The latent structure of diversification from relatedness and network embeddings
Abstract
Graph embeddings complement relatedness measures in models of economic diversification. Analyses of international trade and public procurement show that their predictive contribution varies by scale and target: relatedness is especially informative about entry, while embeddings add information about specialization magnitude.
Type
Publication
PLOS ONE, 21(10), e0344958
Figure: Entity clusters in the embedding space, comparing firms and countries. Sturm, Candia, Damásio and Pinheiro (2026), Figure 2. CC BY 4.0. Original figure.
Cristian Candia
Associate Professor, Data Science Institute, School of Engineering, Universidad del Desarrollo, Chile. Director of CRiSS-LAB.
Cristian Candia is a computational social scientist studying human and collective behavior with large-scale data, network science, experiments, and AI.
