averages of the third-market variables instead.26 The third-market variables
are jointly significant according to the F-tests in Table 3. Yet, the exclusion of
third-market variables from our original specifications does not exert a qualita-
tive impact on the bilateral determinants of interest. This is not as surprising,
since third-market averages of the explanatory variables do not vary much in
the time dimension. Controlling for all time-invariant variables in our fixed
country-pair effects framework reduces their correlation with the bilateral ef-
fects substantially, so that a possible bias in the parameter estimates is avoided.
How do the results relate to those in previous research? Similar to the above
mentioned evidence regarding U.S. outbound FDI (see, e.g., Grubert and Mutti,
1991, 2000; Hines and Rice, 1994; Mutti and Grubert, 2004; Desai, Foley, and
Hines, 2005), parent (host) country tax rates exert a positive (negative) impact
on outward FDI in our large panel of countries. Our point estimates of 0.850
(for parent’s corporate tax rate) and -1.428 (for host’s corporate tax rate) for
the full sample in Table 2 translate into an elasticity of 0.3 and -0.5, respectively.
These values are at the lower bound of the range of results in previous studies
(see de Mooij and Ederveen, 2003). However, one should be careful with such a
direct comparison, since we control for withholding tax rates and depreciation
allowances beyond the statutory corporate tax rates. The present results clearly
indicate that it is decisive to consider these variables in addition to the statutory
tax rates when analyzing the impact of company taxation on FDI.
7 Conclusions
This paper analyzes the role of corporate taxation for outbound FDI. In doing
so, the paper pays particular attention to the fact that FDI flows among the
developed economies are not subject to country-specific but rather to country-
pair-specific taxation. This follows from the prevalence of tax treaties (or the
Parent-Subsidiary-Directive within the EU) among these countries, where devi-
ations from unilateral taxation principles are the rule rather than the exception.
One could think of bilateral effective tax rates as nonlinear aggregates of their
components: parent and host country statutory corporate tax rates, where the
26Note that market interdependencies are assumed to be inversely related to distance, since
complex multinationals engage in trade (see Blonigen, Davies, Waddell, and Naughton, 2004,
for an application). There, we apply the weights e-DISTkl to compute the respective averages.
DISTkl denotes the great circle distance between the economic centers of countries k and l.
In this way, our specification simultaneously accounts for interdependencies among parent
countries and among host economies (see Blonigen, Davies, Waddell, and Naughton, 2004,
2005, for a separate treatment of these interdependencies).
28