The concept of “high-quality development,” formally introduced during the 19th National Congress of the Communist Party of China in 2017, marks a pivotal strategic shift in China’s economic paradigm. It signifies a transition from a phase emphasizing the “quantity” of high-speed growth to a new stage prioritizing the “quality” of development. Classical growth theories posit that economic expansion is driven by factor inputs and technological progress. For a considerable period following the reform and opening-up, China’s economic growth was predominantly fueled by intensive factor inputs, characterized by high investment and relatively low efficiency. However, against the backdrop of a diminishing demographic dividend and the unsustainability of persistently high capital investment, this extensive growth model faces inherent constraints. Consequently, high-quality development necessitates a fundamental transformation in the driving force of China’s economic growth—shifting from reliance on factor accumulation to dependence on technological advancement. The primary agents of technological progress are manufacturing enterprises within the real economy. Notably, technological advancement in China’s manufacturing sector has often exhibited characteristics of being embedded within or dependent on capital investment. Therefore, embodied technological progress, where advanced technology is materialized in new capital goods, becomes crucial for achieving high-quality development. The application of industrial robots, which represent the fusion of physical capital and cutting-edge technology, may thus present a significant opportunity for China to realize its high-quality development objectives.
According to the International Federation of Robotics (IFR), an industrial robot is an automatically controlled, reprogrammable, multipurpose manipulator, programmable in three or more axes, which can be either fixed in place or mobile for use in industrial automation applications. Data reveals a consistent upward trend: from 2015 to 2019, China’s industrial robot production and sales volumes grew steadily, with annual sales reaching 140,500 units in 2019, making China the world’s largest market for industrial robots. As a major developing economy undergoing a critical structural transformation, the proliferation of China robot applications is poised to profoundly influence productivity and is intrinsically linked to the attainment of high-quality development.
This leads to several core questions. Can the adoption of industrial robots facilitate China’s high-quality development? As an embodiment of technological progress, robot integration is expected to increase the contribution share of technology to economic growth, aligning with the goals of high-quality development. However, China’s economic landscape is characterized by significant regional disparities in development levels and growth patterns. Consequently, the impact of robot adoption on economic outcomes may not be uniform across different regions. Finally, the effectiveness of this embodied technology likely depends on a region’s absorptive capacity, suggesting that the channel through which robots influence high-quality development might involve enhancing this local capability.
The academic discourse on industrial robots, both internationally and within China, has primarily focused on two major themes: the impact on employment and the effect on productivity. Regarding employment, a debate exists between the “displacement effect,” where robots replace human labor across a broader range of tasks due to their flexibility and intelligence, and the “creation effect,” where robot adoption generates new business models, industries, and, ultimately, new jobs. Research by Acemoglu and Restrepo (2017) supports the displacement narrative, while others point to job polarization or new opportunities in complementary sectors.
On productivity, consensus is stronger. Studies at various levels consistently find a positive impact. At the micro level, research on firms indicates that computerization and automation boost productivity, with effects amplifying over time. At the industry level, evidence from multiple countries shows that increased robot density enhances labor productivity and output. Within the China robot context, studies using firm-level data confirm that robot adoption significantly raises the labor productivity of Chinese manufacturing enterprises. At the macro level, growth models incorporating automation shed light on its long-term growth implications.
Nevertheless, few studies have specifically examined the impact of industrial robot usage on regional high-quality development across Chinese provinces. This paper aims to address this gap. We employ technical efficiency, derived from an input-output framework using stochastic frontier analysis (SFA), as a proxy indicator for high-quality development at the provincial level. Utilizing provincial panel data, we investigate the influence of industrial robot penetration on this measure of development quality and explore the underlying mechanisms.
Theoretical Framework and Model Specification
To analyze the impact of industrial robot adoption on regional technical efficiency, we construct a conceptual model. We posit that a region’s production capability can be represented by a modified production function where industrial robots act as a distinct, productivity-enhancing capital input. The core hypothesis is that higher penetration of this advanced capital good improves the efficiency with which a region transforms conventional inputs (labor and traditional capital) into output.
We begin with a standard stochastic frontier production function to model provincial output and measure technical inefficiency. The model is specified as follows:
$$ \ln Q_{it} = \beta_0 + \beta_1 \ln L_{it} + \beta_2 \ln K_{it} + v_{it} – u_{it} $$
where, for province \(i\) in year \(t\):
- \(Q_{it}\) represents the real output (Gross Domestic Product).
- \(L_{it}\) denotes labor input (urban employment).
- \(K_{it}\) stands for capital input (fixed asset investment stock).
- \(v_{it}\) is a symmetric random error term, representing statistical noise.
- \(u_{it} \geq 0\) is a non-negative random term associated with technical inefficiency.
The technical efficiency (\(TE\)) of province \(i\) at time \(t\) is then defined as:
$$ TE_{it} = \exp(-u_{it}) $$
This measure ranges between 0 and 1, with 1 indicating full technical efficiency (production on the frontier).
Our primary variable of interest is the intensity of industrial robot usage within a province. Following the methodology adapted from Acemoglu and Restrepo (2017), we define “Robot Penetration” (\(RP\)) as:
$$ RP_{it} = \frac{RB_{it}}{L_{it}} \times 10,000 $$
where \(RB_{it}\) is the stock of industrial robots in province \(i\) at time \(t\), and \(L_{it}\) is the total employment. This yields a measure of robots per 10,000 workers, representing the capital intensity of this specific technology in the production process.
To empirically test the relationship, we specify the following panel data regression model:
$$ \ln TE_{it} = \alpha_0 + \alpha_1 \ln RP_{it} + \alpha_2 \mathbf{C_{it}} + \epsilon_{it} \quad \text{(Model 1)} $$
where:
- \(\ln TE_{it}\) is the natural logarithm of the calculated technical efficiency for province \(i\) in year \(t\).
- \(\ln RP_{it}\) is the log of robot penetration (our core explanatory variable).
- \(\mathbf{C_{it}}\) is a vector of control variables that may influence technical efficiency.
- \(\epsilon_{it}\) is the idiosyncratic error term.
A positive and statistically significant coefficient \(\alpha_1\) would indicate that greater penetration of China robot technology is associated with higher regional technical efficiency, supporting the hypothesis that robot adoption promotes high-quality development.
The control variables \(\mathbf{C_{it}}\) include:
- Foreign Direct Investment (\(FDI\)): Captures potential technology spillovers from abroad.
- Research and Development Expenditure (\(R\&D\)): Represents indigenous innovation efforts and absorptive capacity.
- Per Capita GDP (\(GDPpc\)): Controls for the overall level of economic development.
- Number of Industrial Enterprises (\(ENT\)): Accounts for the scale and structure of the regional industrial base.
Measuring High-Quality Development and Robot Penetration in Chinese Provinces
Our empirical analysis covers 29 provincial-level administrative regions in Mainland China over the period 2017-2019. Data for output, labor, and capital are sourced from provincial statistical yearbooks. Given the lack of direct provincial-level data on robot installations, we utilize customs import data as a proxy for the provincial stock of industrial robots. Specifically, we aggregate the import value of ten categories of industrial robots under specific HS codes to construct the variable \(RB_{it}\). While this measures the flow/value of robot imports rather than the physical stock, it serves as a reasonable indicator of a province’s investment and deployment intensity in this advanced technology.
The calculated technical efficiency scores reveal notable provincial disparities. As shown in Table 1, the average efficiency over the three-year period ranges from a high of 0.9961 (Liaoning) to a low of 0.5230 (Beijing). A majority of provinces exhibit efficiency scores between 0.70 and 0.80. A concerning trend across all regions is a slight year-on-year decline in technical efficiency during this short period, with the smallest decrease observed in the most efficient province. This overall decline underscores the urgency of finding new drivers for efficiency improvement, such as China robot adoption.
| Province Group | Average TE | Range (Min-Max) | Annual Change Trend |
|---|---|---|---|
| High-Efficiency Group (TE > 0.80) | 0.856 | 0.809 – 0.996 | Slow decline |
| Medium-Efficiency Group (0.70 ≤ TE ≤ 0.80) | 0.741 | 0.703 – 0.786 | Moderate decline |
| Low-Efficiency Group (TE < 0.70) | 0.634 | 0.523 – 0.697 | Relatively faster decline |
| All Provinces | 0.724 | 0.523 – 0.996 | General decline |
The distribution of robot penetration is even more polarized, as detailed in Table 2. Shanghai is an extreme outlier with a penetration rate nearly an order of magnitude higher than the next province, Jiangsu. This stark heterogeneity justifies segmenting the sample into “High-Penetration Regions” (the top 15 provinces by average RP) and “Low-Penetration Regions” (the remaining 14 provinces). The gap between the lowest-ranked province in the high group and the highest-ranked in the low group is significant, creating a natural break point for heterogenous effect analysis. The growth trends also vary, with some provinces increasing their China robot penetration while others saw decreases, potentially influenced by industrial policy and upgrading cycles.
| Rank | Province | Avg. RP (Robots per 10k Workers) | Category |
|---|---|---|---|
| 1 | Shanghai | 99.81 | High-Penetration Regions |
| 2 | Jiangsu | 13.75 | |
| 3 | Tianjin | 7.73 | |
| 4 | Guangdong | 7.16 | |
| 5 | Heilongjiang | 6.09 | |
| 6 | Beijing | 5.03 | |
| 7 | Hebei | 4.19 | |
| 8 | Jilin | 3.94 | |
| 9 | Chongqing | 3.01 | |
| 10 | Zhejiang | 2.43 | |
| 11 | Shandong | 2.10 | |
| 12 | Guangxi | 1.97 | |
| 13 | Liaoning | 1.65 | |
| 14 | Fujian | 1.00 | |
| 15 | Anhui | 0.72 | |
| 16 | Qinghai | 0.38 | Low-Penetration Regions |
| 17 | Sichuan | 0.29 | |
| 18 | Hubei | 0.26 | |
| 19 | Hunan | 0.22 | |
| 20 | Jiangxi | 0.15 | |
| 21 | Shaanxi | 0.13 | |
| 22 | Shanxi | 0.10 | |
| 23 | Henan | 0.05 | |
| 24 | Gansu | 0.01 | |
| 25 | Inner Mongolia | 0.01 | |
| 26 | Hainan | 0.01 | |
| 27 | Ningxia | 0.00 | |
| 28 | Guizhou | 0.00 | |
| 29 | Yunnan | 0.00 |
Empirical Results: The Impact of Robot Penetration on Technical Efficiency
We now present the results from estimating Model 1. The descriptive statistics for all variables are summarized in Table 3, providing an overview of the data used in the regression analysis.
| Variable | Obs | Mean | Std. Dev. | Min | Max |
|---|---|---|---|---|---|
| ln(TE) | 87 | -0.320 | 0.130 | -0.652 | -0.004 |
| ln(RP) | 87 | 0.000136 | 0.000272 | 0 | 0.001587 |
| ln(FDI) | 87 | 9.056 | 1.235 | 6.254 | 11.811 |
| ln(R&D) | 87 | 4.627 | 1.071 | 2.339 | 7.064 |
| ln(GDPpc) | 87 | 3.907 | 0.201 | 3.343 | 4.305 |
| ln(ENT) | 87 | 8.889 | 1.200 | 5.820 | 10.922 |
Baseline Regression Findings
The baseline regression results are reported in Table 4. Columns (1) and (2) present the findings for the full sample. The simple regression in column (1) shows a positive and significant coefficient for robot penetration. After including the full set of control variables in column (2), the coefficient remains positive and statistically significant at the 1% level. The estimated coefficient of 0.3079 implies that a 1% increase in a province’s robot penetration is associated with a 0.3079% increase in its technical efficiency, ceteris paribus. This provides initial evidence supporting the positive role of China robot adoption in fostering high-quality development.
Columns (3) to (6) reveal crucial heterogeneity. For Low-Penetration Regions (columns 3 & 4), the effect in the simple regression is negative but insignificant. However, once controls are added, the effect turns positive and highly significant, albeit with a smaller coefficient (0.1306) compared to the full sample. For High-Penetration Regions (columns 5 & 6), the effect is positive, significant, and substantially larger in magnitude. The coefficient in the full specification is 0.4941, suggesting that the marginal benefit of an additional robot is much greater in regions that have already achieved a certain threshold of adoption. This may be due to network effects, deeper integration with supply chains, or the presence of complementary skills and infrastructure.
| Variable | Full Sample | Low-Penetration Regions | High-Penetration Regions | |||
|---|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | (6) | |
| ln(RP) | 0.7631** (0.330) |
0.3079* (0.176) |
-0.3836 (0.310) |
0.1306*** (0.022) |
0.1237*** (0.034) |
0.4941* (0.235) |
| ln(FDI) | -0.0027*** (0.001) |
-0.0018*** (0.001) |
-0.0085*** (0.002) |
|||
| ln(R&D) | 0.0008 (0.001) |
0.0004 (0.001) |
0.0003 (0.002) |
|||
| ln(GDPpc) | -0.0139*** (0.002) |
-0.0117*** (0.003) |
-0.0156*** (0.004) |
|||
| ln(ENT) | 0.0003 (0.001) |
0.0018 (0.002) |
-0.0017 (0.002) |
|||
| Constant | -0.3212*** (0.001) |
-0.2481*** (0.014) |
-0.2827*** (0.000) |
-0.2352*** (0.017) |
-0.3581*** (0.001) |
-0.2574*** (0.025) |
| Year/Region FE | Yes | Yes | Yes | Yes | Yes | Yes |
| R-squared | 0.0910 | 0.7812 | 0.0002 | 0.8076 | 0.2385 | 0.6739 |
| Observations | 78 | 78 | 42 | 42 | 45 | 45 |
Note: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. Robust standard errors in parentheses.
Addressing Endogeneity: Instrumental Variable Approach
A potential concern in the baseline analysis is reverse causality: provinces with higher inherent technical efficiency might be more likely to invest in advanced technologies like robots. To mitigate this endogeneity issue, we employ an instrumental variable (IV) approach. We need an instrument correlated with robot penetration but uncorrelated with the error term in the technical efficiency equation. We use the annual number of academic publications on industrial robots where the author’s affiliation is in a given province. The rationale is that a higher level of local research activity in robotics is likely correlated with greater industrial adoption and policy focus (relevance condition), but the sheer number of academic papers is unlikely to directly affect aggregate provincial production efficiency outside of its correlation with actual adoption (exclusion restriction).
The IV regression results, presented in Table 5, confirm the robustness of our baseline findings. The coefficient on \(\ln(RP)\) remains positive and statistically significant across all samples—full, low-penetration, and high-penetration regions. The magnitudes are consistent with the baseline OLS estimates, lending strong credibility to the causal interpretation that an increase in China robot penetration leads to an improvement in regional technical efficiency.
| Variable | Full Sample | Low-Penetration Regions | High-Penetration Regions |
|---|---|---|---|
| ln(RP) | 0.5701** (0.269) |
0.2171** (0.105) |
0.1916*** (0.042) |
| Controls | Yes | Yes | Yes |
| Year/Region FE | Yes | Yes | Yes |
| Observations | 78 | 42 | 45 |
Exploring the Mechanism: The Role of Absorptive Capacity
The heterogenous effects suggest that simply deploying robots may not be sufficient. The benefits likely depend on a region’s ability to effectively absorb and utilize the technology. We hypothesize that a region’s absorptive capacity, proxied by its R&D investment, interacts with robot penetration to enhance technical efficiency. To test this, we augment Model 1 by including an interaction term between robot penetration and R&D expenditure:
$$ \ln TE_{it} = \alpha_0 + \alpha_1 \ln RP_{it} + \alpha_2 \ln R\&D_{it} + \alpha_3 (\ln RP_{it} \times \ln R\&D_{it}) + \alpha_4 \mathbf{C’_{it}} + \epsilon_{it} $$
where \(\mathbf{C’_{it}}\) includes the other controls. A positive and significant \(\alpha_3\) would indicate that the effect of robot penetration is stronger in provinces with higher R&D intensity.
The results, shown in Table 6, provide strong support for this mechanism. The interaction term \(\ln(RP) \times \ln(R\&D)\) is positive and statistically significant at the 1% level for all three samples. This confirms that the positive impact of China robot adoption on technical efficiency is magnified by a region’s indigenous R&D efforts or absorptive capacity. Notably, for Low-Penetration Regions, the standalone coefficient on \(\ln(R\&D)\) is negative (though insignificant), while the interaction is large and positive. This implies that in regions with low initial robot adoption, R&D spending alone may not boost efficiency, but when directed towards complementing and mastering robot technology, it becomes highly effective. For High-Penetration Regions, both the direct effect of robots and the synergistic effect with R&D are strongly positive.
| Variable | Full Sample | Low-Penetration Regions | High-Penetration Regions |
|---|---|---|---|
| ln(RP) | 0.3201* (0.180) |
0.1036* (0.086) |
0.2541* (0.115) |
| ln(R&D) | 0.0008 (0.001) |
-0.0004 (0.001) |
0.0025* (0.001) |
| ln(RP) × ln(R&D) | 0.0225*** (0.007) |
0.0899*** (0.039) |
0.0440*** (0.009) |
| Other Controls | Yes | Yes | Yes |
| Year/Region FE | Yes | Yes | Yes |
| R-squared | 0.7948 | 0.8108 | 0.8343 |
Conclusions and Policy Implications
This study investigates the relationship between industrial robot adoption and high-quality development, measured by provincial technical efficiency, in China. The empirical analysis yields several key conclusions with important policy implications.
1. Industrial Robots as a Driver for High-Quality Development: The evidence robustly indicates that increased penetration of industrial robots significantly enhances regional technical efficiency. This validates the premise that embodied technological progress through China robot adoption is a viable pathway toward achieving the national strategic goal of high-quality development. Policymakers should therefore consider and reinforce fiscal, tax, and financial policies that encourage the development of the domestic robotics industry, facilitate the import of advanced robotic systems, and incentivize enterprises to accelerate equipment renewal, replacing traditional machinery with modern, efficient industrial robots.
2. Heterogeneous Effects Require Differentiated Policies: The impact of robot penetration is not uniform. The effect is significantly larger in High-Penetration Regions compared to Low-Penetration Regions. This can be attributed to the high fixed costs and network effects associated with advanced automation. For Low-Penetration Regions, the initial hurdle is greater. Policy support here should be more focused on lowering the entry barrier, potentially through targeted subsidies for first-time adopters, establishing regional demonstration centers, and developing financing schemes for small and medium-sized enterprises to invest in robotics. For High-Penetration Regions, policy should shift towards fostering the adoption of next-generation, more complex robots and integrating them with AI and IoT to move up the value chain and sustain efficiency gains.
3. The Critical Role of Absorptive Capacity: The mechanism tests reveal that the productivity benefits of robots are not automatic; they are contingent upon a region’s absorptive capacity, proxied by R&D investment. The significant positive interaction term shows that robot adoption and local R&D are complementary. This implies that policy must be two-pronged: promoting both the diffusion of the technology (the hardware) and the development of the local capability to use it effectively (the software and skills).
4. The R&D Gap in Lagging Regions: The finding that in Low-Penetration Regions, R&D spending alone has an insignificant (or even slightly negative) direct effect, but a powerfully positive interaction effect with robots, is crucial. It suggests that R&D resources in these regions may not be optimally allocated or are disconnected from frontier technological applications like robotics. Therefore, a specific policy recommendation for these regions is to channel a greater portion of public and private R&D investment towards fields directly related to automation, robotics integration, and workforce re-skilling, to build the necessary absorptive capacity to leverage China robot investments when they occur.
In summary, the strategic integration of industrial robots into China’s economic fabric presents a substantial opportunity to enhance production efficiency and steer the economy onto a high-quality development path. However, realizing this potential requires nuanced, region-specific policies that address both the quantity of robot deployment and the quality of the complementary human capital and innovation ecosystem needed to absorb this transformative technology.