[Paper Review] Network structure of inter-industry flows

James McNerney, Brian D. Fath+1 more|IIASA PURE (International Institute of Applied Systems Analysis)|Apr 18, 2012

Business Strategy and Innovation4 citations

TL;DR

This paper analyzes inter-industry money flow networks across 20 national economies using input/output (I/O) tables, revealing a universal network structure characterized by a Weibull distribution of link weights, exponential industry size distribution, and a hierarchical community structure. The key finding is a consistent five-community hierarchy—food, chemicals, manufacturing, services, and extraction industries—indicating a shared macroeconomic network topology across diverse economies.

ABSTRACT

We study the structure of inter-industry relationships using networks of money flows between industries in 20 national economies. We find these networks vary around a typical structure characterized by a Weibull link weight distribution, exponential industry size distribution, and a common community structure. The community structure is hierarchical, with the top level of the hierarchy comprising five industry communities: food industries, chemical industries, manufacturing industries, service industries, and extraction industries.

Motivation & Objective

• To identify common structural patterns in inter-industry money flow networks across diverse national economies.
• To investigate whether industrial networks exhibit universal topological features despite institutional and sectoral differences.
• To assess the suitability of complex systems and statistical physics models for capturing macroeconomic network dynamics.
• To examine the distribution of flow weights, industry sizes, and community structures in national input/output tables.
• To determine whether a typical network structure underlies the diversity of national economic systems.

Proposed method

• Utilizes input/output (I/O) tables from national statistical agencies as a source of inter-industry money flow data.
• Constructs directed, weighted networks where nodes represent industries and links represent monetary transactions between them.
• Applies statistical fitting to test whether flow weight distributions follow a Weibull or lognormal distribution.
• Employs community detection algorithms to identify clusters of industries with dense internal flows.
• Uses hierarchical clustering to reveal multi-level community structures, with the top level comprising five major industry communities.
• Compares model fits (Weibull vs. lognormal) using delta log-likelihood to determine best-fitting distributions.

Experimental results

Research questions

• RQ1 Do inter-industry money flow networks across different national economies exhibit a common structural topology?
• RQ2 What probability distribution best describes the distribution of monetary link weights between industries?
• RQ3 How do industry sizes and flow volumes distribute across national economies?
• RQ4 What is the hierarchical community structure of industrial networks, and how does it relate to economic sectors?
• RQ5 To what extent can a universal network model explain the macroeconomic structure of diverse national economies?

Key findings

• The distribution of inter-industry money flow weights across all 20 economies is best described by a Weibull distribution, with a pooled fit yielding a shape parameter of 0.456 and scale parameter of 1.08×10⁻⁴.
• Industry sizes across economies follow an exponential distribution, indicating a power-law-like concentration of economic activity in a few large industries.
• The network structure exhibits a hierarchical community structure with five primary communities: food, chemical, manufacturing, service, and extraction industries.
• The Weibull distribution provided a better fit than the lognormal in 15 out of 20 countries, with the pooled data strongly favoring the Weibull model (Δ log-likelihood = 545.6).
• Denmark, Finland, Germany, Hungary, Netherlands, Norway, Poland, Spain, and the United States showed better lognormal fits, but the Weibull remained dominant overall.
• The consistent five-community hierarchy across economies suggests a universal macroeconomic network architecture, supporting the use of statistical physics models to study economic systems.