The R&D subsidiary highlights the challenges of power distribution and resource allocation in the innovation processes of BGs, yet existing studies have not thoroughly analyzed this issue. In the field of organizational management theory, BGs can be managed through two primary modes: centralized and decentralized (Chi et al, 2021). In a centralized mode, decision-making and management authority are highly concentrated in the hands of upper management, with lower-level managers and employees responsible solely for implementing decisions and orders from the upper management. In a decentralized mode, decision-making and management authority are distributed across various levels and departments, allowing managers and employees at all levels to exercise corresponding authority within their roles. In the theory of internal capital markets, information asymmetry within BGs is minimized, thereby enhancing the efficiency of internal resource allocation (Belenzon and Tsolmon, 2016; Manikandan and Ramachandran, 2015; Mukherjee et al, 2018). Innovations in BGs, unlike general business activities, are characterized by high risk and significant investment, necessitating more effective mechanisms for power distribution and resource allocation.

Existing studies have usefully discussed the challenges of power distribution and resource allocation in the production and operation processes of BGs. Centralization has both advantages and disadvantages. Its advantage lies in reconciling conflicts of interest and incompatibilities within BGs, thereby improving the efficiency of BG’s overall resource use (Keupp et al, 2011). However, centralization can create strong bureaucratic control, weaken employee initiative, and reduce the efficiency of knowledge transfer, adversely affecting BGs (Kastl et al, 2013). Similarly, decentralization presents a range of potential benefits and challenges. It allows local resources to be combined with the power held by subsidiaries, enhancing decision-making efficiency (Geleilate et al, 2020) and stimulates subsidiary initiative and contribution (Birkinshaw and Hood, 1998). However, decentralization can make curbing opportunistic behavior in subsidiaries difficult (Aghion and Tirole, 1997). Beyond these theoretical studies, scholars have examined concrete issues in power distribution and resource allocation. Strategic power decentralization negatively impacts firm performance (Efferin and Hopper, 2007), whereas operational power decentralization positively influences firm performance (Gammelgaard et al, 2012). However, these studies have not focused on power distribution and resource allocation in the innovation processes of BGs. This study adopts the R&D subsidiary as a research lens to investigate the impact of R&D independence on the innovation of BGs, thereby contributing to the existing body of research on power distribution and resource allocation within BGs, and expanding our understanding of the drivers of innovation.

The dilemmas of power control and resource competition are prevalent within BGs, negatively impacting innovation. First, from the perspective of management models, BGs typically adopt a high level of power control. A clear principal-agent conflict exists between the management of parent company and that of subsidiaries. The management of parent company aims to maximize the group’s value, whereas the subsidiary management seeks to maximize its own value. To mitigate the moral hazard and adverse selection associated with subsidiary management, the parent company enhances supervision and incentives, which manifests as increased control over the subsidiary’s management, personnel, and finances. However, innovation, being a highly flexible process, requires a significant degree of autonomy for innovator (Kastl et al, 2013). Consequently, an increase in power control may have a detrimental impact on innovation within BGs. Second, from a resource allocation perspective, intense competition for resources is prevalent within BGs. To achieve rapid scaling, various business units within BGs compete for resources from the headquarters. This competition frequently results in resource mismatches and reduces allocation efficiency. However, innovation, requiring substantial investment, demands that innovative entities receive significant resource support (Ederer and Manso, 2013). Therefore, increased resource competition can detrimentally affect innovation within BGs.

The R&D subsidiary serves as a crucial mechanism for enhancing R&D independence within BGs, addressing challenges related to power control and resource competition during the innovation process. First, the R&D subsidiary functions as a “power aggregator”. On one hand, the allocation of decision-making power within BGs is influenced by the minimization of agency and information costs, the latter encompassing both the costs of information transfer and the opportunity costs due to information scarcity (Geleilate et al, 2020). Unlike general business operations, innovation activities incur high knowledge transfer costs. This becomes challenging for BGs to exert centralized control over the R&D subsidiary. Conversely, the R&D subsidiary enhances R&D efficiency by retaining a degree of decision-making autonomy while ensuring strategic coherence (Kastl et al, 2013). Therefore, the R&D subsidiary is better suited to possess greater decision-making autonomy. Consequently, the power aggregation effect of the R&D subsidiary can effectively mitigate the power control challenges within BGs’ innovation processes, thereby fostering innovation. Second, the R&D subsidiary acts as a “resource aggregator”. Subsidiaries embody the strategy of the parent company within BGs, with the parent company channeling resources to the subsidiaries in alignment with this strategy, thereby establishing a “strategic infrastructure” (Gulati and Gargiulo, 1999). The R&D subsidiary represents the delegation of R&D activities by BGs. By reallocating internal R&D resources to the R&D subsidiary, BGs achieve resource integration, reduce the costs associated with resource transfer and internal communication, and meet the resource demands of the innovation process. Consequently, the resource aggregation effect of the R&D subsidiary can effectively alleviate resource competition challenges, thereby promoting innovation within BGs. Based on the preceding analysis, we present the following hypothesis:

Hypothesis 1: The R&D subsidiary can promote innovation within BGs.

The R&D subsidiary can foster innovation within BGs through mechanisms for internal power decentralization and external resource acquisition.

The R&D subsidiary can facilitate internal power decentralization, thereby fostering innovation within BGs. The key powers influencing innovation within BGs include management, personnel, and financial power. First, management power pertains to the authority to make decisions (Finkelstein, 1992). The R&D subsidiary’s management possesses more market information pertinent to the firm’s R&D activities than the parent company’s management. Decentralizing management power allows the R&D subsidiary’s management to align R&D activities with market information, leading to more efficient and effective decision-making. This decentralization signifies the parent company’s trust and recognition of the R&D subsidiary’s management, which can motivate R&D activities and enhance organizational incentives (Eklund, 2022). Second, personnel power involves the authority to hire, appoint, and dismiss employees. Integrating human capital with corporate resources is central to innovation within BGs (Belloc, 2012). Compared to the parent company’s management, the R&D subsidiary’s management is more familiar with its human resources, enabling more effective decisions regarding job allocation, hiring, and compensation. Decentralizing personnel power allows the R&D subsidiary to recruit higher-quality R&D personnel and implement stronger incentive policies, thereby fostering innovation within BGs. Third, financial power pertains to the ability to manage financial resources. The high investment and risk associated with R&D activities necessitate that R&D entities have sufficient and flexible financial resources. R&D funding is crucial for ensuring high-quality innovation activities, and a lack thereof can significantly diminish the R&D subsidiary’s motivation. Decentralizing financial power can enhance the R&D subsidiary’s ability to utilize cash effectively and improve capital efficiency, thereby promoting innovation within BGs. Thus, the R&D subsidiary can assist BGs in optimizing internal power distribution, thereby fostering innovation.

The R&D subsidiary can enhance external resource acquisition, thereby fostering innovation within BGs. External resource acquisition primarily involves acquiring government and market resources. Regarding government resources, there is strong government support for BGs to engage in innovative activities, including R&D-related subsidies and tax incentives. Acquiring government resources can lower enterprises’ R&D costs and reduce the financial burden on corporate funds for R&D activities (Howell, 2017). Additionally, government resource acquisitions have a signaling effect, conveying positive messages about the enterprise’s R&D activities, thereby facilitating further market resource acquisition. In terms of market resources, the R&D subsidiary enables enterprises to access additional equity and debt financing. Due to the specialized nature of innovation activities and the tendency for firms to withhold information on developing products and technologies to prevent knowledge leakage, there exists significant information asymmetry in enterprise innovation (Chan et al, 2001). The R&D subsidiary can reduce information asymmetry for equity investors, enabling them to better assess the value of BGs’ innovations and enhancing their expectations for future performance, thereby attracting more equity financing (Hu and Jefferson, 2004). Besides equity financing, the R&D subsidiary also facilitates greater access to debt funding. Creditors are more inclined to offer low-cost debt financing to the R&D subsidiary due to its robust cash flow, supported by group headquarters, making the R&D subsidiary more financially stable. Thus, the R&D subsidiary aids BGs in securing more external resources, ultimately advancing their innovation. Drawing upon the preceding analysis, we propose the following hypothesis:

Hypothesis 2a: The R&D subsidiary can foster innovation within BGs through the internal power decentralization mechanism.

Hypothesis 2b: The R&D subsidiary can foster innovation within BGs through the external resource acquisition mechanism.

The R&D subsidiary is more effective in fostering innovation within BGs when the economic policy uncertainty is high. On one hand, heightened economic policy uncertainty can subject BGs to significant challenges due to dramatic fluctuations in the external environment and continual changes in economic policies (Shi et al, 2020). The centralized management model often lacks the flexibility needed for BGs to effectively navigate a complex and evolving economic environment. To address the increased economic policy uncertainty, BGs are incentivized to establish an R&D subsidiary. This subsidiary reduces information transmission and opportunity costs through internal power decentralization, enhance the decision-making efficiency of R&D teams, and consequently promote innovation within BGs. On the other hand, increased economic policy uncertainty exacerbates information asymmetry between BGs and external stakeholders, potentially hindering access to external resources (Nagar et al, 2019). The enhanced innovation efficiency provided by the R&D subsidiary can facilitate greater access to government resources, including R&D subsidies and tax incentives. Additionally, the R&D subsidiary improves the level of R&D-related information disclosure and shorten the communication chain between BGs and external stakeholders. These improvements can mitigate the resource mismatches caused by information asymmetry, thereby assisting BGs in acquiring more market resources and fostering innovation. Based on the above analysis, we propose the following hypothesis:

Hypothesis 3: The role of the R&D subsidiary in fostering innovation within BGs becomes more pronounced when the economic policy uncertainty is high.

The R&D subsidiary is more effective in fostering innovation within BGs when the market competition is fierce. On one hand, fierce market competition accelerates the rate of product iteration, necessitating that BGs efficiently align with market demands (de Bettignies et al, 2023). Centralized management models, characterized by long decision chains and slow market responses, are unable to adapt effectively to rapidly changing market demands. The R&D subsidiary facilitates the decentralization of internal power by distributing management, personnel, financial, and other authorities to innovation teams closer to the market. This enables rapid adaptation to market changes, fostering innovation and providing BGs a competitive edge. On the other hand, heightened market competition exacerbates financial constraints, posing greater challenges for enterprises competing for limited resources (Zhang and Zhou, 2022). As a tangible embodiment of the parent company’s innovation strategy, the R&D subsidiary exhibits greater specialization and independence. This enhances their credibility in capital markets, attracting more equity and debt funding to support innovation within BGs. Building on the preceding analysis, we propose the following hypothesis:

Hypothesis 4: The role of the R&D subsidiary in fostering innovation within BGs becomes more pronounced when the market competition is fierce.

The R&D subsidiary is more effective in fostering innovation within BGs when the institutional distance between the parent company and the subsidiary is significant. On one hand, a substantial institutional distance signifies that the subsidiary encounters an external environment, policies, regulations, and market rules that are markedly different from those of the parent company. In such situations, R&D activities directed by the group headquarters tend to suffer from reduced information transfer speed and decision-making efficiency, complicating the management of institutional differences among various subsidiaries (Ciabuschi et al, 2012). The R&D subsidiary can assist BGs in adopting a more flexible organizational structure and decision-making process to address the diverse innovation needs arising from institutional differences among subsidiaries. This approach accelerates the innovation process, promotes technological breakthroughs, and enhances the market competitiveness of BGs. On the other hand, a greater institutional distance often leads to increased information asymmetry, which not only complicates external financing for BGs but also results in inefficient resource allocation (Liu et al, 2020). The R&D subsidiary, with their focus on technological innovation, is more likely to cultivate strong relationships with government bodies, financial institutions, and other external stakeholders, thereby securing additional resources to support the innovation efforts of BGs. Based on the preceding analysis, we propose the following hypothesis:

Hypothesis 5: The role of the R&D subsidiary in fostering innovation within BGs becomes more pronounced when the institutional distance between the parent company and the subsidiary is significant.

The theoretical framework is illustrated in Fig. 1.

On January 1, 2008, the Enterprise Income Tax Law of the People’s Republic of China, together with its implementing regulations, came into effect. This legislation introduced tax incentives for enterprise innovation, such as preferential tax rates for high-tech enterprises and additional deductions for R&D expenses, which significantly stimulated innovation among enterprises. Since then, tax policies favoring enterprise innovation have been gradually standardized and systematized. Therefore, to mitigate the impact of these tax policy shocks, we selected a sample of A-share listed BGs from the Shanghai and Shenzhen Stock Exchanges, covering the period from 2008 to 2021. BGs are enterprises that have at least one subsidiary. Data on corporate innovation and subsidiary information were obtained from the Chinese Research Data Services (CNRDS) Platform, while additional data were sourced from the China Stock Market & Accounting Research (CSMAR) Database.

The sample selection process is as follows: (1) We initially obtained 40,242 firm-year observations by selecting A-share listed BGs on the Shanghai and Shenzhen Stock Exchanges from 2008 to 2021. (2) To mitigate the impact of industry-specific characteristics, we excluded 920 observations in the financial industry. (3) To avoid the influence of abnormal business conditions, we removed 1789 observations of *ST, ST, and PT companies. (4) We excluded 3502 observations with missing data on key variables. Finally, 34,031 observations were obtained after the above processing. Table 1 provides details of the sample selection process.

(1) Dependent variable. The dependent variable in this study is the innovation. Following the methodologies of previous studies (Chen et al, 2015; Yuan and Wen, 2018), we employ the natural logarithm of total patent applications plus one as a proxy for total innovation within BGs (TPatent). Additionally, we differentiate between invention innovation (QPatent) and non-invention innovation (NPatent), which are measured by the natural logarithm of invention patent applications plus one, and the natural logarithm of non-invention patent applications (including utility model and design patents applications) plus one, respectively. The quality of invention innovation is superior to that of non-invention innovation.

(2) Independent variable. The dependent variable in this study is the R&D subsidiary (RDSubsidiary). We employ a dummy variable to indicate whether an enterprise has established an R&D subsidiary, using this as a proxy for the R&D subsidiary. The variable RDSubsidiary is assigned a value of 1 if the enterprise has at least one R&D subsidiary in the year of observation, and 0 otherwise. The identification of the R&D subsidiary is primarily based on the name and business scope of the R&D subsidiary. If the name and business scope include keywords such as “R&D”, “invention”, “innovation”, “development”, “experiment”, etc., the subsidiary is classified as an R&D subsidiary.

(3) Control variables. Referring to previous studies (Chang et al, 2015; Galasso and Simcoe, 2011; Wang and Zhao, 2015), the following variables are controlled: enterprise size (Size), financial leverage (Lev), return on assets (ROA), fixed assets ratio (Fixed), cashflow ratio (Cashflow), enterprise age (Age), enterprise growth (Growth), nature of ownership (SOE), percentage of independent directors (Indep), management ownership (Mshare), institutional ownership (Ishare), and book-to-market ratio (BM). In addition, we also control for year (Year FE) and industry (Industry FE) fixed effects. Continuous variables are winsorized at the 1st and 99th percentiles to address outliers. Table 2 provides a comprehensive description of the primary variables.

To analyze the impact of the R&D subsidiary on innovation within BGs, we developed a one-stage lagged empirical model to address potential bidirectional causality issues, as follows:

(1)${\text{ TPatent }}_{it}/{\text{ QPatent }}_{it}/{\text{ NPatent }}_{it}=\alpha +\beta {\text{ RDSubsidiary }}_{it-1}$ $+ \gamma {\text{ Controls }}_{it-1}+\text{Year}+\text{Industry}+{\epsilon }_{it}$

where, i and t denote the firm and year, respectively; TPatent, QPatent, and NPatent correspond to total innovation, invention innovation, and non-invention innovation, respectively; RDSubsidiary refers to the R&D subsidiary; Controls denotes control variables; Year and Industry indicate year and industry fixed effects, respectively; $\epsilon$ represents the random error term.

Table 3 presents descriptive statistics of the main variables in this study. The mean values for TPatent, QPatent, and NPatent are 2.413, 1.663, and 1.922, respectively, indicating that the level of total innovation within BGs is low, with invention innovation being even lower compared to non-invention innovation. The standard deviations for TPatent, QPatent, and NPatent are 1.751, 1.508, and 1.652, respectively, indicating significant variability in innovation within BGs. The mean value for RDSubsidiary is 0.101, suggesting that approximately 10.1% of the sample has an R&D subsidiary.

Before performing the multiple regression analysis, we first conducted univariate and correlation tests, with the results shown in Table 4. Panel A of Table 4 displays the outcomes of the univariate tests. For total innovation (TPatent), the group with an R&D subsidiary exhibits a mean value 1.035 higher than the group without one, and this difference is statistically significant at the 1% level. Regarding invention innovation (QPatent), the group with an R&D subsidiary shows a mean value 0.953 higher than the group without one, with the difference also statistically significant at the 1% level. Similarly, for non-invention innovation (NPatent), the group with an R&D subsidiary reports a mean value 0.870 higher than the group without one, and this difference is likewise statistically significant at the 1% level. These statistical results suggest that BGs with an R&D subsidiary demonstrate a substantial increase in both the quantity and quality of innovation.

Correlation tests were conducted on the dependent and independent variables, and the results are presented in Panel B of Table 4. The correlation coefficients of RDSubsidiary with TPantnt, QPatent, and NPatent are 0.178, 0.190, and 0.159, respectively, all of which are significantly positive at the 1% level. This indicates a significant positive correlation between the R&D subsidiary and overall innovation, invention innovation, and non-invention innovation. The R&D subsidiary significantly enhances both the quantity and quality of innovation within BGs.

Table 5 displays the regression outcomes concerning the impact of the R&D subsidiary on corporate innovation. Columns (1) and (2) present the results for overall innovation, whereas columns (3) and (4) detail the outcomes for invention and non-invention innovations. Columns (1) and (2) indicate that the regression coefficients of RDSubsidiary on TPatent are 0.711 and 0.311, respectively, both significant at the 1% level. This suggests that the R&D subsidiary exert a substantial positive influence on overall innovation within BGs. Enterprises with an R&D subsidiary exhibit a 31.1% increase in overall innovation compared to those without such a subsidiary. Columns (3) and (4) reveal that the regression coefficients of RDSubsidiary on QPatent and NPatent are 0.343 and 0.256, respectively, both significant at the 1% level. This demonstrates that the R&D subsidiary enhance both invention and non-invention innovation within BGs. Enterprises with an R&D subsidiary experience a 34.3% increase in invention innovation and a 25.6% increase in non-invention innovation compared to those without an R&D subsidiary. These results indicate that the R&D subsidiary enhances both the quantity and quality of innovation within BGs, as demonstrated by a marked increase in total patent applications and a substantial rise in invention patent applications.

After conducting the analysis, several issues have been identified that may impact the reliability of our findings. First is the issue of omitted variables. Although we have controlled for basic firm-level characteristics, corporate governance, industry fixed effects, and year fixed effects, there may still be significant factors influencing innovation and the establishment of the R&D subsidiary that are unaccounted for, leading to the omitted variable issue. Second is the issue of sample self-selection. There may be inherent differences between BGs that establish an R&D subsidiary and those that do not, leading to a sample self-selection issue. Third is the issue of bidirectional causation. More innovative BGs may have greater incentives to establish an R&D subsidiary to maintain their leading innovation status, thereby achieving higher innovation outputs. Fourth is the issue of variable measurement bias. Although patent applications can comprehensively reflect corporate innovation output, this indicator may be subject to management manipulation, thereby impacting the reliability of the empirical results. Therefore, we conduct a series of robustness tests.

(1) Instrumental variable method. To address the issues of omitted variables and bidirectional causality, we employ an instrumental variable approach. We construct the instrumental variable for R&D subsidiaries by utilizing the increase in China’s high-tech enterprise recognition standards as an exogenous shock. In 2016, China’s Ministry of Science and Technology, Ministry of Finance, and State Administration of Taxation issued the Measures for the Administration of the Recognition of High-tech Enterprises, which raised the standards for recognizing high-tech enterprises. High-tech enterprises benefit from a reduced corporate income tax rate of 15%, in contrast to the 25% rate for general enterprises. When the recognition standards for high-tech enterprises are elevated, BGs have increased incentives to establish R&D subsidiaries. This is because setting up an R&D subsidiary facilitates meeting the recognition standards, thereby allowing BGs to access greater tax incentives. Simultaneously, the elevation of high-tech enterprise recognition standards is an exogenous event for these BGs. To identify the increase in recognition standards for high-tech enterprises, we construct an interaction variable (Treat$\times$Post), where Treat is a group variable assigned a value of 1 for non-high-tech enterprises in 2015 and prior, and 0 otherwise; Post is a time variable, assigned a value of 1 for the years 2016 and beyond, and 0 otherwise.

The regression results using the instrumental variable method are presented in columns (1) to (4) of Table 6. The weak instrumental variables test results indicate that the Cragg-Donald Wald F statistic is 30.947, exceeding the 10% critical value of 16.38. This suggests that the increase in the recognition standard of high-tech enterprises is not a weak instrumental variable. Column (1) presents the first-stage results, where the coefficient of Treat$\times$Post on RDSubsidiary is 0.038 and is statistically significant. This indicates that the increased recognition standard for high-tech enterprises encourages the establishment of R&D subsidiaries within BGs, aligning with our theoretical expectations. Columns (2), (3) and (4) present the second-stage results, where the regression coefficients of RDSubsidiary on TPatent, QPatent, and NPatent are positive and significant at the 1% level. These findings suggest that our conclusions remain robust after addressing the omitted variable issue and the bidirectional causation problem.

(2) Treatment effect model. To address the issue of sample self-selection, we employ a treatment effect model. In the initial stage, we introduce an exogenous instrumental variable regarding the increase in high-tech recognition standards (Treat$\times$Post). The results of the treatment effect model are presented in columns (5) through (8) of Table 6. Column (5), using the probit estimation method, indicates that the estimated coefficient of Treat$\times$Post on RDSubsidiary is significantly positive, suggesting that an increase in high-tech recognition standards enhances the likelihood of BGs establishing R&D subsidiaries. Columns (6) through (8) demonstrate that the regression coefficients of RDSubsidiary on TPatent, QPatent, and NPatent are all significant at the 1% level. These results suggest that, after mitigating the issue of sample self-selection, the R&D subsidiary continue to enhance both the quantity and quality of innovation within BGs.

(3) Propensity score matching method. To further address the issue of sample self-selection, we employ the propensity score matching method. Control variables are used as matching variables, and we conduct 1:2 nearest neighbor matching with replacement. The results from the propensity score matching method are presented in Table 7, columns (1) through (3). The regression coefficients of RDSubsidiary on TPatent, QPatent, and NPatent are all significantly positive. These findings suggest that the conclusion regarding the R&D subsidiary’s role in promoting innovation within BGs remains valid even after applying the propensity score matching method.

(4) Time-varying DID approach. To more accurately identify the causal relationship between the R&D subsidiary and innovation within BGs, we employ a time-varying difference-in-differences (DID) approach to enhance the robustness of our results. We construct the following model:

(2)$\mathrm{ T}Paten{t}_{it}/QPaten{t}_{it}/NPaten{t}_{it}=\alpha +\beta D{u}_u\times D{t}_t$ $+ \gamma Control{s}_{it}+Year+Industry+{\epsilon }_{it}$

where, Du$\times$Dt refers to the R&D subsidiary. Du is a group variable that equals 1 if an enterprise establishes an R&D subsidiary and 0 otherwise. Dt is a time variable that equals 1 for the year an enterprise first establishes an R&D subsidiary and all subsequent years, and 0 otherwise. The Du$\times$Dt represents the interaction between the group variable (Du) and the time variable (Dt), used to assess the long-term impact of the R&D subsidiary on innovation within BGs.

The results of the time-varying DID analysis are presented in columns (4) through (6) of Table 7. The variable RDSubsidiary is positive and significant at the 1% level for TPatent, QPatent, and NPatent. These findings indicate that our conclusions remain valid after employing the time-varying DID approach.

(5) Multidimensional fixed effect. To mitigate the influence of firm-specific intrinsic factors on our empirical results, we incorporate firm fixed effects, with the outcomes presented in columns (1) to (3) of Table 8. After accounting for firm fixed effects (Firm FE), the regression coefficients of RDSubsidiary on TPatent, QPatent, and NPatent are 0.097, 0.087, and 0.145, respectively. All coefficients are statistically significant, indicating that the R&D subsidiary exerts a substantial positive influence on overall innovation, invention innovation, and non-invention innovation within BGs. These findings suggest that our conclusions remain robust despite the inclusion of firm fixed effects.

(6) Alternative measures of dependent variable. To address the issue of variable measurement bias, we refer to Bereskin et al (2016), which evaluates innovation through patent acquisition. The following metrics are constructed: TPatent_acq, calculated as the natural logarithm of one plus the total number of patent acquisition; QPatent_acq, determined by the natural logarithm of one plus the number of invention patent acquisition; and NPatent_acq, measured by the natural logarithm of one plus the number of non-invention patent acquisition. The results of the alternative measures for key variables are presented in columns (4) through (6) of Table 8. The regression coefficients of RDSubsidiary on TPatent_acq, QPatent_acq, and NPatent_acq are 0.342, 0.374, and 0.298, respectively, all of which are statistically significant, indicating that the R&D subsidiary enhances the quantity and quality of innovation within BGs. These findings indicate that our conclusions remain robust even after accounting for the issue of variable measurement bias.

The decentralization of internal power induced by the R&D subsidiary primarily manifests in the distribution of management, personnel, and financial power. For management power, we reference the studies of Fisman and Wang (2010) and Gu (2017), using the ratio of related-party transactions to operating revenue as a measure of management power decentralization (AD_ma). A lower ratio indicates greater decentralization of managerial power. For personnel power, the natural logarithm of one plus the number of R&D staff serves as a proxy for personnel power decentralization (AD_st). A larger indicator suggests greater decentralization of personnel power. For financial power, the ratio of cash held by subsidiaries is used as a proxy for financial power decentralization (AD_ca). This indicator is calculated as one minus the ratio of monetary cash in the parent company’s financial statements to that in the consolidated financial statements. A larger indicator suggests greater decentralization of cash power.

The results of the internal power decentralization mechanism are presented in columns (1) through (3) of Table 9. Column (1) indicates that the regression coefficient of RDSubsidiary on AD_ma is –0.026, significant at the 1% level, suggesting that the R&D subsidiary induces decentralization in management power. Column (2) indicates that the regression coefficient of RDSubsidiary on AD_st is 0.314, significant at the 1% level, suggesting that the R&D subsidiary causes decentralization in personnel power. Column (3) indicates that the regression coefficient of RDSubsidiary on AD_ca is 0.038, significant at the 1% level, suggesting that the R&D subsidiary causes decentralization in financial power. These decentralizations of internal power enable BGs to engage in innovative activities with greater flexibility.

The acquisition of external resources facilitated by the R&D subsidiary is primarily evident in two areas: government resources and market resources. In terms of government resource acquisition, our focus is primarily on government subsidies and tax incentives. Following Chen et al (2010), we construct two indicators: government subsidies (RA_gs) and tax incentives (RA_ti). Government subsidies are quantified using the natural logarithm of one plus the government subsidies received by BGs, while tax incentives are measured as the ratio of tax rebates received by BGs to the sum of tax rebates and taxes paid by BGs. Regarding market resource acquisition, we concentrate on financing constraints. We reference the study by Kaplan and Zingales (1997) and Yu et al (2024), utilizing the Kaplan and Zingales (KZ) index to evaluate the firm’s financing constraints (RA_fc).

The outcomes of the external resource acquisition mechanism are presented in columns (4) through (6) of Table 9. Column (4) indicates that the regression coefficient of RDSubsidiary on RA_gs is 0.222, significant at the 1% level, suggesting that the R&D subsidiary facilitates BGs in acquiring more government subsidies. Column (5) reveals that the regression coefficient of RDSubsidiary on RA_ti is 0.020, significant at the 1% level, indicating that the R&D subsidiary enhances BGs’ receipt of tax incentives. Column (6) demonstrates that the regression coefficient of RDSubsidiary on RA_fc is –0.078, significant at the 1% level, indicating that the R&D subsidiary alleviates BGs’ financing constraints. These external resource acquisitions enhance the capacity of BGs to engage in innovative activities.

To analyze the impact of economic policy uncertainty, we employ the methodology of Baker et al (2016) to construct the Chinese economic policy uncertainty index as a proxy indicator of economic policy uncertainty (EPU). We perform group regressions based on the median EPU, with the results presented in Table 10. Columns (1) and (2) indicate that the regression coefficient of RDSusidiary on TPatent is significantly greater in the high economic policy uncertainty group compared to the low economic policy uncertainty group (difference = 0.256, p-value = 0.000). Columns (3) and (4) demonstrate that the regression coefficient of RDSusidiary on QPatent is significantly higher in the high economic policy uncertainty group than in the low economic policy uncertainty group (difference = 0.270, p-value = 0.000). Columns (5) and (6) reveal that the regression coefficient of RDSusidiary on NPatent is significantly higher in the high economic policy uncertainty group compared to the low economic policy uncertainty group (difference = 0.247, p-value = 0.000). These results suggest that the R&D subsidiary are more effective in promoting innovation within BGs under conditions of high economic policy uncertainty. This implies that the R&D subsidiary can aid BGs in effectively responding to the challenges posed by high economic policy uncertainty.

To assess the impact of market competition, we reference the study by Giroud and Mueller (2010), which utilized the Herfindahl-Hirschman index to quantify market competition (MC). We conduct a group regression analysis based on the median MC, with results presented in Table 11. Columns (1) and (2) indicate that the regression coefficient of RDSusidiary on TPatent is significantly greater in the intense market competition group compared to the stable market competition group (difference = 0.410, p-value = 0.000). Columns (3) and (4) demonstrate that the regression coefficient of RDSusidiary on QPatent is significantly higher in the intense market competition group than in the stable market competition group (difference = 0.417, p-value = 0.000). Columns (5) and (6) reveal that the regression coefficient of RDSusidiary on NPatent is significantly higher in the intense market competition group compared to the stable market competition group (difference = 0.392, p-value = 0.000). These findings suggest that the R&D subsidiary is more effective in promoting innovation within BGs under conditions of intense market competition. This implies that the R&D subsidiary can aid BGs in effectively responding to the challenges posed by intense market competition.

To examine the effect of institutional distance between parent and subsidiary, we reference Gaur et al (2007) and Xu et al (2004). We utilize the marketization index of each region in China as a proxy for institutional factors, calculate the absolute difference between the marketization indices of the regions where the parent and each subsidiary are located, and use the average of these differences as a measure of institutional distance (ID). We conduct group regressions based on the median ID, with results presented in Table 12. Columns (1) and (2) indicate that the regression coefficient of RDSusidiary on TPatent is significantly higher in the large institutional distance group compared to the small institutional distance group (difference = 0.390, p-value = 0.000). Columns (3) and (4) demonstrate that the regression coefficient of RDSusidiary on QPatent is significantly higher in the large institutional distance group than in the small institutional distance group (difference = 0.353, p-value = 0.000). Columns (5) and (6) reveal that the regression coefficient of RDSusidiary on NPatent is significantly higher in the large institutional distance group compared to the small institutional distance group (difference = 0.363, p-value = 0.000). These results suggest that the R&D subsidiary is more effective in promoting innovation within BGs when the institutional distance is large. This implies that the R&D subsidiary can assist BGs in effectively addressing the challenges posed by significant institutional distance between parent and subsidiary.

This study investigates the influence of R&D independence on innovation within BGs, focusing on the R&D subsidiary as the research lens. The findings indicate that the R&D subsidiary enhances both the quantity and quality of innovation within BGs. The mechanism analysis reveals that the R&D subsidiary facilitates internal power decentralization and bolster external resource acquisition, thereby fostering innovation within BGs. Internal power decentralization primarily involves the distribution of management, personnel, and financial authority, while external resource acquisition mainly encompasses the procurement of government and market resources. The cross-sectional analysis indicates that the influence of the R&D subsidiary on promoting innovation within BGs is more pronounced under conditions of high economic policy uncertainty, intense market competition, and significant institutional distance between parent and subsidiary. These findings suggest that R&D independence optimizes power distribution and resource allocation within BGs, thereby enhancing their innovation.

Our study serves as a decision-making reference for BGs aiming to enhance their innovation efficiency. First, BGs should proactively establish R&D subsidiaries to fully leverage the positive effects of R&D independence on innovation. BGs frequently encounter the challenge of power concentration and resource competition during the innovation process. By establishing R&D subsidiaries, BGs can enhance their R&D independence. This strategy maximizes the power and resource aggregation effects, thereby enhancing the innovation capacity and capability of BGs, and ultimately improving both the quantity and quality of their innovations. Second, BGs should effectively leverage the internal power decentralization and external resource acquisition mechanisms of R&D subsidiaries to further enhance efficiency. Regarding internal power, BGs should adopt an appropriate decentralization of management, personnel, and financial authority to grant R&D subsidiaries sufficient autonomy, thereby stimulating their motivation and willingness to innovate. With respect to resource acquisition, BGs should capitalize on the advantages of R&D subsidiaries to secure government resources, such as R&D subsidies and tax incentives, as well as market resources from investors and creditors, enhance the innovation capabilities of R&D subsidiaries. Third, BGs must fully exploit the advantages of R&D subsidiaries in addressing high economic policy uncertainty, intense market competition, and significant institutional distance between parent and subsidiary. Confronted with the realities of economic globalization and business diversification, BGs should actively adjust their innovation strategies. Under conditions of high economic policy uncertainty, fierce market competition, and significant institutional distance between parent and subsidiary, a centralized management approach reduces the R&D efficiency of BGs, hindering timely adjustments to R&D strategies in response to policy changes, market demands, and institutional shifts. Additionally, in such scenarios, the resource acquisition capabilities of BGs are also diminished. R&D subsidiaries can assist BGs in promoting internal power decentralization and bolstering external resource acquisition to better address the challenges posed by high economic policy uncertainty, intense market competition, and significant parent-subsidiary institutional distance.