Financial centers and the geography of capital flows

We examine an assumption common in empirical work on bilateral portfolio capital flows that the countries the flows are attributed to are also the countries of the security's issuer, seller, or ultimate buyer. We do this by estimating U.S. investors' holdings of debt and equities in over 40 countries and, for the same countries, foreign investors' holdings of U.S. debt and equities. A comparison of our estimates with data from benchmark surveys provides insight into U.S. data on international debt and equity transactions. We find that, contrary to the common assumption, the data do not track the location of U.S. investment or the location of investors in U.S. assets very well. Because the U.S. portfolio flow data collection system was designed to measure cross-border transactions with foreign counterparties who are often intermediaries, the majority of the flows are attributed to financial centers. By aggregating our country-level estimates, we find that U.S. data accurately portray net inflows into U.S. equities and net outflows into foreign bonds. However, the data substantially overcount net inflows into U.S. bonds and may undercount net outflows into foreign equities. We conclude with a discussion of the implications of our findings for research on capital flows.