📈 economics

Scale Meets Networks: Interactions and Multi-path Transmission Mechanisms between Urban Economic Scale and Innovation Networks in China

Using patent data from 296 Chinese cities and a generalized structural equation model, this study reveals a converging superlinear relationship and positive feedback cycle between urban economic scale and innovation networks, where economic scale drives network formation more strongly through labor matching, while industrial agglomeration creates negative lock-in effects and policy pathways remain context-dependent.

Original authors: Ying Zhou, Wensheng Zheng

Published 2026-07-23

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Ying Zhou, Wensheng Zheng

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the world of cities as a giant, bustling playground where two big forces are constantly dancing together. On one side, you have Urban Economic Scale, which is basically just a fancy way of saying "how big and rich a city is." Think of it like the size of a city's wallet and the number of people living there. On the other side, you have Innovation Networks, which are like invisible webs of friendship and teamwork connecting people in different cities. These webs are made of shared ideas, like when a scientist in Beijing teams up with an engineer in Shanghai to invent something new.

For a long time, experts argued about which dancer leads the dance. Does being a big, rich city make you a better friend to other cities? Or does having a huge web of friends make your city richer and bigger? This paper dives into that question, looking at how these two forces pull and push against each other. It's like asking: "Does having a big house help you make more friends, or does having a huge group of friends help you build a bigger house?" Understanding this helps us figure out how to make sure no city gets left behind while others get super successful.

The Big Discovery: The "Rich Get Richer" Dance

This study, written by Ying Zhou and Wensheng Zheng, took a massive look at 296 cities in China between 2018 and 2023. They didn't just guess; they looked at a staggering 1,790,452 patent cooperation records. Think of a patent as a certificate for a new invention, and a "cooperation record" as proof that two different cities worked together to create it. By mapping out who worked with whom, they built a giant map of the innovation network.

They used a special math tool called a Generalized Structural Equation Model (GSEM) to figure out the cause-and-effect. Imagine trying to untangle two knots of string that are tied together; this tool helps you see which string is pulling which.

Here is what they found:

1. The Super-Linear Superstar Effect
The researchers discovered a "superlinear" relationship. In plain English, this means that as a city gets bigger, its connections to other cities don't just grow a little bit—they explode! If a city's economy doubles, its innovation network grows by more than double. It's like a snowball rolling down a hill; the bigger it gets, the faster it picks up more snow. However, they noticed this "explosion" effect is slowly calming down, suggesting the system is trying to find a more balanced rhythm.

2. The One-Way Street is Stronger
While big cities and big networks help each other (a positive feedback loop), the paper found that economic scale is the stronger leader. A city's size and wealth have a bigger impact on how many friends it makes in the innovation world than having friends has on making the city richer. It's like having a big, fancy house attracts more party guests than having a lot of party guests helps you build the house.

3. The Secret Ingredient: People on the Move
The study tested three "secret ingredients" (pathways) that might explain how a city's size helps it make friends.

The Good News: Labor Matching (people moving around) is a superstar. When big cities attract skilled workers who move between cities, it creates a perfect match for ideas. This pathway works really well!
The Bad News: Industrial Agglomeration (having too many factories of the same type in one spot) actually hurts the network. The paper found a "negative lock-in effect." It's like a club that's so exclusive and full of the same people that they stop talking to outsiders. If a city is too focused on just one type of industry, it stops reaching out to new partners.
The "It Depends" News: Policy Support (government giving money) didn't work as a magic bullet on its own. Just because the government gives a city money doesn't automatically mean that city will become a hub of innovation.

4. The Reverse Dance: How Networks Change Cities
When the researchers looked at how having a big network helps a city grow, they found some surprises too.

Innovation Output: Being in the center of the network helps a city invent more things (more patents), which is good.
Population: Here is the twist. At the national level, big networks seem to attract people. But when you look at specific groups of cities, being a "network hub" actually seems to drain people from smaller, weaker cities. It's like a magnet: the big hubs suck the talent away from the smaller towns, making the rich cities richer and the poor towns poorer.
Industrial Upgrading: The one thing that worked consistently was that networks helped cities move toward higher-value industries (like moving from making simple toys to designing complex robots). This happened in all types of cities, big and small.

5. The Four Types of Cities
The authors sorted the cities into four groups based on their size and how well they were connected:

Network Driven: Big cities with huge networks (like the superstars).
Leapfrogging: Small cities with surprisingly huge networks (the underdogs doing great).
Resource Mismatch: Big cities that are surprisingly lonely (big wallets, few friends).
Double Lagging: Small cities with few connections (the ones left behind).

Most cities stayed in their group over the years, showing that it's hard to break out of your starting position. However, some "mismatch" cities managed to become "network driven," and some "double lagging" cities managed to "leapfrog."

What This Means for the Future

The paper suggests that we can't just assume "bigger is better" or "more connections are always good."

For the Big Cities: They need to be careful not to get too stuck in their own industries. They need to keep their doors open to new ideas and people, or they might stop growing.
For the Small Cities: They can't just wait for the government to give them money. They need to focus on making it easy for talented people to move in and out, and they need to upgrade their industries to higher-value work.
The Big Picture: The system is currently reinforcing inequality. The big cities are getting bigger and more connected, while the smaller ones struggle to catch up. The authors suggest that to fix this, we need to help the smaller cities break out of their "lock-in" and make sure that the flow of people and ideas doesn't just drain the small towns dry.

In short, the dance between city size and innovation networks is real, it's powerful, and it's currently favoring the big dancers. But with the right steps—like helping people move freely and avoiding getting stuck in one industry—maybe everyone can learn to dance better.

Technical Summary: Scale Meets Networks

Problem Statement

Under the context of innovation-driven development and highly mobile production factors, the spatial organization of urban systems in China is undergoing profound restructuring. A critical tension exists between two competing spatial logics: the traditional "central place" hierarchy, where urban economic scale dictates factor agglomeration and resource allocation, and the "space of flows" paradigm, where urban competitiveness depends on network embeddedness and cross-scalar linkages.

While existing literature has examined the unidirectional influence of urban scale on network position or the impact of network centrality on urban growth, there is a lack of systematic understanding regarding their bidirectional interaction mechanisms. Furthermore, the multidimensional transmission pathways through which these two dimensions influence one another remain insufficiently decomposed. Specifically, it is unclear whether the expansion of innovation networks reinforces existing hierarchical inequalities or generates synergistic effects, and what specific mechanisms (e.g., labor matching, industrial agglomeration, policy bias) govern these interactions.

Methodology

The study employs a robust quantitative framework utilizing data from 296 prefecture-level and above cities in China spanning the period 2018–2023.

Data Sources:

Innovation Networks: Constructed using 1,790,452 joint invention patent applications disclosed by the China National Intellectual Property Administration (CNIPA).
Economic Scale: Measured by Gross Domestic Product (GDP) from the China Urban Statistical Yearbook.
Mediating Variables: Derived from Gaode (Amap) intercity migration data (population flows), statistical yearbooks (industrial agglomeration, policy expenditure), and calculated indices (local innovation output, industrial structure upgrading).

Analytical Approach:

Power-Law Identification: The study first identifies the relationship between urban economic scale and innovation network centrality using the Urban Scaling Law framework. It tests for superlinear ( $\beta > 1$ ), linear ( $\beta = 1$ ), or sublinear ( $\beta < 1$ ) relationships. Cities are subsequently classified into four types based on residuals and economic scale: Network-driven, Leapfrogging, Resource mismatch, and Double lagging.
Generalized Structural Equation Modeling (GSEM): A GSEM is constructed to simultaneously estimate:
- Bidirectional Causality: The direct feedback loop between economic scale (lnGDP) and innovation network centrality (lnNC).
- Mediation Pathways: The decomposition of indirect effects through specific mechanisms:
  - Scale $\to$ Network: Industrial agglomeration, labor matching (population flows), and policy bias (R&D expenditure).
  - Network $\to$ Scale: Population agglomeration, local innovation capacity, and industrial structure upgrading.
- Heterogeneity Analysis: The model is estimated separately for the four city types to identify conditional effects.
- Estimation: Maximum Likelihood (ML) estimation in Stata 17 with clustered robust standard errors at the city level to account for serial correlation.

Key Results

1. Superlinear Relationship and Spatial Patterns

Converging Superlinearity: A power-law relationship exists where innovation network centrality scales superlinearly with economic scale ( $\beta > 1$ ). However, the exponent $\beta$ declined from 1.479 (2018) to 1.338 (2023), suggesting a gradual convergence toward a more balanced state.
Spatial Heterogeneity: "Network-driven" cities are concentrated in major urban agglomerations (Yangtze River Delta, Pearl River Delta, Beijing-Tianjin-Hebei), while "Double lagging" cities are prevalent in western and northeastern China.
Path Dependence: Approximately 68% of cities remained in the same category over time, indicating strong path dependence in the co-evolution of scale and networks.

2. Bidirectional Interaction Mechanisms

Asymmetric Feedback: A positive feedback loop exists, but it is asymmetric. The effect of economic scale on innovation networks (coefficient: 0.405) is significantly stronger than the reverse effect of networks on economic scale (coefficient: 0.226). This suggests that economic hierarchy remains the dominant driver of network structure.
Mediating Pathways (Scale $\to$ Network):
- Labor Matching (Positive): Intercity population flows significantly mediate the relationship, enhancing network embeddedness.
- Industrial Agglomeration (Negative): Contrary to traditional agglomeration theories, excessive industrial concentration produces a negative "lock-in" effect, reducing network openness and inter-city collaboration.
- Policy Bias (Insignificant): At the national level, policy resource allocation does not significantly translate into stronger network positions.
Mediating Pathways (Network $\to$ Scale):
- Population Agglomeration (Context-Dependent): While positive at the national level, this pathway becomes significantly negative within specific city types, indicating that network intensification may exacerbate population outflows from peripheral cities within homogeneous groups.
- Local Innovation (Insignificant): Increased patent output driven by networks does not consistently translate into sustained economic growth at the national level.
- Industrial Upgrading (Positive): Innovation networks consistently promote economic growth by facilitating structural upgrading across all city types, even if this effect is masked in aggregate national data.

Key Contributions

Theoretical Integration: The study moves beyond unidirectional causality to empirically validate a bidirectional, asymmetric feedback loop between urban hierarchy and innovation networks, confirming that while they reinforce each other, economic scale remains the primary shaper of network structure.
Pathway Decomposition: It reveals the heterogeneous and often counter-intuitive nature of transmission mechanisms. Specifically, it challenges the assumption that industrial agglomeration always aids innovation, demonstrating instead that it can induce a "lock-in" effect that hinders network expansion.
Contextual Heterogeneity: By classifying cities into four distinct types, the study uncovers scale-dependent reversal mechanisms. For instance, population agglomeration acts as a growth engine at the national level but as a displacement mechanism within specific city groups, highlighting the complex interplay between integration and polarization.

Significance and Claims

The authors claim that these findings provide critical implications for optimizing China's urban innovation networks and promoting coordinated regional development. The study argues that urban economic scale and innovation networks are not independent drivers but mutually constitutive forces embedded in a hierarchically structured system.

The research suggests that:

Policy Implications: Strategies must move beyond simple top-down resource allocation, which has shown limited conversion efficiency. Instead, policies should focus on improving labor mobility channels to counteract the negative lock-in effects of industrial agglomeration.
Differentiated Strategies: "Network-driven" cities need resource decentralization to avoid diminishing returns, while "Leapfrogging" cities require support to convert network advantages into economic scale. "Resource mismatch" cities need to align traditional industries with innovation networks, and "Double lagging" cities require targeted support as market mechanisms alone are insufficient to close the gap.
Structural Upgrading: Innovation networks serve as effective vehicles for industrial structure upgrading, suggesting that fostering cross-regional collaboration is vital for moving cities toward higher value-added economic structures.

Ultimately, the paper posits that without addressing the asymmetric nature of the scale-network relationship and the specific transmission pathways, regional development strategies risk reinforcing existing spatial inequalities rather than fostering genuine coordinated growth.