Abstract
Fifty six bilateral country relationships combining 7 home countries from the EU and the US, and 8 Central and East European host countries (CEECs) of foreign direct investment (FDI) from 1995-2003 are used in a panel gravity-model setting to estimate the role of taxation as a determinant of FDI. While gravity variables explain most of the variation of FDI inflows, the bilateral effective average tax rate (beatr) is roughly equally important to other cost-related factors. The semi-elasticity of FDI with respect to taxes is about -4.3. This value is above those of earlier studies in absolute terms and can partly be attributed to using the beatr instead of the statutory tax rate. Our results indicate that tax-lowering strategies of CEEC governments seem to have an important impact on foreign firms location decisions.
Acknowledgement
We are grateful to the Austrian Science Fund for providing financial support by grant F 2008.
Notes
1The home countries are Austria, France, Germany, Italy, The Netherlands, The United Kingdom and The United States. The host countries are Bulgaria, Croatia, the Czech Republic, Hungary, Poland, Romania, Slovakia and Slovenia (CEEC-8).
2Having described the general relationship between FDI and lower taxes, the story could, however, be different on a more disaggregated level. We are grateful to the referee for raising this problem. First, footloose manufacturing sectors may react differently to service sectors. Second, different tax incentives etc. may apply on the sectoral level. Recently this issue was analysed empirically by Stöwhase (Citation2005). He shows that the tax sensitivity of FDI depends crucially on the sector receiving the capital flow. This raises the possibility of an over- or underestimation of sectoral tax elasticities on the aggregate level. Given the difficulties of obtaining appropriate data at a reasonable scale at present, we have not followed this strategy. As more data will become available in the future, sectoral analyses will show the size of the bias incurred.
3Rather, in many cases it is the other way round, i.e. a listed acquired firm is de-listed from the stock exchange after the acquisition by a foreign MNE, in order to gain 100% ownership.
4Should the coefficient carry a positive sign, this could be an indication of an omitted variable problem, as in this case labour costs may capture effects of an increasing level of skill in the host country.
5We define unsystematic outliers as data points which do not represent heterogeneity between the host countries. For example, using box plots the ulc for Slovenia are shown to be extreme values throughout the sample period. Hence, these data represent heterogeneity between the host countries which we exploit in our analysis.
6These variables may be called ‘gravity specific’.
7We perform two types of Hausman-tests. First, if no serial correlation and heteroskedasticity seem to be present and if the other requirements of the original Hausman-test are fulfilled (e.g. the difference between fixed effects and random effects variance matrices is invertible) we use the original Hausman-test. Second, in case of non-spherical errors or a nonpositive definite difference in the fixed effects and random effects variance matrices we perform a regression based Hausman-test with cluster robust SE (Wooldridge, Citation2002).
8We also used a specific to general approach to asses the robustness of the tax-rate elasticity with respect to single location factors additionally included in the empirical model. The results show the robustness of our estimate in this respect but are not reported here. For details see Bellak and Leibrecht (Citation2005).
9Dropping pp does not change the results of our analysis. Details can be found in Bellak and Leibrecht (Citation2005).