A GDP-driven model for the binary and weighted structure of the International Trade Network

TL;DR

This paper introduces a GDP-driven model that unifies the prediction of both the topology and trade volumes of the International Trade Network, advancing understanding of economic shock propagation.

Contribution

It develops a maximum-entropy model linking GDP to both binary and weighted network properties, bridging the gap between gravity and fitness models.

Findings

The model accurately reproduces the ITN's topology and trade volumes.

GDP strongly correlates with model parameters, enabling predictive power.

Binary network properties can be inferred without weighted data.

Abstract

Recent events such as the global financial crisis have renewed the interest in the topic of economic networks. One of the main channels of shock propagation among countries is the International Trade Network (ITN). Two important models for the ITN structure, the classical gravity model of trade (more popular among economists) and the fitness model (more popular among networks scientists), are both limited to the characterization of only one representation of the ITN. The gravity model satisfactorily predicts the volume of trade between connected countries, but cannot reproduce the observed missing links (i.e. the topology). On the other hand, the fitness model can successfully replicate the topology of the ITN, but cannot predict the volumes. This paper tries to make an important step forward in the unification of those two frameworks, by proposing a new GDP-driven model which can simultaneously reproduce the binary and the weighted properties of the ITN. Specifically, we adopt a maximum-entropy approach where both the degree and the strength of each node is preserved. We then identify strong nonlinear relationships between the GDP and the parameters of the model. This ultimately results in a weighted generalization of the fitness model of trade, where the GDP plays the role of a `macroeconomic fitness' shaping the binary and the weighted structure of the ITN simultaneously. Our model mathematically highlights an important asymmetry in the role of binary and weighted network properties, namely the fact that binary properties can be inferred without the knowledge of weighted ones, while the opposite is not true.