater access to resources and information, which can be obtained through partnerships with the foreign private equity firms that typically possess a more substantial pool of resources and information.
The paper makes theoretical contributions by examining the partner selection process of the SPE and highlighting the important roles played by the heterogeneous cooperation, the institutional environment, and investment preferences. Furthermore, the study employs the ERGM to control for network structure dependencies, yielding effective estimation results. The findings reveal that the state-owned private equity, operating in regions characterized by a high degree of marketization, can enhance cooperation performance by engaging in partnerships with both domestic and foreign private-owned PEs. Additionally, when state-owned PE firms have a substantial presence in high-tech sectors, they are advised to collaborate with foreign investors who possess industry-specific information and expertise. These insights have practical implications and can serve as a guiding framework for state-owned PEs in their decision-making processes.
This article is different from the existing literature because I tested the joint investment behavior of the state-owned private equity, which exists in a few countries in the world. The study applies heterogeneity theory to analyze the partner selection process of the SPE and confirms that the investment environment and preferences serve as the boundary conditions that impact coinvestment by the state-owned private equity. In addition, the applications of the exponential random graph models overcome the problem of independent residuals in traditional models. The behavior of the state-owned private investment can be deeper understood by examining its joint investment patterns. The study reveals that the SPE is more inclined to engage in coinvestment with heterogeneous partners rather than with other SPE firms. This preference is driven by the need to acquire information and resources and to enhance the incentives of the team.
This paper does not differentiate between the social selection model and the social influence model. This paper does not answer the question of whether the state-owned private equity is influenced by its neighbors or whether it has made its own choice. The simulation investigation for empirical network analysis (SIENA) methodology can be employed in the study of the SPE. The study does not employ SIENA due to the limited size of the matrix that it can handle. Future research could focus on analyzing a particular industry with a smaller network scale, allowing for more detailed analyses to distinguish between the social selection and social influence processes.
Data Availability
I used data from the Zero to One database (https://www.pedata.cn/). The dependent variables were formed through the dataset that shows which companies are invested by PEs every year. The independent variables include two datasets. One shows the basic characteristics of PEs, including the ownership structure of private equities and the location of PEs. And another dataset shows the basic characteristics of companies that PEs invested including whether the companies are high-tech companies.
Conflicts of Interest
The author declares that there are no conflicts of interest.
Acknowledgments
The author appreciates the suggestions from Xibao Li, Lu Zheng, and Likun Cao. This research was funded by the Fundamental Research Funds for the Central Universities under Grant no. FRF-TP-22-063A1.
Supplementary Materials
In the first part of the supplementary materials, I gave a detailed explanation of ERGM and used Table S1 as an example to explain the differences between ERGMs and traditional regression models. In the second part of supplementary materials, the full ERGMs (Model 1 to Model 5) with all control variables were given in Table S2 for reference. (Supplementary Materials)