Interpreting Economic Complexity

TL;DR

This paper reveals that the Economic Complexity Index and Product Complexity Index are spectral clustering measures that explain economic development patterns and regional specialization through network analysis.

Contribution

It demonstrates the equivalence of ECI and PCI to spectral clustering and links these measures to dimensionality reduction, providing a new interpretation of their empirical success.

Findings

ECI and PCI are equivalent to spectral clustering algorithms.

High ECI countries tend to specialize in high PCI products.

ECI and PCI reveal meaningful regional economic patterns.

Abstract

Two network measures known as the Economic Complexity Index (ECI) and Product Complexity Index (PCI) have provided important insights into patterns of economic development. We show that the ECI and PCI are equivalent to a spectral clustering algorithm that partitions a similarity graph into two parts. The measures are also related to various dimensionality reduction methods and can be interpreted as vectors that determine distances between nodes based on their similarity. Our results shed a new light on the ECI's empirical success in explaining cross-country differences in GDP/capita and economic growth, which is often linked to the diversity of country export baskets. In fact, countries with high (low) ECI tend to specialize in high (low) PCI products. We also find that the ECI and PCI uncover economically informative specialization patterns across US states and UK regions.