Research on the Reverse Technology Spillover Effect from China’s CVC Overseas Investments

Abstract

China’s corporate venture capital (CVC) overseas investment began in the late 20th century and has expanded significantly over the years. By 2021, more than 265 Chinese institutions and companies had engaged in cross-border investments, contributing over USD 100 billion. These investments present a unique opportunity to examine the reverse technology spillover effect on China’s technological development. Using a Difference-in-Differences model and regression analysis, we investigate whether China’s CVC overseas investments drive technological progress. Our findings reveal three key insights: (1) these investments have a positive impact on China’s technological advancement, (2) the effect is stronger when the host country has a higher level of technology, and (3) larger investment amounts amplify the impact. This research not only highlights the transformative potential of cross-border CVC investments but also demonstrates how enterprises can leverage reverse innovation spillovers to accelerate China’s technological progress. Additionally, we introduce a novel approach to studying this phenomenon, contributing to the existing scholarship on global innovation dynamics.

1. Introduction

Corporate venture capital (CVC), a specialized form of venture capital, involves direct investments by corporations to foster innovation and strategic growth. In China, cross-border CVC investments date back to the late 20th century and have expanded significantly over the past decades. According to the Zephyr database, the total value of Chinese cross-border CVC investments surged from USD 6.9 billion in 2011 to USD 107.9 billion in 2021, reflecting a compounded annual growth rate of 31.6%.

Leading corporations, particularly Tencent, have played a pivotal role in this expansion. Their share of total investments rose sharply from 12.4% in 2011 to 80.4% in 2021, underscoring the increasing dominance of major firms in shaping the cross-border CVC landscape. This trend highlights their crucial role in driving technological innovation and expanding global market engagement.

Since 2011, China’s cross-border CVC investments have been concentrated in a select group of countries, with the United States, Singapore, Israel, the United Kingdom, and India emerging as the top five destinations. Among these, the United States stands out as the primary target, accounting for 478 investment events—far surpassing other nations. This preference underscores the U.S.’s status as a global hub for innovation and advanced technology, making it a focal point for China’s CVC strategy.

The literature suggests that when corporate venture capital (CVC) investments flow from less developed to more developed countries, the gap in technological potential triggers a reverse technology spillover effect. This effect tilts advanced technology, R&D, and innovation experience toward the less developed countries, enhancing their technological innovation and competitiveness (Kogut & Chang, 1991). With China’s overseas CVC investments increasing significantly in recent years, it becomes crucial to investigate whether Chinese companies are successfully capturing these technology spillover benefits. Understanding this dynamic could provide valuable insights into the role of cross-border investments in driving China’s technological advancement.

Previous studies have primarily examined reverse technology spillover effects at the country level through the lens of outward foreign direct investment (OFDI). In contrast, this paper focuses on reverse technology spillover effects from the perspective of China’s corporate venture capital (CVC) overseas investments. This approach addresses a gap in the literature by offering a detailed and systematic analysis of cross-border CVC investments, providing new insights into their role in fostering technological advancements and innovation in less developed countries.

The rest of the paper is organized as follows: Section 2 reviews the relevant literature, providing the foundation for our study. Section 3 introduces the variables and data used in the analysis, followed by a detailed discussion of the methodology in Section 4. Section 5 presents the empirical results, and Section 6 concludes with the key findings and implications.

2. Literature Review

In this section, we review the literature from four key perspectives: the motivations and functions of cross-border corporate venture capital (CVC) investments, the reverse technology spillover effects of such investments, the factors influencing the relationship between cross-border CVC and technology spillovers, and the recommendations provided by scholars for companies engaging in cross-border CVC investments.

2.1. Motivations and Functions of Cross-Border Venture Capital Investment

2.1.1. From the Perspective of CVC Companies

The literature identifies three primary motivations for cross-border corporate venture capital (CVC) investments: (1) Capital Gains Focus: these investments prioritize maximizing returns through liquidation activities, such as initial public offerings or selling shares to third parties (Chesbrough, 2002); (2) Strategic Goals Focus: these investments primarily aim to complement internal R&D efforts and support the company’s technological development (Basu et al., 2011); and (3) Dual Goals: these investments seek to balance capital gains and strategic benefits.

Yamawaki (1994) and Bringmann (2018) highlighted that many cross-border CVC investments prioritize strategic goals, particularly in advancing the company’s technological capabilities. Similarly, MacMillan et al. (2012) reported that over half of their survey respondents invested in CVC projects primarily for strategic benefits derived from technology spillovers. Only 20% focused exclusively on capital gains, though strategic benefits often accompanied these investments. The remaining 30% pursued an equal balance between strategic and financial objectives, underscoring the multifaceted motivations driving cross-border CVC investments.

2.1.2. From the Perspective of Portfolio Companies

The literature suggests that cross-border corporate venture capital (CVC) investments have both positive and negative impacts on the portfolio companies.

On the positive side, cross-border CVCs are typically large firms with substantial industry experience and established management systems. They can offer valuable advice and support in areas such as business development and organizational growth, helping startups navigate early-stage challenges (Y. Wang et al., 2015). Additionally, CVC investments help alleviate financing difficulties for startups, enabling them to focus on technology research and development (Park & Steensma, 2012).

On the negative side, CVC investors may exploit their advantageous position to gain access to valuable information and technology developed by the startups for their own benefit. This opportunistic behavior can create tensions and challenges in the relationship between the CVC investors and the portfolio companies (Katila et al., 2008; Dushnitsky & Shaver, 2009). Startups, being in a weaker position, are more vulnerable to such opportunism (J. Y. Kim et al., 2019). While entrepreneurs may attempt to protect their technologies using patents, trade secrets, or the timing of financing rounds, these measures are not always successful (Lindsey, 2008; Diestre & Rajagopalan, 2012). Furthermore, since many early-stage technologies are not yet ready for patents, startups may struggle to legally protect their innovations (Katila et al., 2008). Additionally, CVC investors can influence the strategic and R&D directions of startups, potentially obtaining significant technological spillovers by shaping the development of these early-stage firms (J. Kim & Park, 2017).

2.2. Reverse Technology Spillovers

Technology spillover effects can be categorized into forward and reverse spillovers based on the direction of the flow. Forward technology spillover refers to the direct promotion of technological advancement in the portfolio companies through CVC investments in foreign markets. In contrast, reverse technology spillover refers to the process where the home country absorbs technology from the investing country through overseas investment channels, such as mergers and acquisitions, joint ventures, or the establishment of branches (Du & Lin, 2018). This process allows domestic enterprises to acquire technological factors and leverage feedback mechanisms to benefit from technology spillovers. These mechanisms include reverse technology transfer, shared R&D capital, talent flow, and the transformation of research achievements (De Mello, 1997; Lytras et al., 2002; L. Li, 2016).

Empirical studies support the existence and significance of reverse technology spillovers. (Basant & Fikkert, 1996) found that Indian companies saw high returns from purchasing overseas technology, with much higher returns than from their own R&D investments. Driffield et al. (2008) categorized outward foreign direct investment (OFDI) into “technology acquisition” and “efficiency seeking”, finding that both types of investments—particularly those in high R&D-intensive countries—contributed to increased productivity growth in the UK. V. Chen et al. (2012) argued that OFDI from emerging markets to developed countries primarily aims to seek knowledge and technology, further confirming reverse technology spillovers. Hu et al. (2021) used knowledge structure theory to explore reverse knowledge spillovers, showing that both the breadth (diversity of industries and technologies) and depth (degree of knowledge specialization) of the spillover promote technological progress in investing companies. Their findings suggest that in the early stages of reverse knowledge spillover, depth has a greater impact, while breadth becomes more influential in later stages.

2.3. Factors Influencing Cross-Border CVC Investment

(1)

Macro perspective

The effectiveness of cross-border corporate venture capital (CVC) investments is influenced by several macro factors, including the economic environment, legal and cultural distance, geographical distance, and technological level, among others.

For instance, Akcigit et al. (2024) argued that the degree of intellectual property rights (IPRs) protection in host countries plays a crucial role in determining the ease of cross-border CVC investment. Stronger IPRs protection facilitates the technology learning process for investing companies, making the investment more effective.

H. Cai et al. (2020) suggested that high technological similarities between investing and host countries could create restrictions on CVC investments. Investors may be concerned that key technologies could be leaked, leading to the potential theft of business by market competitors. This concern can limit the willingness of companies to engage in cross-border CVC activities, as the risk of intellectual property loss or competitive disadvantage increases with technological overlap.

Recent studies have increasingly focused on how geographical distance impacts cross-border capital flows. Tykvová and Schertler (2014) argued that although investors generally prefer to invest in close proximity, following the “one-hour drive rule” for venture capital investments, the number, frequency, and geographic diversity of distant venture capital deals have increased over time. Using gravity models, Bringmann (2018), found that geographical distance is a critical factor in cross-border venture capital portfolio selection. Despite innovations in transportation and communication, greater geographical distance hinders investors’ ability to participate in management and learn from the portfolio companies, thus affecting the reverse technology spillovers from cross-border corporate venture capital (CVC) investments.

The technological level of the host country also influences the efficiency of technology spillovers from cross-border investment. On the one hand, the technological level of the host country affects the availability and quality of potential investment targets (Black & Gilson, 1998). Higher technological levels in the host country increase the quantity and quality of investment opportunities, which is vital for companies seeking technologies that are difficult to develop solely through local cooperation (Dang & Cha, 2011). On the other hand, leveraging the scientific and technological resources of developed countries plays a crucial role in promoting reverse technology spillovers from foreign investments (Park & Steensma, 2012).

(2)

Micro perspective

The literature suggests that several factors at the company and personnel levels influence cross-border corporate venture capital (CVC) and its technology spillover effects, including the investor’s choice of investment model, investment stage, and the background of entrepreneurs.

For example, Y. Wang et al. (2015) and W. Chen et al. (2022) argued that the impact of cross-border CVC on the innovation of venture firms depends on the characteristics of the investors and the dynamics of the investment process. When investing alone, CVC typically favors late-stage companies due to information asymmetry. However, joint ventures with local firms can help mitigate this issue, encouraging CVCs to invest in early-stage or high-tech companies. Di Lorenzo and Sabel (2023) explored how the timing of CVC investments affects the income and R&D intensity of the portfolio companies. They found that late-stage investments boost income but reduce R&D intensity, as these companies are more likely to transfer technology to generate revenue (Kang et al., 2021).

From the perspective of entrepreneurs, their access to international knowledge sources plays a key role in the performance of the portfolio companies. Oviatt and McDougall (1994) highlighted the importance of entrepreneurs’ prior international experience, while Bringmann (2018) identified three sources of international knowledge that can benefit the portfolio companies: (1) the entrepreneur’s international experience, (2) access to cross-border CVC, and (3) the recruitment of new managers. Cross-border CVC can serve as a crucial channel for entrepreneurs to access technological information, promoting reverse resource spillovers.

Despite the growing research on foreign venture capital investments in China (Jiang, 2014; L. Li & Li, 2017), most studies have focused on overseas investment by foreign institutions, using national-level data. However, we argue that such aggregated data may not fully capture the impact of technology spillovers. Thus, this paper examines reverse technology spillovers in China from the perspective of Chinese cross-border CVC investments.

2.4. OFDI and Technology Spillovers

Research on reverse knowledge and technology spillovers from outward foreign direct investment (OFDI) has garnered significant attention. Hu et al. (2021) examined the mechanisms through which reverse knowledge spillovers influence the technological progress of investing enterprises, finding that both the breadth and depth of these spillovers contribute to technological advancement. Expanding on this, He (2023) analyzed the impact of OFDI-driven reverse technology spillovers on regional innovation capacity in China, revealing that such spillovers have both direct and spatial positive effects on regional innovation.

Institutional factors also play a crucial role in facilitating international investment. S. Li et al. (2021) explored how developing countries leverage bilateral investment treaties (BITs) to mitigate investment uncertainties arising from informal institutional distance, thereby supporting domestic firms in expanding abroad. Meanwhile, Y. Li et al. (2022) investigated the effects of OFDI-driven reverse technology spillovers on the green total factor productivity (GTFP) of domestic manufacturing. Their findings indicate that, while these spillovers do not significantly enhance manufacturing GTFP at a national level, they have a positive impact in the central and western regions of China. Additionally, absorptive capacity—particularly financial development—moderates the relationship between reverse spillovers and GTFP growth.

Several studies have focused on the role of absorptive capacity in maximizing the benefits of reverse spillovers. Yang et al. (2011) analyzed inter-provincial panel data (2003–2008) and found that while human capital enhances total factor productivity, it has a limited effect on improving absorptive capacity for reverse spillovers. Similarly, Hong et al. (2019) examined how OFDI influences domestic innovation performance and found that investments in developed countries enhance innovation, whereas investments in emerging markets tend to hinder it.

From an energy and environmental perspective, W. Zhang et al. (2022) assessed the impact of Chinese enterprises’ overseas investments on energy intensity using a two-stage least squares (2SLS) approach. Their study, leveraging an instrumental variable based on money supply fluctuations and port-opening history, concluded that reverse technology spillovers significantly reduce energy intensity. This effect is particularly strong in investments associated with the Belt and Road Initiative (BRI) and in technology-intensive, non-resource-intensive industries.

Together, these studies underscore the multifaceted impact of reverse technology spillovers, highlighting their role in technological advancement, regional innovation, institutional adaptation, and environmental sustainability.

2.5. Cross-Border CVC Investment and Reverse Technology Spillovers

Despite the abundant studies on general OFDI investment and reserve technology spillovers, relatively few studies have examined the reverse technology spillover effects of cross-border corporate venture capital (CVC) investments, particularly from the perspective of the investing country. Akcigit et al. (2024) explored foreign corporate investments in Silicon Valley, analyzing how such funding may support U.S. entrepreneurs but also potentially lead to knowledge spillovers to foreign investors and the countries where they are based. Their findings, based on non-U.S. corporate investments in U.S. start-ups from 1976 to 2015, suggest a knowledge of spillovers to foreign investors. However, their focus was on guiding governments to raise costs to deter investments, preventing such technology spillovers.

Kang et al. (2021) found a positive relationship between technology spillovers and capital gains in the biopharmaceutical industry. However, this relationship weakened or became negative when CVC investments were made for financial purposes or in early-stage startups. Their research emphasized the benefits to the corporate investor, rather than focusing on reverse technology spillovers at the country level.

R. Zhang and Zhang (2022) examined CVC’s external spillover effects, noting that CVC investments improved innovation in peer companies and local enterprises. However, their research was limited to the company level, as well, rather than addressing national-level spillovers. In contrast, our paper collects data at the company level and then aggregates it to the national level to study reverse technology spillover effects in China from the perspective of Chinese CVC investments abroad. By doing so, it fills a critical gap in the literature, providing insights into how cross-border CVC investments contribute to technological advancement in the investing country.

3. Hypotheses and Variables

In this section, we discuss the hypotheses and present the variables in our empirical analysis.

3.1. Hypotheses

Hu et al. (2021) defined OFDI reverse knowledge spillover as the process where domestic enterprises invest in foreign markets and acquire local knowledge and skills through mechanisms such as mergers and acquisitions, joint ventures, and the establishment of branches. This process includes various elements like reverse technology transfer, R&D capital sharing, the flow of R&D talent, technology transfer, and the transformation of research achievements (Lytras et al., 2002; L. Li, 2016). Y. Li et al. (2022) further demonstrated the positive impact of OFDI reverse technology spillovers on China’s innovation performance using a threshold model. Since corporate venture capital (CVC) represents a form of OFDI, these insights lead to Hypothesis 1, as stated below:

H1. 

China’s CVC overseas investment will promote the development of China’s technological innovation.

Hu et al. (2021) emphasized that an enterprise’s ability to absorb and learn from reverse technology spillovers is influenced by its internal factors, such as human resources and technology gaps. These factors act as regulatory mechanisms for reverse knowledge spillovers. Similarly, Park and Steensma (2012) noted that the advanced scientific and technological resources of developed countries—such as cutting-edge technology, state-of-the-art research equipment, and high-quality talent—form the foundation for domestic companies to benefit from reverse technology spillovers. Building on these insights, we argue that the technological level of the host country plays a significant role in facilitating the cross-border CVC technology learning process. Therefore, the higher the technological level of the host country, the more favorable the conditions for effective reverse technology spillover, ultimately enhancing the technology transfer and innovation of the investing country.

H2. 

The higher the technological level of the host country, the greater the reverse technological spillover effect will be for China’s CVC overseas investment.

H. Wang et al. (2019) argued that controlling for other factors increased investment in R&D, leading to higher technological innovation efficiency and faster economic development. Similarly, Gans and Stern (2003) and Park and Steensma (2012) highlighted that CVC investments create joint value opportunities for both the invested and investing companies. These collaborative R&D efforts contribute to the generation of relationship-specific value, enhancing technological progress for both parties. Kang et al. (2021) further emphasized that such investments help investors obtain technological spillover effects by supporting the technological advancements of the investee, creating a mutually beneficial environment for innovation. This collaborative dynamic reinforces the role of CVC in facilitating knowledge transfer and innovation.

Accordingly, we propose Hypotheses H3.

H3. 

The higher the investment level of the home country in one technology, the greater the reverse technology spillover effect will be from that technology.

3.2. Data

Investment data in the primary market, particularly for cross-border venture capital events, are often not widely published, posing significant challenges for data collection. To address this, we conducted a comprehensive analysis of available data sources, comparing the Zephyr and PitchBook databases. After considering various factors, such as data integrity, downloadability, and completeness, we chose to primarily rely on the PitchBook database for our data collection. PitchBook, founded in 2007 by John Gabbert, is a leading SaaS company that provides data, technology, and investigative reports for the private capital markets, including venture capital (VC), private equity (PE), and mergers and acquisitions (M&A). To supplement this data, we also utilized the United States Patent and Trademark Office (USPTO), the China State Intellectual Property Office, and the World Bank databases, ensuring a robust and comprehensive data collection process for this study.

Data selection bias is a potential limitation in this study, as our dataset may not fully capture the entire scope of Chinese CVC overseas investments. Publicly available data often excludes smaller or private transactions, which may result in an overrepresentation of larger, more transparent firms. Additionally, variations in geographic and industry distribution could affect the generalizability of our findings.

To mitigate these concerns, we have cross-verified data from multiple sources and acknowledged these limitations in our analysis. Future research could benefit from broader datasets or alternative methodologies to enhance the robustness of findings.

3.3. Variables

3.3.1. Dependent Variable

Our dependent variable is “Patent”, which is defined as the number of new patents of Chinese enterprises in the technology category, c, in year t. We use this variable to measure the progress of various technologies in our country. This data came from the State Intellectual Property Office of China.

3.3.2. Explanatory Variables

(1)

Variable to measure reverse technology spillover from CVC overseas investments

In our study, we tried to investigate whether the Chinese companies’ overseas investments can feed back into China’s technological progress, so we created this variable to measure the existence of the reverse technology spillover as follows: first, we created the investment time point dummy variable “post”, which equals 0 for the time before the investment and equals 1 for the year of investment and afterward. We then created an investment object variable, “Treated”, which equals 1 if the technology is invested during the sample period and 0 otherwise. So, the interaction term “Post×Treated” is used to measure whether there is an effect of reverse technology spillover impact from the CVC overseas investments.

(2)

Variables to measure the technology level of the host country

We created two variables to measure the technology level of the host country. The variable “Publications” refers to the number of scientific papers published by the host country, and the variable “f_patents” measures the number of patent applications in the host country. Both datasets came from the World Bank.

(3)

“Amount”: Investment from the CVCs

The variable “Amount” represents the total amount of cross-border venture capital investment made by Chinese enterprises in technology category c in year t. The data were sourced from the PitchBook database.

In cases where the investment is syndicated (i.e., involving multiple investors from different countries), the PitchBook database only provides the total investment value for the entire deal rather than the individual contributions from each investor. To address this, we adopted the methodology proposed by Tykvová and Schertler (2011) to allocate the investment amount among different investors. Specifically, we allocated the total transaction value to each country based on the proportion of investors from that country in the syndicated deal.

3.3.3. Intermediate Variable

(4)

“Cor_patents”: the technology level of the portfolio companies

In this research, we hypothesize that higher levels of investment in cross-border CVC will provide greater research support to the portfolio companies, which will, in turn, stimulate the company’s technological progress and enhance the effect of reverse technology spillovers. To assess the technological development of the portfolio companies, we created the variable “cor_patents”, which is defined as the number of patent applications submitted by the invested company in each year during the sample period.

The data for “cor_patents” was sourced from the PitchBook database. Given that obtaining R&D investment and financial data for unlisted companies can be challenging, patent data serves as a reliable proxy for technological progress. In this paper, we use the number of patents filed by a company as an indicator of the company’s technological advancement in our study.

3.3.4. Control Variables

(1)

Economic development in the host country

In this research, we created two key variables to measure the economic development of the host country, which are essential for understanding the broader context of the reverse technology spillover effect.

(a)

“GDP”: This variable is defined as the logarithm of the Gross Domestic Product (GDP) of country f in year t. It is used to measure the general economic level of the host country. Drawing on S. Chen (2017), we propose that the larger the host country’s economic scale, the more developed its financial market will be, making it easier for entrepreneurs to access funds. Therefore, we argue that a larger GDP size in the host country increases the likelihood of high-quality projects available for venture capital investment, which in turn enhances the potential for reverse technology spillover.

(b)

“GDPP”: This variable is defined as the GDP per capita, which we use to gauge the labor cost of the host country. We argue that a lower labor cost in the host country increases the attractiveness of the country to overseas investors, thereby facilitating cross-border investment. This factor is likely to be more conducive to reverse technology spillover, as it encourages investments in high-tech or innovation-driven sectors at a lower cost.

These two variables help us examine the relationship between the economic environment in the host country and the potential for reverse technology spillovers resulting from cross-border CVC investments.

The data for both variables were obtained from the World Bank.

(2)

Opening level of the host country

In our research, we also used two variables to measure the level of openness of the host country, as it is crucial in understanding the potential for reverse technology spillovers.

(a)

“open_ivs”: This variable is defined as the percentage of net foreign investment inflows relative to the GDP of the host country. Following D. Cai and Liu (2012), we consider this as a good indicator of the country’s openness to foreign investment. A higher percentage suggests a more open economic environment that is conducive to attracting foreign capital and fostering cross-border technology spillovers. The data for this variable is sourced from the World Bank.

(b)

“open_inex”: This variable is calculated as the proportion of the country’s import and export volume relative to its total GDP. It serves as an indicator of the country’s dependency on foreign trade, which is often correlated with its openness to international economic activities. According to S. Chen and Guo (2021), the greater the dependency on foreign trade, the higher the openness to external influences, including foreign investments, which can enhance reverse technology spillover. Data for this variable also came from the World Bank.

Both of these variables help us examine how the host country’s level of openness can influence the effectiveness of cross-border CVC investments and their potential to generate reverse technology spillovers.

(3)

Host country resource endowment

In our research, we also incorporated the variable “Source(f,c,t)”, which measures the host country’s resource endowments. Specifically, this variable is calculated as the percentage of ore and metal exports out of the total commodity exports for the host country. According to S. Chen and Guo (2021), countries with abundant natural resources tend to attract more foreign investments due to their valuable resources, which create an environment conducive to investment opportunities.

We argue that the presence of rich natural resources not only draws foreign capital but also enhances the potential for reverse technology spillovers. As foreign investments flow into these resource-rich countries, the associated technological advancements and R&D activities can have spillover effects that benefit the investing countries. Thus, “Source(f,c,t)” is an important factor in our analysis of how resource endowments influence the reverse technology spillover effect.

Data for this variable was sourced from the World Bank, ensuring that the information is accurate and up to date for each country and year in our sample.

(4)

Geographical distance

We included the variable “Distance”, which measures the straight-line distance between the capitals of the host and investing countries. Geographical proximity is crucial for foreign investments, as shorter distances allow investors to engage more closely with the portfolio companies, enhancing management participation and reducing operational costs. This proximity may also improve reverse technology spillovers. The distance was calculated using Google Maps (L. Li & Li, 2017).

(5)

Investment stage

The variable “CVC_stage” is a dummy variable indicating the stage of the corporate venture capital (CVC) investment. It equals 0 for early-stage investments and 1 for later-stage investments. This data was sourced from the Pitchbook database.

(6)

Patent type of the invested country (country–patent fixed effect)

α(f,c) refers to the fixed effect for a specific type of patent in a particular country within the research sample. It is included as a fixed effect in the panel data analysis to account for country and patent-type specific variations.

(7)

Year (year fixed effect)

μ(t) refers to the year fixed effect in our panel data analysis.

Table 1. Descriptions of the variables in our analysis.

Variable Type	Symbol	Description
Dependent Variable
	Patents(China,c,t)	In year t, the number of new patents issued by Chinese enterprises in technology category c in year t. This variable is used to indicate the technology progress level in China. We use the ln of the patent number in our empirical analysis to minimize the impacts of the outliers.
Explanatory Variable
	Post(China,c,t)	Dummy variable, which equals 1 for the investment year and the 5 years afterwards, and 0 otherwise.
	Treated(China,c,t)	Dummy variable which equals 1 if China has invested in overseas companies in technology category c and 0 otherwise.
	Post(China,c,t) × Treated (China,c,t)	Interactive variable, which is used to measure the existence of technology spillover effects.
	Publications(f,c,t)	In year t, the number of articles in scientific and technological journals in country f, in which Chinese companies have invested in c patents. We use this variable to indicate the invested countries’ technology level.
	f_patents(f,c,t)	In year t, the number of patent applications in country f that has invested in patent c by Chinese companies.
	Amount(f,c,t)	In year t, the amount of cross-border venture investments by Chinese enterprises in a company in country f that masters technology category c.
Middle Variable
	cor_patents(f,c,t)	In year t, the number of new patent applications by the companies located in country f that were invested in technology c by Chinese companies.
Control Variable
	GDP(f,c,t)	GDP of country f that has invested in patent c by Chinese enterprises in year t. This variable is used to indicate the host country’s economic development level.
	GDPP(f,c,t)	Per capita GDP of country f in year t, where Chinese companies have invested in patent c. This variable is used to indicate the labor cost.
	open_ivs(f,c,t)	The percentage of net foreign investment inflows out of the GDP for country f in year t. This variable is used to indicate the open level of the host country.
	open_inex(f,c,t)	The proportion of import and export volume in country f’s total GDP in year t. This variable is used to indicate the opening level of the host country, as well.
	Source(f,c,t)	The percentage of the ore and metal exports of the total commodity exports in the country f in year t.
	Distance(f,c,t)	In year t, the straight-line distance between the geographical center of country f, which has invested in patent c by Chinese enterprises, and the geographical center of China.
	CVC_stage	The investment stage of the company in country f.
	α(f,c)	Country–patent fixed effects (technology effect).
	μ(t)	Year fixed effect.

provides a detailed summary of the variables in our analysis. In this table, we present a detailed explanation of each variable used in our analysis.

4. Methodology

In this section, we discuss our methodology.

4.1. Exploring the Existence of Reverse Technology Spillover Effects

(1)

Difference-in-Difference model (DID model)

To address the endogeneity issue in estimating the impact of China’s cross-border venture capital (CVC) investments on its innovation, we use a Difference-in-Differences (DiD) approach. This method allows us to examine the evolution of China’s innovation in specific technology categories after it invests in foreign startups. The key challenge is that cross-border CVC investments are endogenous—Chinese companies may invest in foreign startups because of pre-existing technological needs, and these investments may reflect technological upgrades rather than being a result of the learning process.

Our empirical strategy involves comparing the innovation activities of two groups: the experimental group, which includes technology categories where China first invested in foreign startups, and the control group, consisting of technology categories with no such investments. By comparing changes in innovation between these two groups, we can isolate the effect of cross-border CVC investments on China’s technological innovation, controlling for the potential reverse causality and technological shocks. This analysis provides insights into how cross-border CVC investments contribute to China’s innovation activities.

(2)

The setting of the experimental group and the control group

To compare the impact of China’s cross-border CVC investments on its technological innovation, we match the experimental group (technology categories with CVC investments) with a control group (categories without such investments). We use two pre-investment measures of innovation for matching: the number of granted patents in China and the number of foreign patents cited by China in each technology category over the five years before the investment. We calculate the squared differences in both the number of patents and patent of citations between each potential pair of experimental and control groups, then average these differences. The matched pairs that minimize this average squared distance are selected, allowing for a valid comparison of the effects of CVC investments on China’s innovation.

Once the sample is selected, we use the following methodology to investigate the existence of the reverse technology spillover effect.

Ln_Patentsc,t = β0 + β1Postt + β2Postt × Treatedc + αc + μt + εc,t

(1)

where Ln_Patentsc,t is the squared difference of the patents produced by China between the experimental technology category and control technology category during the past five years. $Postt$ is the dummy variable, which equals 1 for the investment year and 5 years afterward, and 0 otherwise. $Treatedi$ is also a dummy variable, which equals 1 if the Chinese CVCs invested in the foreign companies in technology category c, so its coefficient $\beta 2$ of the interactive variable (Post × Treated) can be used to measure the existence of the reverse technology spillover effect. $\alpha i$ is the technology fixed effect, and $\mu t$ is the year fixed effect.

4.2. Exploring the Factors Affecting Reverse Technology Spillover Effects

We also explore the factors that affect the reverse technology effects from China’s CVC overseas investments using the following model.

Ln_Patentschina,c,t + i = β0 + β1ln_techf,c,t + β3ln_amountf,c,t + βCountrolsf,c,t + αf,c + μt + εf,c,t

(2)

where ln_techf,c,t refers to either ln_Publications or ln_f_patents used in different regression models. ln_amountf,c,t refers to the investment amount in country f technology t from China. The control variables include ln_GDP, ln_GDPP, open_inx, open_inex, source, ln_distance, etc.

To address potential delays in technological learning and the circular causality between cross-border CVC investments and reverse technology spillovers, we account for the following considerations:

• The learning process may be delayed due to the economic environment, openness, resource endowments, technological level of the host country, and the technological progress of the invested company in the year of the investment. This delay could affect the timing of technological spillovers.
• The reverse technology spillover from cross-border investment may influence the investment decisions of the investing company, which could create circular causation between the dependent and explanatory variables.

To resolve these issues, we lag the dependent variables by one to four periods and conduct regression analysis for each lag period separately, allowing us to examine the impact of investments on technological learning over time, while mitigating potential endogeneity.

5. Empirical Analysis

In this section, we present our empirical results.

5.1. Data Collection

Our data sample includes 840 completed CVC investment transactions, each with available transaction amounts. For these investments, we ensured that the investor is based in mainland China, the investee is based outside mainland China, Hong Kong, Macao, and Taiwan, and the investee company holds at least one patent. This resulted in data on 646 companies and a total of 26,354 patent records.

To determine which patents were involved in each transaction, we manually downloaded patent data for these companies. If the transaction data does not specify which patents were invested in, we assigned the patent type with the largest share in the invested company as the type of patent receiving the investment. Additionally, the investment’s time and amount were recorded.

The ultimate goal of this study is to identify the patent types in which China has made overseas venture investments and compare them to the patent types that did not receive investments from China, allowing for a meaningful analysis of the reverse technology spillover effect.

Next, we calculated the changes in the number of domestic patents in various technology categories. We used the number of newly granted patents in each category every year as an indicator of innovation progress in that technology area. The patents are classified into nine main departments (A-H, Y) and 136 categories under the CPC patent classification. According to the official website of China’s State Intellectual Property Office, registration data is available for 128 of these categories. We obtained the data for newly granted patents from 2000 to 2021, yielding a total of 2816 patent observations.

In addition to analyzing the impact of CVC investments on reverse technology spillovers, we also gathered data on the host country’s technological level, development status, openness, resource endowments, and geographical distance to understand how these factors influence the spillover effects.

5.2. Distribution Statistics

Since 2000, China’s overseas CVC investments have targeted patent types across seven departments and 45 major categories. The department with the highest amount of investments and the largest number of investments is Department G (Physics). Within this department, patent type G06 has received the largest investment, totaling USD 35.229 billion. This accounts for 92.44% of the total investment in Department G and 54.68% of all patent investments. G06 also has the highest number of investment events, representing 92.44% of investments in Department G and 26.42% of all patent investment events. The details are shown in

Table 2. Investment patent type distribution (department).

Patent Classification (Department)	Investment Amount (USD 100 Million)	Number of Investments
G-Physics	381.11	304
A-Necessities of Human Life	79.18	177
C-Chemistry; Metallurgy	71.31	110
B-Work; Transportation	64.03	47
H-Electricity	47.15	128
F-Mechanical Engineering; Lighting; Heating; Weapons; Blasting	1.44	5
E-Fixed Structure	0.02	1

.

5.3. Existence of Reverse Technology Spillover Effects

5.3.1. Difference-in-Differences Method (DID)

In this section, we present the results using the Difference-in-Differences (DID) method to study reverse technology spillover effects. The DID method relies on the critical assumption of parallel trends—meaning that, in the absence of policy intervention, the time trends of the treatment and control groups should align. To investigate reverse technology spillover effects, we employ a multi-period DID approach. First, we centralize the investment time by subtracting the investment time from each period. We then run the regression model as follows:

$$\mathit{ln}\_Patentc,t=\beta 0+{\sum }_{\tau =-M}^N\beta \tau Eventc,t-\tau +\alpha c+\mu t+\epsilon c,t$$

(3)

where $\mathit{ln}\_Patentc,t$ refers to the difference in numbers of patents between the experimental group and the control group in a log format, and $Eventc,t-\tau$ is a dummy variable which equals 1 if China made CVC overseas investments in technology type c, and 0 otherwise (M and N, respectively, represent the number of periods before and after the investment time point). So, $\beta 0$ measures the reverse technology spillover effect in the current period of investment, $\beta -M$ to $\beta -1$ measure the effect in the 1-M period before the investment, and $\beta 1$ to $\beta N$ measure the effect in the 1-N period after the investment. To avoid complete collinearity problems, the samples in the first period (that is, the five periods before the investment, pre5) are removed as the benchmark group. The regression results are shown in

Table 3. Test of parallel trend test.

Variables	y
pre4	−0.031
	(−0.16)
pre3	0.027
	(0.13)
pre2	0.118
	(0.59)
pre1	0.207
	(1.01)
current	0.538 ***
	(2.61)
post1	0.756 ***
	(3.60)
post2	0.695 ***
	(3.25)
post3	0.671 ***
	(3.08)
post4	0.718 ***
	(3.23)
post5	0.675 ***
	(2.97)
Constant	4.938 ***
	(62.89)
Observations	484
R-squared	0.904

t-statistics in parentheses.*** p < 0.01.

.

In Table 3, pre1–pre4 represent the four periods prior to the investment year, while current refers to the investment period itself, and post1–post5 represent the five periods following the investment. As shown in

Table 4. Results of regression analysis.

Variables		Variables		Variables
post_treated	0.540 ***	95.number	1.185 ***	2010.year	−0.816 *
	(4.72)		(4.24)		(−1.68)
post	0.097	100.number	−0.266	2011.year	−0.788
	(0.79)		(−1.11)		(−1.62)
11.number	−2.307 ***	107.number	1.513 ***	2012.year	−0.595
	(−8.32)		(5.48)		(−1.22)
12.number	0.132	108.number	1.820 ***	2013.year	−0.420
	(0.49)		(6.79)		(−0.86)
13.number	1.313 ***	109.number	1.002 ***	2014.year	−0.390
	(4.66)		(3.75)		(−0.79)
15.number	−0.013	114.number	−0.301	2015.year	−0.030
	(−0.04)		(−1.12)		(−0.06)
16.number	0.147	115.number	1.152 ***	2016.year	1.020 **
	(0.53)		(4.17)		(1.97)
29.number	0.496 *	116.number	−1.334 ***	2017.year	1.888 ***
	(1.85)		(−4.97)		(3.64)
33.number	0.298	117.number	0.823 ***	2018.year	2.484 ***
	(1.10)		(3.00)		(4.72)
42.number	−0.003	119.number	−1.870 ***	2019.year	2.878 ***
	(−0.01)		(−6.97)		(5.42)
44.number	0.333	121.number	3.162 ***	2020.year	3.300 ***
	(1.44)		(11.71)		(6.13)
46.number	−2.447 ***	122.number	1.045 ***	2021.year	3.558 ***
	(−9.12)		(3.83)		(5.90)
52.number	0.179	123.number	1.008 ***	Constant	4.724 ***
	(0.60)		(4.17)		(9.21)
60.number	1.340 ***	125.number	1.843 ***
	(5.00)		(7.31)
62.number	−1.680 ***	126.number	2.362 ***
	(−6.22)		(8.52)
64.number	−0.198	128.number	1.419 ***
	(−0.72)		(5.15)
65.number	−3.929 ***	2004.year	−0.177
	(−14.27)		(−0.32)
79.number	−3.721 ***	2005.year	−0.389
	(−13.51)		(−0.79)
86.number	−0.426	2006.year	−0.428
	(−1.55)		(−0.87)
89.number	−0.878 ***	2007.year	−0.435
	(−3.63)		(−0.90)
91.number	−1.305 ***	2008.year	−0.695
	(−4.82)		(−1.44)
92.number	−0.073	2009.year	−0.759	Observations	484
	(−0.30)		(−1.57)	R-squared	0.904

t-statistics in parentheses and *** p < 0.01, ** p < 0.05, * p < 0.1.

, the coefficients for the periods before the investment (pre1–pre4) are not significantly different from zero, suggesting no significant difference between the experimental and control groups prior to the investment. In contrast, the coefficients for the periods after the investment (post1–post5) are significantly positive at a 99% confidence level, indicating that after the investment, the patents involved indeed exhibit a reverse technology spillover effect on China.

Figure 1 visually presents the coefficients and the 5% confidence intervals for all study periods before and after the investment events. The graph clearly shows a change in the parallel trend after the technology investment, which reinforces our earlier finding of the existence of a reverse technology spillover effect. The significant positive shift after the investment period further supports the impact of cross-border CVC investments on China’s technological progress.

This figure shows the result from the parallel trend test between invested and uninvested technology. It charts the coefficients of different periods along with the 95% confidence level. The solid line represents the actual coefficient values while the dotted line represents the 95% confidence level.

5.3.2. Regression Analysis of Technology Spill-Over Effect

In addition to the Difference-in-Differences approach, simple regression analysis is also employed to examine the reverse technology spillover effect for each technology category invested in, as indicated in Equation (1). The regression model tests whether the coefficient for the interaction term, Post × Treated, is significantly positive at a 1% significance level. The results of this regression analysis are presented in Table 4.

This table shows the regression result with a difference in patents between the matched experimental group and the control group as the dependent variable. The explanatory variables include Post and interaction variable Post × Treated. αc is the technology type fixed effect and μt is the year fixed effect.

As shown in Table 4, the coefficient for the interaction term Post × Treated is positive and significant at the 1% level. This result supports the existence of a reverse technology spillover effect from China’s CVC overseas investments, confirming Hypothesis 1. The positive coefficient suggests that China’s investments in foreign startups have contributed to technological learning and innovation within the country, thereby enhancing its domestic technological progress.

5.3.3. Robustness Check

To validate the robustness of our results, we performed a placebo test on the DID model. The essence of the placebo test is to simulate a “fake” policy occurrence by applying the analysis to a period before the actual investment, testing if the same policy effects appear. If a significant effect is observed during this placebo test, it would suggest that the findings from the baseline regression are unreliable, potentially due to unobservable factors other than the policy of interest.

In our placebo test, we selected the years prior to the actual investment as the placebo policy. Specifically, for each experimental–control pair, we advanced the investment year by six years, setting the placebo investment year to t−6. As a result, the pre-investment period was adjusted to be from t−12 to t−7, and the post-investment period was shifted to be from t−5 to t−1. The results of this placebo test are presented in

Table 5. Regression results for placebo test.

ln_patents	Coef.	Std.Err.	t	p > |t|	[95% Conf. Interval]
Post	−0.495	−0.068	−0.720	0.469	−0.184	0.085
post_treated	−0.011	0.064	−0.170	0.864	−0.137	0.115
Number of obs	483
R-squared	0.965

, where we analyze whether any significant effects were observed.

As shown in Table 5, the p-value for the interaction term coefficient Post × Treated is positive but not statistically significant. This indicates that, in the absence of any actual investment, there is no significant difference in the number of patents between the experimental and control groups. The lack of a significant result in the placebo test confirms that the observed reverse technology spillover effect is not due to any unobservable factors or pre-existing trends but rather reflects the actual impact of China’s CVC overseas investments. This further strengthens the evidence for the existence of the reverse technology spillover effect.

5.4. Factors Affecting Reverse Technology Spillover Effects

Regression Analysis of Factors

In this section, we extend the previous analysis by exploring the factors that influence the reverse technology spillover effect. We investigate how various factors, such as the host country’s economic development, technological level, openness, and resource endowments, as well as the characteristics of the investments, may impact the effectiveness of reverse technology spillover from China’s CVC overseas investments. The regression results, presented in

Table 6. Regression results showing the factors that influence the reverse technology spillover effect.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Lagging Periods	0	0	1	1	2	2	3	3	4	4
ln_publications	0.861 ***		0.845 ***		0.894 ***		1.016 ***		1.109 ***
	(22.44)		(21.93)		(25.4)		(26.07)		(25.12)
ln_f_patents		0.629 ***		0.615 ***		0.651 ***		0.766 ***		0.847 ***
		(23.35)		(23.11)		(26.21)		(25.57)		(25.8)
ln_amount	0.136 ***	0.133 ***	0.126 ***	0.114 ***	0.079 ***	0.061 **	0.068 **	0.078 ***	0.04	0.031
	(3.93)	(3.92)	(3.75)	(3.54)	(2.71)	(2.13)	(2.42)	(2.81)	(1.18)	(0.94)
ln_GDP	−0.223 *	−0.257 **	−0.077	−0.108	−0.015	−0.018	−0.119	−0.124	0.173	0.095
	(−1.93)	(−2.27)	(−0.65)	(−0.95)	(−0.15)	(−0.18)	(−1.21)	(−1.27)	−1.58	−0.89
ln_GDPP	0.541 ***	0.490 ***	0.634 ***	0.567 ***	0.501 ***	0.505 ***	0.261 **	0.259 **	0.311 **	0.250 *
	(3.62)	(3.35)	(4.16)	(3.87)	(3.82)	(3.94)	(2.17)	(2.15)	(2.33)	(1.91)
open_ivs	−3.410 ***	−3.193 ***	−2.656 ***	−2.531 ***	−2.764 ***	−2.504 **	−1.740 *	−1.566	0.121	−0.262
	(−3.56)	(−3.42)	(−2.68)	(−2.66)	(−2.75)	(−2.55)	(−1.66)	(−1.50)	(−0.11)	(−0.25)
open_inex	0.021 **	0.023 ***	0.005	0.007	0.005	0.005	0.016 **	0.017 **	−0.009	−0.003
	(2.59)	(2.98)	(0.6)	(0.9)	(0.7)	(0.67)	(2.13)	(2.28)	(−0.99)	(−0.38)
Source	5.726 **	5.057 **	1.318	1.424	0.501	0.75	−0.12	0.124	−0.779	−0.688
	(2.35)	(2.12)	(0.57)	(0.64)	(0.25)	(0.39)	(−0.07)	(0.07)	(−0.43)	(−0.39)
ln_distance	−1.150 ***	−1.084 ***	−0.824 ***	−0.725 ***	−0.853 ***	−0.855 ***	−0.829 ***	−0.813 ***	−0.545 **	−0.522 **
	(−4.34)	(−4.19)	(−3.03)	(−2.77)	(−3.66)	(−3.75)	(−3.86)	(−3.79)	(−2.37)	(−2.32)
Constant	3.927	6.990 ***	−0.364	2.494	0.6	3.581 *	2.448	5.298 **	−3.825	0.437
	(1.57)	(2.9)	(−0.15)	(1.05)	(0.28)	(1.7)	(1.14)	(2.51)	(−1.63)	(0.2)
Observations	450	450	401	401	366	366	306	306	234	234
R-squared	0.777	0.787	0.774	0.791	0.836	0.842	0.872	0.872	0.89	0.894

t-statistics in parentheses and *** p < 0.01, ** p < 0.05, * p < 0.1.

, provide insight into the relative importance of these factors in driving the observed spillover effect.

This table shows the regression results using Ln_Patentschina,c,t+I as the dependent variables. Our explanatory variables include ln_Publications or ln_f_patents and ln_amountf,c,t. The control variables include VC_stage, ln_GDP, ln_GDPP, open_inx, open_inex, Source, ln_distance, etc. Models (1) to (10), respectively, represent the regression results when the dependent variable is lagged by 0–4 periods. The purpose of this regression analysis is to test Hypothesis 2: the technological level of the host country has a significant positive impact on the reverse technology spillover effect of cross-border CVC, and Hypothesis 3: the investment level of the home country has a significant positive impact on the reverse technology spillover effect. The number in the parentheses shows the t value.

As shown in Table 6, the coefficients for the variables “Publications” and “f_patent” are significantly positive at a 99% confidence level across all models, confirming that higher technological levels in the host country, as measured by these two factors, lead to a more pronounced reverse technology spillover effect. This supports Hypothesis 2. Additionally, the variable “ln_amount” shows positive coefficients at a 95% confidence level for lag periods 1 to 3, suggesting that larger investment amounts by China’s CVCs contribute to stronger reverse technology spillovers, thus confirming Hypothesis 3. However, when considering a lag of four periods, the coefficient for “ln_amount” becomes insignificant, indicating that the impact of investment amounts diminishes with longer time lags.

Furthermore, the negative and significant coefficients for “ln_distance” across all models suggest that greater geographical distance between the invested country and China negatively affects the reverse technology spillover. This aligns with the general understanding that longer distances hinder CVC overseas investments, thereby reducing spillover effects.

Interestingly, the coefficient for “ln_GDP” is significantly negative, while “ln_GDPP” is significantly positive when no lag is applied. This suggests that higher economic levels in the invested country might reduce reverse technology spillover to China, while higher labor costs in those countries seem to increase the spillover effect. This counterintuitive result could be due to China’s limited investments in highly developed countries. In such regions, the presence of other investors might dilute China’s contribution, leading to a weaker reverse technology spillover effect.

6. Conclusions

In this study, we investigate the reverse technology spillover effects of China’s overseas corporate venture capital (CVC) investments on domestic technology development. Our findings confirm that such investments have a significant and positive impact on China’s technological progress, which supports our Hypothesis 1.

We further examine the factors influencing this spillover effect, with the following key insights:

• Technological Development in Host Countries: we find that the higher the technological level of the host country, measured by indicators such as the number of scientific publications and patent applications, the stronger the reverse technology spillover effect on China’s innovation, which is consistent with our Hypothesis 2
• Investment Amount and Lag Effect: while larger investments initially seem to reduce the impact of reverse technology spillovers, within a certain lag period, the greater the investment amount and investment level positively correlate with more pronounced spillover effects, which partially support our Hypothesis 3

Our paper makes several key contributions to the literature and to practice. First, cross-border CVC investment, which has been underexplored in venture capital research, is systematically examined in the context of reverse technology spillover effects. This research fills a significant gap by offering new insights into how China’s overseas CVC investments influence domestic technological progress. Second, from an enterprise perspective, understanding the factors that shape reverse technology spillovers provides Chinese companies with valuable guidance on how to strategically deploy overseas investments to foster innovation and improve technological capabilities at home.

Third, our findings contribute to policymaking by demonstrating how cross-border CVC investments can serve as a tool to enhance national technological development. This has broader implications for shaping policies that encourage international collaboration and innovation, especially in emerging economies. Additionally, our research expands the scope of reverse technology spillover studies beyond traditional outward foreign direct investment (OFDI) models by introducing a novel methodology using company-level data on CVC investments. This granular level of analysis allows for more accurate and actionable insights into how CVC investments influence the technological advancement of both the investing and host countries.

Our study also contributes to the growing field of technology innovation by offering a deeper understanding of the interplay between foreign investments, technological spillovers, and national innovation systems. By identifying factors that amplify or dampen reverse spillover effects—such as the technological level of host countries and investment size—we provide a clearer roadmap for companies seeking to maximize the returns on their overseas investments.

Lastly, we introduce a fresh perspective on how investment stages and industries affect the spillover process, which helps explain the complex dynamics of international venture capital. The use of detailed, company-level patent data in our study sets it apart from previous research that mainly relied on national-level data, offering richer and more precise insights into the drivers of technological progress.

Looking ahead, we plan to expand this research by exploring how the heterogeneity of CVC investments—across different industries and stages—affects spillover outcomes. We also aim to extend the analysis to examine the impact of CVC investments on technology innovation at the individual company level, offering a more nuanced understanding of how investments translate into tangible innovation outcomes.