currently I am trying to link patent data to company data from Bureau van Dijk's Amadeus database. Based on the matching result provided by Peruzzi et al. (2014)*, which contains a linking table from the person_id in Patstat to the corresponding identifiers in the Amadeus database, I tried to combine the patent data from Patstat with the corporate data from Amadeus. However, when merging the Amadeus data with Patstat data from the TLS206 table, which also contains the person_name, I realized that none of the corporate names provided in Amadeus was even close to be similar to those provided in the variable person_name in Patstat. Just as a plausible check, I even looked up some of the linked firms in the internet, in order to check whether they would be somehow related - however, I could not find any connection at all, which lead me to conclude that my match is not only erroneous but entirely failed.
Peruzzi et al. based their matching on Patstat data from version of October 2011. In the Patstat Data Catalog from Autumn 2016 I found out that the person_id was stable from 2013 onwards. Therefore, I was wondering whether the Patstat version from 2011 contained different person_ids which might explain the above described, bad results by applying the matching table provided by Peruzzi et al. to the person_ids from the Patstat version of 2016.
Also, did anyone experience a similar problem or knows about a common issue that could be applicable in this case too?
Thank you very much in advance for your time and help!!
Best Regards,
Jan
* Peruzzi, Zachmann, Veugelers (2014) "Remerge: Regression-based record linkage with an application to Patstat"