Examining Diverse Investors in the Clean Energy and Environmental Technology Sector: A Network Analysis from Japan

Abstract

Startups in the clean energy and environmental technology (CEET) sector can develop sustainable innovations, but mobilizing private finance has been difficult. As the venture capital (VC) investment model was found to be not well-suited for the CEET startups, diverse types of investors have received more attention. However, since previous studies have been dominated by a VC-centric perspective in the US and have overlooked collaborative relationships, the roles of various CEET investors have not been systematically analyzed. This study aims to analyze the diverse investors in the CEET investor network formed through co-investment syndication, using Japan as an underexplored regional context. Based on Japan’s comprehensive data from 2008 to 2022, this study examines the evolution, structure, and communities of the network. The analysis identified the development stages of the investor network: the formation stage (2008–2012), the expansion and diversification stage (2013–2017), and the stable growth stage (2018–2022). The results confirmed the strong influence of VCs, while a community analysis suggested the bridging role of governmental venture capital. The findings based on the CEET investor network contribute to expanding both the theoretical understanding and practical implications for overcoming the financing difficulties of CEET startups to address the climate change crisis.

1. Introduction

To accelerate the rapid diffusion of decarbonization technologies aimed at achieving the net-zero goal, startups in the clean energy and environmental technology (CEET) sector can play significant roles. These CEET startups are expected to develop sustainable innovations that improve sustainability performance [1] by reconciling economic, environmental, and social goals [2]. However, mobilizing private finance for the CEET sector has historically been difficult due to its specific technological and market barriers [3,4,5]. Nevertheless, while the IPCC’s discussions have focused on macro-finance for mature technology transfer [6], the issue of the serious funding gap for early-stage CEET has not been fully addressed. To understand the financing mechanisms that foster a sustainable innovation ecosystem, this study focuses on startup investors in the CEET sector.

A fundamental mismatch between the CEET and the venture capital (VC) investment model was identified because of technological features and the structure of VC funds [7]. As this issue became more widely recognized, more attention has been paid to non-VC investors as alternative sources of funding, such as patient capital [8] and governmental venture capital (GVC) [9]. However, empirical knowledge of these diverse investors has remained limited. Because previous studies have been concentrated on the US market [1,7,10,11,12], the VC-centric perspective originating in Silicon Valley [13] has tended to be generalized beyond its regional context. Understanding diverse investors in novel regional contexts expands the theoretical space for securing funding for the CEET sector.

Furthermore, co-investment syndication is particularly important when it is difficult for a single investor to take risks and provide sufficient capital [14,15], as is typical of deals in the CEET sector. Although diverse investors cooperate to address the CEET funding gap, previous studies tend to focus on investment performance and individual investors. There is a need to place these investors in the context of collaborative relationships. Therefore, this study uses social network analysis (SNA) [16,17] to understand the role of diverse investors in the CEET investor network.

We selected Japan as a unique regional case to complement and enrich previous research. Japan has established innovation capabilities, as evidenced by its leading position in the number of inventions in the manufacturing and CEET sectors [18]. This has fostered the diversity of investors, including non-VC investors whose investment amounts are comparable to those of VCs [19].

To examine the network evolution, structure, and roles of the diverse investors, we use SNA and a comprehensive dataset of CEET investments in Japan. The remainder of this study is organized as follows: Section 2 provides a literature review on CEET investors. Section 3 outlines the datasets and methodological approach. Section 4 presents the analytical results, including (1) the historical evolution of the investor networks, (2) the roles of various investor types, and (3) the characteristics of governmental venture capital (GVC). In Section 5, based on the results, we discuss the findings and draw practical implications. By advancing knowledge of the diverse investors in the CEET sector, this research contributes to developing funding strategies to support sustainable innovation.

2. Literature Review

The main challenge in the CEET sector is to secure stable private investment to address the barriers [5]. While private investors usually have a maximum time horizon of 12 years [20], energy technologies tend to have a long research and development (R&D) period and a large capital investment [4]. Because of these factors, investors failed to realize the expected returns during the previous clean tech bubble of 2006–2011, and as the capital cycle stagnated, VCs withdrew from the clean energy market [12,21]. A similar bubble-bust pattern was observed in the recent climate tech boom during and after the post-COVID economic recovery [22,23]. Fluctuations in the CEET investment cycle have led to changes in the composition and relationships among CEET investors. However, previous research has not fully examined how the investor network evolves over this temporal transition.

As explained in the introduction, current research has highlighted the diversity of the financial system in the CEET sector to improve stability and resilience [24,25]. The CEET investors in each stage of sustainable innovation are typically categorized as follows [26,27]: grants and strategic partnerships in the basic research stage, accelerators and angel investors in the applied R&D stage, VCs in the early growth stage, and private equity funds and conventional finance in the late growth stage. GVCs have recently attracted attention due to their important capacity to support the CEET sector [28,29] through direct investments in startups and indirect investments in VCs [30]. However, previous research has failed to consider how different types of investors with distinct investment strategies are involved in financing the CEET sector.

Syndication is a means of sharing and reducing the investment risk and complementing resources [15]. Syndication forms social networks among investors, promoting risk sharing, knowledge exchange, and value creation [14,31]. While syndication research began as a part of the financial study on investor behavior, SNA was used to deepen the understanding from a network perspective [32,33,34]. Recently, the understanding of the structural properties and roles of VC networks has been expanded by the development of the SNA methodology [35,36,37,38,39]. However, the CEET sector has not been specifically investigated in most studies, with the exception of a study on renewable energy [40]. Furthermore, previous studies have tended to overlook the diversity of investors within investor networks. Given the need to pay more attention to non-VC investors, especially in CEET, understanding investor diversity is beneficial.

Based on the literature review, this study sets the following research directions. First, we aim to examine how the CEET investor network has evolved. This study provides insights into the dynamics of investor networks over time, given the rapidly changing market and policy environment in the CEET sector. Second, we aim to analyze the characteristics of diverse types of investors in the CEET investor network. This study quantifies the role and influence of VC and non-VC investors in the CEET sector. Third, we aim to understand the specific role of GVCs as the sole public investment actor. A deeper understanding of GVCs also leads to useful insights for policymakers on how to foster startup ecosystems and stimulate other private investors.

3. Materials and Methods

3.1. Research Context

This section reviews the research context for Japan to analyze how the CEET startup ecosystem has evolved through the involvement of diverse investors. The institutional framework has encouraged companies and financial institutions to enter the energy market, driven by policy changes [41]. The liberalization of regulated energy markets started in 1995 and was completed in 2016 [42]. This deregulation of the utility market transformed the traditional, regulated energy system into a more favorable market for new business entrants [43]. This trend has led to the rise of startups and their investors, allowing smaller companies to enter the previously regulated market. After the Fukushima nuclear accident in 2011, the Feed-in Tariff (FIT) started to encourage the installation of renewable energy in 2012 [44]. These drastic market changes increased the need for energy companies to create new businesses and fueled the trend of creating CVCs [45]. Next, an important policy was the Green Transformation (GX) initiative, which comprehensively supports the transition of large corporations and promotes startup innovations [46]. This growing GX policy momentum has expanded the investment activity of CVCs and financial institutions in decarbonization technologies.

Next, we review the context of Japan’s overall startup ecosystem. According to historical investment data, the total startup investment was 73.7 billion yen, and 896 startups received investment in 2009 [47]. The total startup investment increased drastically to 877.4 billion yen, and the number of funded startups rose to 2224 in 2022 [48]. This rapid growth can be attributed to the recent strengthening of startup policies. The Japanese government introduced the Startup Development Five-Year Plan in 2022 [49]. This plan stated the ambitious goal of increasing the amount of startup investment to more than 10 times the current level (to 10 trillion yen) in five years (by 2027). This policy trend has also stimulated the activities of GVCs. Since these policies target startup investors, the investment environment has improved dramatically. The following section explains the dataset and methodology.

3.2. Dataset and Data Cleaning

In this study, we constructed a dataset using the following steps. The procedure is consistent with the approach taken in previous studies of cleantech startups [11,50,51]. First, we selected primary data sources. After carefully considering several databases, we selected the Japanese data providers Speeda Startup Information Research [52] and Startup DB [53] in terms of quality and coverage, considering the language barriers. These databases and their published reports have been used in government reports and research articles on Japanese startups [49,54,55]. Speeda Startup Information Research was last updated on 30 October 2023, investor information was last updated on 27 February 2024, and deal information was last updated on 26 March 2024. Startup DB was last updated on 17 April 2023. In the following process, Speeda Startup Information Research was used as the primary data and Startup DB as supplemental data. To the best of our knowledge, these databases provide the most comprehensive and accurate data on Japan’s CEET startups.

Second, we created a dataset of deals and investors using the following methods. We first created lists of target startups. To compare the characteristics of deep tech, we selected drug discovery as a capital-intensive and R&D-intensive control group and artificial intelligence (AI) as a software-based and asset-light control group. Based on the specifications in the startup ecosystems [19,23,56], we identified startups in the three sectors using the tag information of startups. Based on this list of startups, we constructed a dataset of their investment deals and identified the investors involved in these deals. This dataset did not include other forms of financing, such as grants and subsidies, bank debt, project finance, and social lending.

Third, we categorized the types of investors in this study. Based on the descriptions of Japan’s startup ecosystem [19,48], we categorized them as follows: (1) VC, investment funds that focus mainly on investing for financial returns; (2) corporate venture capital (CVC), business operating companies and their investment vehicles; (3) GVC, investment arms of the government, ministries, and local governments; (4) Other Financials, banks, securities companies, insurance companies, and non-VC financial investors. The “Others” type consists of all investors not included in the above types, such as business angels, consulting firms, non-profit organizations, etc. In total, we identified 525 CEET investors across five categories (VC: 119, CVC: 164, GVC: 11, Other Financials: 37, Others: 194). Finally, we manually verified the datasets with publicly available information, such as government reports, web articles, company websites, and other databases.

The annual investment amounts and deals in the CEET sector are shown in Figure 1. The investment activity was relatively low in the 2000s and began to rise in 2015, peaking in 2017 and growing steadily since then. To ensure sufficient reliability in collecting accurate data, the aggregation period was set to 31 December 2022, providing a sufficient sample size for a detailed network analysis. Therefore, this analysis mainly focuses on the period from 2008 to 2022.

3.3. Social Network Analysis

This study analyzes the CEET investor network using SNA, primarily with Python 3.9.13 and its NetworkX library. SNA is a methodological framework for analyzing complex network structures and relationship patterns [16,17], which is appropriate for the objective of this study to understand the roles of diverse investors based on the investor syndication network.

The investor networks in this study were based on the syndication relationships of each investment deal. This definition did not include relationships with prior shareholders. Nodes represent investors, and edges represent syndication relationships between investors. To gain a better understanding of network structures, each edge was weighted based on the number of syndications, and isolated nodes were removed.

To understand the temporal changes in the structural characteristics of the investor networks, we calculated network metrics. The annual number of new investors was calculated as the number of investors who made their first investment in the target sectors in each year, and the continuing investor rate was calculated using the Jaccard index, both starting from 2000.

When calculating network metrics and visualizing graphs, investor networks were weighted by the number of syndications. The network metrics were calculated using rolling five-year time windows. In addition, sensitivity analyses of this setting were conducted by changing the time windows to 4 and 6 years to check the consistency of observations.

The edge density was calculated as the ratio of the total weight of edges between nodes to the maximum possible number of edges.

$$D={\displaystyle \frac{2E}{N(N-1)}}$$

E is the total edge weight, and N is the number of nodes.

The clustering coefficient indicates the degree to which neighboring nodes are linked to each other, taking edge weights into account [57]. The average clustering coefficient is calculated as follows:

$$<C>={\displaystyle \frac{1}{N}}\sum _{i=1}^N{C}_i^w$$

The average path length is the distance between any two nodes. Here, we use the largest connected component of the networks.

$$<d>={\displaystyle \frac{1}{N(N-1)}}\sum _{i\ne j}d(i,j)$$

$d(i,j)$ is the weighted shortest path length between nodes i and j. To calculate the average path length, the inverse of the edge weight is used as the distance in the calculation, since more syndication means closer distances between investors.

We calculated the small-world-ness S [58], by comparing the observed network with multiple random networks that preserve the same numbers of nodes and edges. Small-worldness is characterized by high clustering coefficients and short path lengths, describing networks where the distances between randomly selected nodes are small [16,59].

To understand the roles of different investors in the investor networks, we calculated the scale-free parameter α and centrality measures for each investor type. The value of α falls in the range 2 < α < 3 in typical scale-free networks [60]. In such networks, a small number of hub nodes have highly concentrated edges.

Centrality measures were calculated to understand the roles of the different investors. Firstly, to select the centrality measures for our purpose, several candidate measures were selected [16,61,62]. Secondly, the centrality measures were calculated for the CEET investor network (2008–2022), and a correlation matrix was generated, followed by hierarchical clustering. Based on the results of this preliminary analysis, we identified the degree centrality, betweenness centrality, closeness centrality, and eigenvector centrality as the representative measures for this investor network.

Third, we calculated centrality measures [63], accounting for the edge weights based on the number of syndications. The weighted degree centrality is calculated by expanding $\sum _{i=1}^n{a}_{ix}$ and summing the edge weights with adjacent nodes. Investors with a high weighted degree centrality can have strong direct relationships with other investors.

Betweenness centrality is based on

$$\sum _{i=1;i\ne x}^n\sum _{j=i+1;j\ne x}^n{\displaystyle \frac{g_{ij}\left(x\right)}{g_{ij}}}$$

$g_{ij}$ is the number of shortest paths from node i to j, and $g_{ij}\left(x\right)$ represents the number of paths that pass through node x. Calculations of betweenness and closeness centrality use the inverse of the edge weight as the distance. Nodes with high betweenness centrality can be important connectors.

Closeness centrality is calculated from

$${\displaystyle \frac{1}{{\sum }_{i=1}^n{d}_G(x, i)}}$$

Here, $d_G(x, i)$ is the shortest distance from node x to node i, considering the edge weight. Nodes with high closeness centrality can efficiently gather information and interact with each other.

Eigenvector centrality is a measure that extends degree centrality by considering the qualitative influence of its neighbor nodes. This is calculated based on

$${\displaystyle \frac{1}{{\lambda }_{max}\left(A\right)}}\sum _{j=1}^n{a}_{jx}{v}_j$$

where $a_{jx}$ represents the edge weight from node j to x, and v is the eigenvector corresponding to the maximum eigenvalue ${\lambda }_{max}\left(A\right)$ of the adjacency matrix A. Investors with high eigenvector centrality can be connected to other influential investors.

3.4. Community Analysis

By analyzing community structures of networks [64], we gain a deeper insight into the role of GVCs. The Louvain method [65] was used to identify the community structure. First, the Louvain method was applied to the CEET investor network (2008–2022) with the resolution parameter set to the default value of 1.0, and the result with the highest modularity was used. To ensure the consistency of community detection, we checked the Normalized Mutual Information [66].

Second, in order to quantitatively assess the role of each type of investor in the community structures, this study used two indicators. To assess the bridging capabilities between communities in the investor networks, we developed a novel metric: the cross-community bridging score. The cross-community bridging score $B_{i,c}$ for an investor type i in a community $C$ is defined in the following equation. It is calculated per community and per investor type.

$$B_{i,c}={\displaystyle \frac{1}{\left|V_{i,c}\right|}}\sum _{v\in V_{i,c}}\left|\right(v,u)\in E: u\in V\backslash C_c|$$

There is a set of nodes $V_{i,c}$ in a specific investor type i and a specific community c. For each node $v\in V_{i,c}$, we count the edges $(v,u)$ that point to nodes u outside the given community c. We sum these counts overall $v\in V_{i,c}$ and divide them by the number of nodes |$v\in V_{i,c}$|. This cross-community bridging score is calculated for one community and one investor type.

The participation coefficient was also used in this study, as a measure of how well the edges of the node are distributed across different communities [67]. A higher participation coefficient indicates that the edges are more evenly distributed across multiple communities. Nodes with high participation coefficients are associated with inter-community fluxes. These two indicators offer complementary perspectives; while the cross-community bridging score focused on the degree to which nodes of a given investor type within a given community were connected to other communities, the participation coefficient focused on the structural characteristics of the node layers as a macro understanding of the dispersion of edges across communities. These indicators offer a comprehensive understanding of the role of GVCs in the community structures within the network.

4. Results

4.1. Historical Evolution of the CEET Investor Network

In this section, we analyzed the historical evolution of the CEET investor network. The growth of the network was analyzed in terms of the annual number of new investors and the continuing investor rate (Figure 2). The annual number of new investors was low until 2012 and increased between 2013 and 2018. The continuing investor rate decreased between 2013 and 2017 and increased in 2018. Both declined through 2020 before rising again.

The results of the network metrics are shown in Figure 3. The edge density decreased continuously during the observation period, indicating that the network became decentralized. The average path length remained stable until 2013, increased until 2018, and entered a phase of growth after 2019. The average clustering coefficient remained relatively stable until 2016 and then declined until 2019. These observations suggest high investor turnover due to the continuous entry and exit of investors in the CEET sector.

The small-world-ness S of the CEET sector remained relatively stable from 2013 to 2017, before increasing rapidly after 2019 (Figure 4). Looking at the other sectors, the small-world-ness S of the drug discovery sector has been growing since the mid-2010s, and the AI sector has been growing since the late 2010s, indicating general trends in the Japanese startup ecosystem.

Furthermore, to confirm the consistency of the observations, sensitivity analysis was conducted by changing the time window from five years to four and six years (Supplementary Figure S1). The network metrics showed consistent main patterns. For the small-world-ness S, although a decrease was observed in 2022 in the four-year time window, the main pattern was consistent.

The above discussion is supported by the three-window CEET investor network visualization (Figure 5). These temporal changes suggest the increasing maturity and complexity of the CEET investor network, from an early, closely connected network to a more decentralized network with many investors. In the next section, we analyze the role of various investors.

4.2. The Characteristics of the Diverse Investors

In this section, we aim to understand the characteristics of the CEET investor network and the diverse investors there. To examine the interconnections among the diverse investors, we analyzed the basic structural properties of the investor networks. The results showed a decreasing degree distribution from the upper left (a) and a scattered and increasing degree correlation distribution from the lower left (b) (Figure 6). Across all the tech sectors, we observed a hierarchical structure consisting of a small number of highly connected nodes and a large number of small-degree nodes.

The scale-free parameter α was estimated for a detailed understanding of the structural characteristics (

Table 1. Estimation results of the scale-free parameter α in the tech sectors.

Tech Sector	Number of Nodes	Number of Edges	α Estimate (Scale-Free Parameter)	95% Confidence Interval for α	Mean KS Statistic	Mean p-Value	Optimal Xmin
CEET	525	1716	2.941	[2.460, 3.778]	0.870	p < 0.001	14.0
Drug Discovery	250	1035	2.767	[2.094, 4.507]	0.924	p < 0.001	30.0
AI	303	891	2.641	[2.126, 4.248]	0.650	p < 0.001	5.0

). The results revealed that the α values for all three sectors (CEET (2.94), drug discovery (2.77), and AI (2.64)) were within the range of typical scale-free networks (2–3), although they deviated from the theoretical power-law distribution in the Kolmogorov–Smirnov test. In particular, the number of high-degree nodes was smaller than the number predicted by the power-law model.

Next, we focus on the CEET investor network to understand the roles of the diverse investors related to these network characteristics. We calculated the centrality measures for each investor type (

Table 2. Descriptive statistics of centrality measures by investor type.

Investor Types	VC	CVC	GVC	Other Financials	Others
Degree Centrality
Mean	16.395	5.530	7.091	7.649	4.758
Median	8.000	4.000	6.000	6.000	3.000
Min	1.000	1.000	1.000	1.000	1.000
Max	104.000	22.000	21.000	24.000	38.000
Betweenness Centrality
Mean	0.010	0.001	0.002	0.001	0.001
Median	0.000	0.000	0.000	0.000	0.000
Min	0.000	0.000	0.000	0.000	0.000
Max	0.138	0.022	0.008	0.014	0.017
Closeness Centrality
Mean	0.312	0.210	0.216	0.236	0.167
Median	0.308	0.262	0.267	0.266	0.225
Min	0.004	0.002	0.002	0.002	0.002
Max	0.504	0.397	0.398	0.414	0.430
Eigenvector Centrality
Mean	0.041	0.007	0.014	0.012	0.007
Median	0.013	0.002	0.001	0.002	0.001
Min	0.000	0.000	0.000	0.000	0.000
Max	0.427	0.067	0.088	0.088	0.087

). The results showed that VCs consistently exhibited the highest values for all centrality measures. GVCs had the second highest mean values for betweenness centrality and eigenvector centrality. CVCs and Other Financials showed moderate values. The “Others” type showed comparatively low mean values.

Statistical tests of these centrality measures revealed significant differences among investor types (

Table 3. Statistical comparison of centrality measures by investor type in the CEET investor network.

Centrality Measure	H-Statistic	p-Value	Effect Size (ε2)	Significant Differences (p < 0.05)
Degree Centrality	59.476	p < 0.001	0.107	VC-CVC, VC-Others
Betweenness Centrality	88.642	p < 0.001	0.163	VC-CVC, VC-Other Financials, VC-Others, CVC-Others, GVC-Others
Closeness Centrality	76.118	p < 0.001	0.139	VC-CVC, VC-Other Financials, VC-Others
Eigenvector Centrality	73.702	p < 0.001	0.134	VC-CVC, VC-Other Financials, VC-Others

). Neither the normality nor the homogeneity assumptions were met; therefore, we employed the Kruskal–Wallis test. The results revealed significant differences among investor types for all centrality measures (p < 0.01). The effect size, as measured based on the square of epsilon, was moderate, explaining about 0.107 to 0.163 of the variability in the centrality measures. The results of Dunn’s test, which analyzes differences between pairs of investor types, revealed that VCs were significantly different from CVCs and Others in all centrality measures and that VCs were different from Other Financials in all measures except degree centrality. In addition, GVCs were found to be significantly different from Others in betweenness centrality. The statistically significant differences between VCs and other investor types in many cases indicate that VCs have a structurally unique position in the CEET investor network, suggesting different roles for each type of investor.

4.3. Community Analysis to Understand the Role of GVCs

In this section, we conducted a community analysis to gain a deeper understanding of the role of GVCs. A total of 65 communities were detected in the CEET investor network (2008–2022). The distribution of community sizes is shown on the vertical axis, which represents the number of investors in each community (Figure 7). A small number of large communities and a large number of small communities were observed. Figure 8 shows the composition of the community members of the CEET investor network

To understand the distribution of GVCs in communities, the size of GVC communities was analyzed (

Table 4. Comparison of GVC and non-GVC community sizes.

Metric	CEET
Number of GVC Communities	10
Number of Non-GVC Communities	55
Average Size of GVC Communities	28.70
Average Size of Non-GVC Communities	4.33
Median Size of GVC Communities	29.00
Median Size of Non-GVC Communities	2.00
Mann-Whitney U Test Result	Significant
U Statistic	466.5
p-value	p < 0.001
Epsilon-Squared (ε2)	0.348

). The results revealed that GVC communities were larger than non-GVC communities. All communities were categorized as GVC communities (with at least one GVC investor) and non-GVC communities; 10 GVC communities and 55 non-GVC communities were identified in CEET. GVC communities were much larger than non-GVC communities, with a higher mean (28.70 vs. 4.33) and median (29.00 vs. 2.00). Nine communities contained one GVC, while there was one community that contained two GVCs. The results of a non-parametric Mann–Whitney U test comparing GVC and non-GVC communities revealed a statistically significant difference in the sizes of GVC and non-GVC communities in CEET (U = 466.5, p < 0.001).

To understand the role that investor types play in different communities, we calculated the cross-community bridging score in CEET (Figure 9,

Table 5. Summary statistics of cross-community bridging scores in the CEET investor network.

Investor Types	Number of Communities	Mean	Median	Min	Max
VC	14	2.588	2.0	0.0	7.600
CVC	41	0.378	0.0	0.0	2.625
GVC	10	2.250	0.5	0.0	8.000
Other Financials	17	1.368	0.5	0.0	9.000
Others	53	0.201	0.0	0.0	2.286

). This score was calculated for each investor type and for each community; a higher score indicates that a given investor type has more connections with nodes outside its own community. VCs (mean = 2.588, median = 2.0) and GVCs (mean = 2.250, median = 0.5) had relatively high scores. The results of the non-parametric Kruskal–Wallis test in Table S1 showed significant differences among investor types (p < 0.01). Dunn’s test also showed significant differences between VC and CVC (p = 0.003) and between VC and Others (p < 0.001), as well as between Other Financials and Others (p = 0.037).

To further analyze the bridging roles of investor types, we calculated the participation coefficient (Figure 10 and

Table 6. Summary statistics of participation coefficient in the CEET investor network.

Investor Types	Count	Mean	Median	Min	Max
VC	119	0.354	0.408	0.0	0.822
CVC	164	0.173	0.000	0.0	0.684
GVC	11	0.309	0.444	0.0	0.680
Other Financials	37	0.205	0.000	0.0	0.710
Others	194	0.104	0.000	0.0	0.720

). This coefficient is computed for each node. The value ranges from 0 to 1, and a higher value means that the node’s edges are evenly distributed across all communities. VCs showed the highest value (mean = 0.354, median = 0.408), followed by GVCs (mean = 0.309, median = 0.444). The results of the Kruskal–Wallis test showed that there were significant differences among investor types (p < 0.01) (Table S2). The results of Dunn’s test showed that VC was significantly different from CVC (p < 0.001), Others (p < 0.001), and Other Financials (p = 0.016). CVC was marginally significantly different from Others (p = 0.056). In the results of cross-community bridging scores and participation coefficients, VCs had the highest scores, and GVCs followed with relatively high scores, suggesting that these investor types have a role in connecting different communities in the CEET investor network.

5. Discussion and Conclusions

The purpose of this study was to explore diverse investors in the CEET investor network. Using Japanese CEET startup investors as a case study, we investigated the evolution of the investor network, the characteristics of the various investor types, and the role of GVCs. The understanding of the diverse investors contributes to expanding knowledge about mobilizing private capital and enhancing sustainable innovation.

From the analysis of CEET investor networks from 2008 to 2022, the development process of investor networks can be divided into three stages. In the context of Japan, the background of the changes in each stage is described as follows. This evolutionary path suggests that Japan’s CEET network had gradually evolved into a mature structure combining diversity and efficiency. This case highlights the importance of understanding investor networks from a temporal perspective.

• Formative stage: This stage corresponds to the period from 2008 to 2012. A close-knit network was formed, characterized by a comparatively small number of investors. From 2010 onwards, the entry of investors continued, and the first signs of network cohesion appeared. During this period, energy deregulation could be associated with an increase in startup entry [68]; however, the energy market had not yet been sufficiently open to investor entry.
• Expansion and diversification stage: This stage corresponds to the period from 2013 to 2017. With the rapid growth of new investors, the CEET investor network has expanded and diversified. This growth trend was likely related to the total expansion of the Japanese startup investment market [19]. In addition, the rise of capital-intensive deep-tech startups may have increased the need for mutual collaboration among investors to allocate risk and secure funding.
• Stable growth stage: This stage corresponds to the period 2018–2022. The investor network was growing steadily, and the efficiency of the network was continuously improving. During this period, not only was the Japanese startup investment market growing, but also the promotion of sector-specific transition policies and the global trend of rapid investment growth in climate tech [69] were likely to be associated with this trend. As VC funds specializing in decarbonization were established and partnered with GVCs and CVCs [70,71], more diverse investors entered the CEET sector.

This study aims to understand the various investor types based on structural characteristics and centrality measures. From the degree distribution and degree correlation results, high-degree nodes tend to preferentially connect with other high-degree nodes, suggesting the existence of hub investors. However, there tended to be fewer high-degree nodes than the theoretical power law distribution would predict. This could be explained by investors’ resource constraints and their strategies to maintain more selective relationships with a more reliable group of colleagues. The results of the centrality measures revealed that VCs have the strongest influence through their connections and ability to spread information in the CEET investor network. The non-VC investors are likely to have different investment motives in addition to the financial returns that VCs typically seek [72]. CVCs tend to emphasize strategic returns, such as business synergies and open innovation with their parent companies [73,74]. Therefore, CVCs are relatively less active in the network than VCs because they make more targeted investments based on corporate strategies. GVCs are influenced by public sector intentions such as environmental policy goals [75]. For this reason, GVCs can play a distinctive role in the network through longer-term investment activities. Other Financials may develop their investment strategies from their organizational missions, which may be why they are similar to CVCs in the investor network. The “Other” type includes a small number of investors with high values, suggesting that this type may include a wide variety of investors.

To understand the roles of GVCs, the findings from the community analysis indicated that GVCs were associated with larger community sizes. Furthermore, the results of the community bridging analysis suggested that both VCs and GVCs exhibited greater connecting capability than the other investor types. The findings regarding GVCs could be explained by two mechanisms: (1) GVCs’ expected long-term commitment and substantial financial resources make them attractive to private investors, creating incentives to cooperate with GVCs; and (2) GVCs provide signals to visualize and guarantee feasibility and reliability through investment, encouraging private investors to participate in their investments. GVCs benefit from syndication with private investors [76], but on the other hand, GVCs can mediate connections between private investors across the communities. Although the limitations of the sample size (11 GVCs) should be considered when interpreting the findings, the characteristics of GVCs in the investor network differ from those of other private investors.

From these observations, we can draw practical implications. The CEET investor network has changed over time, and different roles have been found for different investors. The investor network formed by the accumulated decisions of diverse investors is likely to be more resilient than more homogeneous networks. For example, even if VCs stop investing during financial recessions, other investors can fill the funding gap and prevent a slowdown in sustainable innovation. Policymakers should therefore encourage the diversification of investors and their cooperation with each other. This could include tax incentives for corporate investment, expanding the types of GVCs at each stage of innovation, and introducing mutual matching programs for new investors.

The study has several limitations. First, the scope of financing methods was limited to equity financing. Future studies would benefit from exploring other financing methods, such as public grants, crowdfunding, and venture debt. Second, the regional context focused on Japan. An analysis in different countries and regions would deepen the findings, especially the impacts of startup promotion policies and the role of GVCs. Third, our analysis did not include investment outcomes. Since the number of initial public offerings (IPOs) and mergers and acquisitions (M&As) in Japan was still limited [19,48] in the CEET sector, it would also be interesting to examine non-financial investment indicators and the innovation outcomes of portfolios. These are important future research spaces.

This study provides the first comprehensive analysis of the CEET investor network, using Japan as a case study, and clarifies the roles of its diverse investors. Since previous investor network studies have focused on VCs and sectors other than CEET, this study also contributes to extending these findings. The findings of this study provide valuable insights to advance both the theoretical understanding and practical implications for overcoming the CEET financing difficulties to address the climate change crisis. This study offers empirical evidence to support sufficient funding for sustainable innovation.