Evidence on the Determinants of Foreign Direct Investment: The Case of Three European Regions

Table 8: LOMBARDIA

Dependent variable: FDIP

Method of estimation: POLS (with White correction)

Values in brackets: Standard Error

***Level of significance 1 %, ** 5%. *10%

Tables 6 to 8 summarize the empirical results obtained for each region: Catalunya (Table 6),
Baden-Württemberg (Table 7) and Lombardia (Table 8). We selected eight specifications. In the first
four, we regress FDIP (Foreign Direct Investment per capita by region and by sector) on a few
macroeconomic variables including GDP per capita and regional productivity measures at sector level.
The next three specifications focus on determinants regarded as representative of the level of
technology and innovation. The eighth specification includes macroeconomic determinants and human
capital. ⁶

Looking at the results across regions, two comments can be made. First, there is a determinant
that is statistically significant in all specifications and in all the three regions. This determinant is GDP
per capita by region and by sector. This confirms the findings of the literature on FDI at national level.
Our regressions show that the positive relationship seems to be robust even at the regional and sector
level. This result can also be interpreted as an evidence of the horizontal nature of FDI in these three
regions, where market size predominates as a determinant.

Second, the dummies by sector are also always significant. In each specification, we introduce
a dummy associated with the distribution by sector of the FDI inflows. Raw data points out that the
distribution of FDI across sectors is not the same across our regions. A region may display either a
distribution strongly orientated toward a specific sector or a changing distribution over time. The
Dummy DUM reflects the high share of manufacturing in the FDI distribution in Lombardia, while the
dummy DAVERAGE gives higher weight to the sectors that receive a high proportion of FDI (namely

⁶ We ran also regressions (available upon request) including one-period lag of the explanatory variables and the results we
obtain are similar. For Baden-Württemberg and Catalunya, POLS estimations (available upon request) replicate the same
results as LSDV estimations but with a lower R-square.

18

<<   <   15   16   17   18   19   20   21   >   >>