About the possibility of recovering a trade network from Bill of Lading data

Abstract

Importers and exporters play a central role in international trade and therefore are important economic agents. Customs data is an interesting source of company transactions and relationships, that may help to understand economic behaviour at the company, country or product level. We compare several approaches to clean a dataset of US Bill of Lading including a set of techniques from record linkage, text classification and network measures in order to improve the number of exporting companies with a country and product category which can be identified in this dataset. These experiments depicts our initial attempts to reconstruct a network of import-export companies from transactional data.