Table 5 presents results when we estimate our empirical specification for US MNE activity in European OECD countries by individual economic sectors categorized by the BEA. Publicly available data on US FDI activity has limitations in how disaggregated such data can be reported and these categories represent the finest level of disaggregation for which we can get public data by country and industry on an annual basis. Our time period, sample countries, and control variables match those for the European OECD regression above. Given the importance suggested above, we also include country dummies in all specifications. The number of observations for each of the sector estimations varies due to missing data when the BEA suppresses the data point out of confidentiality concerns. As one may expect, there is substantial heterogeneity in estimates across sectors though these differences are mainly in the magnitude of estimated parameters, not their sign. Importantly, we find stronger evidence of export-platform activity in the European-OECD sample when adopting these more disaggregated sector-level data. Five of the eleven sectors show a sign pattern that is consistent with export-platform motivations for FDI—a positive coefficient on the market potential variable and a negative spatial lag. For two of these, “food and kindred products” (Column 2) and “primary and fabricated Metals“ (Column 4), both of these spatial variables are also statistically significant. The coefficient on market potential is non-negative in nine of the 11 sectors and significantly positive 4 times. The spatial lag is likewise non-positive 10 out of eleven times and significantly negative three times. The one exception to this is “Petroleum,” (Column 1). However, given that a great deal of this industry is likely tied to the geographic proximity of oil fields, such may not be surprising.
In summary, by disaggregating the data (as much as public data allow) and focusing on a fairly homogeneous group of countries distributed evenly across space, we get stronger evidence for an FDI motivation that we would expect in the European sub-sample ——export-platform FDI. This highlights how important sample selection is in estimating empirical FDI models, particularly those with spatial terms, if one wants to be able to relate such results back to FDI theory.