While regional innovation centers (RICs) can encourage corporate green innovation, firms may opportunistically exaggerate their green innovation activities under RICs’ spatial influence, thereby engendering green innovation bubbles. We theorize that geographic proximity to RICs shapes these bubbles by influencing firms’ capability and motivation to exaggerate green outputs. The results indicate that (1) proximity to RICs exhibits an inverted U-shaped relationship with corporate green innovation bubbles; (2) green strategic orientation (GSO) positively moderates this relationship, steepening the inverted U-shaped curve among firms with higher GSO; and (3) political connections shift the peak of the curve, with bubble formation peaking and declining at greater geographic distances. Overall, this study enhances our understanding of geographic influences on green innovation from a strategic response perspective. Moreover, it highlights organizational heterogeneity’s role in shaping how firms interpret and respond to geographic influences.
Regional science and technology innovation centers (RICs) have played a pivotal role in optimizing the regional innovation systems and activities in China (Tao et al., 2025). RICs are defined as cities or regions with a high concentration of science and technology resources and innovation activities (National Development and Reform Commission, 2021). They significantly influence surrounding areas through industrial, financial, and technological spillovers. As core cities within urban agglomerations that concentrate innovation resources and institutions, RICs have become critical innovation hubs in the country (Husted, Jamali, et al., 2016; Wang et al., 2025).
Meanwhile, China’s recent policy emphasis on green and low-carbon development has positioned green innovation as a central performance indicator in regional innovation systems (Hunjra et al., 2024). However, not all innovations necessarily translate to improved environmental sustainability and efficiency (Arun & Ozmutlu, 2025). A growing focus on patent-based innovation indicators has unintentionally driven firms to exaggerate their green outputs through excessive filings to gain policy, reputational, or investment advantages (Liu et al., 2025; Ruan et al., 2024). This behavior has created a corporate green innovation bubble, wherein the quantity of reported green innovations far exceeds their technological or environmental substance (Geng et al., 2024). Such bubbles not only distort innovation resource allocation and mislead policymakers into overestimating regional innovation performance, but also erode public trust in the credibility of green transformation efforts. Paradoxically, RICs’ spatial influence, intended to promote genuine innovation, may accelerate green innovation bubble formation.
While research has illuminated how proximity to RICs shapes substantive green innovation (Hu & Xu, 2023; Sheng & Ding, 2024), little is known about how distance to RICs influence green innovation bubble formation. Building on their resource-supporting role (Husted, Jamali, et al., 2016), proximity to RICs may enhance firms’ capability for symbolic innovation by improving access to policy signals and professional intermediaries that help package and promote green achievements. Thus, such proximity can encourage symbolic patenting and superficial environmental activities which inflate green innovation bubbles. Neglecting these location-induced innovation distortions can lead scholars and policymakers to overestimate RICs’ true effectiveness in fostering substantive green innovation. To address this, we examine how proximity to RICs shapes corporate green innovation bubbles, offering a more fine-grained understanding of the underlying spatial mechanisms.
However, despite strengthening firms’ bubble-generating capability, proximity to RICs also entails stronger oversight. This restrains overly symbolic innovation behavior (Liao & Zhang, 2024). Building on Oliver’s (1991) view of firms as active interpreters of institutional pressures, we theorize that proximity to RIC conditions firms’ motivation and capability to engage in symbolic green innovation, thereby influencing the extent of their green innovation bubbles. Firms near RIC cores possess stronger symbolic capabilities but limited motivation to inflate their outputs. Meanwhile, those farther away exhibit stronger motives to exaggerate but insufficient capacity to effectively do this. The interaction between rising motivation and declining capability across distance produces a potentially inverted U-shaped relationship between proximity to RICs and green innovation bubbles. Understanding this non-linear pattern can help identify high-risk zones for green innovation bubbles within RICs, thus optimizing the spatial allocation of innovation resources and regulatory efforts.
Economic geography research has largely attributed proximity effects on innovation to the spatial configuration of local resources and stakeholders, emphasizing how location shapes firms’ access to knowledge, funding, and oversight (Boschma, 2005). However, little is known about how such geographic influences are conditioned by firm-level characteristics. Proximity effects likely depend on how firms interpret and respond to institutional and resource environments. Accordingly, we examine two salient firm-level moderators: green strategic orientations (GSO) and political connections.
GSO reflects a firm’s long-term commitment to improve environmental performance (Hong et al., 2009). Firms with a stronger GSO are more responsive to policy incentives and reputational pressures (Chen et al., 2024; Zhang et al., 2025). Political connections reflect firms’ embeddedness in political networks, providing access to policy resources and shielding them from competitive pressures (Acemoglu & Robinson, 2006; Goldman et al., 2013). Such institutional embeddedness may shift the overall firm motivation and capability for innovation bubbles, thereby shifting the inverted U-curve’s position. Examining these two moderators can explain why firms located at similar distances from RICs can exhibit markedly different levels of green innovation bubbles, revealing the contingent character of geographic influences. Thus, we further bridge economic geography with strategic management and institutional perspectives, showing that the proximity effects are shaped by organizational attributes differences.
Our study contributes to the literature in three ways. First, we extend research on geography and green innovation by advancing the understanding of how proximity to RICs not only stimulates green innovation but may also generate its inflated, symbolic counterpart: green innovation bubbles. Studies have examined how urban geography enhances green innovation efficiency, quantity, and quality (Hu & Xu, 2023; Li et al., 2022; Wang et al., 2023). Yet, few consider that such innovation may also involve inflated outputs. Through a geographic proximity lens, we examine how RICs may inflate green innovation bubbles among nearby firms, revealing the distorted consequences of their spatial influence.
Second, economic geography often treats green innovation as a knowledge spillover outcome whose effects diminish with distance (Döring & Schnellenbach, 2006; Obschonka et al., 2023; Xu et al., 2022). However, few consider its nature as a strategic choice shaped by firm-specific motivations and capabilities. By theorizing how firms’ motivation and capability function as latent mechanisms that inversely vary with distance to RICs, we propose and verify an inverted U-shaped relationship between proximity and green innovation bubbles. Thus, we respond to the call of Sorenson and Baum (2003) for research linking geography with organizational strategy.
Third, research on the geography of innovation often adopts an agglomeration-based perspective, treating firms as homogeneous recipients of knowledge spillovers (Dong et al., 2025; J. Hu et al., 2025). Meanwhile, we show that GSO and political connections moderate how firms interpret and act upon such influences. GSO reflects a firm’s strategic commitment and value alignment with environmental issues (Waqas et al., 2024; J. Xiao et al., 2025). By demonstrating that green-oriented firms exhibit steeper inverted U-shaped patterns with proximity, we show how a value-based strategic logic amplifies the tension between capacity and motivation for green innovation bubbles. Meanwhile, political connections represent a form of institutional embeddedness that grants firms privileged access to resources while reducing competition (Zhong & Zheng, 2025). By changing the baseline motivation and capability levels for green innovation bubbles, political connections shift the inverted U-shaped curve’s peak away from RIC urban cores. Overall, these findings deepen our understanding of how political ties may alter the effects of geographic influences, yielding unintended outcomes such as green innovation bubbles.
Theoretical background and hypotheses developmentSpatial influence of regional innovation centersA regional science and technology innovation center refers to “a city or region where science and technology resources and innovation activities are highly concentrated, innovation capacity is strong, and which plays a leading and radiating role in regional innovation development in terms of industry, talent, capital, technology, and information” (National Development and Reform Commission, 2021). The literature typically defines such centers as growth poles within urban agglomerations or core cities, characterized by dense innovation resources, concentrated innovation activities, significant outcomes, and broad influence (Tao et al., 2025). While international innovation centers focus on global competition and resource allocation, RICs emphasize domestic or regional leadership, bridging global innovation center and local nodes (Feng et al., 2021). Compared to innovative provinces and cities, which stress government coordination of innovation inputs, RICs function more as hubs and growth poles, promoting innovation factors’ concentration and diffusion within urban agglomerations (Chen, 2023).
Since RICs are essentially urban nodes with dense innovation resources and strong policy support, understanding their spatial influence on corporate innovation activity is crucial. This influence typically operates through three main channels: R&D collaboration (Frigon et al., 2025; Yang et al., 2025), talent mobility (He et al., 2025; Qiu et al., 2024), and patent transfer (Ma et al., 2024). Studies show that diverse external collaboration networks, particularly those linking firms to universities and research institutions, serve as critical conduits for material and energy eco-innovations (García-García et al., 2025). Beyond knowledge flows, RICs’ spatial influence is amplified by the concentration and circulation of complementary innovation factors, such as financial resource and professional service (Sheng & Ding, 2024; Tang et al., 2025).
RICs exert spatial influence through three primary modes. The technology-driven mode leverages national-level research platforms to advance basic research and achieve breakthroughs in key technologies, reflecting strong original innovation capacity (Wang et al., 2020; Wu et al., 2024). The policy-driven mode relies on institutional pilots and integrated policy frameworks to create environments conducive to policy diffusion and institutional replication (Liu et al., 2024; Zou et al., 2022). In demand- or context-driven innovation models, cities act as place-based platforms enabling the real-world application and scaling of technological solutions. By integrating users, firms, and research actors into shared urban environments, cities foster experimentation and collaborative problem-solving (Abi Saad & Agogué, 2024; Fastenrath et al., 2023).
Capability–motivation mechanisms in green innovation bubblesFollowing Geng’s et al. (2024) operationalization, we define corporate green innovation bubbles as “situations in which firms’ reported green innovation efforts significantly exceed their actual environmental value.” Such bubbles emerge when firms possess both the capability and motivation to inflate signals. Here, bubble-generating capability does not stem from technological or financial resources for genuine innovation (Byrski & Wang, 2025; Z. Zhang et al., 2024), but from firms’ symbolic and procedural capacity to interpret, imitate, and manipulate green innovation’s evaluative criteria. Conversely, bubble-generating motivation originates from external pressures and incentives which reward visible green outputs, such as policy benefits, performance evaluations, or reputational gains (Chen & Dagestani, 2023; S. Hu et al., 2025; Lu et al., 2025).
The bubble-generating capability depends on the firm’s access to regulatory and evaluative knowledge regarding how green innovation is recognized and rewarded. Firms develop the operational know-how through policy exposure, intermediary interactions, benchmarking practices, and peer observation (Tian et al., 2023; Xu et al., 2024). With richer access to these knowledge and resource providers, they can more accurately imitate prevailing green innovation templates (Husted, Montiel, et al., 2016). Further, they can strategically package low-quality innovation activities as high-value outputs aligned with policymakers’ and investors’ evaluative criteria (Huang et al., 2023). Hence, capability reflects not the firm’s technological strength per se, but its learned proficiency in producing institutionally legible signals.
Meanwhile, bubble-generating motivation stems from the external pressures and incentives to display visible green achievements. These pressures arise from subsidy competition, local government performance systems, ESG-oriented financing channels, and broader legitimacy concerns in markets (Lee & Raschke, 2023; Lian et al., 2022; Wang & Zhang, 2024; W. Zhang et al., 2024). As evaluative pressures or competition for environmental legitimacy intensify, firms become motivated to amplify their reported green activities (Lee & Raschke, 2023). Conversely, bubble-generating motivation declines when regulatory scrutiny or transparent monitoring heightens the exposure risks of symbolic inflation (Li & Gan, 2025). Accordingly, the diminished marginal benefits can weaken firms’ incentives to overstate their environmental performance.
Firms with strong capability but limited motivation have little reason to inflate signals. Meanwhile, those with strong motivation but insufficient capability cannot translate intent into credible outputs. Therefore, green innovation bubbles are most likely when firms possess both the know-how to construct symbolic signals and external incentives rewarding such behavior.
Proximity to regional innovation centers and green innovation bubblesBuilding on this capability-motivation framework, we argue that geographic distance to RICs shapes both mechanisms. Spatial proximity influences how firms access information, resources, institutional oversight, and signaling incentives, thereby distinctly conditioning their capability and motivation. These spatially contingent mechanisms may create a non-linear relationship between distance and green innovation bubbles.
As Figure 1 illustrates, firms situated near RICs’ urban cores benefit from clustered innovation resources, such as university-industry collaborations, green finance platforms, and patent consulting services (Nguyen et al., 2025; Rossi et al., 2024; Song et al., 2020; Yin et al., 2023). These advantages enhance their likelihood to produce both substantive and symbolic green innovation outputs. Further, with stronger locational advantages, such firms have low motivation to engage in symbolic inflation, as proximity to urban areas grants easier access to policy resources and financial support (Husted, Jamali, et al., 2016). This reduces their need to overstate innovation for legitimacy, visibility, or investor signaling. Additionally, firms closer to RICs are subject to more stringent local enforcement (Liao & Zhang, 2024), as competitive innovation environments foster whistleblower mechanisms and reputational scrutiny, thereby dampening the bubbles formation.
Meanwhile, firms located farther from RICs are less embedded in the institutional and informational networks (Sheng & Ding, 2024; Zhong & Li, 2024), which define prevailing green innovation standards. Limited exposure to policy signals, professional intermediaries, and peer imitation effects weakens their capability to exaggerate green innovation outputs. Meanwhile, being distant from policy hubs and competitive visibility strengthens their motivation to signal innovation through symbolic means (i.e., excessively low-quality patent filings) to attract external recognition, collaboration, or subsidies. Furthermore, geographic remoteness is often associated with weaker regulatory enforcement and institutional oversight, creating loopholes for opportunistic behavior (Liao & Zhang, 2024). However, despite stronger motivations, firms in remote locations exhibit relatively low levels of green innovation bubbles. Locational disadvantages constrain their ability to translate such motivation into effective symbolic signaling, illustrating a capability-limited response pattern at the far end of the distance curve.
Together, firms located at moderate distances from RICs are most likely to generate green innovation bubbles. On the one hand, compared to firms in more remote areas, they have better access to resources such as policy information, technology transition, and funding opportunities. This enhances their bubble-generating capability. On the other hand, compared to firms located near urban cores, they face lower regulatory scrutiny and exposure risk, yet experience pressures to offset locational disadvantages through inflated innovation outputs. Consequently, these moderately located firms possess both motivation and capability, making them the most likely to engage in green innovation bubble formation.
Hypothesis 1 Geographic proximity to RICs and green innovation bubbles have an inverted U-shaped relationship.
Although geographic distance influences firms’ capability–motivation configurations, firms are not passive recipients of geographic influences. Their strategic orientation actively shapes how they interpret and respond to external environments in pursuing organizational outcomes (Usman Khizar et al., 2022; S. F. Xiao et al., 2025). Among various strategic orientation types (Theodosiou et al., 2012), GSO reflects a firm’s long-term commitment to planning and implementing business activities that improve environmental performance (Hong et al., 2009). Thus, GSO may alter how firms leverage innovation resources and respond to institutional pressures related to environmental issues, engendering different levels of green innovation bubbles under similar geographic conditions.
Firms with high GSO exhibit stronger internalized environmental values and a greater sensitivity to green-related opportunities. Strategically, these firms are more committed to substantive environmental improvements guided by deeply rooted values and beliefs (Calic & Mosakowski, 2016; Varadarajan, 2017). The sustainability orientation fosters a stronger internal motivation to pursue and demonstrate substantive green innovations (Cheng, 2020). It also increases sensitivity to the reputational risks associated with perceived symbolic actions that lack credible implementation (Jagani & Saboori-Deilami, 2025). Meanwhile, GSO promotes organizational learning and cross-functional integration, which enhances the firm’s knowledge to identify, acquire, and apply green knowledge from external sources (Adams et al., 2016; Claudy et al., 2016). This capability allows high-GSO firms to capitalize on opportunities to display environmental commitment, amplifying their visible green outputs under legitimacy or performance pressures.
We argue that the expression of green innovation bubbles among high GSO firms may vary across geographic contexts. As firms move from remote areas toward RICs, they gain greater access to innovation-related resources. High-GSO firms, with a stronger capacity to integrate external knowledge and collaboration networks (Dangelico et al., 2013), can better capitalize on these resources and foster green innovation. At low-to-moderate proximity to central urban areas, where regulatory and social scrutiny remain relatively weak (Liao & Zhang, 2024), high-GSO firms can leverage their resource mobilization capabilities to project stronger environmental performance signals. Their strategic orientation toward environmental legitimacy makes them better equipped to amplify visible green outputs in response to performance and legitimacy pressures. Consequently, their level of innovation bubbles may rise more sharply than low-GSO firms at the same distance.
Meanwhile, firms located in close geographic proximity, typically embedded in or adjacent to RICs’ urban cores, face stronger external scrutiny, regulatory oversight, and institutional pressures (Bertels & Peloza, 2008; Husted, Jamali, et al., 2016). Firms with a stronger sustainability orientation respond to such scrutiny through self-regulation (Jagani & Saboori-Deilami, 2025), shifting their innovation efforts from quantity to quality. Consequently, for firms in very close proximity to RICs, the intensity of green innovation bubbles declines more sharply as they move closer to the urban centers. In summary, high-GSO firms exhibit greater variance in green innovation bubbles across geographic distance, manifesting as a steeper inverted U-shaped curve than their low-GSO counterparts.
Hypothesis 2 Compared with firms with a low GSO, those with a high GSO exhibit a steeper inverted U-shaped relationship between proximity to RICs and green innovation bubbles.
Political connections refer to the relationships between firms and political actors (Faccio, 2010; Faccio et al., 2006). They typically manifest as politicians’ presence on the board or in the management ranks (Tihanyi et al., 2019). Political connections provide firms with a spatially independent channel for accessing policy resources (Zhong & Zheng, 2025), and expose them to greater scrutiny (Marquis & Qian, 2014). The resource provision and bureaucratic oversight reshape firms’ baseline levels of bubble-generating capability and symbolic motivation, thus shifting the inverted U-shaped proximity–bubble curve’s turning point.
Political connections enable firms to access both institutional resources, such as insider policy information, regulatory approvals, and licenses (Goldman et al., 2013), and direct financial support, including government subsidies (Czarnitzki & Toole, 2007; Fang et al., 2022), preferential loans (Haveman et al., 2016), and government contracts (Jia, Huang, et al., 2018). These resource access mechanisms can enhance firms’ ability to repackage, signal, and amplify green innovation outputs through accessible policy and financial channels. This holds for firms far from RICs, where access to urban-centered R&D resources is typically limited. When motivation remains unchanged, firms without previous means to symbolically signal green innovation performance are now enabled to do so. Consequently, green innovation bubbles are more likely to emerge at greater distances, thus shifting the peak of bubble formation away from the urban core.
Simultaneously, political connections suppress firms’ motivation to engage in green innovation bubbles across geographic locations. With more secure resources access (Fang et al., 2022; Goldman et al., 2013) and reduced uncertainty (Sutton et al., 2021), politically connected firms are less reliant on symbolic green achievements to gain government favor or public recognition. Moreover, political connections often bring heightened scrutiny from governments and communities (Jia, Shi, et al., 2018; Marquis & Qian, 2014), further discouraging opportunistic behavior. This motivation-suppressing effect is particularly salient near the urban core, where local regulatory pressures are already intense. Consequently, politically connected firms may reduce green bubbles even at greater distances from RICs. Thus, the inverted U-shaped curve’s turning point to shift toward greater distances from RICs.
Together, political connections reshape green innovation bubble formation by altering firms’ bubble-generating capability and symbolic motivation across geographic space. By providing alternative policy and financial channels, they strengthen firms’ ability to symbolically project green innovation performance even in remote areas, while suppressing their motivation to exaggerate near RICs due to stronger oversight. Consequently, the bubble peak shifts to more distant locations, repositioning the curve’s inflection point farther from urban cores.
Hypothesis 3 Political connections moderate the relationship between proximity to RICs and green innovation bubbles such that the turning point of the inverted U-shaped occurs at greater distances for firms with stronger political connections.
To examine the nonlinear effect of proximity to RICs on firms’ tendencies to generate green innovation bubbles, we constructed a panel dataset from multiple sources comprising Chinese A-share listed companies over the 2006–2023 period. Firms’ geographic coordinates and financial control variables (e.g., return on assets (ROA), leverage, and cash flow) were obtained from the China Stock Market & Accounting Research database. Green patent data, including application and grant counts for invention, utility model, and design patents, were retrieved from the State Intellectual Property Office database (Geng et al., 2024). GSO indicators were manually coded from annual reports and corporate social responsibility (CSR) disclosures. Meanwhile, political connection data were extracted from executive profiles disclosed in annual reports. Appendix A documents the full variable definitions, item sources, and measurement scales.
Following this, we retained observations with no missing values across all variables, yielding an initial sample of 24,708 firm-year observations. We then refined the sample by excluding firms labelled as ST or *ST, which are companies flagged by the Shanghai and Shenzhen Stock Exchanges for sustained financial distress; these are typically those reporting losses for two or more consecutive years, or facing delisting risk. The final sample comprises 23,893 firm-year observations, spanning diverse industries and regions.
Table 1 reports our sample’s characteristics, including firm size, firm age, industry, region and ownership structure. The sample spans a wide range of firms across China. Firm size and age display substantial variation, from smaller and younger firms to large and long-established enterprises. This broad distribution provides meaningful heterogeneity and supports the generalizability of the empirical analysis.
Samples characteristics.
Notes:Total assets are measured in billions of RMB (B = billion, 1B = 10⁹).
Table 2 presents the descriptive statistics. The green innovation bubble index (CGI_bub) has a mean of –0.007 and ranges from –10.406 to 40.040, indicating substantial variation in firms’ tendencies to inflate green innovation signals. The average geographic distance to the nearest RIC is 9.998 (range: 5.806 to 12.602), reflecting broad spatial dispersion across firms. GSO has a mean of 1.375 (SD = 1.619), with values ranging from 0 to 7, suggesting heterogeneity in firms’ environmental strategic commitment. Political connection level (PCLev) averages 0.937, also showing notable variation in firms’ degree of political embeddedness.
Descriptive statistics of main variables.
Green innovation bubbles (CGI_bub) is the dependent variable. Studies generally consider green patent applications as an indicator of the scale or intensity of a firm’s green innovation activities. This reflects its willingness or effort to innovate in environmentally friendly technologies. Meanwhile, patent grants are viewed as a more reliable proxy for innovation quality, as they have undergone substantive examination and approval processes. A large discrepancy between applications and grants suggests a potential overstatement of innovation outcomes, potentially forming “innovation bubbles.” Following Geng et al. (2024), we measured corporate green innovation bubbles (CGI_bub) as the standardized difference between the numbers of green patent applications and patents granted. We included invention, utility model, and design patents (both joint and sole applications). A larger gap (score) indicates a greater likelihood of overstatement in innovation output, thus reflecting a higher level of green innovation bubbles.
Independent variableProximity to RICs is the key independent variable in our analysis, capturing the spatial relationship between firms and regional innovation systems. Geographic proximity can facilitate knowledge exchange, inter-organizational interaction, and institutional pressures (Boschma, 2005; Storper & Venables, 2004). Following Davis and Henderson (2008), we determined firm location using the geographic coordinates of its actual headquarters. Consistent with established practices in economic geography (e.g., Sheng & Ding, 2024; Ter Wal, 2014), we measured proximity as the natural logarithm of the shortest geographic distance from each firm’s headquarters to the nearest RIC. This log transformation helps normalize the distribution and captures distance’s diminishing marginal effect.
While no universally accepted criterion exists for defining RICs, we adopted the classification provided by the National Development and Reform Commission, and drew on prior research (Tao et al., 2025), prioritizing cities that serve as key nodes within urban agglomerations, and act as drivers of regional development and innovation systems. Based on these principles, we identified 36 cities as RICs. Besides China’s four first-tier cities (Beijing, Shanghai, Guangzhou, and Shenzhen), the list also included all provincial capitals, and economically advanced cities such as Suzhou, Dalian, Ningbo, Qingdao, and Xiamen.
Moderating variableGSO (GSO) is the first moderator, representing a firm’s strategic commitment to environmental sustainability. While GSO is typically measured via survey-based constructs (e.g., Casidy et al., 2024; Waqas et al., 2024), such methods are less feasible for large-scale archival research. Therefore, we constructed an index based on disclosures in listed firms’ annual and CSR reports. Specifically, we coded seven binary indicators: whether the firm (1) articulates an environmental philosophy, (2) sets environmental goals, (3) establishes an environmental management system, (4) obtains ISO 14000 certification, (5) conducts environmental training, (6) organizes environmental campaigns, and (7) implements the “Three Simultaneities” policy (i.e., concurrent design, construction, and operation of environmental facilities).
The seven indicators aligned with the latent constructs in survey-based GSO measures. For instance, environmental philosophy and goals reflect green vision; training and campaigns indicate internal mobilization; ISO certification and environmental systems imply institutionalization of green practices; and the “Three Simultaneities” policy captures integration into operational design. The final score ranged from 0 to 7, with higher values indicating stronger green orientation.
Our second moderator is political connection (PCLev). We extended the approach of Wu et al. (2012), who identify political connections based on whether a firm’s chairman or general manager (equivalent to CEO in U.S. firms) currently holds or has held a government position (coded as a binary variable), by incorporating Liu et al.’s (2018) framework of political-administrative hierarchy in China. We constructed an ordinal measure of political connection (PCLev) that ranges from 0 to 4 based on the highest administrative level attained by the chairman or general manager in various government bodies or agencies. A value of 1 indicates township/county-level, 2 for prefectural/municipal, 3 for provincial, and 4 for national-level positions. Firms without such political ties were coded as 0.
Control variablesTo enhance our findings’ credibility, we included several firm-level control variables which may influence a firm’s capability or incentive to exaggerate green innovation performance. Following Geng et al. (2024), we controlled for company size (Size, log of total assets), leverage (Lev), ROA (ROA), cash flow intensity (Cashflow), fixed asset ratio (Fixed), growth rate (Growth), board size (Board), state ownership (SOE, coded as 1 if state-owned), and firm age (Age). These variables help account for heterogeneity in firm characteristics that may confound the relationship between proximity to RICs and green innovation bubbles.
We further controlled for unobserved heterogeneity by including time (YEAR), industry (IND), and provincial region (PRVN) fixed effects. These dummies help account for temporal, sectoral, and regional differences (Husted, Jamali, et al., 2016; Wang et al., 2025; Zamir & Saeed, 2020). This ensures that proximity’s estimated effect on green innovation bubbles is not confounded by firm-specific characteristics or structural variations across years, industries, and regions.
Model and estimation methodWe examined the inverted U-shaped effect of proximity to RICs on corporate green innovation bubbles, and moderating roles of GSO and political connections through regression models. Hereafter, CGI_bubi,t represents the green innovation bubbles of firm i in year t. Distancei,t represents the geographical distance between firm i and its nearest RIC in year t. CONTROLSi,t represents firm-level control variables for firm i in year t.
Equation (1) examines the inverted U-shaped impact of proximity to RICs on green innovation bubbles using pooled ordinary least squares regression. Since decisions regarding headquarters location are shaped by long-term strategic considerations (Jiang et al., 2016), endogeneity concerns were minimized. To account for potential correlations within groups, we followed the method of Husted, Jamali, et al. (2016) to adjust standard errors at both the firm and year levels.
To examine the moderating effects of GSO (GSO) and political connections (PCLev) on the curvilinear relationship between proximity and green innovation bubbles, we followed Haans’s (2016) approach and used the following respective equations.
Results
Table 3 presents the correlation coefficients. Both the main and control variables show significant correlations with green innovation bubbles. Notably, green innovation bubbles and geographic proximity (Distance) exhibit a negative correlation. The correlations among the explanatory variables do not exhibit extreme values. Additionally, in our supplemental tests, all variance inflation factors are below three, indicating no significant multicollinearity issues.
Correlation coefficients of main variables.
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1.CGI_bub | 1 | |||||||||||
| 2.Distance | -0.001 | 1 | ||||||||||
| 3.GSO | 0.049⁎⁎⁎ | 0.079⁎⁎⁎ | 1 | |||||||||
| 4.PCLev | 0.026⁎⁎⁎ | -0.005 | -0.032⁎⁎⁎ | 1 | ||||||||
| 5.Size | 0.105⁎⁎⁎ | -0.180⁎⁎⁎ | 0.352⁎⁎⁎ | 0.073⁎⁎⁎ | 1 | |||||||
| 6.Lev | 0.046⁎⁎⁎ | -0.164⁎⁎⁎ | 0.129⁎⁎⁎ | 0.057⁎⁎⁎ | 0.550⁎⁎⁎ | 1 | ||||||
| 7.ROA | 0.021⁎⁎⁎ | 0.088⁎⁎⁎ | 0.028⁎⁎⁎ | 0.004 | -0.079⁎⁎⁎ | -0.341⁎⁎⁎ | 1 | |||||
| 8.Cashflow | 0.019⁎⁎⁎ | 0.086⁎⁎⁎ | 0.095⁎⁎⁎ | -0.009 | 0.025⁎⁎⁎ | -0.155⁎⁎⁎ | 0.381⁎⁎⁎ | 1 | ||||
| 9.Fixed | -0.029⁎⁎⁎ | 0.058⁎⁎⁎ | 0.070⁎⁎⁎ | 0.021⁎⁎⁎ | 0.069⁎⁎⁎ | -0.002 | -0.140⁎⁎⁎ | 0.069⁎⁎⁎ | 1 | |||
| 10.Growth | 0.002 | 0.003 | -0.024⁎⁎⁎ | 0.015⁎⁎⁎ | 0.037⁎⁎⁎ | 0.046⁎⁎⁎ | 0.206⁎⁎⁎ | 0.041⁎⁎⁎ | -0.089⁎⁎⁎ | 1 | ||
| 11.Board | 0.040⁎⁎⁎ | -0.050⁎⁎⁎ | 0.115⁎⁎⁎ | 0.074⁎⁎⁎ | 0.277⁎⁎⁎ | 0.171⁎⁎⁎ | 0.010* | 0.004 | 0.072⁎⁎⁎ | 0.008 | 1 | |
| 12.SOE | 0.039⁎⁎⁎ | -0.186⁎⁎⁎ | 0.164⁎⁎⁎ | -0.029⁎⁎⁎ | 0.351⁎⁎⁎ | 0.297⁎⁎⁎ | -0.074⁎⁎⁎ | -0.004 | 0.144⁎⁎⁎ | -0.039⁎⁎⁎ | 0.265⁎⁎⁎ | 1 |
| 13.Age | 0.003 | -0.055⁎⁎⁎ | 0.167⁎⁎⁎ | -0.068⁎⁎⁎ | 0.262⁎⁎⁎ | 0.182⁎⁎⁎ | -0.131⁎⁎⁎ | 0.011⁎⁎ | 0.019⁎⁎⁎ | -0.077⁎⁎⁎ | 0.013⁎⁎ | 0.130⁎⁎⁎ |
Notes: N = 23893,
Table 4 reports the regression results. Model 1 includes year, industry, and province fixed effects to isolate the baseline effects of proximity. Model 2 adds firm-level control variables to account for potential confounders. Models 3 and 4 sequentially introduce the linear and quadratic terms of geographic distance, respectively, to capture both the linear and curvilinear effects. In Model 4, the coefficient on the squared distance term Distance2 is significantly negative (β = –0.014, p < 0.01), strongly supporting the proposed inverted U-shaped relationship. Thus, Hypothesis 1 is supported.
Inverted U-shaped effect of geographic proximity on green innovation bubbles.
| Variables | Model 1 | Model 2 | Model 3 | Model 4 |
|---|---|---|---|---|
| Distance | -0.004* | 0.005 | 0.004 | |
| (-0.991) | (1.183) | (0.880) | ||
| Distance2 | -0.007⁎⁎⁎ | -0.014⁎⁎⁎ | ||
| (-4.959) | (-7.918) | |||
| Size | 0.109⁎⁎⁎ | 0.111⁎⁎⁎ | 0.113⁎⁎⁎ | |
| (7.680) | (7.600) | (7.658) | ||
| Lev | 0.013 | 0.013 | 0.009 | |
| (0.548) | (0.528) | (0.344) | ||
| ROA | 0.221⁎⁎⁎ | 0.232⁎⁎⁎ | 0.237⁎⁎⁎ | |
| (3.990) | (3.995) | (4.065) | ||
| Cashflow | 0.081 | 0.084 | 0.086 | |
| (1.161) | (1.158) | (1.190) | ||
| Fixed | -0.117⁎⁎⁎ | -0.121⁎⁎⁎ | -0.116⁎⁎⁎ | |
| (-5.866) | (-5.701) | (-5.541) | ||
| Growth | -0.081⁎⁎⁎ | -0.085⁎⁎⁎ | -0.088⁎⁎⁎ | |
| (-3.647) | (-3.671) | (-3.782) | ||
| Board | 0.002 | 0.002 | 0.003 | |
| (0.503) | (0.443) | (0.531) | ||
| SOE | 0.029* | 0.029* | 0.036⁎⁎ | |
| (1.698) | (1.669) | (2.051) | ||
| Age | -0.006 | -0.003 | 0.005 | |
| (-0.309) | (-0.152) | (0.213) | ||
| Constant | -0.068 | -2.375⁎⁎⁎ | -2.372⁎⁎⁎ | -2.398⁎⁎⁎ |
| (-1.598) | (-7.554) | (-7.589) | (-7.632) | |
| Year dum | YES | YES | YES | YES |
| Indus dum | YES | YES | YES | YES |
| Prvn dum | YES | YES | YES | YES |
| Adj. R² | 0.004 | 0.030 | 0.030 | 0.032 |
Notes: t-statistics in parentheses, N = 23893,
Figure 2 illustrates the estimated inverted U-shaped relationship between geographic distance to RICs and green innovation bubbles. The curve peaks at a log-distance of approximately 10.1 (approximately 24.3 kilometers). Thus, firms located at moderate distances from RICs exhibit the highest levels of green innovation bubbles. Thereafter, the level of green innovation bubbles declines, suggesting diminishing marginal effects of geographic proximity. The 95% confidence intervals, indicated by capped vertical bars, reinforce the robustness of the relationship.
The moderation of green strategic orientation
Table 5 presents GSO’s moderating role in the relationship between proximity to RICs and green innovation bubbles. Model 1 includes Distance along with control variables. Model 2 adds Distance2 to capture potential nonlinearity. Model 3 incorporates the interaction between GSO and distance, Distance*GSO, which is not statistically significant (β = 0.003, p > 0.1). Thus, GSO does not significantly moderate the linear effect of distance. Meanwhile, Model 4 includes the interaction between GSO and Distance2, Distance2*GSO, which is significantly negative (β = –0.005, p < 0.01). Thus, GSO strengthens the inverted U-shaped relationship, making the inverted U-curve steeper. That is, firms with stronger GSO exhibit greater sensitivity to changes in geographic proximity, creating more pronounced green innovation bubbles at moderate distances from RICs.
Results for moderating effect of green strategic orientation.
| Variables | Model 1 | Model 2 | Model 3 | Model 4 |
|---|---|---|---|---|
| Distance | 0.005 | 0.004 | 0.003 | 0.004 |
| (1.163) | (0.862) | (0.823) | (1.004) | |
| Distance2 | -0.014⁎⁎⁎ | -0.014⁎⁎⁎ | -0.014⁎⁎⁎ | |
| (-7.957) | (-7.913) | (-7.902) | ||
| Distance*GSO | 0.003 | 0.003 | ||
| (1.435) | (1.606) | |||
| Distance2*GSO | -0.005⁎⁎⁎ | |||
| (-4.483) | ||||
| GSO | 0.000 | 0.000 | -0.000 | -0.000 |
| (0.036) | (0.049) | (-0.059) | (-0.003) | |
| Size | 0.112⁎⁎⁎ | 0.114⁎⁎⁎ | 0.114⁎⁎⁎ | 0.114⁎⁎⁎ |
| (7.098) | (7.155) | (7.141) | (7.138) | |
| Lev | 0.011 | 0.007 | 0.006 | 0.005 |
| (0.443) | (0.259) | (0.256) | (0.216) | |
| ROA | 0.228⁎⁎⁎ | 0.233⁎⁎⁎ | 0.231⁎⁎⁎ | 0.221⁎⁎⁎ |
| (3.872) | (3.941) | (3.915) | (3.732) | |
| Cashflow | 0.090 | 0.092 | 0.090 | 0.094 |
| (1.236) | (1.265) | (1.248) | (1.292) | |
| Fixed | -0.122⁎⁎⁎ | -0.117⁎⁎⁎ | -0.117⁎⁎⁎ | -0.117⁎⁎⁎ |
| (-5.690) | (-5.532) | (-5.535) | (-5.562) | |
| Growth | -0.083⁎⁎⁎ | -0.085⁎⁎⁎ | -0.086⁎⁎⁎ | -0.084⁎⁎⁎ |
| (-3.518) | (-3.631) | (-3.639) | (-3.596) | |
| Board | 0.002 | 0.002 | 0.002 | 0.003 |
| (0.327) | (0.418) | (0.437) | (0.508) | |
| SOE | 0.030* | 0.037⁎⁎ | 0.037⁎⁎ | 0.037⁎⁎ |
| (1.701) | (2.089) | (2.087) | (2.120) | |
| Age | -0.002 | 0.006 | 0.005 | 0.004 |
| (-0.107) | (0.261) | (0.240) | (0.200) | |
| Constant | -2.431⁎⁎⁎ | -2.443⁎⁎⁎ | -2.450⁎⁎⁎ | -2.458⁎⁎⁎ |
| (-6.520) | (-6.541) | (-6.530) | (-6.543) | |
| Year dum | YES | YES | YES | YES |
| Indus dum | YES | YES | YES | YES |
| Prvn dum | YES | YES | YES | YES |
| Adj. R² | 0.031 | 0.032 | 0.032 | 0.032 |
Notes: N = 23590, t-statistics in parentheses,
Figure 3 graphically illustrates GSO’s moderating effect. Firms with high GSO exhibit a clear inverted U-shaped pattern: Green innovation bubbles increase with proximity up to a certain point and then decline. In contrast, low-GSO firms display a mostly flat curve, showing little variation in bubbles across distances. Figure 3 suggests that proximity’s nonlinear effect is more pronounced among firms with stronger GSO, consistent with Hypothesis 2.
The moderation of political connections
Table 6 reports the results for political connections’ (PCLev) moderating effect on the relationship between proximity to RICs and green innovation bubbles. Model 1 includes all control variables and the linear term Distance. Model 2 adds the quadratic term Distance2. In Model 3, the interaction between distance and political connections, Distance* PCLev, is significantly positive (β = 0.008, p < 0.01). Model 4 adds the interaction between PCLev and squared distance, Distance2* PCLev, which is not significant (β = -0.001, p > 0.1). Meanwhile, the interaction with the linear distance term remains significant and positive (β = 0.008, p < 0.01). Following prior work (Haans et al., 2016), these results suggest that stronger political connections shift the inverted U-shaped curve rightward. Thus, firms with higher political connections experience the peak of green innovation bubbles at a greater distance from RICs.
Results for moderating effects of political connections.
| Variables | Model 1 | Model 2 | Model 3 | Model 4 |
|---|---|---|---|---|
| Distance | 0.004 | 0.003 | 0.002 | 0.002 |
| (1.089) | (0.758) | (0.520) | (0.523) | |
| Distance2 | -0.015⁎⁎⁎ | -0.015⁎⁎⁎ | -0.014⁎⁎⁎ | |
| (-7.961) | (-7.948) | (-7.879) | ||
| Distance* PCLev | 0.008⁎⁎⁎ | 0.008⁎⁎⁎ | ||
| (2.686) | (2.688) | |||
| Distance2* PCLev | -0.001 | |||
| (-0.938) | ||||
| PCLev | 0.013⁎⁎ | 0.014⁎⁎⁎ | 0.014⁎⁎⁎ | 0.014⁎⁎⁎ |
| (2.550) | (2.638) | (2.640) | (2.636) | |
| Size | 0.110⁎⁎⁎ | 0.112⁎⁎⁎ | 0.112⁎⁎⁎ | 0.112⁎⁎⁎ |
| (7.705) | (7.761) | (7.738) | (7.738) | |
| Lev | 0.013 | 0.009 | 0.009 | 0.008 |
| (0.536) | (0.348) | (0.349) | (0.323) | |
| ROA | 0.234⁎⁎⁎ | 0.239⁎⁎⁎ | 0.240⁎⁎⁎ | 0.239⁎⁎⁎ |
| (3.953) | (4.026) | (4.031) | (4.036) | |
| Cashflow | 0.091 | 0.093 | 0.096 | 0.096 |
| (1.248) | (1.278) | (1.316) | (1.318) | |
| Fixed | -0.124⁎⁎⁎ | -0.118⁎⁎⁎ | -0.119⁎⁎⁎ | -0.119⁎⁎⁎ |
| (-5.668) | (-5.514) | (-5.513) | (-5.496) | |
| Growth | -0.081⁎⁎⁎ | -0.084⁎⁎⁎ | -0.083⁎⁎⁎ | -0.084⁎⁎⁎ |
| (-3.545) | (-3.661) | (-3.652) | (-3.656) | |
| Board | 0.001 | 0.002 | 0.002 | 0.002 |
| (0.283) | (0.373) | (0.440) | (0.452) | |
| SOE | 0.036* | 0.043⁎⁎ | 0.044⁎⁎ | 0.044⁎⁎ |
| (1.959) | (2.344) | (2.361) | (2.366) | |
| Age | -0.005 | 0.004 | 0.002 | 0.002 |
| (-0.212) | (0.166) | (0.113) | (0.113) | |
| Constant | -2.388⁎⁎⁎ | -2.400⁎⁎⁎ | -2.404⁎⁎⁎ | -2.403⁎⁎⁎ |
| (-7.097) | (-7.118) | (-7.111) | (-7.113) | |
| Year dum | YES | YES | YES | YES |
| Indus dum | YES | YES | YES | YES |
| Prvn dum | YES | YES | YES | YES |
| Adj. R² | 0.031 | 0.032 | 0.033 | 0.033 |
Notes: N = 23590, robust t-statistics in parentheses,
Figure 4 illustrates political connections’ moderating effect. For firms with weak and strong political connections, green innovation bubble levels initially rise as firms move closer to RICs and decline after peaking. However, for firms with stronger political connections (higher PCLev), the peak shifts rightward. Thus, the decline in bubble formation occurs at greater distances. Hence, Hypothesis 3 is supported.
Robustness checkRobustness check for the inverted U-shaped relationship
Subsequently, we conduct several additional tests to test the robustness of the hypothesized relationships. Table 7 reports the results. First, we replace the dependent variable with an alternative measure, CGI_bub_proxy, which captures the extent of green innovation bubbles solely based on independently completed green patents (i.e., excluding joint applications). This alternative indicator allows us to test whether the main findings remain valid when focusing on a firm’s own innovation output. In Model 1, we include only the linear term of distance, while Model 2 incorporates the squared term. The significantly negative coefficient on the squared term (β = –0.014, p < 0.01) supports the inverted U-shaped effect’s robustness.
Robustness test on the inverted U-shaped relationship.
| Variables | Model 1 | Model 2 | Model 3 | Model 4 |
|---|---|---|---|---|
| Alternative Dependent Variable | Sequential estimation | random effect model | ||
| Distance | 0.004 | 0.003 | 0.002 | 0.002 |
| (0.884) | (0.604) | (0.317) | (0.254) | |
| Distance2 | -0.014⁎⁎⁎ | -0.011⁎⁎⁎ | -0.014⁎⁎⁎ | |
| (-8.287) | (-3.673) | (-3.955) | ||
| Constant | -2.358⁎⁎⁎ | -2.385⁎⁎⁎ | 0.023⁎⁎⁎ | -1.576⁎⁎⁎ |
| (-7.273) | (-7.330) | (3.580) | (-3.088) | |
| Controls | YES | YES | YES | YES |
| Year dum | YES | YES | YES | YES |
| Indus dum | YES | YES | YES | YES |
| Prvn dum | YES | YES | YES | YES |
| N | 23893 | 23893 | 24636 | 23893 |
| Adj. R² | 0.027 | 0.028 | — | — |
Notes: t-statistics in parentheses,*p < 0.10;**p < 0.05;
Second, Model 3 applies a sequential estimation method, which is appropriate for examining the effects of time-invariant variables such as geographic proximity (Kripfganz & Schwarz, 2019). The coefficient on the squared distance term remains significant and negative (β = –0.011, p < 0.01), supporting H1. Third, following Husted, Montiel, et al. (2016), Model 4 adopts a random-effects model. Again, Distance2 remains significantly negative (β = –0.014, p < 0.01), supporting the robustness of the hypothesized inverted U-shaped relationship.
Robustness check for the moderating effectsWe further examine the robustness of the moderating effects of GSO (GSO) and political connections (PCLev). As shown in Table 8, Models 1-3 employ an alternative measure of GSO (GSO2), constructed as a composite index based on four key indicators: whether the firm articulates an environmental philosophy, sets environmental goals, establishes an environmental management system, and obtains ISO 14000 certification. The interaction term between GSO2 and the squared distance remains significantly negative (β = –0.008, p < 0.01), supporting GSO’s moderating role.
Robustness test on the moderating effect of green strategic orientation.
| Variables | Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | Model 6 |
|---|---|---|---|---|---|---|
| Alternative Moderator | Alternative Dependent Variable | |||||
| Distance | 0.003 | 0.003 | 0.004 | 0.003 | 0.003 | 0.003 |
| (0.813) | (0.796) | (1.012) | (0.628) | (0.591) | (0.722) | |
| Distance2 | -0.014⁎⁎⁎ | -0.014⁎⁎⁎ | -0.015⁎⁎⁎ | -0.015⁎⁎⁎ | -0.015⁎⁎⁎ | -0.015⁎⁎⁎ |
| (-7.954) | (-7.866) | (-7.852) | (-8.355) | (-8.290) | (-8.276) | |
| GSO2 | 0.007 | 0.007 | 0.007 | |||
| (1.549) | (1.442) | (1.364) | ||||
| Distance × GSO2 | 0.003 | 0.003 | ||||
| (0.953) | (1.163) | |||||
| Distance2 × GSO2 | -0.008⁎⁎⁎ | |||||
| (-4.434) | ||||||
| GSO | 0.000 | -0.000 | 0.000 | |||
| (0.071) | (-0.034) | (0.011) | ||||
| Distance × GSO | 0.003 | 0.003 | ||||
| (1.184) | (1.306) | |||||
| Distance2 × GSO | -0.004⁎⁎⁎ | |||||
| (-3.735) | ||||||
| Constant | -2.387⁎⁎⁎ | -2.391⁎⁎⁎ | -2.404⁎⁎⁎ | -2.427⁎⁎⁎ | -2.434⁎⁎⁎ | -2.441⁎⁎⁎ |
| (-6.655) | (-6.633) | (-6.649) | (-6.473) | (-6.460) | (-6.469) | |
| Controls | YES | YES | YES | YES | YES | YES |
| Year dum | YES | YES | YES | YES | YES | YES |
| Indus dum | YES | YES | YES | YES | YES | YES |
| Prvn dum | YES | YES | YES | YES | YES | YES |
| Adj. R² | 0.032 | 0.032 | 0.032 | 0.028 | 0.028 | 0.028 |
Notes: N = 23590, t-statistics in parentheses,*p < 0.10;**p < 0.05;
Models 4-6 apply the alternative dependent variable (CGI_bub_proxy), which focuses on independently completed green patents. GSO’s moderating effect remains significant and negative (β = –0.004, p < 0.01), further supporting its role in amplifying the inverted U-shaped relationship between proximity to RICs and green innovation bubbles.
Table 9 presents robustness checks using an alternative moderator and dependent variable to further validate political connections’ moderating role. Models 1-3 replace the ordinal variable PCLev with a binary indicator (PC), coded as 1 if any top executive or board member currently holds or previously held a political position, and 0 otherwise. The interaction term between Distance and PC, Distance × PC, remains significant and positive (β = 0.017, p < 0.01). Meanwhile, Distance2 × PC is not significant (β = 0.017, p < 0.01), thus supporting Hypothesis 3.
Robustness test on the moderating effect of political connections.
| Variables | Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | Model 6 |
|---|---|---|---|---|---|---|
| Alternative Moderator | Alternative Dependent Variable | |||||
| Distance | 0.003 | 0.002 | 0.002 | 0.003 | 0.001 | 0.001 |
| (0.722) | (0.559) | (0.557) | (0.650) | (0.396) | (0.400) | |
| Distance2 | -0.015⁎⁎⁎ | -0.015⁎⁎⁎ | -0.014⁎⁎⁎ | -0.014⁎⁎⁎ | -0.015⁎⁎⁎ | -0.014⁎⁎⁎ |
| (-7.960) | (-7.956) | (-7.864) | (-7.924) | (-7.921) | (-7.737) | |
| PC | 0.038⁎⁎⁎ | 0.037⁎⁎⁎ | 0.037⁎⁎ | |||
| (2.618) | (2.592) | (2.569) | ||||
| Distance × PC | 0.017⁎⁎ | 0.017⁎⁎ | ||||
| (2.154) | (2.165) | |||||
| Distance2 × PC | -0.002 | |||||
| (-0.424) | ||||||
| PCLev | 0.014⁎⁎⁎ | 0.014⁎⁎⁎ | 0.014⁎⁎⁎ | |||
| (2.684) | (2.686) | (2.684) | ||||
| Distance × PCLev | 0.008⁎⁎⁎ | 0.008⁎⁎⁎ | ||||
| (2.710) | (2.708) | |||||
| Distance2 × PCLev | -0.002 | |||||
| (-1.263) | ||||||
| Constant | -2.412⁎⁎⁎ | -2.413⁎⁎⁎ | -2.413⁎⁎⁎ | -2.435⁎⁎⁎ | -2.439⁎⁎⁎ | -2.439⁎⁎⁎ |
| (-7.104) | (-7.103) | (-7.104) | (-7.330) | (-7.324) | (-7.324) | |
| Controls | YES | YES | YES | YES | YES | YES |
| Year dum | YES | YES | YES | YES | YES | YES |
| Indus dum | YES | YES | YES | YES | YES | YES |
| Prvn dum | YES | YES | YES | YES | YES | YES |
| Adj. R² | 0.032 | 0.032 | 0.032 | 0.034 | 0.035 | 0.035 |
Notes: N = 23590, t-statistics in parentheses,*p < 0.10;
Models 4-6 replace the dependent variable with a green innovation bubble measure solely based on invention patents, emphasizing higher-quality innovation. The interaction term between the linear distance and political connections Distance × PCLev remains significant (β = 0.008, p < 0.01). Meanwhile, the interaction with the squared distance Distance2 × PCLev is not significant (β = -0.002, p > 0.1). This reinforces the hypothesis of an apex-shifting moderation effect of political connections.
Heterogeneity analysisWe further conduct a heterogeneity analysis to examine whether the inverted U-shaped relationship varies across firm subgroups, classified by industry pollution level (heavily polluting versus non-polluting), technological orientation (high-tech versus non-high-tech), and ownership structure (SOE versus non-SOE). These groupings reflect differences in regulatory exposure, innovation capacity, and institutional embeddedness (see Table 10).
Further analysis of the inverted U-shaped relationship.
| Variables | Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | Model 6 |
|---|---|---|---|---|---|---|
| Heavily polluted industry | Hi-tech industry | State-owned enterprise | ||||
| YES | NO | YES | NO | YES | NO | |
| Distance | -0.001 | 0.006 | 0.018⁎⁎ | -0.003 | 0.003 | 0.002 |
| (-0.178) | (1.146) | (2.124) | (-1.122) | (0.236) | (0.539) | |
| Distance2 | -0.008 | -0.015⁎⁎⁎ | -0.019⁎⁎⁎ | -0.009⁎⁎⁎ | -0.022⁎⁎⁎ | -0.010⁎⁎⁎ |
| (-1.257) | (-7.680) | (-5.716) | (-4.266) | (-6.260) | (-5.767) | |
| Constant | -0.887⁎⁎⁎ | -2.875⁎⁎⁎ | -3.507⁎⁎⁎ | -0.917⁎⁎⁎ | -3.747⁎⁎⁎ | -1.760⁎⁎⁎ |
| (-3.621) | (-7.321) | (-6.943) | (-3.859) | (-5.992) | (-5.104) | |
| Controls | YES | YES | YES | YES | YES | YES |
| Year dum | YES | YES | YES | YES | YES | YES |
| Indus dum | YES | YES | YES | YES | YES | YES |
| Prvn dum | YES | YES | YES | YES | YES | YES |
| N | 4812 | 19081 | 13611 | 10196 | 6578 | 17315 |
| Adj. R² | 0.006 | 0.040 | 0.048 | 0.029 | 0.058 | 0.023 |
Notes: t-statistics in parentheses,*p < 0.10;
First, we find that the inverted U-shaped relationship between proximity to RICs and green innovation bubbles is significant in non-polluting industries (β = -0.015, p < 0.01) and not significant in heavily polluting sectors (β = -0.008, p > 0.1). Firms in non-polluting sectors generally face lower environmental scrutiny. This makes their green signaling more discretionary, and thus, more responsive to changes in RIC proximity. Conversely, heavily polluting firms face uniformly high regulatory pressure, making opportunistic behavior less elastic to location. Their motivation is suppressed across the board, thus weakening the proximity effect.
Second, the inverted U-shaped relationship is stronger among high-tech firms (β = -0.019, p < 0.01) and remains significant but flatter among non-high-tech firms (β = -0.009, p < 0.01). With greater embeddedness in innovation ecosystems, high-tech firms can better absorb spillovers from RICs. Hence, their bubble-generating capability is more responsive to geographic proximity. Their stronger alignment with policy priorities also curbs symbolic behaviors under heightened scrutiny near RICs. Conversely, non-high-tech firms face greater constraints in leveraging spatial advantages, weakening proximity effects.
Third, the inverted U-shaped relationship is stronger for SOEs (β = -0.022, p < 0.01) and weaker yet still significant for non-SOEs (β = -0.010, p < 0.01). As state-affiliated entities, SOEs have greater access to policy resources and are more politically sensitive. Hence, they are more inclined to undertake symbolic green innovation, especially at moderate distances where oversight is weaker. However, their stronger exposure to political accountability near institutional centers suppresses such opportunism, causing a sharper decline after the peak.
Conclusion and discussionsConclusionDrawing on institutional theory, the resource-based view, and strategic response theory, we investigate how RICs influence green innovation bubble formation. Based on a large sample of Chinese listed firms from 2006 to 2023, we provide robust evidence showing that proximity to RICs influences green innovation bubble formation among nearby firms. Firms located farther from RICs are more likely to engage in green innovation bubbles as they move closer to urban centers. Meanwhile, firms already in close proximity tend to reduce such behavior as the distance decreases.
Additionally, we examine the boundary conditions shaping this nonlinear relationship. Firms with a higher GSO, characterized by long-term environmental commitment and heightened reputational sensitivity, exhibit a steeper inverted U-curve between proximity to RICs and green innovation bubbles. These firms are more likely to amplify bubble behaviors at low-to-moderate proximity, where resources are accessible but external scrutiny is weak. However, as they move closer to RICs, where regulatory oversight and stakeholder visibility intensify, they curb such opportunistic behaviors. Conversely, low-GSO firms show weaker responsiveness to both the resource advantages and institutional pressures of nearby RICs.
Furthermore, political connections moderate the relationship between proximity to RICs and green innovation bubbles by shifting the peak of the inverted U-shaped curve. Politically connected firms benefit from privileged access to institutional resources regardless of location. Hence, those farther from RICs can translate their enhanced capabilities into symbolic innovation signaling despite limited local resources. Meanwhile, heightened regulatory scrutiny associated with political ties curbs such opportunistic behaviors near RICs. Together, these effects suppress opportunistic behavior in close proximity while facilitating bubble formation at greater distances, thereby shifting the peak of green innovation bubbles toward urban peripheries.
Finally, we conducted a heterogeneity analysis based on industry pollution level (heavily polluting versus non-polluting), technological orientation (high- versus non-high-tech firms), and ownership structure (SOE versus non-SOE). We find that the inverted U-shaped relationship between proximity to RICs and green innovation bubbles is significant in non-polluting industries, but not significant in heavily polluting sectors. The pattern is stronger for high-tech firms and remains significant albeit flatter for non-high-tech firms. Finally, the relationship is stronger for SOEs and weaker yet still significant for non-SOEs.
Theoretical implicationsFirst, this study enhances the understanding of geographic influences on green innovation from a strategic response perspective. While green innovation research often highlights the spillover effects of cities, it typically treats firms as passive recipients of such geographic influences (Dong et al., 2025; Hu & Xu, 2023). However, firms are strategic actors who may respond to urban resources and pressures with opportunistic behaviors. This may distort the intended effects on green innovation. We highlight how RICs, as policy-intensive and resource-concentrated urban spaces, may inadvertently foster symbolic innovation behaviors among nearby firms. This perspective views spatial proximity both as a knowledge diffusion channel and condition that can provoke symbolic green actions.
Second, we show how such distorted innovation behaviors systematically vary with proximity to RICs. While the literature often assumes that spatial effects monotonically decline with distance (Döring & Schnellenbach, 2006; Obschonka et al., 2023; van der Wouden & Youn, 2023), we find an inverted U-shaped relationship between proximity to RICs and green innovation bubbles. By distinguishing and integrating the two latent mechanisms (i.e., bubble-generating capability and symbolic motivation), we advance the understanding of how firms at moderate proximity to RICs exhibit the highest green innovation bubbles. Meanwhile, those too close or too far display lower levels due to either reduced motivation or limited capability. These findings move beyond conventional spatial spillover models by introducing a strategic response logic that links geographic influences to green innovation bubbles.
Third, organizational heterogeneity shapes how firms interpret and respond to geographic influences. While research on the geography of innovation largely assumes that firms uniformly respond to location-based advantages or constraints (Dong et al., 2025), we show that internal strategic orientations and political ties condition these responses in distinct ways. Firms with strong GSO exhibit a steeper inverted U-shape, as they respond more sensitively to both spatial incentives and resource constraints. Meanwhile, politically connected firms display an outward-shifted curve: Political ties reduce dependence on local resources, while heightened scrutiny suppresses opportunistic behavior near RICs. Clearly, we need to move beyond the traditional economic geography perspective that treats firms as universal passive recipients of geographic influences, and recognize how organizational values and institutional positions shape firms’ strategic responses under similar geographic proximity.
Managerial implicationsOur findings provide actionable implications for both managers and policymakers. First, RICs do not necessarily engender substantive green innovation and may inadvertently foster green innovation bubbles. Hence, policymakers need to update green innovation evaluation frameworks by placing greater emphasis on outcome-oriented indicators rather than input- or activity-based measures. Firms located near RICs should remain cautious and invest in strengthening their internal capabilities so that proximity-driven incentives translate into genuine rather than superficial innovation outcomes.
Second, the inverted U-shaped relationship indicates that inflated green innovation is most likely in intermediate proximity zones. Hence, governance practices should more explicitly differentiate across spatial contexts, with clearer expectations and more consistent monitoring in areas where risks tend to concentrate. Firms located at moderate distances from RICs should particularly strengthen self-regulation to prevent channeling resources into symbolic green innovations.
Third, the moderating roles of strategic orientation and political connections highlight how organizational attributes shape firms’ responses to geographic proximity. For policymakers, a “firm profiling” approach could enable more tailored governance of green innovation. For example, green-oriented firms should be incentivized to transform environmental values into substantive outcomes. Meanwhile, politically connected firms warrant closer oversight to prevent them from leveraging institutional privileges to bypass regulations.
Limitations and future research directionsFirst, although we identify an inverted U-shaped relationship between proximity to RICs and green innovation bubbles, supplemental analyses reveal regional variation in curve steepness across eastern, central, and western China. Thus, our findings’ generalizability beyond the current spatiotemporal context may be limited. Given disparities in socio-economic structures, institutional environments, and innovation capacities, we do not know whether the observed non-linear pattern holds in smaller cities, less-developed regions, or other national settings. Here, future research can perform cross-regional or cross-country comparisons. Second, our proposed mechanisms—firms’ motivation and capability—are inherently latent and measured through indirect proxies. Although we rely on established theoretical logic to interpret the inverted U-shaped relationship, limited direct behavioral or perceptual data constrain the precision of our mechanism identification. Future research can incorporate qualitative or experimental methods (e.g., interviews or surveys), or analyze corporate annual reports using natural language processing tools, thereby better capturing firms’ opportunistic motivations.
Funding sourcesThis study was supported by Social Science Foundation of Jiangsu Province (22GLD020), Fundamental Research Funds for the Central Universities (B220201057).
Ethical approvalAll procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Availability of data and materialAnyone can ask the correspondence author for the data.
CRediT authorship contribution statementTeng Wang: Writing – review & editing, Writing – original draft, Visualization, Resources, Methodology, Investigation, Funding acquisition, Data curation, Conceptualization. Jiaoren Lu: Writing – original draft, Visualization, Methodology, Formal analysis, Data curation. Xiaotong Li: Writing – review & editing, Writing – original draft, Supervision, Project administration.
None.
