<p>This paper uses Input-Output data to search for trading communities in the world trade network both in final and intermediate goods and it then uses a structural gravity model to conduct counterfactual analyses to tease out the main drivers behind such communities. The main findings are twofold (i) global trade is divided into communities broadly corresponding to regional (continental) areas which are driven entirely by bilateral characteristics such as geography, trade policy and cultural similarities; (ii) the trade network is significantly less modular than the corresponding random networks and this is driven by individual characteristics such as productivity, comparative advantage or size.</p>