Revealing persistent structure of international trade by nonnegative matrix factorization

Abstract

The international trade network is analyzed to identify latent relations between nations and traded commodities. Nonnegative matrix factorization (NMF) is used to decompose the international trade network from 2000 to 2010 into groups of nations and commodities. Groups in consecutive years are compared, and it is shown that there are persistent characteristics within these groups. We identify economically interpretable results, such as intra-regional trade, accounting for the largest component and the recent growth of Chinese exports of commodities.