Abstract:
The paper presents a new methodology, based on tensor decomposition, to map dynamic trade networks and to assess their strength on spreading economic fluctuations at different periods of time in Asia. Using the monthly merchandise import and export data across 33 Asian economies, together with the US, EU and UK, we detect the modularity structure of the evolving network and we identify communities and central nodes inside each of them. Our findings show that data are well represented by two communities, in which the People's Republic of China and Japan play the major role. We then analyze the synchronisation between GDP growth and trade, and apply our model to the prediction of economic fluctuations. Our findings show that the model leads to an increase in predictive accuracy, as higher order interactions between countries are taken into account.