A microscopic study of the fitness-dependent topology of the world trade network

• We study the hidden variable correlation matrix of the world trade network.
• These correlations resemble the previously found disassortativity of world trade.
• A new fitness definition that leads to more robust results is proposed.
• The impact of a changed data source on first order results is illustrated.

Previous studies have suggested that the world-trade network belongs to the class of static hidden variable models. In this article we investigate the microscopic structure of the world trade network, that is the hidden variable correlation matrix of the network. The hidden variable is defined as a rank ordering of gross domestic products. This choice significantly reduces the noise in the statistical analysis found in previous studies. The hidden variable correlation matrix, that expresses the probability that a trade relationship between two countries of given fitness exists, suggests an attachment kernel that at least partially favours trading pairs or dissimilar fitness rather than the purely multiplicative one found previously. Additionally, we provide an in-depth look at the data source and reveal that first-order results, such as the degree distribution, exhibit significant qualitative differences depending on the data provider. Furthermore, we shed light on the intertemporal activity of international trade and point out that fluctuations occur mostly between countries with strong dissimilarities of fitness and connectivity.