Repairing a Historical Mistake in Bilateral <scp>FDI</scp> Statistics: A New Dataset Covering 2001–2022

ABSTRACT The statistics on foreign direct investment (FDI) used to be important quantifiers of ownership‐based control over foreign companies. However, during the 2008 financial crisis, IMF and OECD changed the FDI definition to obtain more information on intra‐company finance activity. They did so by giving cross‐border loans between affiliated subsidiaries (within the same parent firm) the same status as acquiring ownership of foreign firms. The change became effective in 2013. The paper shows that it resulted in a systematic drop of quality and consistency of FDI statistics, while causing massive double counting. It made comparison between pre‐ and post‐2013 FDI statistics impossible. We propose a methodology to repair this historical mistake by emulating the pre‐2013 FDI definition, which was based primarily on FDI assets. The paper provides a full proof‐of‐concept with a dataset holding bilateral FDI between 232 jurisdictions over the period 2001–2022. The dataset is strictly based on reported data; it uses no estimation or imputation. The new dataset is evaluated quantitatively by a comparison with the original source data. The paper also quantifies the dimensions and the country structure of ‘phantom FDI’ that resulted from current double‐counting practices.