CPB Hybrideseminar – Inferring product-level input-output linkages from firm-level trade data

We develop an analytical approach to infer product-level input-output relationships from firm-level trade data. By adapting an Association Rule Mining (ARM) algorithm, we identify systematic co-occurrences between the products that firms typically import and export, which we interpret as potential input-output linkages. We refine this mapping and demonstrate that our inferred linkages broadly resemble the relationships reported in conventional input-output tables and that they perform well in several external validation exercises. Subsequently, we apply the product-level input-output information obtained from Chinese data to study how trade policy shocks propagate along the supply-chain. Doing so, we find that China’s WTO entry in 2001 entailed a significant increase of imported inputs to produce its exports. The second application shows that the US-China trade war in 2018-19 entailed a significant reduction in Chinese imports of goods with high downstream US tariff exposure. Our findings suggest that relatively simple pattern-recognition methods can be applied to infer input-output linkages that enable more granular measurement and evaluations of trade shocks and their propagation along the supply-chain.