Spillovers across industries and regions in China’s regional economic diversification

Jian Gao, Bogang Jun, Alex Pentland, Tao Zhou, César A. HidalgoRegional 2021 · DOI

TL;DR

This paper investigates how industries and regions in China diversify their economies by leveraging knowledge spillovers. It finds that while both industry-to-industry and region-to-region spillovers are crucial, they act as substitutes, meaning the benefit of one diminishes when the other is strong. The study also uses high-speed rail (HSR) as a natural experiment to demonstrate the causal impact of improved transportation on regional economic integration.

What Problem Does This Paper Address?

Economic development often hinges on a region's ability to diversify its industrial base. This diversification isn't random; it's heavily influenced by "spillovers" – the diffusion of knowledge, capabilities, and best practices. These spillovers can occur in two primary ways: inter-industry, where a region leverages expertise from related industries already present, and inter-regional, where a region benefits from the success of neighboring regions. While both have been studied individually, their interaction and how they influence each other, especially in a rapidly developing economy like China, remained unclear. The paper aims to untangle this complex relationship and explore how factors like transportation infrastructure can amplify these effects.

Methods

The authors employ a combination of econometric modeling and a quasi-experimental approach.

  • Econometric Models: They use probit models to estimate the probability of a province entering or maintaining a "revealed comparative advantage" (RCA) in a new industry. This involves constructing measures for:
  • Density of Active Related Industries (w): This quantifies how many industries related to a target industry are already present and successful (have RCA) in a given province.
  • Density of Active Neighboring Provinces (Ω): This measures the extent to which neighboring provinces are successful in a particular industry. They then analyze the individual effects of w and Ω, their interaction (to see if they are complements or substitutes), and control for factors like industry ubiquity and provincial diversity.

  • Difference-in-Differences (DID) with HSR: To address potential endogeneity issues (i.e., regions that are already developing might also be the ones to receive HSR), the study uses the introduction of high-speed rail (HSR) as an instrumental variable. They compare pairs of provinces connected by HSR (treatment group) with those not connected (control group) before and after HSR implementation. This allows them to isolate the causal effect of improved transportation on regional economic similarity.

Datasets

The study primarily uses two datasets:

  1. China's Publicly Listed Firm Data: This dataset covers the period from 1990 to 2015 and includes information on registered addresses, industry categories, revenue, and employee numbers for listed firms. This data is used for the main analysis of inter-industry spillovers.
  2. Chinese Enterprise Data: This dataset, sourced from the China Statistical Yearbook, provides census data on labor and economic output for industrial enterprises above a designated size from 1998 to 2008. It offers a broader coverage of firms and industries at the provincial level.
  3. Brazilian Labour Data: To ensure the robustness of their findings beyond China, the authors also utilize Brazilian labor market data (2006-2013) covering a fine spatial resolution (micro-regions) and a large number of industries.

Key Insights and Ideas

  • Spillovers as Substitutes: Perhaps the most striking finding is that inter-industry and inter-regional spillovers tend to act as substitutes. This means that if a province is already strong in related industries, the benefit it receives from neighboring regions' success in that industry is reduced, and vice-versa. This suggests a dynamic where regions can specialize in leveraging either internal (industry-related) or external (region-related) knowledge, rather than benefiting equally from both simultaneously.
  • HSR as a Catalyst for Regional Integration: The HSR analysis provides strong causal evidence that improved transportation infrastructure significantly increases industrial similarity between connected provinces. This highlights how HSR can break down geographical barriers, facilitating the flow of knowledge and capabilities, and thus promoting regional economic convergence.
  • Path Dependency in Development: The study reinforces the idea that economic development is path-dependent. Provinces are more likely to develop industries that are related to their existing industrial base or are present in neighboring regions, suggesting that historical endowments and geographical proximity play a crucial role in shaping future economic trajectories.

Key Results

  • Inter-Industry Spillovers: The density of active related industries in a province is a strong, positive, and significant predictor of both entering new industries and maintaining existing ones.
  • Inter-Regional Spillovers: Similarly, the density of active neighboring provinces positively and significantly predicts the development of new industries. However, the effect on maintaining existing industries is more complex, showing a tension between competition and spillover benefits.
  • Interaction Effect: The interaction term between industry and regional spillover densities is negative and significant, confirming that these two channels behave as substitutes.
  • HSR Impact: The DID analysis shows that HSR entry leads to a statistically significant increase in industrial similarity between connected provinces, with a DID estimate of 0.029. This effect is robust to controlling for macroeconomic factors.

Limitations and Future Work

The authors acknowledge several limitations:

  • Multilocation Firms: The study doesn't fully account for firms operating in multiple locations, which can lead to within-firm spillovers and diversification across regions.
  • Distinguishing Competition from Spillovers: While the study finds evidence of regional spillovers, it notes that neighboring regions can also be sources of competition, which might dampen the observed spillover effects.
  • Micro-Channels of Spillovers: The paper identifies the existence of spillovers but doesn't delve into the specific micro-mechanisms (e.g., spin-off companies, migrant workers, labor pooling) through which they occur.

Future research could explore these micro-channels, investigate interregional linkages beyond geographical proximity (like scientific collaboration or co-patenting), and utilize larger, more granular datasets to further refine the understanding of these complex economic dynamics. The findings also offer practical implications for policymakers, suggesting that investments in transportation infrastructure can be a powerful tool for fostering regional economic development and diversification.

Summary updated

Abstract

Industrial diversification depends on spillovers from related industries and nearby regions, yet their interaction remains largely unclear. We study economic diversification in China during the period 1990–2015 and present supportive evidence on both spillover channels. We add to the literature by showing that these two channels behave as substitutes when explaining new entries and exits, and by using acceleration campaigns of high-speed rail to address some endogeneity concerns with regional spillovers. Our findings confirm the role of relatedness and geographical distance in the diffusion of economic capabilities and support the idea that improvements in transportation can facilitate the diffusion of productive capabilities.
BibTeX
@inproceedings{gao2021spillovers,
  author = {Jian Gao and Bogang Jun and Alex Pentland and Tao Zhou and César A. Hidalgo},
  title = {{Spillovers across industries and regions in China’s regional economic diversification}},
  booktitle = {Regional},
  year = {2021},
  doi = {10.1080/00343404.2021.1883191},
}
Figure