Asymmetric Relatedness from Partial Correlation

Area of Science:

• Economics
• Network Science
• Data Science

Background:

• Economic complexity relies on sector similarity for development strategies.
• Existing relatedness measures lack explicit consideration of export time correlation structures.

Purpose of the Study:

• To introduce a novel asymmetric relatedness measure incorporating temporal export correlations.
• To apply this measure to a comprehensive database of goods and services exports.
• To analyze the resulting economic activity network structure and properties.

Main Methods:

• Developed an asymmetric relatedness definition using statistically significant partial correlations.
• Generalized a correlation-filtering algorithm (partial correlation planar graph) for bipartite temporal networks.
• Employed bootstrapping for statistical confidence assessment of network edges.

Main Results:

• Constructed a network of economic activities where links signify temporal correlation influence.
• Observed the formation of intuitively related clusters of economic sectors.
• Found strong assortative mixing of sectors based on their economic complexity.
• Identified that hub nodes exhibit more robust connections than peripheral nodes.

Conclusions:

• The new asymmetric relatedness measure effectively captures temporal dynamics in economic activities.
• The network analysis reveals meaningful structures in economic complexity, including clustering and assortativity.
• The findings support the use of time-correlated relatedness for informed economic development strategies.