Digital trade stimulates enterprises’ R&D investment through the innovation and transformation effect, which further promotes green technology innovation and green investment efficiency. First, digital technology has emerged as a crucial force in shaping the core competitiveness of digital trade that is closely linked to organizational paradigms and production modes of enterprises. Second, digital trade development makes trade convenient and liberalized, accelerates the inter-regional flow of resources, intensifies the competition, and speeds up the embedding of international standards (Wen et al., 2023; Xiao et al., 2023). These dynamics require enterprises to increase green R&D investment to comply with international environmental standards. Finally, digital trade development changes the social consumption structure, prompting the rise of green demand (Wang et al., 2023). To meet this demand, enterprises engage in collaborative green initiatives with other innovation entities, strengthen knowledge spillover, and circulate innovative elements, thus improving green innovation capacity and efficiency. Furthermore, digital trade accelerates the R&D investment of enterprises through the innovation and transformation effect; fosters the agglomeration of capital, talent, and other factors of production; speeds up the integration of innovation resources, improves innovation efficiency (Li et al., 2024), and promotes green production technology upgrading, thereby improving the green investment efficiency. From the perspective of enterprise investment decision making, managers do not favor green investment projects because of their long-term nature, high inputs, and uncertain returns. Enterprises cannot obtain returns in the short term. However, the technological upgrading and knowledge increment generated through R&D investment can increase energy utilization and reduce return uncertainty, thereby increasing managers’ attention to green investment.

Based on resource-based view, enterprises’ core competitiveness originates from their ability to acquire and utilize resources effectively. R&D investment is considered a strategic resource for enterprises’ green investment, providing technical support and knowledge accumulation for innovation. Through increasing R&D investment, developing green energy technology, optimizing production processes, and accumulating green knowledge stock, enterprises can secure critical resources, raise investment levels, and improve resource allocation and investment efficiency (Liu and Zhen, 2022a). From a green investment decision-making perspective, R&D investment provides an innovative driving force for green technology development, enabling enterprises to overcome the technological bottlenecks that often hinder traditional green investment projects. The technological upgrading and knowledge learning brought by continuous investment in R&D enhances corporate resource utilization efficiency and mitigates risks associated with the long-term and uncertain nature of green investment projects (Wen et al., 2023). Through the integration and optimization of technology and knowledge resources, enterprises can allocate green resources efficiently, thereby boosting green investment efficiency and fostering a balance between environmental protection and economic benefits (Zhang et al., 2019a). Therefore, we develop the second hypothesis:

Hypothesis 2: Digital trade can stimulate enterprises to increase R&D investment, thereby contributing to their green investment efficiency.

The rapid development of digital trade facilitates industrial structural upgrading, promoting both industrial rationalization and advancement. Firstly, empowered by digital technology, digital trade accelerates the process of industrial digitalization and stimulates the growth of information services and electronic industries such as computing and communications. Secondly, digital trade enhances the circulation, processing, and efficient allocation of data elements through digital technologies. It reshapes the structure of social supply and demand, mitigates imbalances in industrial supply and demand, and thereby promotes industrial structural upgrading (Wang and Li, 2024c). Finally, digital trade encourages industrial scaling, generates agglomeration effects, improves synergy within industrial chains, and reduces costs associated with irrational industrial structures (Li et al., 2024), ultimately optimizing the industrial framework. Furthermore, digital trade accelerates industrial upgrading, driving the gradual phase-out of high-energy-consumption and high-pollution industries while facilitating the transition of enterprises toward low-energy-consumption and low-carbon alternatives. This transformation enhances energy utilization efficiency, optimizes resource allocation, and contributes to improved green investment efficiency (Sun et al., 2024).

Based on the “technology-economy” paradigm theory, enterprises can reshape production modes and cost structures through technological progress (Tang and Lan, 2024). Specifically, in the context of digital trade, the application of digital technologies enables firms to transform traditional high-energy-consumption and high-pollution industries into green industries, thereby providing additional capital reserves and technical support for green investment. Moreover, enterprises can utilize renewable resources more efficiently through green technologies, reduce resource waste, and enhance green investment efficiency alongside industrial upgrading. Additionally, technological revolution has reshaped corporate ecosystems. With the advancement of digital trade, collaboration and integration between upstream and downstream industries become more seamless, accelerating the formation of green industrial chains (Du et al., 2021). Within this ecosystem, resource sharing and the dissemination of green technologies through collaborative innovation can more effectively promote the implementation of green investment projects. From the perspective of corporate investment decision-making, industrial structural upgrading increases firms’ willingness to invest in green projects. According to attention-based theory, when enterprises prioritize low-energy-consumption and low-carbon transition projects, increase investments in green initiatives, and optimize the allocation of green resources, the public can better supervise corporate environmental responsibilities, thereby enhancing green investment efficiency. Therefore, we propose the following hypothesis:

Hypothesis 3: Digital trade enhances green investment efficiency by promoting industrial structure upgrading.

Digital platform development constitutes an indispensable intermediary channel through which digital trade affects firms’ green investment efficiency. As a key vehicle of institutional innovation, digital trade policies provide strong support and development momentum for the construction of digital platform infrastructure. The implementation of policies such as the Cross-border E-commerce Comprehensive Pilot Zone (CEPZ) has promoted the establishment and improvement of online integrated service platforms, such as “single window” systems. These policies substantially reduce the institutional transaction costs for firms to access platforms and engage in digital collaboration (Wang et al., 2025). This platform-based development not only enhances trade facilitation but, more importantly, provides the necessary infrastructural conditions for firms’ green transformation.

Digital platforms influence firms’ green investment efficiency through three interrelated mechanisms. First, by aggregating global green technologies and solutions, digital platforms create efficient technology search and matching mechanisms, significantly reducing firms’ information and trial-and-error costs for acquiring energy-saving and emission-reduction technologies (Shen and Zhang, 2024). Second, the use of digital platforms enables traceability across the entire supply chain, making product carbon footprints and energy consumption information visible and manageable. This transparency, in turn, pressures supply chain partners to jointly implement emission reduction measures through market mechanisms (Zhang et al., 2024). Third, platforms integrate information, logistics, and capital flows in trade processes, providing data support for the precise pricing and delivery of green financial products, effectively alleviating financing constraints for green investment (Li et al., 2023a).

External conditions may moderate the mediating effect of digital platforms. The quality of the institutional environment, the level of market development, and the state of technological infrastructure can influence the magnitude and direction of platform effects (Pardo and Tayi, 2007; Bonina et al., 2021). In particular, in regions with well-developed institutional environments, digital platforms are likely to realize their resource integration and information transmission functions, enhancing green investment efficiency. Therefore, we propose the following hypothesis:

Hypothesis 4: Digital trade enhances green investment efficiency by promoting digital platform development.

The explanatory variable in this study is green investment efficiency. Drawing on Richardson (2006) and Liu et al. (2022b), we estimated this variable in Model (2).

(2)$$G{I}_{it}={\alpha }_0+{\alpha }_{1}I{O}_{it}+{\alpha }_{2}Cas{h}_{it}+{\alpha }_{3}Siz{e}_{it}+{\alpha }_{4}O{R}_{it}$$ $$+{\alpha }_{5}G{I}_{it-1}+{\alpha }_{6}Top{1}_{it}+{\alpha }_{7}SO{E}_{it}+{\alpha }_{8}Dp{i}_{it}$$ $$+{\alpha }_{9}Wate{r}_{it}+{\alpha }_{10}Ai{r}_{it}+{\eta }_{\mathrm{i}}+{\lambda }_{\mathrm{t}}+{\epsilon }_{\mathrm{it}}$$

Where GI is the firm’s green investment. Following Liu et al. (2022b), GI = log (greening fee, sewage fee, and environmental protection-related expenditures in the management fee/operating income). IO represents investment plan, measured by net profit divided by owner’s equity. Cash represents cash flow, measured by money funds divided by total assets. OR represents operating profit, measured by Earnings Before Interest and Taxes (EBIT) divided by total assets. Dpi denotes regional investment in industrial pollution prevention and treatment, measured by completed investment in industrial pollution prevention and treatment divided by industry value-added. Water denotes the Chemical Oxygen Demand (COD) and ammonia nitrogen emissions from industrial wastewater. Air denotes the SO₂ and nitrogen oxide emissions from industrial waste gas. Table 1 lists other variables in Model (2). The residual value estimated by Model (2) represents the deviation of a firm’s actual green investment level from the optimal scale of green investment, indicating the inefficiency in corporate green investment. A negative residual signifies under-investment in green projects by the sample firm, while a positive residual suggests over-investment. Therefore, this study uses the absolute value of the estimated residual to measure corporate green investment efficiency (GIE). A larger absolute value indicates a greater deviation between the actual green investment and the optimal level, reflecting a higher degree of green investment inefficiency and, thus, lower green investment efficiency. Given that the measurement of green investment efficiency requires first-order lagged data (Liao et al., 2023), the data of green investment efficiency in this study spans from 2011 to 2020.

Given the sample’s continuity, we selected a total of 105 comprehensive CEPZ cities in the first five batches (2015–2020) to represent digital trade (DID). According to the announcement time, some cities have less than one year during the year of announcement. Therefore, we counted the cities approved before June as the current year, and those approved after June are deferred to the next year for counting.

Following Liao et al. (2023), we selected a series of control variables. They are Financial leverage (Lev), Total asset growth rate (Totassgrt), Enterprise age (Age), Profitability (ROE), Firm size (Size), Occupancy of funds by major shareholders (Occupy), Top shareholder ownership (Top1), Nature of ownership (SOE), Big Four Audit (Big4), and Inventory ratio (Inv). Table 1 lists and defines the specific variables.

The validity of the difference-in-differences model is that the assumption of parallel trends must be met. That is, the trend of changes in the green investment efficiency for pilot and non-pilot cities should be parallel before the policy’s implementation. Therefore, we construct the following model for testing:

(3)$$GI{E}_{\mathrm{it}}={\beta }_0+\sum _{t=-3}^3{\beta }_{t}DI{D}_{it}+\lambda Control{s}_{it}+{\varphi }_c+{\lambda }_t+{\epsilon }_{it}$$

In Model (3), $DI{D}_{it}$ denotes whether the firm’s city is approved as a pilot city in year t, with a value of 1 if yes, and 0 otherwise. Other variables are consistent with Model (1). To visually observe the parallel trend results, all sample results are reported at 95% confidence intervals.

As shown in Fig. 1, before the implementation of the policy in 2015, the estimated coefficient was not statistically significant, indicating that the green investment efficiency of enterprises in the experimental and control groups was consistent before the policy took effect. Following policy implementation, the estimated coefficient became significantly negative, demonstrating a marked improvement in green investment efficiency among enterprises in the experimental group. This observation satisfies the parallel trends assumption.

Fig. 1.

Parallel trend test result.

To exclude the influence of other unobservable factors on the results, we drew on Liu and Lu (2015) to conduct a placebo test. It randomly generates an experimental group and randomly assigns it a policy time, constructs a new interaction term and replaces the one in Model (1) to re-estimate the regression. The above process is repeated 500 times.

Fig. 2 indicates that the coefficients of digital trade (DID) follow a normal distribution centered around zero and do not intersect with the true estimated coefficient (–0.19). Thus, excluding the effect of other unobservable factors on the benchmark regression results proves that digital trade enhances green investment efficiency.

Fig. 2.

Placebo test result.

To eliminate the influence of anticipated effects on the regression results, we manually advanced the policy implementation time by one, two, and three years; created new interaction terms; and added them in Model (1).

Table 4 shows that the newly added variables are statistically insignificant, whereas the digital trade (DID) is not significantly changed. Therefore, the results are valid.

To avoid other policies affecting green investment efficiency during the research period, we conducted an exhaustive search and categorization of pilot policies potentially impacting green investment efficiency during the research period. We found that the Low Carbon City Pilot and Broadband China may affect green investment efficiency. We controlled for these policies in Model (1), where LC represents whether the enterprise’s city belongs to the low-carbon city pilot in the current year, with a value of 1 if yes, and 0 otherwise. BBC represents whether the enterprise’s city is a demonstration city for “Broadband China” in the current year, with a value of 1 if yes, and 0 otherwise.

Table 5 Column (1)–(3) shows that, after excluding the influence of other policies, the digital trade policy consistently increased firms’ green investment efficiency by approximately 18%–19%. Compared with the baseline regression, no significant changes are observed, confirming the robustness of the policy’s positive effect on optimizing green investment efficiency.

If firms with inherently high green efficiency are likely to relocate to CEPZ cities, the observed effect may be driven by such migration behavior rather than by the policy itself. To rule out the possibility of reverse causality, we used historical data on the registered addresses of listed companies from the CSMAR database. We excluded firms that relocated to pilot cities only after the implementation of the CEPZ policy, leaving only the “native” firms that were already present before the policy and did not move.

As shown in column (4) of Table 5, the coefficient for digital trade (DID) is –0.132, which remains significantly negative. Its absolute value of 0.132 is smaller than that of the baseline regression coefficient (0.19), indicating that by excluding migrating firms, we successfully removed the “selection effect” caused by the policy attracting high-efficiency firms. We retained only the “net treatment effect” reflecting the policy’s actual stimulation of local firms’ transformation. This improvement primarily arises from the policy genuinely promoting local firms’ green transformation and innovation activities, rather than from a selection bias because of attracting high-efficiency firms.

To further alleviate the endogeneity problem caused by sample selection bias, we used the PSM method to screen the sample. Specifically, we first selected control variables in the same year as the covariates for matching. Second, we used the nearest neighbor matching, radius matching, and kernel matching methods to match samples. Finally, we performed regression based on the matched samples. Table 6 reports the results. The digital trade (DID) is significantly negative at the 1% level, supporting the results of this study.

In the era of digital trade, the media serves an essential function in environmental information dissemination and public opinion supervision, and exerts a certain influence on enterprises’ environmental and production behaviors (Ren et al., 2023). The high level of media attention promoted the transparency of information, which drives enterprises to focus on environmental responsibility and carry out green technology innovation and production activities (Yan, 2023; Sun et al., 2024).

Following Liao et al. (2023), we measured the media attention index by the frequency of firms’ appearances in headlines and categorized into high and low groups at median. Columns (1) and (2) of Table 8 present the results regarding media attention. Under conditions of high media attention, the absolute value of digital trade (DID) is 0.217, significantly negative at the 5% level, which to some extent raises is above the national average. Under low media attention, the absolute value of digital trade (DID) is 0.073 and not significant, which is consistent with our expectations. In other words, in an environment with high media attention, digital trade can more effectively enhance firms’ green investment efficiency. In an environment with low media attention, the policy fails primarily because external supervision is lacking. The Fisher combination test of inter-group coefficient differences yields significant results. The effect of digital trade on green investment efficiency varies significantly across firm samples with different levels of media attention, making comparative analysis feasible.

In areas with high environmental quality, the environmental protection concept is deeply rooted in people’s hearts. Implementing green environmental protection strategy not only fulfils enterprises’ environmental responsibilities but also meet the public’s expectations for a better environment in the future. It also shapes their green image, obtains the support of stakeholders, and improves their environmental performance, making enterprises inclined to invest green projects (Derchi et al., 2021). Therefore, in areas with higher environmental quality, enterprises better utilize digital trade to meet the public’s expectation of a better environment, thereby enhancing green investment efficiency.

We used the entropy weight method for three waste emissions to calculate the regional comprehensive environmental quality index for each province and categorized firms into high and low groups at median for regression. Columns (3) and (4) of Table 8 report the results regarding environmental quality. In regions with high environmental quality, the absolute value of digital trade (DID) is 0.314, significantly negative at the 1% level, which notably is above the national average. In regions with low environmental quality, the absolute value of digital trade (DID) is 0.091 and not significant. This result aligns with expectations. Firms in regions with higher environmental quality can better leverage the effects of digital trade to improve their green investment efficiency. In regions with low environmental quality, policy failure is mainly attributable to insufficient institutional foundations. The Fisher combination test of inter-group coefficient differences yields significant results. The effect of digital trade on firms’ green investment efficiency varies significantly across firm samples in regions with different environmental quality, making comparative analysis feasible.