Harmonizing the Harmonized System

International trade research relies heavily on data reported at the product level. Regular classification updates of the “Harmonized System” lead to intertemporal inconsistencies affecting up to 44% of world goods trade. Existing methods to standardize product vintages either drop numerous products over time or bulk updated codes into large synthetic categories. We largely overcome these issues by developing an algorithm that exploits the persistence of trade data to convert trade flows between vintages. Our conversion estimates are robust to year and sample choices. Provided a clear correspondence and persistent data, the algorithm can be applied to other classifications