• Modularity Optimization
• International Trade in Food Commodities
• Economic Networks
• Determinants of Food Trade Blocs
• Community Structure of Complex Networks
• Multilevel Local Search Algorithm
• Probit Model

This study aims at identifying the community structure of the international food trade multi-network and at shedding some light on the factors driving its formation.
Adopting a data-driven methodology of community detection inspired to the Louvain algorithm (Blondel et al., 2008), we will identify relevant community structure of the most globally traded food commodities.
Furthermore, we will highlight the factors underpinning the structure of the trade blocs detected, exploring to what extent variables traditionally employed in the empirical trade literature, and in particular in gravity models, succeed in providing evidence of the co-presence of any possible country pair into the trade communities identified.
Results reveal that the commodity-specific networks considered are characterized by a robust community structure with strong modularity levels and, thus, that countries tend to organize themself into densely connected trading groups, which seem to reflect relevant geo-political and economic patterns. Furthermore our estimation results show that geographical proximity as well as trade agreements’ membership increase the probability for country pairs to belong to the same trade community much more than their economic size and/or income.
In the final section, we will discuss the limitations of the present work, underlining a number of possible directions for improvement and further research.