英语论文网

βn+1 · (Host Country k’s Share of Total Sales in industry j in 1989)
+βn+2 · [Vertical FDI of Industry j ×Macro Risk in Host Country k] + εj,k
where β’s are coefficient estimates. I use dummy variables to control for country and industry
time-invariant unobserved characteristics. For this econometric specification, a negative and statistically
significant βn+2 indicates that FDI activities of industries with higher share of vertical
FDI respond more adversely to macro risks in a host country. The advantage of this setting is that
it makes predictions about within-country differences between industries based on the interaction
between country and industry characteristics. It should also be less subjected to omitted variable
bias and model misspecification.4
I use the 1989, 1994, and 1999 benchmark surveys on US multinational activities compiled
by the US Bureau of Economic Analysis. Table 2 provides a list of countries, together with
their share and level of vertical FDI. Countries in the top half of the table witness, on average,
https://www.51lunwen.org/australiathesis/more than 25 billion USD (in 1995$) of total sales generated by affiliates of US multinationals. In
contrast, the bottom half of Table 2 shares just a little below 10 percent of total sales.5 This feature
of the sample is useful for guiding the robustness check. Although there is a presumption that
horizontal FDI should dominate in large destinations, high-income, industrial countries, I find in
this sample set that the average share of vertical FDI is 19 percent for both large and small hosts
(excluding Barbados, Bermuda, Netherlands Antilles, Nigeria, and UAE as outliers).
3 If the production technology determining vertical FDI does not carry over to all foreign affiliates and changes frequently
over time, the choice of export versus FDI becomes an important consideration. For studies on export versus
FDI, see for example Brainard (1997), Grossman and Helpman (2004), and Helpman et al. (2004). On the insensitivity
of production technology to host-country conditions, see for example Morley and Smith (1977).
4 Rajan and Zingales (1998) use similar empirical approach to study whether industrial sectors that are relatively more
in need of external finance develop disproportionately faster in countries with more-developed financial markets.
5 Using the total sales numbers is admittedly a crude approximation. The immediate concern is transfer pricing.
Swenson (2001) finds an association between foreign corporate tax rates and the reported transaction prices between
parent firms and foreign affiliates. Because of data availability, it is difficult to measure cross industry-country differences
in transfer-pricing manipulation over time. I treat this distortion as a random component.Y. Jinjarak / Journal of Comparative Economics 35 (2007) 509–519 515
disregard close to a third of the observations. With such limitations in mind, I report in panels II
and III of Table 4 the regression results using the 25 billion USD cutoff. We can see that the
vertical-versus-horizontal differential impact of macro risks is not applicable to those smallershare
destinations. The merits of cross-industry vertical versus horizontal FDI in response to
macro-level risks apply only to a subset of medium and large host-country destinations.
To a degree, the results are in line with theory and country-leve