where αj∙ is a regional dummy variable that takes on the value one for countries belong-
ing to region j and zero otherwise (j = 1,...,J where J is the number of regions) and
which adjusts for time invariant regional effects. In this case, xijt is a vector of explana-
tory variables that possibly includes interactions between regional dummies and selected
explanatory variables.
The panel studies with regional dummies (reviewed in Section 3.1) explain regional
differences in FDI inflows by time-invariant regional effects. If one believes that FDI flows
are ultimately driven by arbitrage that leads to the equalisation of marginal productivity
of production factors, see Selaya and Sunesen (2008), then this approach argues that the
uneven distribution of FDI is due to some regional effect that allows the productivity of
production factors in one region to differ systematically from other regions. We could
think of this as "historic agglomeration effects" that have given the region a reputation
or as permanent differences in production factors. If such time-invariant regional effects
turn out to be important, the implication is that a country that is lagging behind today
will stay behind irrespective of its ability to implement policies aimed at strengthening
the institutions that are positively associated with FDI (included in Xijt).
The panel studies with heterogeneous effects (reviewed in Section 3.2) use interactions
between regional dummies and selected explanatory variables to allow for heterogene-
ity in the response to FDI determinants. One reason for such structural differences is
that investors are attracted to different countries according to their motive for investing
abroad.3 If the composition of FDI in this way varies systematically across regions, it
is likely that the flow of FDI to these regions will respond differently to traditional FDI
determinants. Empirically, this means that the vector of explanatory variables should
include interactions between the regional dummy variable and the explanatory variables
thereby allowing parameter estimates to vary across regions.
The second group of studies (reviewed in Section 3.3) bases the analysis on a panel of
countries that belong to the same region and estimates (1) for the region under review.
This estimation method therefore allows for full heterogeneity in both a^ and βi between
regions. One reason for using this approach is that some studies aim at answering ques-
tions, which require the use of region-specific variables that might not be relevant or might
not even exist for other regions. Examining the impact of transition on FDI inflows to
Eastern European countries could be one example.
In the next section we review the large number of empirical studies that have modelled
the regional aspect of FDI explicitly. We do so in order to detect empirical regularities
in the driving forces of FDI that can inform us about the degree of heterogeneity across
regions. Ultimately, this should lead to a greater understanding of what causes regional
differences in the distribution of FDI.
3The literature typically distinguishes between market-seeking, resource-seeking, efficiency-seeking and
asset-seeking FDI.