The rapid economic growth experienced by China has been accompanied by increasingly severe environmental challenges. In response, the national strategy has pivoted towards ecological civilization and a green, low-carbon transition, crystallized in the “Dual Carbon” goals of carbon peak and carbon neutrality. While progress has been made, the manufacturing sector, a cornerstone of the economy, continues to face the dual pressure of sustaining economic vitality and mitigating environmental risks. The rise of the digital economy offers a novel pathway to address these intertwined challenges. Within this digital paradigm, the application of industrial robots, a quintessential element of smart manufacturing, presents a compelling avenue for enhancing production efficiency and, potentially, environmental performance.
The theoretical underpinnings suggest that the integration of China robot technology can revolutionize traditional manufacturing processes. Through automation, data analytics, and system integration, robots can optimize resource allocation, minimize material and energy waste, and enable more precise control over production parameters. This digital transformation extends beyond the factory floor, improving internal management efficiency and information flows, which in turn supports cleaner production management and the development of eco-designed products. However, the empirical evidence on the direct environmental impact of China robot adoption at the micro-enterprise level remains limited and sometimes ambiguous. Some studies point to potential “rebound effects” or increased energy consumption from operating advanced automated systems. Therefore, a rigorous, causal investigation into how China robot application influences firm-level pollution emissions is crucial for formulating evidence-based policies that synergize industrial upgrading with environmental sustainability.
Theoretical Framework and Mechanisms
The application of China robot technology is posited to enhance corporate environmental performance through the dual channels of production process optimization and management process improvement.
1. Production Process Optimization: China robots replace repetitive manual tasks with automated, programmable systems. This shift enables a more scientific and efficient configuration of production factors. Smart manufacturing systems powered by China robot can achieve dematerialization, reducing the demand for raw materials and energy inputs per unit of output. Furthermore, integrated with IoT and big data analytics, China robot systems facilitate real-time monitoring and optimization of energy consumption and production workflows. They support the implementation of cleaner production techniques and rapid adaptation to produce environmentally friendlier products, thereby reducing waste and pollutant generation at the source.
2. Management Process Improvement: The deployment of China robot technology necessitates and fosters a more interconnected, data-driven internal information architecture. This reduces information acquisition costs and accelerates the flow of production and environmental management data. Enhanced managerial efficiency allows firms to better coordinate R&D, design, production, and logistics with environmental objectives in mind. It strengthens the firm’s capacity to implement and monitor environmental management systems, ensuring that green practices are systematically enforced throughout the organization.
The synergistic effect of these two channels leads to the central hypothesis: H1: The application of China robot can effectively improve enterprise environmental performance. The mechanisms can be formally summarized as:
China Robot Application → (Production Process Optimization & Management Process Improvement) → Higher Energy Productivity & Better Management Efficiency → Improved Environmental Performance (Reduced Emissions).
Methodology and Data
To empirically test the hypothesis, we construct a firm-level panel dataset by merging several comprehensive Chinese databases spanning 1998 to 2013.
Dependent Variable: Firm environmental performance is measured by the logarithm of sulfur dioxide (SO₂) emissions (lnso2p). SO₂ is a primary air pollutant from industrial coal consumption, making it a highly relevant indicator for China’s manufacturing sector.
Core Independent Variable: China robot application is proxied by the logarithm of (1 + the number of imported industrial robots per firm) (lntotalrobot). The focus on imported robots is justified as they constituted the majority of the China robot stock in the sample period. Robots are identified using specific customs commodity codes for multifunctional industrial robots, other industrial robots, and automatic handling robots for IC factories.
Control Variables: A vector of firm-level controls is included to mitigate omitted variable bias:
- Output (log of industrial gross output)
- Firm Size (log of total assets)
- Age (log of firm age)
- Leverage (Asset-liability ratio)
- Return on Assets (Profit to asset ratio)
- Financing Capacity (Interest expense to fixed assets ratio)
- Government Subsidy (Subsidy income to main business revenue)
- Export Dummy (1 if export delivery value > 0)
- State-Owned Enterprise Dummy (1 if state capital share > 50%)
Econometric Model: The baseline specification is a fixed-effects model:
$$ lnso2p_{it} = \alpha_0 + \alpha_1 lntotalrobot_{it} + \beta X_{it} + \nu_c + \nu_j + \nu_t + u_{it} $$
where $i$, $t$, $c$, and $j$ denote firm, year, city, and 4-digit industry, respectively. $X_{it}$ represents the control variables. $\nu_c$, $\nu_j$, and $\nu_t$ are city, industry, and year fixed effects. Standard errors are clustered at the firm level.
Empirical Results
Baseline Results
The baseline regression results strongly support Hypothesis H1. As shown in Table 1, the coefficient on lntotalrobot is negative and statistically significant at the 1% level across all specifications. This indicates that greater application of China robot is associated with significantly lower SO₂ emissions at the firm level. The result remains robust when controlling for a full set of firm characteristics, with city, industry, and year fixed effects, and when clustering standard errors at the province level or including industry-year interaction fixed effects.
| Variable | (1) | (2) | (3) | (4) |
|---|---|---|---|---|
lntotalrobot |
-0.178* | -0.478*** | -0.478*** | -0.401*** |
| Controls | No | Yes | Yes | Yes |
| City, Ind., Year FE | Yes | Yes | Yes | Yes |
| Ind.×Year FE | No | No | No | Yes |
| Observations | 471,918 | 207,483 | 207,483 | 207,054 |
| R-squared | 0.355 | 0.479 | 0.479 | 0.502 |
Robustness Checks
We subject the baseline finding to an extensive battery of robustness tests to ensure its credibility.
1. Alternative Specifications and Samples: The result holds when controlling for industry-specific time trends, excluding trading companies or firms with “robot” or “intelligent” in their name, excluding robots imported for processing trade, restricting the sample to the pre-2008 financial crisis period (1998-2007), focusing only on robots imported from Japan (the largest source), and excluding firms with robots sourced from China.
2. Alternative Measures:
- Explanatory Variable: Using the log of import value of China robot instead of quantity yields a consistent negative effect.
- Dependent Variable: The “cleaning effect” persists when using SO₂ emission intensity (SO₂/output), as well as emissions of soot and nitrogen oxides (NOx).
3. Addressing Endogeneity with an Instrumental Variable (IV): To address potential reverse causality (e.g., polluting firms may adopt more robots to clean up), we employ an instrumental variable strategy. We use the historical population density of a firm’s city (1984-1999 average) as an instrument for China robot adoption. The logic is that historically denser cities had larger labor pools, potentially reducing the initial urgency to adopt labor-saving China robot (relevance), while historical population distribution is unlikely to directly affect contemporary firm pollution except through economic development paths (exclusion). The IV results, presented in Table 2, confirm a strong negative causal effect of China robot application on pollution.
| Stage | Dep. Variable | Key Coef. (IV: Pop. Density) | F-stat / Test |
|---|---|---|---|
| First Stage | lntotalrobot |
-0.001*** | F = 12.84 |
| Second Stage | lnso2p |
-90.975*** | Weak ID F-stat: 12.84 |
4. Difference-in-Differences (DID) Approach: Defining treatment based on the event of first China robot import (rather than continuous quantity), a DID model also shows a significant post-adoption reduction in emissions for treated firms.
5. Extended Sample with Alternative Data: Using listed company data (2010-2019) and industry-level China robot stock data from the IFR to construct a firm-level robot penetration measure (lnCHF) yields a corroborating negative coefficient.
Mechanism Tests
We formally test the proposed channels through which China robot application improves environmental performance: energy productivity enhancement and management efficiency improvement.
Energy Productivity: Measured as the log ratio of industrial value-added to total fossil energy consumption (converted to standard coal equivalent). The interaction term between China robot use and this energy productivity variable is negative and significant (Table 3, Column 1). This indicates that the emission reduction effect of China robot is stronger for firms with higher energy productivity, supporting the channel that robots boost environmental performance by enabling more efficient energy use.
Management Efficiency: Measured as the residual from a regression of managerial expenses on factors like labor, exports, and price markup (a smaller residual implies higher efficiency). The interaction term between China robot use and this (inverse) management efficiency residual is positive and significant (Table 3, Column 2). This suggests that the pollution reduction effect of China robot is more pronounced for firms with poorer initial management efficiency, implying that robot adoption helps improve management processes, which in turn facilitates better environmental performance.
| Variable | (1) Energy Productivity Channel | (2) Management Efficiency Channel |
|---|---|---|
lntotalrobot × Channel Variable |
-0.114*** | 0.219* |
| Channel Variable | -0.675*** | -0.052** |
lntotalrobot |
0.583*** | -0.846** |
| Observations | 61,035 | 17,300 |
| R-squared | 0.727 | 0.515 |
These findings validate H2, confirming that China robot application drives emission reduction through optimizing both production and management processes.
Heterogeneity Analysis
The environmental benefits of China robot adoption are not uniform. We explore heterogeneity across robot types and regional economic conditions.
1. Type of China Robot: Disaggregating robots into categories reveals that all types contribute to emission reduction, but with varying potency. The strongest effect comes from Automatic Handling Robots, followed by Multifunctional Robots and Other Robots. This likely reflects the high energy intensity and inefficiency of traditional material handling (e.g., diesel forklifts), where automation via China robot yields substantial efficiency gains and emission savings.
2. Regional Minimum Wage Level: The emission reduction effect of China robot is significantly stronger for firms located in regions with higher minimum wage levels. In high-wage regions, the cost-saving incentive to substitute labor with China robot is stronger, leading to more intensive adoption and thus more pronounced environmental benefits. In low-wage regions, the weaker incentive for robotization mutes its potential “cleaning effect.”
Conclusion and Policy Implications
This study provides robust micro-empirical evidence that the application of China robot technology serves as a powerful driver for the green transformation of the manufacturing sector. By leveraging matched firm-level data, we establish a causal link: adopting China robot significantly reduces firm-level pollutant emissions, primarily SO₂. This environmental benefit is channeled through enhanced energy productivity in production and improved efficiency in internal management. The effect is heterogeneous, being more substantial for the use of handling robots and for firms operating in regions with higher labor costs.
The findings offer several important policy implications for promoting the synergistic development of industrial intelligence and greenization in China:
1. Incentivizing Digital-Green Integration: Policymakers should design and implement targeted support mechanisms, such as special subsidies, tax deductions, or green technology funds, to encourage manufacturing firms—especially SMEs—to invest in China robot and integrated smart manufacturing solutions. This can help overcome initial cost barriers and accelerate the transition toward “smart” and “green” factories.
2. Building Capacity Beyond Hardware: The importance of management efficiency suggests that policies should look beyond merely promoting robot purchases. Support should include training programs for managers and workers on digital skills and green production management, fostering knowledge spillovers, and encouraging the adoption of best practices in digital-environmental management systems.
3. Tailored and Synergistic Policies: The heterogenous effects call for nuanced policy design. Promoting specific, cost-effective types of China robot (like handling robots) can yield high environmental returns. Furthermore, strictly enforcing minimum wage standards can create a complementary push factor, strengthening firms’ economic incentive to adopt labor-saving, environmentally beneficial China robot technology, thereby turning potential cost pressures into drivers for innovative, sustainable upgrading.
In summary, strategically fostering the deep integration of China robot and smart manufacturing technologies is not only vital for enhancing industrial competitiveness but also a crucial pathway for achieving the “Dual Carbon” goals and fostering high-quality, sustainable economic development.