Made in China 2025 Improved Robot Quality. Core Components Still Come From Japan
A Made in China 2025 pilot program significantly improved robotics innovation quality in applications and assembly, but failed to advance core component technologies.
The paper (Does “Made in China (2025)” improve innovation quality in robotics? Evidence from PageRank-based patent network by Yihan Han, Silu Pang, Fuxin Jiang and Tao Wang) talks about the (frankly) surprising success of the Made in China 2025 in boosting the actual quality of China’s robotics industry (well midstream and downstream, not so much the upstream). China overtaken Japan a while ago in advanced industrial robots manufacturing, but one of the key (and cliched) limits is the *quality* of the robots. That is frankly changing. The program mobilized 12 pilot cities and 604 companies generating 398,872 patents, yet achieved contrasting outcomes across the robotics value chain. These results offer evidence for global debates about industrial policy effectiveness.
By the numbers:
China's domestic industrial robot market share rose from marginal to 52.45% by 2023
Robot exports increased 86.4%
8,453 firm-year observations from 604 A-share listed robotics companies (2010-2023)
Innovation quality improved 15.9% overall in pilot cities after 2017
Downstream integrators achieved 23.4% innovation quality gains
Midstream manufacturers saw 17.3% improvement
Upstream component makers showed no statistically significant improvement
Sample includes 209 upstream, 111 midstream, and 284 downstream companies
The methodology: The paper use an enhanced PageRank algorithm to measure innovation quality through patent citation networks. The method evaluates patents based on their influence on subsequent innovations, similar to how Google's algorithm ranks websites by the quality of incoming links.
Four Mechanisms Driving Partial Success
1. Targeted Innovation Subsidies
Pilot cities including Guangzhou, Wuhan, and Nanjing shifted from general corporate support to R&D-specific funding. Innovation subsidies increased 67.2% (statistically significant at 1% level) while general subsidies showed no significant change. Researchers identified 68,354 innovation subsidy records through keyword analysis of government documents.
2. Strengthened Intellectual Property Protection
Patent infringement resolution rates rose 5.1% in pilot zones (p < 0.05). Guangzhou implemented rapid protection systems with enhanced penalties and established monitoring mechanisms for IP violations. The study measured protection strength using the formula: Resolution Rate × ln(1 + Granted Patents), capturing both enforcement effectiveness and innovation activity.
3. Accelerated Robot Adoption
Robot installation density increased 111.8% in pilot regions (p < 0.01). This created feedback loops between manufacturers and users. Suzhou Green Harmonic Drive, for example, developed "P-tooth" structures and third-generation harmonic transmission technology in response to specific demands from Yangtze River Delta manufacturers. The study weighted installations by industry robot intensity and regional labor shares to calculate accurate density measures.
4. Enhanced Human Capital
Firms in pilot cities increased their share of employees with advanced degrees by 6.2% (p < 0.05) and raised per-employee training expenditure by 10.6% (p < 0.01). Programs such as the "Pearl River Talent Plan" attracted high-skilled professionals while universities partnered with companies for collaborative innovation.
Value Chain Analysis Reveals Sharp Contrasts
Successful segments:
Midstream robot body manufacturers (111 companies, 149,632 patents): 17.3% improvement focusing on mechanical components, end effectors, and joints
Downstream system integrators (284 companies, 160,649 patents): 23.4% improvement in welding, assembly, and service robot applications
Failed segment:
Upstream component producers (209 companies, 88,591 patents): No significant improvement in core technologies including reducers, controllers, and sensors
The policy coefficient for upstream firms was 0.089, statistically indistinguishable from zero despite seven years of targeted support. These companies focus on precision reducers, control algorithms, high-performance sensors, and servo systems where Japanese firms Nabtesco and Harmonic Drive maintain dominance.
Validation Through Multiple Approaches
The study's robustness checks confirm the main findings:
Propensity Score Matching: 14.5% effect persists with 3,841 matched observations
Entropy Balancing: Effect increases to 16.6% after balancing covariates
500 placebo tests: Random assignments cluster around zero
Alternative innovation measures confirm improvements: invention patents up 12.1%, utility patents up 23.2%, forward citations up 10.2%
When excluding COVID-19 years (2020-2021), the policy effect strengthens to 17.9%, suggesting temporary pandemic disruption rather than policy failure. The effects remain significant after controlling for concurrent AI Innovation and National Innovative City pilot programs.
Large Firms Captured Disproportionate Benefits
Established companies benefited more than smaller firms:
Large firms (above median assets): 20.0% innovation improvement (p < 0.01)
Small firms: 12.9% improvement (p < 0.01)
Difference statistically significant at 5% level
This 55% stronger effect for large firms contradicts expectations that smaller companies would respond more dynamically to policy incentives. The study attributes this to superior resource access, established R&D infrastructure, and stronger institutional relationships among larger firms.
Shanghai Step Electric: A Case Study
The company exemplifies successful innovation under the policy. Using the PageRank algorithm, researchers tracked how Shanghai Step's patents evolved from baseline scores to 230.3, indicating genuine influence on subsequent control system innovations. The analysis traced specific patent citations through IPC classifications including B25J (manipulators), H02P (control of electric motors), and G01B (measuring instruments), demonstrating knowledge diffusion across technical domains.
Technological Barriers Explain Upstream Failure
Core component development faces fundamentally different challenges:
Development cycles of 5-10 years versus 1-2 years for applications
Requirement for materials science breakthroughs rather than incremental improvements
Need for precision manufacturing capabilities developed over generations
Investment scales exceeding downstream innovation by orders of magnitude
Green Harmonic achieved advances in torque and lifespan but continues to lag behind Harmonic Drive in precision and noise reduction. These gaps reflect accumulated expertise rather than funding shortfalls.
Policy Implications
For upstream technologies: Establish mission-driven teams combining enterprises and universities, support technology acquisition through overseas mergers with tax incentives, create patient capital funds with horizons exceeding 10 years, and focus resources on specific technical bottlenecks.
For quality assessment: Implement three-tier evaluation frameworks with government standards, third-party technical assessment, and enterprise data provision. Apply network-based algorithms to identify breakthrough innovations beyond simple patent counts.
For ecosystem development: Strengthen IP protection infrastructure before expecting innovation returns, facilitate researcher mobility between institutions, create platforms linking technology users and developers, and implement targeted talent programs modeled on successful pilot city initiatives.
Bottomline
The Made in China 2025 initiative demonstrates the potential of industrial policy, and the blindspots that policymakers might want to be aware of. The program successfully improved quality in robot applications and integration, where market feedback loops are tight and development cycles short. However, it failed to advance core component technologies requiring fundamental scientific breakthroughs and generational expertise accumulation.
This divergence suggests that industrial policy works best when enhancing existing capabilities rather than creating entirely new ones. China's 52.45% market share built partially on imported components showcase this: policy can accelerate commercialization and application innovation. It cannot compress the time required for basic technological development, unless you design for basic technological development. Successful industrial policy requires matching support mechanisms to technological characteristics. Resources alone cannot overcome the patient accumulation of expertise that defines leadership in precision manufacturing and materials science.
For American, European, Japanese, Korean, etc policymakers: Another paper on top of the mountain of evidence suggests it's time to grow up and move past consensus about whether governments should have an industrial policy or even funding it (because tariffs and noise making doesn’t exactly seem to provide results ig you aren’t going to put your money where your mouth is).
The question isn't whether industrial policy works, China keeps on just proved it does, you just need to make sure to adapt it on a market by market basis. The real question is whether you're willing to compete in a world where your biggest rival is successfully using tools you've spent 50 years claiming don't work or been downgrading.