2.1. Public green attention and green technological innovation

The rise in PGA has established green consumption as both a societal norm and a key market expectation (Jiang and Gao, Reference Jiang and Gao2023). This shift exerts direct normative pressure on corporate behaviour (Zhou and Ding, Reference Zhou and Ding2023). From the perspective of organizational legitimacy theory, when firms face significant expectations from external social groups, they are motivated to proactively adapt their strategies to conform to prevailing social norms. This alignment is essential for securing critical resources, ensuring long-term operational continuity and maintaining social recognition (Wang et al., Reference Wang, Chen, Long and Li2024a). Public concern over environmental issues stems not only from fundamental demands for health and safety but is also reinforced by continuous information oversight and public opinion pressure (Ren and Ren, Reference Ren and Ren2024).

Furthermore, the focus of public attention on a specific domain fundamentally directs societal resource allocation (Lee, Reference Lee2023). Even when not directly translating into financial investment, concentrated attention from the public and policymakers can be converted into valuable market-based resources such as consumer trust, corporate reputation and preferential policy treatment (Zheng et al., Reference Zheng, Wan, Wang, Chen, Li, Wu and Song2023; Mahmoudian et al., Reference Mahmoudian, Nazari and Poursoleyman2025). Within this context, GTI emerges as a critical strategic response for enterprises. As a verifiable signal of a firm’s commitment to environmental stewardship, GTI serves to enhance its environmental image and solidify its legitimacy. Consequently, GTI represents a pivotal pathway for companies to constructively address the pressures and opportunities presented by PGA.

Based on this theoretical reasoning, the following hypothesis is proposed:

2.1. Analysis of the internal influence mechanism of PGA on GTI

3.1. Definition of variables

3.1.1. Core explanatory variables

PGA reflects societal sensitivity to environmental risks, support for sustainable development and the capacity to exert institutional pressure on corporations through oversight and market feedback. The proliferation of internet technology has transformed this attention from a passive awareness into an active force, enabling continuous monitoring and response to environmental issues within daily information flows. To capture this phenomenon, this study constructs a PGA index by integrating the Baidu Keyword Index with established methodological approaches (Yu and Jin, Reference Yu and Jin2022; Wang et al., Reference Wang, Chen, Wang and Li2025b). Baidu Search dominates the domestic search engine market with a 96.3% penetration rate, as reported in the ‘2023 China Search Engine Industry Research Report’ (CTR Market Research, 2023). Therefore, its keyword index provides a highly representative and reliable metric for assessing PGA (Li et al., Reference Li, Wang, Li and Wang2025b).

The construction of the PGA index followed a series of procedures. First, we identified relevant keywords by reviewing key policy documents, including the Chinese government’s 2015 environmental protection initiatives, the 2022 Communist Party of China National Congress report and the 2023 ‘White Paper on China's Green Development in the New Era’. To ensure a comprehensive and scientifically robust keyword database, we supplemented this list with terms identified in prior research (Ma et al., Reference Ma, Song and Zhang2023; Yang et al., Reference Yang, Ling and Chen2024). These keywords were subsequently categorized into five distinct dimensions (Tao et al., Reference Tao, Wang, Ye, Wu and Zhang2023), as detailed in online appendix Table A1. Second, we used Python to crawl the annual search volume data for these keywords for each city from 2011 to 2022, which were then organized and aggregated into city-year panel data. We applied the entropy method to calculate comprehensive PGA index scores and determine the weights for each dimension. The weights are visualized in Figure 1, while the specific calculation formulas and weight distributions are provided in the online appendix. Finally, we visualized the spatial distribution of the PGA index and its five dimensions, which reveals a strengthening trend from west to east over the time series (Figure 2; planning approval map, GS (2020) No. 4619).

Figure 1.

Weight distribution of keywords in the PGA.

Figure 2.

PGA – the temporal and spatial change trend in 2011–2022.

3.2. Data sources and descriptive statistics

The initial research sample consists of A-share companies listed on the Shanghai and Shenzhen stock exchanges from 2011 to 2022, selected for their market representation. Data processing and analysis were conducted using STATA 18. Data were compiled from multiple sources. The core explanatory variable, PGA, was constructed from the Baidu Keyword Index. Data for GTI were sourced from the National Intellectual Property Administration and the CNRDS database. Firm-level data, as well as city-level control variables, were obtained from the CSMAR database, the China Economic Data Statistics Network, the China Statistical Yearbook, the China Environmental Statistics Yearbook and Hexun.com. The data processing workflow is outlined as follows: (1) exclusion of entities classified as ST, PT, or *ST; (2) removal of firms in the financial and real estate sectors; (3) elimination of observations with missing data for critical variables; (4) removal of firms with only a single year of data within the sample period; and (5) exclusion of firms with negative net assets or a debt-to-equity ratio exceeding 1. All continuous variables were winsorized at the 1st and 99th percentiles to mitigate the influence of outliers. The final dataset comprises 17,645 firm-year observations.

This study presents a descriptive statistical analysis of the variables based on the collected data, with detailed results summarized in Table 1. As shown, GTI1 ranges from 0.000 to 2.456, while GTI2 varies between 0.000 and 4.759, suggesting considerable heterogeneity in the level of GTI across regions. Similarly, FER1 spans from 1.700 to 4.893, and FER2 ranges from 2.590 to 6.002, reflecting substantial regional differences in the stringency of government environmental regulation. In contrast, the cross-regional variation in PGA, though present, appears relatively modest compared to the disparities observed in GTI and environmental regulation (ER). Detailed results for all variables are provided in Table 1.

Table 1.

Variable definitions and descriptive statistics

3.3. Model construction

Drawing on the research by Yang et al. (Reference Yang, Ling and Chen2024), Shen et al. (Reference Shen, Pan and Feng2023) and Hao et al. (Reference Hao, Fu and Albitar2023), the following model was constructed to examine the impact of PGA on GTI:

(1)\begin{equation}{\text{GT}}{{\text{I}}_{it}} = {\beta _0} + {\beta _1}{\text{PG}}{{\text{A}}_{it}} + \sum {\beta _i}{\text{Contro}}{{\text{l}}_{it}} + {u_i} + {\gamma _{ct}} + {\lambda _{int}} + {\varepsilon _{it}},\end{equation}

(2)\begin{equation}{\text{Me}}{{\text{d}}_{it}} = {\alpha _0} + {\alpha _1}{\text{PG}}{{\text{A}}_{it}} + \sum {\alpha _i}{\text{Contro}}{{\text{l}}_{it}} + {u_i} + {\gamma _{ct}} + {\lambda _{int}} + {\varepsilon _{it}},\end{equation}

(3)\begin{equation}{\text{GT}}{{\text{I}}_{it}} = {\vartheta _0} + {\vartheta _1}{\text{PG}}{{\text{A}}_{it}} + {\vartheta _2}{\text{Me}}{{\text{d}}_{it}} + \sum {\vartheta _i}{\text{Contro}}{{\text{l}}_{it}} + {u_i} + {\gamma _{ct}} + {\lambda _{int}} + {\varepsilon _{it}},\end{equation}

(4)\begin{equation}{\text{Eou}}{{\text{t}}_{it}} = {\delta _0} + {\delta _1}{\text{PG}}{{\text{A}}_{it}} + \sum {\delta _i}{\text{Contro}}{{\text{l}}_{it}} + {u_i} + {\gamma _{ct}} + {\lambda _{int}} + {\varepsilon _{it}},\end{equation}

(5)\begin{equation}{\text{Eou}}{{\text{t}}_{it}} = {\theta _0} + {\theta _1}{\text{PG}}{{\text{A}}_{it}} + {\theta _2}{\text{Me}}{{\text{d}}_{it}} + \sum {\theta _i}{\text{Contro}}{{\text{l}}_{it}} + {u_i} + {\gamma _{ct}} + {\lambda _{int}} + {\varepsilon _{it}}.\end{equation}

Among them, model (1) is the benchmark regression model. ${\text{GT}}{{\text{I}}_{it}}$ is the explained variable, which represents the GTI of enterprises, the authorization of green patents ${\text{GTI}}1$ and its application volume measurement, ${\text{GTI2}}{\text{.}}$ ${\text{PG}}{{\text{A}}_{it}}$is the explanatory variable, which represents the PGA; models (1)–(3) are mechanism test models, and models (1), (4) and (5) are economic consequences analysis. ${\text{Eou}}{{\text{t}}_{it}}$ represents enterprise operating performance, operating risks and ESG performance (BR, TFP and ESG); ${\text{Me}}{{\text{d}}_{it}}$is the mediating variable, which represents environmental regulation ( ${\text{FER}}1$ and ${\text{FER}}2$), agency cost ( ${\text{Agcost}}$) and corporate social responsibility ( ${\text{CSR}}$); and ${\text{GTI}}$ in the economic consequences analysis. ${\text{Contro}}{{\text{l}}_{it}}$ represents a series of control variables, ${u_i}$, ${\gamma _{ct}}$ and ${\lambda _{int}}$ are used to control the fixed effects of individual enterprises, city*year and industry*year, respectively. ${\varepsilon _{it}}$ is the random error term, and the subscripts $i$, $t$, $c$ and $in$ represent enterprises, years, cities and industries, respectively. For the possible heteroscedasticity problem, this paper uses the clustered standard error at the enterprise level for statistical analysis.

4.1. Benchmark regression model

Table 2 presents the regression results analysing the impact of PGA on GTI. Columns (1) and (2) report the results without control variables, while columns (3) and (4) include the full set of controls. The results consistently show a statistically significant positive relationship between PGA and GTI, regardless of the inclusion of control variables. Specifically, the estimates in columns (3) and (4) indicate that a one-unit increase in the PGA index is associated with an increase of 30.808 units in green patent applications (GTI1) and 51.692 units in green patent grants (GTI2). This provides strong initial support for Hypothesis H1. Furthermore, to dissect this aggregate effect, we decomposed the composite PGA index into its five constituent dimensions: pollution, demand, perspective, activity and attention. The regression results for each dimension, presented in columns (5)–(11), reveal that all five components exert a positive and significant influence on both GTI. This granular analysis underscores the critical role of comprehensive public engagement across different environmental topics in motivating corporate green innovation.

Table 2.

Benchmark regression

4.2. Heterogeneity analysis

To further investigate the varying impact of PGA on GTI, this study conducts heterogeneity tests across regional and firm-level dimensions. At the regional level, we examine how differences in regulatory environments, development stages and resource endowments influence this relationship. Cities are categorized into specific types, including resource-based cities, old industrial bases, transportation hubs and key environmental protection cities, to assess the effect of different local governance contexts on PGA’s efficacy. At the firm level, the analysis introduces two dimensions: ownership structure (state-owned versus non-state-owned enterprises) and environmental impact (whether the firm operates in a heavily polluting industry). This approach elucidates the distinct role public scrutiny plays across firms with different inherent characteristics. The test results, detailed in online appendix Table A3 and summarized in Figure 3, indicate that the positive effect of PGA on GTI is significantly more pronounced in resource-based cities, old industrial areas, key environmental protection cities and transportation hubs.

Figure 3.

Visualization of heterogeneity test results.

Note: Significance levels: *p < 0.1, **p < 0.05, ***p < 0.01.

We propose several explanations for these findings. First, in resource-based and old industrial cities, where environmental issues are historically prominent, PGA acts as a catalyst. Increased PGA in these regions more effectively amplifies pressure on local governments and firms, making authorities more responsive to public demands and prompting them to encourage corporate green responsibility through various channels. Second, in key environmental protection areas, a complementary effect emerges between the informal pressure of PGA and formal regulatory systems. While stringent government policies already create high barriers to green transformation, PGA functions as an amplifier and accelerator. It does so by highlighting corporate environmental misconduct and pushing firms to proactively manage risks and adopt green development strategies. Finally, in transportation hubs and other central nodes characterized by high information flow, concentrated innovation resources and intense market competition, PGA fosters opportunities for innovative behaviour. In these dynamic environments (Esposito and Jalal, Reference Esposito and Jalal2025), GTI not only helps mitigate public opinion risks but also serves as a strategic imperative for firms to survive and thrive in a competitive marketplace.

4.3. Robustness test

To ensure the reliability of the baseline regression results, this study conducted a series of robustness checks, including endogeneity treatments and sensitivity analyses. The findings consistently confirm that the positive effect of PGA on GTI is robust (see online appendix Tables A4–A7).

To address potential endogeneity concerns, we employed two primary strategies. First, a two-stage least squares estimation was conducted using the ‘number of internet users’ as an instrumental variable. After controlling for fixed effects, the coefficient for PGA remains positive and statistically significant. The instrumental variable satisfies relevance conditions, and its estimated coefficient is larger than the baseline OLS estimate, suggesting that the original model may have underestimated the true effect of PGA. Second, to account for exogenous shocks from concurrent environmental policies such as the low-carbon city pilot and carbon emission trading pilot, we performed grouped regression analysis. The results confirm that the positive impact of PGA on GTI remains significant in both pilot and non-pilot cities, aligning with the baseline conclusions.

For additional robustness checks, this study performed several sensitivity tests. These included: replacing the baseline model with a Tobit mo introducing lags of one to three periods for the core explanatory variable (PGA); adjusting the measurement of GTI using patent grant data and application adjustment coefficients; modifying the measurement of core variables by taking the logarithm of PGA and distinguishing between green invention patent applications and grants; and incorporating additional city-level control variables, such as economic development level and industrial structure. Across all these alternative specifications, the significantly positive relationship between PGA and GTI remained unchanged, further underscoring the robustness of our core findings.

5.1. Mechanism test

This study posits that PGA influences GTI by optimizing the external institutional environment and enhancing internal corporate governance. The results of a three-step mediation test confirm that all hypothesized pathways are statistically significant, indicating that PGA’s positive effect on GTI is partially transmitted through these indirect channels.

5.2. Results and analysis

The results of the mediation analysis are presented in Tables 3 and 4, while the bootstrapped indirect effects with 95% confidence intervals are illustrated in Figure 4. The results suggest two main findings.

Figure 4.

Mechanism test.

Notes: Panel (a) presents the 95% confidence intervals of indirect effects across all paths, illustrating their significance and variation; panel (b) enlarges the significant paths whose confidence intervals do not cross zero; and panel (c) displays the distribution of indirect effects and standard errors, allowing comparison of effect magnitude and estimation stability. Overall, most confidence intervals do not cross zero, indicating significant mediating effects. Specifically, indirect pathways through environmental regulation (FER1, FER2 and ER) and CSR are significantly positive, while pathways through agency costs (Agcost1 and Agcost2) are negative, suggesting that reductions in internal governance costs also play an important role.

Table 3.

External environment optimization

Table 4.

Internal growth effect

First, the PGA exerts influence by strengthening the external institutional environment through ER. Empirical results indicate a significant mediating role for ER. Specifically, a one-unit increase in PGA leads to an increase of up to 32.434 units in regulatory stringency. This heightened regulation, in turn, significantly promotes GTI; for example, a one-unit increase in punitive regulations (FER1) can increase green patent grants (GTI2) by up to 42.331 units. When both regulatory types are included in the model, PGA’s direct effect on GTI remains positive, though slightly attenuated. This suggests that: (1) PGA pressures the government to translate public concern into concrete policy tools; (2) both incentive-based and punitive policies independently stimulate corporate green behaviour; and (3) the synergistic perception of these policies is a crucial driver for corporate green transformation. Economically, this pathway demonstrates that external policy incentives and pressures shape a stable institutional environment, reducing the uncertainty and cost of innovation for firms.

Second, the PGA transmits external pressures through CSR. The findings confirm that CSR plays a significant mediating role. A one-unit increase in PGA corresponds to a 3.445-unit increase in CSR scores. Subsequently, enhanced CSR significantly promotes GTI, with a one-unit increase in CSR leading to a maximum increase of 37.611 units in GTI. This indicates that PGA drives firms to bolster their environmental commitment, which is realized through concrete GTI activities (Sousa et al., Reference Sousa, Nogueira and Gomes2025). From an economic perspective, public opinion and reputation mechanisms create endogenous constraints, making social responsibility investment a strategic asset for building a green competitive advantage.

Finally, the PGA influences innovation by optimizing internal governance through reduced agency costs (Agcost). The results demonstrate that PGA significantly lowers Agcost, with a one-unit increase in PGA reducing agency costs by up to 0.556 units. This improvement in governance efficiency significantly promotes GTI, as a one-unit decrease in Agcost leads to a maximum increase of 37.405 units in GTI. Lower agency costs signify more efficient resource allocation and reduced managerial opportunism, thereby freeing up resources for long-term innovation. Economically, public scrutiny enhances transparency and managerial accountability, internally activating a dynamic where ‘external pressure’ is converted into ‘internal incentive’ for green innovation.

5.3. Analysis of economic consequences

Against the backdrop of persistent global economic pressures, the performance and sustainability of corporate development are of paramount concern to both governments and the public. To better understand the tangible economic outcomes of PGA-driven GTI, this study examines its impact from three perspectives (Sun et al., Reference Sun, Ma and Ma2024): BR, TFP and corporate ESG performance. As shown in Table 5, the results indicate that PGA, by promoting GTI, significantly enhances corporate ESG performance, improves TFP and BR. First, GTI stimulated by PGA enables firms to comply with increasingly stringent environmental standards, thereby mitigating compliance risks such as heavy fines or operational suspensions. This proactive approach stabilizes operational expectations and safeguards against environmental liabilities. Second, GTI often focuses on resource-efficient technologies like clean production, which directly lowers production costs. Beyond technological upgrades, this innovation can spur broader organizational and managerial improvements, thereby optimizing production processes and boosting overall productivity and operational performance. Finally, in an era where ESG metrics are critically important to capital markets, a demonstrated commitment to GTI helps firms attract environmentally conscious investors. This enhances corporate reputation, lowers financing costs and helps achieve a balance between economic and environmental objectives. Thus, PGA-driven GTI facilitates a win-win outcome, advancing both corporate development and societal well-being.

Table 5.

Analysis of economic consequences

6.1. Limitation and perspective

Like previous studies, this paper has several limitations that suggest directions for future research. First, although the Baidu Search Index provides timely information on public environmental interest, its representativeness is limited by platform algorithms, user demographics and regional differences in internet use. Moreover, search behaviour does not necessarily reflect actual green actions. Future studies may construct more comprehensive measures of PGA using multi-source data such as social media content. Second, the empirical evidence is drawn from Chinese firms and regions, which may restrict its applicability to other institutional settings. Cultural factors and organizational characteristics, which may also influence GTI, are not explicitly considered. Future work could incorporate these elements and refine measurement approaches as conditions evolve. Third, the study does not explore how emerging digital technologies, including artificial intelligence and big data, interact with environmental requirements to influence green innovation. Examining these dynamics offers a promising direction for future research. Despite these limitations, this paper provides important evidence on the role of PGA in advancing corporate green innovation.