FDI and domestic investment in Germany: crowding in or out?

Abstract

This paper estimates the effects of outward FDI on domestic business investment in Germany at the industry level for a panel of 19 industry and 10 services sectors. We pay particular attention to the different motivations behind FDI, and distinguish between FDI to high-versus low-wage countries, to Europe versus the rest of the world, and FDI in services and industry sectors.We find that, in industry, FDI to low-wage countries crowds out domestic investment, whereas FDI to high-wage countries outside Europe crowds in domestic investment. In services, FDI to Western Europe crowds in domestic investment.

Keywords:

• domestic capital formation
• foreign direct investment
• cost seeking
• market seeking
• Germany

Acknowledgements

The authors are grateful to Dorothea Stiebler and Dietmar Scholz at the Deutsche Bundesbank for data support, and to Markus Leibrecht, Christian Bellak, and two anonymous referees for helpful comments. Support from the FWF Project Nr. F2012 is acknowledged. The usual disclaimer applies.

Notes

1. Western Europe (Belgium, Denmark, France, Ireland, Italy, Luxembourg, Netherlands, Austria, Portugal, Sweden, Finland, Spain, United Kingdom, Greece, Norway, Switzerland, Iceland), 2. Emerging East (Poland, Romania, Bulgaria, Slovakia, Czech Republic, Hungary, Turkey, Estonia, Latvia, Lithuania, Albania, Macedonia, Bosnia-Herzegovina, Serbia, Croatia, Slovenia, Ukraine, Moldova, Belarus, Russia), 3. High Wage Rest (USA, Canada, Australia, New Zealand, Japan), 4. Low Wage Rest (Africa, America excluding USA and Canada, Asia excluding Japan, Australia and Oceania excluding Australia & New Zealand)

2. Bundesbank kindly compiled and provided the aggregated data by sectors and country groups; however only for a limited period of 1998–2005. Obviously, the issue of rising FDI as well as stagnating domestic investment has been simultaneous developments, which date back well before the 2000s; however, due to data availability, our period of analysis will cover only this later phase of FDI.

3. Data for the sectors of ‘wearing apparel, dressing and dyeing of fur (18)’ and ‘electricity, gas and water supply (40)’ are not available at the level of our detailed destination country classification due to confidentiality reasons because of the limited number of firms engaged in outward FDI in each country group in these sectors. Furthermore, in the sectors for Financial Intermediation and Insurance and Pension Funding, there are very high year-to-year changes of value added (above 50%), which might be due to changes in classification of several firms in different years. Therefore, we used one-digit level data for the aggregate sector ‘financial intermediation, insurance, pension funding, and auxiliary activities’ (65–67).

4. The figures correspond to the sectors for which the econometric estimations are done.

5. The estimator has three steps: In a first step a fixed-effects estimation with the autocorrelation parameter, rho, is estimated. The autocorrelation parameter is obtained by a Prais-Winsten estimation, which has the advantage over Cochrane-Orchutt, that the first observation does not become lost (Wooldridge 2006, 426). Then the data is transformed in order to remove the AR(1) component based on a method by Baltagi and Wu (1999). In the third step a fixed effects estimator on the transformed data is run. We lose one more observation in this stage of the estimation procedure. Nevertheless this estimator is very efficient with short time series and comparably larger number of cross-sections compared with other estimation procedures that reduce autocorrelation.

6. Although FDI data are only available until 2005, since the data for investment and value added are available for the later years, our estimation period can be extended until 2006 given the lag structure. We lose one observation due to differencing, two observations due to the lags and one observation due to the three step procedure to deal with autocorrelation as discussed above; therefore the estimation period starts in 2002.

7. d ^IND has an entry 1 if a sector belongs to industry and zero otherwise. d ^SRV has an entry 1 if a sector belongs to services and zero otherwise.

8. Since Y is used on the right-hand side both in first difference and in the denominator of FDI/Y, the effect of a change in Y ceteris paribus, i.e. for a given FDI, has to be calculated by summing up the coefficient of Y as well as the indirect effect through the decline in FDI/Y for a given FDI.

9. The results are available upon request.