Generating a Time-Consistent Manufacturer ID (MID) in Census Import Data

Previous work has shown that the foreign exporter code in the U.S. import data (the Manufacturer ID, or MID) is not always reliable, leading to multiple MIDs referring to the same exporter. While this earlier work has proposed a string match algorithm to group similar MIDs together, this approach is extremely time-consuming due to the need to compare every MID to every other MID in the import data. This project proposes a methodology that constructs restricted sets of potential matches based on economic information. We construct three restricted sets of potential MID matches: i) using only MIDs that set similar prices, ii) using only MIDs that sell to the same buyer, and iii) using only MIDs that ship via the same U.S. port. We also examine the “brute force†approach of comparing all MIDs from the same exporter country, which turned out to be slow but feasible. We combine the matches resulting from the four approaches and generate a time-consistent, “grouped ID†which combines similar MIDs into one. Our new concordance can easily be merged onto the LFTTD and allows researchers to work with this grouped ID going forward. The number of grouped IDs we obtain matches relatively well with external counts of exporters from the World Bank Exporter Dynamics Database.