Network Impact on the Investment Strategy and Performance of Cross-Border Venture Capital Institutions in China

Abstract

In this paper, we investigate whether the VC institutions’ position in the network affects their investment strategies and performance. Our results show that network location has a negative correlation with geographic concentration index, industry concentration index, and stage concentration index, which indicates that the higher the network position of the cross-border VC institutions, the higher the degree of geographical diversification, the wider the industry diversification, and the higher the investment stage diversification of their investments. In addition, our results show that the relationship between network position and the VC’s performance is not simply linear but rather an inverted U-shaped correlation. When the network location is lower than the critical value, higher network location positively improves the VC’s investment performance; however, when the network location is higher than the critical value, then the relationship is reversed.

1. Introduction

Since the first introduction of cross-border venture capital investment in China, it has undergone significant growth and until 2003, China has become the world’s largest net cross-border venture capital inflow country (Aizenman and Kendall 2012). With an annual growth rate of 20.76% from year 2003 to year 2016, the cross-border VC investment even has a growth rate higher than the GDP growth rate in China. (Yin and Zeng 2013; Yang et al. 2023).

Despite their rapid growth in China, cross-border venture capital institutions face significant challenges, such as agency costs caused by different geographical locations of their home countries, communication problems, and supervision problems caused by the institutional and cultural environment, etc. (Zhang et al. 2020). Forming syndication with joint investment has become one of the popular strategies for the cross-border VC institutions to reduce the uncertainty and improve their success rate. Over the past 20 years, the number of joint investments by cross-border VCs and local VCs has continued to grow, which leads to the changes of the cross-border VC positions in the network, and the change, in turn, will have an impact on the investment strategies and investment performance of cross-border venture capital institutions.

In this paper, we investigate whether the VC institutions’ positions affect their investment strategies and performance. Our research aims to bridge the gap in the literature regarding the network of the cross-border VC investments from the following perspectives:

First, most studies on venture capital networks focus on network formation and the relationship between networks and performance and thus miss the discussion of network segmentation. In addition, few researchers have studied the investment preferences and cross-border venture capital investment strategies from the perspectives of the network location.

Secondly, there are inconsistent empirical results regarding network impact on investment performance, i.e., most research results show that networks have a positive impact on investment performance, but there are also arguments that the impact is not significant or even negative.

In addition, our research will help explain the behavioral patterns of the cross-border VC institutions and provide guidance for joint investments and development of the local VC industry development.

The rest of the paper is organized as follows: in Section 2, we review the relevant literature, in Section 3, we discuss methodology and variables, we then present empirical results in Section 4 and draw conclusions in Section 5.

2. Literature Review

In this section, we review the relevant literature, mainly from two perspectives: cross-border venture capital investment and research on investment network locations.

Cross-border VC investment refers to projects invested in by venture capital institutions overseas. Traditionally, most of the cross-border VC investments are from mature market countries to emerging markets. Since 2000, cross-border investment transactions have accounted for one-third of global venture capital transactions (Schertler and Tykvová 2011). For a long time, foreign investment capital such as Sequoia Capital, SoftBank, and IDG have been the leading force in the development of China’s venture capital industry. After Baidu, Alibaba and Tencent received financial support from the above-mentioned foreign investment funds; these companies grew with a rapid speed and they were successfully listed in Hong Kong, China, and the United States security markets (Zhan et al. 2016).

The venture capital industry generally faces significant information asymmetry problems, which greatly increases their investment risk. Cross-border venture capital institutions encounter more risks due to the disadvantages of outsiders (Dai et al. 2012). Therefore, the investment strategies of cross-border venture capital institutions are slightly different from those of local investment institutions.

For example, some research shows that joint investment is a common strategy for cross-border venture capital institutions to reduce risks, especially joint investment with local institutions (Dimov et al. 2012; Pollock et al. 2015), which helps cross-border venture capital to contact local investors, collect resources and information, build connections, etc., thereby increasing the probability of successful investment exit (Hochberg et al. 2007). Lerner (1994) also argues that VCs tend to syndicate their investments with other VCs rather than investing alone and are, therefore, bound by their current and past investments into webs of relationships with other VCs, which result in the formation of VC syndication networks.

VC syndication has been widely studied in the literature. For example, Bygrave (1988) mapped how venture capital firms are linked together in a network by their joint investments and showed how VC institutions exchange resources with one another through connections in that network. Hochberg et al. (2007) argue that syndication networks improve the quality of deal flow, through the selection of the most promising companies, and the ability to nurture investments and thus add value to portfolio companies. Wilson (1968) and Sah and Stiglitz (1986) claim that VC institutions benefit from investment network by screening each other’s willingness to invest in potentially promising deals and this will help them pool correlated signals and thus select better investments in situations of often extreme uncertainty. Sorenson and Stuart (2001) argue that syndication also helps diffuse information across sector and industry boundaries and thus provide more diversification benefits for the VC institutions. Hochberg et al. (2007) find that better-networked VC firms experience significantly better fund performance, as measured by the proportion of investments that are successfully exited through an IPO or a sale to another company. Similarly, the portfolio companies of better-networked VCs are significantly more likely to survive subsequent financing and eventual exit. They also provide evidence about the VC network evolution.

Researchers argue that building VC networks with the local investors provides significant value to the cross-border VC institutions. Cumming and Johan (2017) and Drover et al. (2017) argue that joint investment with local venture capital institutions played an important role in improving the investment performance of cross-border venture capital institutions. They also found out that as the number of cross-border venture capital investment events increases, the network positions of the cross-border VC companies also gradually change from the edge to the middle, and their investment strategies and behaviors also converge with those of local investment institutions in the host country; for example, their investment will have a wider regional coverage and industry coverage than the local VC institutions (Dai et al. 2012; Nahata et al. 2014). In addition, some cross-border venture capital institutions have begun to penetrate early or mid- to late-stage companies, the same as local VC institutions (Nahata et al. 2014).

Researchers also found that the VC network can help VC institutions to improve their investment performance. For example, Humphery-Jenner and Suchard (2013) studied 4753 investment companies from China and found that cross-border institutional investors are more likely to exit when they invest at a later stage and have extensive investment experience in the entire industry. Hochberg et al. (2007) found that better-networked VC firms experience significantly better fund performance, as measured by the proportion of investments that are successfully exited through an IPO or a sale to another company. Esposito et al. (2022) characterize the performance of venture capital-backed firms based on their ability to attract investment, and they found out that success has a strong positive association with centrality measures of the firm and of its large investors and a weaker but still detectable association with centrality measures of small investors and features describing firms as knowledge bridges. They further indicate that success is not associated with firms’ and investors’ spreading power (harmonic centrality), nor with the tightness of investors’ community (clustering coefficient) and spreading ability (VoteRank).

In addition, researchers argue that rich historical investment experience can also help achieve better exit performance. For example, Yang et al. (2023) and other studies found that the local expertise and domestic investment experience of cross-border venture capital are important factors affecting the performance of cross-border investment capital, the investment cycle, and the exit path of cross-border investment capital.

On the contrary, other studies show that there is no significant relationship between joint investments with local investment and the cross-border VCs’ investment performance. For example, research by Bertoni and Groh (2014) shows that joint ventures made by the cross-border investors provide more exit opportunities for cross-border VC institutions, but the statistical results for IPO exits are not significant. Nguyen et al. (2023) examine how the geographical structure of social networks shapes venture capital (VC) investment decisions and they found that VC firms invest more in portfolio companies in socially connected regions. They further found out that social connectedness lowers the likelihood of a successful exit since it induces VC firms to undertake suboptimal investment decisions. Gu et al. (2023) investigated the VC network communities from the investors’ movement across VC network communities’ perspective and examined how this movement behavior impacted their VC performance, measured by the enterprises’ innovation. They found out that the investors’ movement across network communities has an inverted U-shape effect on enterprises’ innovation. In addition, the experience of the VCs and the higher level of clustering of the network community make the inverted U-shape curve steeper and make the turning point to the left. Huang (2022) investigated the significant role of networks, particularly alumni networks, in venture capital markets. He found that the introduction of new alumni connections prompts an 8.21% increase in investments of startups founded by alumni. But at the same time, these alumni ties trigger a 22.93% increase in failure rates, a 17.53% decline in acquisition rates, and 66.45% decline in IPO rates. He argued that despite venture capitalists gaining better information from networks, the inherent preference for alumni startups offsets these advantages, leading to potential capital misallocation and thus a lower success rate. Hu and Zhou (2018) believe that joint investment by venture capital institutions will reduce their investment performance due to defects in pre-selected projects and the failure of value-added activities after investment. Li and Liang (2017) found that during the joint investment process, the performance of follow-up investors was better than those from the original venture capital institutions. Wang et al. (2018) found that network position of the VC institutions negatively regulates the relationship between VC institutions’ perception of network resource and their investment performance.

In summary, the literature has not reached consistent conclusions regarding the impact of VC network on the VC institutions’ investment strategies and performance. In this study, we focused on the network of cross-border VC institutions in China and studied how the network positions influence their investment strategies and performance. Our research aims to bridge the gap and solve the puzzle in the literature regarding the network of the cross-border VC investment strategies and performance.

3. Hypotheses

In this section, we present our hypotheses.

First, researchers argue that when cross-border venture capital firms invest in industries with high risks and uncertainties, they tend to increase their investment in joint investments (Wang et al. (2015), and as the number of joint investments increases, the status of cross-border venture capital institutional networks continues to improve (Chen and Zhang 2020). Jin et al. (2021) argue that cross-border venture capital institutions with top network positions have more resources and information. Based on this, we argue that the higher the network position, the broader the investment channels, and the more dispersed the investment target industries will be. Based on this, we propose Hypothesis 1:

H1. 

The higher the network position of a cross-border venture capital institution is, the more diversified across industries their investment will be.

Researchers also argue that venture capital institutions have geographic preferences (Christensen 2006; Hochberg et al. 2007). China has a vast territory and the development and performance of the venture capital industry in different regions can vary significantly. Research by Peng and Tang (2020) shows that China’s venture capital institutions are mainly concentrated in developed eastern regions such as Beijing, Shanghai, and Guangdong, but the distribution of investment target companies is relatively dispersed.

Hu and Zhou (2018) found that as the network position of venture capital institutions increases, the wider the scope of their geographic investment and the greater the probability of successful exit of the investment will be. We argue that the higher network position of the cross-border venture capital institutions will provide more resources and information to the VC institutions, and thus they will have fewer regional constraints when they invest. Based on this logic, we propose the Hypothesis 2:

H2. 

The higher the network position of a cross-border venture capital institution, the higher the degree of geographic diversification of its investment will be.

According to the China Investment Database, the development stages of startups are usually divided into four stages, i.e., an early stage, a developing stage, an expansion stage, and a mature stage. It is generally believed that VC institutions face higher market risks, technical risks, and management risks when investments in the early and developing stages are greater than those in the later stages; for example, Sun et al. (2020) found out that investments in the early stages are less likely to result in a successful exit for VC institutions, and the earlier the stage of the investment in the target company, the greater the uncertainty of the company’s success will be. Therefore, when investing in early-stage companies, VC institutions need more information, resources, and experiences to evaluate the project, and they are more inclined to co-invest (Cumming and Dai 2010).

Based on the above information, we argue that as the network position of cross-border venture capital institutions improves, VC institutions are more likely to invest in the early stage of the invested companies because they can obtain more information and resources to help us evaluate the project. Accordingly, we propose Hypothesis 3, as follows:

H3. 

The higher the network position of a cross-border venture capital institution, the higher the diversification of its investment stages.

Researchers also found that the network positions of the cross-border VC institutions also help them improve investment performance (Buchner et al. 2018; Cumming and Johan 2017; Cumming et al. 2016; Drover et al. 2017). For example, much research found that VC firms with better network centrality tend to have better investment performance (Luo et al. 2016; Xiong et al. 2020). They argue that a more centralized network position means that venture capital institutions have greater market influence, more information resources, and can obtain information faster, and all these may contribute to better investment performances.

Other researchers, however, argue the opposite. For example, Hu and Zhou (2018), Li and Liang (2017), and Wang et al. (2018) show that the investment performance of an investment institution, which is in a core position in joint investment, is inferior to those that are not in the core position, and the higher the network location centrality, the worse its effect on investment performance.

We argue that a VC company’s network positions might have two different impacts on the cross-border VC institutions’ investment performance. On one hand, with the improvement of network position, cross-border VC institutions are more likely to obtain valuable information and resourses and thus are more likely to find better investment opportunities. However, when the network position exceeds the critical value, the probability of information redundancy increases and thus reduces the efficiency of the investment decision-making process. As a result, we argue that the relationship between VC institutions’ network position and their performance might not be linear, but rather inverted U-shaped, i.e., when the network position is lower than the critical value, the higher the network position, the better the investment performance of cross-border venture capital institutions; however when the network position exceeds the critical value, then the higher network position might lead to worse investment performance. Accordingly, we propose the Hypothesis 4 as below:

H4. 

There is an inverted U-shaped relationship between the network location of cross-border venture capital institutions in China and their investment performance.

4. Data, Variables, and Methodology

4.1. Data

In this study, we focused on the cross-border venture capital institutions in China, so we selected our research targets from the China Venture Source database (China Venture Source, referred to CV Source) based on the following criteria: (1) headquartered outside China (including the Cayman Islands and the British Virgin Islands); (2) conduct business around the world; (3) have investment experience in China; (4) funding sources are from outside China; (5) institutional types are VC, PE, and early-stage venture capital institutions. We then conducted manual screening based on their specific businesses and the available data, and eventually we identified 167 companies as our research targets.

We then screened the investments and exit events of those 167 cross-border venture capital institutions in China from 2001 to 2020 from the CV Source database, supplemented with the private equity database Zero2IPO and IT Orange database, etc., to complete the data sample. Among the 83,569 investment events with investment types of VC, PE in the early stage from 2001 to 2020, 41,756 were joint investment events. We then screened the data sample to make sure at least one of the joint investors was a cross-border venture capital institution. This resulted in 24012 individual observations.

We then followed Abell and Nisar (2007) and used Ucinet6.0 network analysis software to construct investment networks with a three-year rolling window period from 2001–2020. Afterwards, the measurement index of network position was calculated. In order to facilitate subsequent expressions, the last year of the rolling window period was used to identify the rolling time window. For example, 2003 represents the 3-year time window of 2001–2003, 2004 represents the time window of 2002–2004 etc.

Table 1. Sample Information for Rolling Periods.

Rolling Time Window	Number Co-Investment Events (Separated by the Institutions)	Number of Investment Events with Cross-Border Venture Capital Institutions
2001–2003	301	53
2002–2004	486	122
2003–2005	624	126
2004–2006	958	238
2005–2007	1227	260
2006–2008	1414	460
2007–2009	1231	507
2008–2010	1203	447
2009–2011	1282	277
2010–2012	1494	479
2011–2013	1538	526
2012–2014	2229	489
2013–2015	3820	523
2014–2016	5244	1217
2015–2017	6688	2080
2016–2018	8650	1947
2017–2019	9644	2661
2018–2020	9868	2941

shows the investment events and the relevant institutions in each 3-year rolling period.

4.2. Variables

4.2.1. Dependent Variables

We have multiple dependent variables to measure the influence of the VC network positions.

(1) 

Industrial Concentration Index (ICI)

The industry concentration index (ICI) is used to measure the industry diversification level of cross-border VC investments. We followed the research of Chen (2011) to create the ICI, which presents a higher degree of diversification when the ICI value is low and vice versa. When the ICI equals 1, it means that all investment capital is concentrated in one specific industry, i.e., there is no diversification at all. To create the ICI, we first classified the industries based on the industry affiliation information in the investment database into 4 general groups, including pan-tech (Industry = 1), general medicine (Industry = 2), general consumption (Industry = 3), and others (Industry = 4). We then created the ICI as follows:

$$ICI={\sum }_{i=1}^n{\left(X_i/Q\right)}^2$$

(1)

where $X_1,X_2,\cdots ,X_n$ represents the investment amount in each industry group and $Q={\sum }_{i=1}^n\left(X_i\right)$

(2) 

Geographic Concentration Index (GCI)

GCI is used to indicate the degree of geographic diversification level of the cross-border VC investments, and the larger the GCI, the lower the geographic diversification level is. Similar to the calculation of ICI, GCI is calculated as follows:

$$GCI={\sum }_{i=1}^n{\left(Y_i/Q\right)}^2$$

(2)

where $Y_1,Y_2,\cdots ,Y_n$ represents the investment in each geographic area and $Q={\sum }_{i=1}^n\left(Y_i\right)$.

In this study, we classify the geographic regions based on the economic development, as follows: western region (Region = 1), eastern region (Region = 2), central region (Region = 3), and northeastern region (Region = 4).

(3) 

Investment Stage Concentration Index (SCI)

The investment database classifies the investment stages into four levels, namely seed stage (Stage = 1), start-up stage (Stage = 2), expansion stage (Stage = 3), and mature stage (Stage = 4).

The stage concentration index indicates the degree of investment stage diversification of cross-border VC investments, and the higher the SCI value, the lower the diversification of investment stage and vice versa. When SCI equals 1, it means no diversification, i.e., all investment projects of venture capital institutions are concentrated in the same stage.

$$SCI={\sum }_{i=1}^n{\left(Z_i/Q\right)}^2$$

(3)

where $Z_1,Z_2,\cdots ,Z_n$ represent the investment in each investment stage and $Q={\sum }_{i=1}^n\left(Z_i\right)$, which represents the sum of investments in all stages.

(4) 

Investment Performance (Performance)

Among the methods to measure the performance of cross-border co-investments, the most common one is to measure whether venture capital institutions can successfully exit through IPOs or mergers and acquisitions (Cumming et al. 2016; Dai et al. 2012). We define the investment performance as the number of subsequent successful exits of the institution in the current period divided by the total investment events of the venture capital institution during the period.

$$Performance\left(i,t\right)={\displaystyle \frac{\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r} \mathrm{o}\mathrm{f} \mathrm{s}\mathrm{u}\mathrm{c}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{s}\mathrm{f}\mathrm{u}\mathrm{l} \mathrm{e}\mathrm{x}\mathrm{i}\mathrm{t}\mathrm{s} \mathrm{o}\mathrm{f} \mathrm{t}\mathrm{h}\mathrm{e} \mathrm{i}\mathrm{n}\mathrm{v}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t} \mathrm{o}\mathrm{f} \mathrm{V}\mathrm{C} \mathrm{i} \mathrm{i}\mathrm{n} \mathrm{p}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}\mathrm{d}(\mathrm{t}-2,\mathrm{t})}{NumberofalltheinvestmentsofVCiinperiod\left(\mathrm{t}-2,\mathrm{t}\right)}}$$

(4)

4.2.2. Independent Variables

The independent variables in this study are the measurement of the network position of cross-border venture capital institutions in the joint network. In summary, we used four indicators to measure venture capital network positions, including degree centrality, closeness centrality, betweenness centrality, and eigenvector centrality.

(1) 

Degree Centrality (Degree)

Degree centrality refers to the number of joint investments from a certain institution in the investment network during the measuring period. The higher the degree centrality, the higher the institution’s network centrality is, which we viewed as the higher the network position. The degree centrality is calculated as below:

$$\mathrm{D}\mathrm{e}\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}i,t={\sum }_j{X}_{ij,t}$$

(5)

where $X_{ij}=1$ indicates the existence of joint investment between institution i and institution j and 0 represents no joint investments between the two.

(2) 

Closeness Centrality (Closeness)

Closeness centrality refers to the ratio of the shortest distance between a certain node network and other nodes, which can be used to describe the quality of the joint investment network where venture capital institutions are located. The closer the venture capital institution is, the greater the closeness centrality and the higher the network position is. Variable “closeness” is calculated as follows:

$$Closenessi,t={\left[{\sum }_{j=1}^{g}d\left(n_i,n_j\right)\right]}^{-1}$$

(6)

where $d(n_i,n_j)$ represented the network distance between n_i and n_j.

(3) 

Betweenness Centrality

Betweenness centrality refers to the ability of a venture capital institution to serve as an intermediary to connect two other institutions. If a venture capital institution can serve as a medium to provide investment opportunities for other institutions, it means that this institution has a high network status. The betweenness centrality is calculated as follows:

$$Betweenness i,t={\sum }_{j<k}\left({\displaystyle \frac{g_{jk}\left(n_i\right)}{g_{ik}}}\right)$$

(7)

where $g_{ik}$ presents the number of paths from venture capital i to venture capital k, where $g_{ik}$(n_i) refers to the path from institution i to institution k through institution i.

(4) 

Eigenvector Centrality

Eigenvector centrality measures the influence of a node on the network. A larger eigenvector centrality means that the venture capital institution has joint investment relationships with more institutions:

Eigenvector centrality (Eig_cen)_i ∈ N is defined by:

λ(Eig_cen)_i = ∑ aji (Eig_cen)_i

(8)

• for all i ∈ N,
• where λ is a scalar and λ > 0.
• a_ji = 1 if there is connection between j and i and 0 otherwise.

4.2.3. Control Variables

Following the literature (Hochberg et al. 2007, etc.), we also created the control variables as follows:

(1) 

Network Scale

Network scale refers to the number of all nodes included in the network. In this study, we define this variable as the number of institutions participating in venture capital during the time window period.

(2) 

Network Density

Network density refers to the closeness between nodes in the network. It is an important indicator used to measure network cohesion and can reflect the overall structure of the network. Density is calculated by using the ratio of the actual number of relationships in the network to the maximum number of relationships that exist in the network. For a network of the same size, the more connections between nodes, the greater the density is.

(3) 

Network Average Distance

Network average distance (Average Distance) represents the average shortest distance that network members reach each other. The larger this indicator is, the larger the span between network nodes is, which means that the network cohesion is lower. The network average distance data in this article is calculated using Uicnet6.0.

(4) 

Cultural Distance

This variable is represented by the Hofstede index, which is used to control the cultural differences between China and the country the cross-border venture capital institutions belong to. This indicator uses five dimensions to measure the cultural differences between the two countries, including individualism (I); power distance (PD); uncertainty avoidance (UA); masculinity (M); and long-term orientation (LOI). The calculation formula is as follows:

$$\mathrm{H}\mathrm{o}\mathrm{f}\mathrm{s}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{e} \mathrm{C}\mathrm{u}\mathrm{l}\mathrm{t}\mathrm{r}\mathrm{u}\mathrm{e} \mathrm{D}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{c}\mathrm{e}={\displaystyle \frac{{\left({\sum _{i=1}^5(C_{x,i}-C_{y,i})}^2\right)}^{\frac{1}{2}}}{5}}$$

(9)

where C_i refers to the value of the five dimensions mentioned above, x and y refer to the two countries in the study.

(5) 

Age of the Institutions

The variable “Age” is defined as the establishment year of the institution minus the last year of the window period. If the difference is less than or equal to 0, it is recorded as 0.

(6) 

Investment Year

Investment year refers to the last year of the window period when the transaction occurs, which spans from 2003 to 2020. During the regression analysis, we created 18 dummy variables to represent the 18 investment years.

(7) 

Number of Years the Institution Has Been in China

This variable is defined as the difference between the year of the investment event and the year when the cross-border venture capital institution first invested in China. If the difference is less than 1 year or the institution has not entered China at the time of investment, it will be recorded as 0.

(8) 

VC Institution (Investor)

The variable “Investor” is created to indicate the ID of the VC institutions, which is coded from 1 to 167.

Table 2. Description of the variables. In this table, we present the variables in our analysis and the detailed descriptions.

	Variable	Symbol	Description
Dependent Variables	Industrial Concentration Index	ICI	The degree of industry concentration of the venture capital institutions’ investment, calculated according to the Formula (1).
	Geographical Concentration Index	GCI	The degree of geographic concentration of the venture capital institutions’ investment, calculated according to the Formula (2).
	Investment Stage Concentration Index	SCI	The degree of investment stage concentration of the venture capital institution’s investments, calculated by Formula (3).
	Investment Performance	performance	This variable is used to measure the performance of a venture capital institution, which is defined as the number of successful exits of invested companies divided by the total number of investments by venture capital institution i during the period. The detailed formula is provided by Formula (4).
Independent Variables	Degree Centrality	Degree	This variable refers to the number of direct connections between an individual and other individuals. Please refer to Formula (5) for detailed calculation.
	Closeness Centrality	Closeness	This variable is to measure the quality of the joint investment relationship in which a certain venture capital institution participates. The detailed calculation is presented in Formula (6).
	Betweenness Centrality	Betweenness	This variable measures the connection ability of a venture capital institution in the network. Formula (7) shows the detailed calculation.
	Eigenvector Centrality	Eig_cen	Measures the impact of nodes on the network, Formula (8) shows the detailed calculation.
Control Variable	Network Scale	Scale	This variable is defined as the number of institutions participating in venture capital during the window period.
	Network Density	Density	Measures the closeness of the relationship between nodes in the network.
	Network Average Distance	Aver_Dist	Represents the average shortest distance from all members in the network to other members.
	Cultural Distance	Cul_Dist	Refers to the Hofstede cultural distance index, the calculation is shown in Formula (9)
	Age of the VC Institutions	Inv_Age	Measures the age of the VC institutions when the investment is made, which is calculated as the difference between the year of the investment event and the year that the VC institution was established.
	Investment Year	Year	Variable to indicate the year of the investment, which is the last year of the window period.
	Number of Years in China	Ent_Age	The difference between the year of investment event and the year of first investment of the cross-border VC institutions in China.
	VC Institutions’ Identification	investor	The identification of the cross-border venture capital institution, coded from 1 to 167

provides a detailed summary of the variables and their descriptions.

After creating the variables, we used the following generalized model to investigate the impact of network position at year i − 1 on VCs’ investment performance or investment strategies for year i.

$$\mathrm{Y}\left(\mathrm{i}\right)=\mathsf{\alpha }+{\beta }_1{x}_1\left(i-1\right)+{\beta }_2{x}_2\left(i-1\right)+\dots +{\beta }_n{x}_n\left(i-1\right)+\epsilon \left(i-1\right)$$

(10)

In the analysis, we have four different dependent variables in the regressions, namely industry concentration index (ICI), geographic concentration index (GCI), investment stage concentration Index (SCI), and investment performance, and our explanatory variables are different measures of the VCs’ network positions, i.e., degree centrality, closeness centrality, betweenness centrality, and Engin vector centrality. We also included multiple control variables in our empirical tests, including age of the VC institution in investment, investor identification, investment year, etc.

Please note that we used panel data in this research to illustrate cross-sectional impacts of network position on the VC institutions’ investment strategies. “Year” is included as the fixed effect to control the time difference of the data. We are not using a dynamic GMM or system GMM model in our analysis as we are not investigating the evolvement of the investment strategies of the VC institution, and thus the lagged dependent variable is not included in our panel analysis.

5. Empirical Results

5.1. Descriptive Statistics

Table 3. Network statistics of cross-border VC institutions in each window period. In this table, we present the network statistics in each of the three-year rolling windows.

	Number of Nodes	Number of Connections	Network Density	Network Average Distance
2001–2003	169	406	0.0143	5.762
2002–2004	234	604	0.0111	5.349
2003–2005	344	834	0.0071	5.596
2004–2006	514	1274	0.0048	6.178
2005–2007	643	1632	0.004	6.345
2006–2008	717	1838	0.0036	6.327
2007–2009	659	1592	0.0037	6.61
2008–2010	691	1536	0.0032	8.042
2009–2011	794	1802	0.0029	6.995
2010–2012	900	2158	0.0027	6.928
2011–2013	919	2250	0.0027	6.556
2012–2014	1212	3110	0.0021	6.433
2013–2015	1893	5210	0.0015	6.313
2014–2016	2480	7000	0.0011	6.172
2015–2017	3168	8982	0.0009	6.209
2016–2018	3931	11,336	0.0007	6.432
2017–2019	4255	12,624	0.0007	6.714
2018–2020	4251	12,598	0.0007	6.625

shows the statistics of VC network in each of the three-year rolling time windows.

From Table 3, we see that the number of network nodes has increased from 169 in 2003 to 4251 in 2020, with an increase of nearly 25 times; the number of network relationships has also increased from 406 groups in 2003 to 12,598 groups in 2020, with an increase of nearly 30 times. As the network scale expands, network density decreases to a certain extent. On the contrary, with the expansion of network scale, the average network distance has only increased slightly, from 5.762 in 2003 to 6.625.

Table 4. Descriptive Statistics of Variables.

Variables	Number of Obs	Mean	Std	Min	Max
ICI	3006	0.717	0.393	0.003	1
GCI	3006	0.734	0.383	0.005	1
SCI	3006	0.738	0.378	0.006	1
Performance	3006	0.032	0.057	0	1
Degree	3006	2.858	3.771	0	25
Closeness	3006	0.089	0.2	0	2
Betweenness	3006	0.114	0.26	0	2.029
Eig_cen	3006	0.014	0.053	0	0.781
Scale	3006	1543	1386.726	169	4255
Density	3006	0.004	0.004	0.001	0.014
Aver_Dist	3006	6.421	0.57	5.349	8.042
Cul_Dist	3006	40.912	9.227	16.198	45.997
Inv_Age	3006	22.026	23.122	0	197
Year	3006	2011.5	5.189	2003	2020
Ent_age	3006	4.706	5.209	0	28
Investor	3006	84	48.216	1	167

shows the descriptive statistics of the variables.

From Table 4, we see that the average values of the industry concentration index (ICD), geographic concentration index (GCI), and stage concentration index are 0.717, 0.734, and 0.738 respectively, all exceeding 0.7, indicating that the cross-border VC’s investments are very concentrated, with low diversification. The average investment performance is 0.032, indicating that the investment exit ratio of the venture capital institutions is also low, which is common in the venture capital industry. The average network size of the entire network is 1543, and the standard deviation is 1386.726, indicating that there is a large gap in network size each year. The maximum value of cultural distance is 45.997 and the minimum value is 16.198, which shows that the cultural gap between the countries where the foreign venture capital institutions in the sample are headquartered and China is quite different. The maximum investment period of an investment institution is 197 years, the average is 22.026 years, and the standard deviation is 23.122. As for its investment period in China, the maximum is 28 years, the average is 4.706 years, and the standard deviation is 5.209. This indicates a significant variation in terms of the ages and investment periods of the cross-border VC institutions in China.

Table 5. Pearson Correlation Among Variables.

	ICI	GCI	SCI	Performance	Degree	Closeness	Betweenness	eig_cen	Scale	Density	aver_dist	cul_dist	inv_age	ent_age	Year	Investor
ICI	1.000
GCI	0.991 ***	1.000
SCI	0.985 ***	0.993 ***	1.000
performance	−0.603 ***	−0.573 ***	−0.565 ***	1.000
degree	−0.006	−0.012	−0.015	0.016	1.000
closeness	−0.471 ***	−0.465 ***	−0.454 ***	0.315 ***	0.198 ***	1.000
betweenness	−0.608 ***	−0.621 ***	−0.624 ***	0.274 ***	0.011	0.270 ***	1.000
eig_cen	−0.339 ***	−0.335 ***	−0.333 ***	0.172 ***	0.272 ***	0.517 ***	0.278 ***	1.000
scale	−0.186 ***	−0.177 ***	−0.178 ***	0.045 **	−0.133 ***	−0.038 **	0.111 ***	−0.006	1.000
density	0.188 ***	0.181 ***	0.178 ***	−0.088 ***	0.214 ***	0.011	−0.141 ***	−0.031 *	−0.648 ***	1.000
aver_dist	−0.090 ***	−0.085 ***	−0.084 ***	0.046 **	−0.005	−0.043 **	0.085 ***	0.032 *	0.145 ***	−0.542 ***	1.000
cul_dist	−0.005	−0.007	−0.006	−0.006	0.067 ***	0.022	0.036 **	0.028	0.000	−0.000	−0.000	1.000
inv_age	−0.122 ***	−0.117 ***	−0.116 ***	0.062 ***	0.006	0.070 ***	0.064 ***	0.140 ***	0.195 ***	−0.177 ***	0.075 ***	0.070 ***	1.000
ent_age	−0.249 ***	−0.246 ***	−0.244 ***	0.051 ***	−0.048 ***	0.164 ***	0.222 ***	0.227 ***	0.648 ***	−0.514 ***	0.185 ***	0.016	0.278 ***	1.000
year	−0.206 ***	−0.197 ***	−0.195 ***	0.067 ***	−0.170 ***	−0.031 *	0.134 ***	0.006	0.903 ***	−0.841 ***	0.369 ***	0.000	0.214 ***	0.672 ***	1.000
investor	0.368 ***	0.359 ***	0.359 ***	−0.236 ***	0.114 ***	−0.041 **	−0.245 ***	−0.031 *	0.000	−0.000	−0.000	0.136 ***	−0.197 ***	−0.067 ***	0.000	1.000

Note: *, ** and *** represent significance level at 10%, 5%, and 1% respectively.

shows the Pearson correlation among our variables, and

Table 6. VIF value of explanatory variables.

Variables	VIF	1/VIF
degree	1.20	0.833114
closeness	1.54	0.648328
betweenness	1.26	0.793291
performance	1.23	0.814041
eig_cen	1.55	0.644170
scale	7.94	0.125952
density	5.14	0.194703
aver_dist	1.69	0.590976
cul_dist	1.04	0.963956
ent_age	2.19	0.456057
inv_age	1.15	0.866940
year	15.47	0.064655
investor	1.21	0.824640

presents the VIF value for all the independent variables.

From Table 5, we see all three concentration indexes (ICI, GCI, and SCI) are significantly and positively correlated, and they all have a significant negative correlation with performance. This implies that the less diversification (higher concentration level) is correlated with the worse investment performance. We also see that the variables “Closeness”, “Betweenness” and “Engin_vector” are all related to the three concentration indexes in a negative and significant way but correlated with performance positively. The variable “Degree”, however, does not show a significant correlation with the four dependent variables.

In Table 6, we show the VIF value of the explanatory variables in our regression analysis. Except for the variable “year”, the VIF value for all other variables are all smaller than 10, thus indicating that we have a very low multi-collinearity problem in our regression analysis, and the variable “year” is used as a fixed effect in our panel data analysis.

5.2. Regression Analysis

Following the previous research, we implement three different methods to control the endogeneity problem. First, we chose the fixed effects models for empirical analysis as they can eliminate endogeneity problems caused by time-varying factors and some of the effects from missing variables. Second, we construct a three-year window period, and the sample is rolled through the window period to reduce its endogeneity. Third, all the explanatory variables are lagged by one period to reduce the endogeneity problems as well.

Table 7. Regression results of cross-border VC network location variables and ICI. In this regression, we investigate the impacts of network location variables on the dependent variable ICI. We include four main network locations measures: degree, closeness, betweenness, and eig_cen as our explanatory variables in the regression, and our control variables include scale, density, aver_dist, inv_age, etc. For detailed explanations of variables, please refer to Table 2.

ICI	1	2	3	4
degree	−0.00574 ***
	(−3.353)
closeness		−0.822 ***
		(−21.70)
betweenness		−0.602 ***
			(−27.47)
eig_cen				−1.089 ***
				(−6.677)
scale	−3.02 × 10−5 ***	−3.55 × 10−5 ***	−2.66 × 10−5 ***	−3.08 × 10−5 ***
	(−2.791)	(−3.540)	(−2.764)	(−2.865)
density	3.021	−1.570	−1.989	−0.183
	(0.898)	(−0.512)	(−0.676)	(−0.0555)
aver_dist	0.00415	−0.0254 **	0.00295	−0.00137
	(0.344)	(−2.295)	(0.279)	(−0.116)
inv_age	−0.0133 *	0.00112	−0.0174 ***	−0.0126 *
	(−1.934)	(0.175)	(−2.855)	(−1.854)
ent_age	0.0385 ***	0.0260 ***	0.0347 ***	0.0372 ***
	(12.49)	(8.920)	(12.61)	(12.12)
year	−0.0208 ***	−0.0265 ***	−0.0132 **	−0.0212 ***
	(−2.794)	(−3.844)	(−1.989)	(−2.873)
constant	42.66 ***	54.26 ***	27.56 **	43.62 ***
	(2.877)	(3.941)	(2.086)	(2.959)
Adj-R2	0.127	0.249	0.308	0.137
F statistics	13.86	10.99	8.548	11.22
p value	0	0	0	0

Note: the numbers in the parentheses are t values and *, **, and *** represent the significance level at 10%, 5%, and 1% respectively.

,

Table 8. Regression results of cross-border VC network location and GCI. In this regression, we investigate the impacts of network location variables on the dependent variable GCI. We include four main network location measures: degree, closeness, betweenness, and eig_cen as our explanatory variables in the regression, and our control variables include scale, density, aver_dist, inv_age, etc. For detailed explanations of variables, please refer to Table 2.

GCI	5	6	7	8
degree	−0.00588 ***
	(−3.469)
closeness		−0.782 ***
		(−20.68)
betweenness		−0.627 ***
			(−29.24)
eig_cen				−1.020 ***
				(−6.305)
scale	−2.77 × 10−5 ***	−3.29 × 10−5 ***	−2.40 × 10−5 **	−2.84 × 10−5 ***
	(−2.586)	(−3.285)	(−2.547)	(−2.664)
density	4.005	−0.520	−1.175	0.815
	(1.201)	(−0.170)	(−0.408)	(0.250)
aver_dist	0.00653	-0.0221 **	0.00538	0.000757
	(0.547)	(-1.998)	(0.521)	(0.0647)
inv_age	−0.0118 *	0.00183	−0.0162 ***	−0.0113 *
	(−1.738)	(0.286)	(−2.705)	(−1.667)
ent_age	0.0350 ***	0.0232 ***	0.0310 ***	0.0339 ***
	(11.47)	(7.968)	(11.55)	(11.11)
year	−0.0189 **	−0.0243 ***	−0.0110 *	−0.0193 ***
	(−2.563)	(-3.533)	(−1.697)	(−2.630)
Constant	38.84 ***	49.82 ***	23.13 *	39.68 ***
	(2.643)	(3.627)	(1.791)	(2.713)
Adj-R2	0.113	0.226	0.316	0.122
F Statistics	13.04	10.15	7.921	10.52
p value	0	0	0	0

Note: the numbers in the parentheses are t values, and *, **, and *** represent the significance level at 10%, 5%, and 1%, respectively.

,

Table 9. Regression results of cross-border VC network location and SCI. In this regression, we investigate the impacts of network location variables on the dependent variable SCI. We include four main network locations measures: degree, closeness, betweenness, and eig_cen as our explanatory variables in the regression, and our control variables include scale, density, aver_dist, inv_age, etc. For detailed explanations of variables, please refer to Table 2.

SCI	9	10	11	12
degree	−0.00616 ***
	(−3.667)
closeness		−0.739 ***
		(−19.57)
betweenness		−0.628 ***
			(−29.69)
eig_cen				−1.053 ***
				(−6.572)
scale	−3.26 × 10−5 ***	−3.77 × 10−5 ***	−2.90 × 10−5 ***	−3.34 × 10−5 ***
	(−3.075)	(−3.772)	(−3.117)	(−3.162)
density	3.635	−0.867	−1.659	0.307
	(1.100)	(−0.284)	(−0.584)	(0.0949)
aver_dist	0.00355	−0.0242 **	0.00209	−0.00251
	(0.301)	(−2.194)	(0.205)	(−0.217)
inv_age	−0.0120 *	0.000881	−0.0163 ***	−0.0114 *
	(−1.775)	(0.138)	(−2.769)	(−1.702)
ent_age	0.0352 ***	0.0241 ***	0.0312 ***	0.0340 ***
	(11.63)	(8.281)	(11.76)	(11.27)
year	−0.0175 **	−0.0226 ***	−0.00954	−0.0179 **
	(−2.393)	(−3.285)	(−1.491)	(−2.461)
Constant	36.02 **	46.34 ***	20.25	36.89 **
	(2.474)	(3.380)	(1.587)	(2.546)
Adj-R2	0.113	0.215	0.320	0.122
F Statistics	12.81	9.926	7.758	10.38
p value	0	0	0	0

Note: the numbers in the parentheses are t values, and *, **, and *** represent the significance level at 10%, 5%, and 1%, respectively.

and

Table 10. Regression results of cross-border VC network location on the cross-border VC institutions’ performance. In this regression, we investigated the impacts of network locations variable on the dependent variable performance. In this regression, we not only include the four main network location measures: degree, closeness, betweenness, and eig_cenm, but also their squared formats as our explanatory variables, as we hypothesize that the impact of network locations might not impact the cross-border VCs’ performance in a linear way, and our control variables include scale, density, aver_dist, inv_age, etc. For detailed explanations of variables, please refer to Table 2.

Performance	13	14	15	16
degree	0.00158 **
	(2.165)
degree2	−4.24 × 10−5
	(−1.015)
closeness		0.282 ***
		(21.04)
closeness2		−0.141 ***
		(−15.93)
betweenness		0.0971 ***
			(8.925)
betweenness2			−0.0680 ***
			(−7.156)
eig_cen				0.218 ***
				(3.991)
eig_cen2				−0.299 ***
				(−2.999)
scale	2.53 × 10−6	4.22 × 10−6 **	2.29 × 10−6	2.78 × 10−6
	(1.309)	(2.345)	(1.201)	(1.440)
density	−1.022 *	0.319	−0.317	−0.448
	(−1.648)	(0.579)	(−0.543)	(−0.758)
aver_dist	−0.00309	0.00271	−0.00147	−0.00182
	(−1.421)	(1.363)	(−0.699)	(−0.859)
inv_age	0.00206 *	0.000699	0.00213 *	0.00194
	(1.673)	(0.609)	(1.759)	(1.582)
ent_age	−0.00762 ***	−0.00415 ***	−0.00706 ***	−0.00745 ***
	(−13.80)	(−7.719)	(−12.87)	(−13.44)
year	0.00298 **	0.00193	0.00247 *	0.00301 **
	(2.232)	(1.550)	(1.881)	(2.264)
Constant	41.47 ***	31.46 **	22.23 *	42.22 ***
	(2.792)	(2.420)	(1.778)	(2.890)
Adj-R2	0.078	0.204	0.102	0.080
F Statistics	4.514	1.465	2.482	3.514
p value	0	0	0	0

Note: the numbers in the parentheses are t values, and *, **, and *** represent the significance level at 10%, 5%, and 1%, respectively.

show the regression results.

To reduce the multicollinearity problem among the control variables, we eliminated two control variables cul_dist and the investor in our regression analysis.

In the regression model 1 to 4 on Table 7, we studied the impact of network location variables on the industry diversification levels of cross-border venture capital investments. We see that all four network location variables have negative coefficients on ICI and are significant at the 1% level. These results indicate that network location variables have positive impacts on the industry diversification of cross-border venture capital institutions’ investment strategies in China (the lower the IDI, the higher the investment with industry diversification). This supports our Hypothesis 1, i.e., the higher the network position of a cross-border venture capital institution, the more industry-diversified their investments will be.

We also see that many control variables, such as network size, VC establishment years, etc., all have significant coefficients, which implies that they significantly impact the VC institution’s ICI value, while other variables, such as network density and average network distance, on the contrary, are not significant. For example, variable “scale” has a negative and significant coefficient, which implies that larger VC institutions tend to be more diversified in industries (less ICI). The variable ent_age, on the contrary, has a positive and significant coefficient, which implies that VC institutions with more investment in China (larger Ent_age) tend to be less diversified in industries. We argue that as the investment experiences grow in China, the cross-border VC institutions have more confidence to become specialized in industries and are thus less diversified.

In Table 8, we use models 5 to 8 to investigate the impact of network locations on the VC companies’ geographic concentration index. We see that all four network location variables have negative and significant (at 1% level) coefficients. This implies that the better the network location of the cross-border VC institutions, the higher the geographic diversification level of their investments (the lower the GDI index), which supports Hypothesis 2.

Table 9 shows that all the network location variables have a negative and significant (at 1% level) coefficients on SCI. This indicates that the higher the network positions of the cross-border venture institutions, the higher the investment stage diversification level of their investments (the lower the SCI value), which is consistent with our Hypothesis 3.

As we see from Table 10, all four network location variables had positive and significant (at 1% level) coefficients for the first moment, but significant and negative coefficients for the second moment (squared variables). This implies the impact of network locations on the cross-border VC institutions’ performance might not be linear, but rather inverted U-shaped, which confirmed our Hypothesis 4.

5.3. Robustness Check

The following two methods were implemented for a robustness check: (1) Winsorize the variables at both 1% and 99% levels; (2) implement the random effect model for panel data analysis. Results are consistent with our previous regression results and available upon request.

6. Conclusions

In this paper, we investigated whether the network positions that a cross-border VC institution has in the network will have an impact on their investment strategies and investment performance, including industry diversification level, geographic diversification level, and investment stage diversification levels. Our empirical results show that all four network measures, including degree centrality, closeness centrality, betweenness centrality, and Engin vectors, have significant and positive impacts on the cross-border VC institutions’ diversification levels (lower concentration index). We also found that the network position does not have a linear impact on the cross-border VC institutions’ performance, and the impact is, rather, inverted U-shaped. A higher network position of the VC institutions will first help increase their investment returns within a certain range, but when their network status exceeds the threshold, investment capabilities will decline due to the high complexity of the information involved and the high level of investment dispersion.

Our research contributes to the literature in many perspectives, as follows:

(1)

Our research provides a systematic and detailed analysis of how the cross-border VC’s network positions impact their investment strategies and performance, which provides more evidence in the literature regarding the impact of network location of cross-border venture capital institutions in their host countries on their investment strategies and investment performance.

(2)

Our research provides significant guidance to the cross-border venture capital institutions in that they need to recognize the agency costs caused by cultural differences to reasonably establish a sound venture capital network in the host country. To reduce agency costs and maintain competitiveness in the host country, cross-border venture capital institutions should try to benefit from VC networks with local investments. Establishing local offices and setting up RMB funds are two of the most useful strategies (Yang et al. 2023).

(3)

Our research also provides guidance to the local government in terms of promoting development of the VC industry. For example, the government should be aware of the geographical restrictions of cross-border venture capital institutions, i.e., not only promote regional development in Beijing, Shanghai, and other more economically developed areas, but should also continue to attract cross-border venture capital investment in other geographic areas of China, especially remote areas such as the central and western regions. Government guidance funds can also be established, and joint venture investments can be conducted through local venture capital institutions as well, to promote the VC industry development in the host country.

In the future, we are planning to extend our research in the following areas:

• In this paper, we studied the impacts of four network position measures on three investment strategies of cross-border VC institutions (ICI, GCI, SCI), and in the future, we are planning to expand the investigations to more network measurements and more perspectives of the investment strategies of the VC institutions.
• VC networks involve three objects: cross-border venture capital institutions, local venture capital institutions, and invested companies. In this paper, we only investigated the impacts of network position from the VC’s perspectives, and for future research, we can extend our analysis to the perspectives of the local investment institutions or even invested companies.