An empirical study of enterprise investment relationship networks in China

Lai Jiang, Jiale Wang, Chunxiao Li, Yaohui Jin, Yanyan XuHumanities 2025 · DOI

Abstract

Enterprise investment relationships serve as a microcosm reflecting macroeconomic vitality, which is restricted and affected by geographic locations. This study develops the Chinese Enterprise Investment Relationship Dataset and integrates complex network analysis to scrutinize spatial agglomeration patterns and industry concentration trends. The research reveals the scale-free characteristic of investment networks, which follow power-law distributions across different spatial scales. Additionally, the study connects these networks with urban development variables such as GDP, population flow, and AQI, examining the manifestation of urban scaling laws within the realm of enterprise investment.

Introduction

While traditional economic analysis often focuses on macro indicators like GDP or micro-level company reports, there is a significant gap in understanding the "meso-level" network of investment. Investment is not merely a series of isolated transactions; it is a spatial web. This paper addresses the lack of a comprehensive, national-level dataset connecting enterprise-level investment ties with geographic and environmental factors in China.

Methodology and Data

The centerpiece of this research is the Chinese Enterprise Investment Relationship Dataset, constructed from the National Enterprise Credit Information Publicity System (2021). It contains 406,769 enterprises and 321,827 investment records, geocoded to provincial, city, and county levels.

The authors employed several advanced techniques:

  • Complex Network Analysis: Treating enterprises as nodes and investments as edges to calculate degree distributions.
  • XGBoost & SHAP: Using machine learning to identify industry drivers of economic indicators.
  • Urban Scaling Law Modeling: Applying the power-law equation $Y = Y_0N^\beta$ to analyze how investment scales with city population.

Spatial Clustering and Network Topology

Despite the digital age, geography remains a powerful determinant of capital flow. The study identifies a strong "geographical spatial clustering effect," with the majority of investment occurring within the same province or between established economic hubs like the Yangtze River Delta and the Pearl River Delta.

Spatial distribution of enterprise investment networks across China, illustrating the concentration of capital in major economic clusters.

The network is fundamentally "scale-free," meaning a small number of hub enterprises and cities facilitate a disproportionate amount of investment activity. This structure makes the economic system robust but highly dependent on these central nodes.

Key Results and Scaling Laws

Statistical analysis confirms that the investment network follows a power-law distribution. The power-law exponent ($\gamma$) was found to be 3.27 at the enterprise level and 1.71 at the county level, highlighting the inhomogeneous, hub-centric nature of the system.

Log-log plots showing the power-law distribution of investment frequency relative to degree at the enterprise (c) and county (d) levels.

Key findings include:

  • Super-linear Efficiency: Investment activity scales super-linearly with population ($\beta > 1$). Larger cities are exponentially more efficient at generating investment than smaller ones.
  • Economic Correlation: Investment quantity shows a strong correlation with GDP (0.88–0.90) and population inflow (0.68–0.72).
  • The AQI Paradox: Active investment regions often exhibit better air quality, suggesting a shift from heavy industry to high-tech and service sectors in China's economic hubs.

Conclusion and Future Directions

This study provides a comprehensive map of China's economic "circulatory system." Future research will aim to incorporate time-series data to observe network evolution and transition to weighted graphs that account for specific monetary amounts and shareholding percentages.

Reference: Jiang, L., Wang, J., Li, C., Jin, Y., & Xu, Y. (2025). An empirical study of enterprise investment relationship networks in China. Humanities & Social Sciences Communications, 12:1712. https://doi.org/10.1057/s41599-025-05969-4

Summary updated

Abstract

Enterprise investment relationships serve as a microcosm reflecting macroeconomic vitality, which is restricted and affected by geographic locations of these enterprises. The study of enterprise investment relationships and their relevance to geographic spatial factors holds significant implications for decision-making and economic planning. Scholars have conducted empirical investigations into global investment networks and venture capital networks. However, fewer studies have employed a network perspective to scrutinize corporate investment relationships, and even fewer have focused specifically on China. In this paper, we develop the Chinese Enterprise Investment Relationship Dataset and innovatively integrate complex network analysis techniques into the study of enterprise investment relationships in China, establishing different scales of geographic networks. Employing statistical methodologies, we scrutinize spatial agglomeration patterns and industry concentration trends, and reveal the scale-free characteristic of the investment networks. That is, they follow the power-law distributions at different spatial scales. Additionally, we connect the investment networks with some urban development variables, including GDP, population flow, and AQI. Finally, we examine the manifestation of urban scaling laws within the realm of enterprise investment relationships. Centered on China, this work presents an well-organized investment networks data and an example of analyzing investment relationships with complex network methods.
BibTeX
@inproceedings{jiang2025empirical,
  author = {Lai Jiang and Jiale Wang and Chunxiao Li and Yaohui Jin and Yanyan Xu},
  title = {{An empirical study of enterprise investment relationship networks in China}},
  booktitle = {Humanities},
  year = {2025},
  doi = {10.1057/s41599-025-05969-4},
}
Figure