Download FDI in the post-EU accession Baltic Sea Region: A global or a

FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?
89
FDI in the post-EU accession Baltic Sea Region:
A global or a regional concern?
H. Richard Nakamura, Mikael Olsson and Mikael Lönnborg1
Abstract
This paper investigates the dynamics of FDIs in the Baltic Sea Region (BSR) by applying
the Poisson Pseudo-Maximum Likelihood estimation method on a gravity model. In particular, we analyze the influence of macro and spatial factors on investment stock changes and
discuss whether the origin of these investments and the 2004 EU enlargement have had any
effects on BSR FDIs.
Our results suggest that EU enlargement has been significant for FDI activity in the region,
and that FDI is basically a regional issue as it tends to be bilateral within the region. However,
the same results also suggest that geographic distance is not a significant factor. We conclude
that while being traditional in nature, the BSR FDI pattern is undergoing changes towards a
lesser degree of geographic bias.
Keywords: Baltic Sea Region, FDI, Trade, Gravity Model, Poisson Pseudo-Maximum Likelihood method.
JEL classification: C21, C23, F21, F23, R12
1. Introduction
Since the EU accession of Estonia, Latvia, Lithuania and Poland, political rhetoric has envisioned a bright future for the overall development and integration of the Baltic Sea Region
(BSR). However, the integration process had been under way long before the Baltic states
and Poland became full EU members, and the integration of these countries into the European
common market has recently been impeded by the global recession, where the Latvian experience embodies the vulnerability of small open economies in times of international financial
crises.
Earlier research (e.g., Petri, 1994; UNCTAD, 1996) has shown that trade and foreign direct
investment (FDI) intensity are closely associated. However, on the global level, investment
distribution is uneven and the BSR is no exception. As seen in Figure 1, investment flows
in the BSR are up until 2008 primarily directed from the top industrialized countries on the
western shores of the Baltic Sea to the eastern BSR countries, which regionally have cost advantages in all or some input resources for industrial or service production, or being emerging
consumer markets close to larger EU countries with a population gaining increasing purchasing power.
1
Centre for Baltic and East European Studies (CBEES), Södertörn University, Södertörn University,
SE-141 89 Huddinge, Sweden. Tel: +46-8-608 5008, Fax: +46-8-608 4170. Main author, H. Richard Nakamura,
e-mail: [email protected]. The present study and article was prepared, analyzed and written by the main author.
Baltic Journal of Economics 12(2) (2012) 89-108
90
Figure 1. Annual inward FDI in the BSR between 1992-2008; % of total world net inflows.
20
18
16
14
BSR-W
12
BSR-E
10
17,9
6,9
5,1
1,1
2,1
1,5
1,7
1,5
2007
0,9
2006
0,8
2005
0,8
2003
0,8
2004
1,2
2002
1,2
2001
1,3
1998
0,4
1997
1995
7,3
7,6
5,7
1996
1994
1993
1992
0,8
0,9 1,2
0 0,5 0,9
12,7
11,9
5,8
3,6
3,1
2
9,1
2000
07,8
4
9,3
1999
6
4,4
1,3
2008
8
Note: BSR-W: Germany, Denmark, Sweden and Finland. BSR-E: Estonia, Latvia, Lithuania and Poland. Source:
UNCTAD and own calculations.
The statistics between 1992 and 2008, as summarized in Figure 1, indicate a development
from an initial boom to a more modest flow of inward investment, suggesting that the Baltic
states are losing their comparative advantage in production factor costs. However, this is
not the only explanation. Concurring with increased loss of production cost advantages, the
purchasing power of host country consumers has increased substantially, which means that
the investment motives of the multinational corporations (MNCs) in the Baltic states changed
over that period. This in turn might lead to market-oriented FDIs, which, by requiring less
investment funds in absolute terms, affects total net FDI inflows downwards. At the very end
of the period under analysis in this paper, the global financial crisis occurred, which also affected international FDI flows negatively.
Figure 2. Cumulative inward FDI per capita between 1992-2008 (USD in current prices).
35 000
32 058
30 000
25 000
20 000
18 138 17 811
11 632 10 689
4 257
3 573 3 297
Lithuania
Germany
Estonia
BSR-8 /avg./
Finland
BSR-W /avg./
Denmark
Sweden
0
5 454
Poland
7 342
5 000
Latvia
10 000
BSR-E /avg./
13 707
15 000
Note: The averages reported for the region as such (BSR-8) and the two constituent parts (BSR-W and BSR-E) are
unweighted averages. Source: UNCTAD and own calculations.
FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?
91
Decomposition of FDI statistics into country level data will reveal that the per capita value of
foreign investment in the BSR area (in cumulative terms) is spread unevenly between mature
economies and former transition economies. From Figure 2, it is possible to conclude that the
largest value of investment stocks tends to be found in the established mature economies of
the western BSR (Germany scoring relatively less in the Figure 2 ranking since the population size is substantially larger than the rest of the BSR countries). Another illustration of
the size of FDI in the region is our compilation of all cross-border mergers and acquisitions
(M&As) over USD 1 million in the BSR between 1998 and 2010 (See Appendix Tables A1
to A6), which gives an impression made by a few “pure” intra-BSR cross-border M&As (for
example, 35% of all cross-border M&As taking place in the BSR between 1998-2010 were
between BSR firms [source: Zephyr M&A database]) to the list of major international M&A
deals in the region. This diverse picture is indicative of the unclear knowledge we have on
contemporary FDI activity in the Baltic Sea Region.
The general purpose of this paper is to investigate and analyze the dynamics of FDIs in the
BSR. In particular, this paper verifies the origin and target country directions of FDIs in
the region. We will do so by an econometric analysis of the influence of gravity factors on
contemporary investment flows, proxied by the inward FDI stocks of the 8 BSR countries of
the region, and will discuss whether the geographic origin of these investments and the EU
accession of the Eastern Baltic countries are of any importance when discussing regional
integration and investment promotion in the BSR. Since EU accession is a central event for
this analysis, we limit our study to the period between 2000 and 2008, that is, four years prior
to and four years after the accession year of Estonia, Latvia, Lithuania and Poland. Thus, the
data obtained cover the immediate pre-accession period to the start of the global financial
crisis of 2008-2009.
2. Theoretical framework and earlier research
Transaction costs in international trade have been a factor traditionally dealt with more or less
pragmatically in order to lift them off the basic models of international trade (e.g., Dixit and
Norman, 1980; Jovanović, 2006; Eicher et al., 2009). This has led to trade models with strong
assumptions, where the purpose has been to theoretically argue for the basic benefits of international trade rather than to let the models describe a factual reality. If transaction costs and
other institutionally related costs are introduced into the theoretical models, the geographic
bias observed in trade statistics can also be better explained theoretically, provided the assumption holds that the longer the geographical distance, the more expensive the transport
costs. If we presume the mainstream argument that trade and investment follow each other
closely (see, for example, Petri, 1994; UNCTAD, 1996), a regional bias similar to those found
in trade patterns should also be expected for investment flows.
Earlier studies (e.g., Petri, 1994; see also Blomström and Kokko, 2003) suggest that trade
flows are more volatile than FDI, which is less surprising since trade is more sensitive to
short-term institutional changes in the world trade environment such as exchange rates, war,
trade conflicts and trade partners’ domestic political factors, as compared to FDI, which tends
to be a more medium- or long-term commitment. Then what are the determinants of FDI?
This question can be considered a classic area in the FDI literature, and in their early contributions, Penrose (1959) and Hymer (1960 [1976]) argue that opportunity-seeking behavior
predominates internationalizing firms’ investment decisions, forwarding the idea that firms
92
Baltic Journal of Economics 12(2) (2012) 89-108
make their first international expansion when the home market has become mature and the
competition has driven profits toward zero – in other words, a situation well accommodated
within the framework of neo-classic theory. More recent literature on FDI determinants (e.g.,
Dunning, 1992; Shapiro and Globerman, 2001) focuses on features connected to host country
market attraction and production cost advantages, such as the size of the market, strong purchasing power, education level, infrastructure standards, trade policies, exchange rates and
political and macroeconomic stability.
In general, increased global competition, liberalization of FDI regimes, technological and
logistic advances have changed the conditions for firms of any origin in doing FDI (Sauvant
et al., 2009). As a natural consequence of this development, the outward investment behavior of emerging market MNCs has increased dramatically, possibly making the validity of
the mainstream assumptions of FDI determinants questionable. For example, Buckley et al.
(2007) argue that capital market imperfections, diversification and strategic asset-seeking in
institutional environments resembling home country conditions are strong influential factors
for Chinese firms’ foreign investment decisions. Similar outward FDI determinants have also
been found for Indian firms (Pradhan, 2004).
Even if we are discussing primarily macro aspects of FDIs in this paper, some words on motives influencing FDI decisions are necessary for our theoretical discussion. Dunning (1992)
has made important contributions to this literature, and has spelt out four different motives
for outward FDIs: (Natural) Resource-seeking, Market-seeking, Efficiency-seeking and Strategic Asset-seeking. The first FDI group aims to acquire all forms of input – physical as well
as intangible – needed for production. The second FDI group aims for sale and supply of
goods and services to other markets. This type of FDI motive is typically triggered as a result
of strategic decisions by investing firms to deepen the commitment in the markets they have
served earlier primarily by exports (see, for example, Johansson and Vahlne, 1977), by, e.g.,
starting host country production. The third type of FDI motive, that is, Efficiency-seeking, is
denoted by Dunning (1992) as a desire among established firms to utilize economies of scale
and scope to, for example, concentrate production in locations associated with internationally
competitive production costs, factor endowments, institutional arrangements, and the like.
Finally, the fourth FDI motive aims to sustain the investing firm’s long-term strategic objectives of being globally competitive (see Barney’s [1991] discussion of “sustained competitive
advantage” of firms), through measures such as acquisition of brands and technology (including intellectual property rights).
If we assume the beneficial effects from trade firstly in terms of improved terms-of-trade
and secondly in terms of improvement of real exchange rates which increases the purchasing
power of the importing country, the benefits to the importing country as a market-seeking
FDI destination are obvious. The situation for efficiency-seeking FDI might be the opposite.
All other things being equal, the initial advantages of locating production in countries with
low production costs can over time, as the overall wealth of the host country increases and
spillover effects materialize, switch to a situation where production turns disadvantageous
(as it has done in, e.g., Estonia). On the other hand, if upgrading of host country production
factors occurs, a production location in the host country might still be advantageous from
the efficiency viewpoint, due to a shift to skill-intensive production (as observed by, e.g.,
Görg and Strobl, 2003). Furthermore (and closely connected to market-seeking type FDI),
FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?
93
the lifespan of efficiency-seeking FDI is also affected by which end markets the products
produced are sold in, i.e., whether these products are sold domestically in the host country or
exported (or both). Thus, the sustainability of efficiency-seeking FDI is heavily dependent
on how the foreign investor manages reinvestment in existing FDI stocks in the host country.
For resource-seeking FDI, the relationship might be similar to efficiency-seeking FDI that
aims for exports, as the investor home country typically imports resources extracted in the
FDI host country and in this way increases trade between host and home countries. Finally,
strategic asset-seeking FDI implies a home country (long-term) claim on the host country,
thus contributing to an improved balance-of-payment position of the importing home country
vs. the exporting host country.
The theoretical literature exemplified by, e.g., Lipsey’s (2006) discussion, argues that facilitated access to a country’s main markets and resulting increase in trade flows should lead to
increased growth and wealth, which in turn makes the country an attractive host for FDIs.
This discussion should also be contrasted to the discussion by, e.g., Hummels (1999), who
argues that geography is still an important factor for today’s trade and investment. Taken
together, there are strong theoretical reasons to believe that economic integration of neighboring regions might foster increased FDI activity.
As stated above, a gravity model is employed in this paper for the empirical analysis. Originally suggested in the 1960s by, e.g., Tinbergen (1962), the gravity model has since been used
in International Trade studies for analyses of influences of various impeding factors on international trade flows (e.g., Bergstrand, 1985; Broadman, 2005; Flam and Nordström, 2010;
Paas, 2000; Paas and Tafenau, 2005; Santos Silva and Tenreyro, 2003). Inspired by these
developments in the trade analysis field, a number of researchers have also proposed and
attempted to transfer the same analysis method to studies on FDI flows and stocks. The geographic focus of these gravity analyses on FDI flows and stocks has, e.g., been studies on FDI
and trade flows between Europe and Latin America (e.g., Africano and Magalhães, 2005),
post-cold war Eastern European economic integration (e.g., Christie, 2003), or FDI flows in
China (e.g., Marchant and Peng, 2004) and Southeast Asia under the ASEAN regime (e.g.,
Stone and Jeon, 1999). However, to the best of the authors’ knowledge, no similar studies
have been made so far for FDIs in the BSR region. By emphasizing the overall importance of
distance in bilateral economic relationships between countries, these attempts to emulate use
of trade gravity models onto FDI analyses have in general confirmed theoretical discussion
by, e.g., Krugman (1991) of the so-called new economic geography school. In other words,
the FDI gravity model estimation results of these studies have suggested the significance of
“resisting” factors influencing overall investment between two countries vis-à-vis the total
FDI sample. In the following sections, we will explain our adaptation of the traditional gravity model to an FDI context.
3. Model
Following the traditional application of the gravity model (e.g., Paas and Tafenau, 2005;
Santos Silva and Tenreyro, 2003, 2006), an FDI gravity model that included GDP, GDP per
capita, trade and distance as independent variables was adopted. Furthermore, in order to address the time trend inherited in the cross-section time series, a time variable is also included
in the regression model.
94
Baltic Journal of Economics 12(2) (2012) 89-108
Formalizing the main gravity model used in this study, we obtain
lnFDIijt = β0 + β1 lnGDPit + β2 lnGDPjt + β3 lnGDPCit + β4 lnGDPCjt +
(1)
β5 lnTradeijt + β6 lnDistanceijt + β7 Time + β8 DEU + β9 DBSR + lnεijt
where
lnFDI = Total bilateral FDI stocks between home country i and host country j transformed into natural logarithms
lnGDP = Gross Domestic Product transformed into natural logarithms
lnGDPC = GDP per capita transformed into natural logarithms
lnTrade = Total bilateral trade between home country i and host country j transformed
into natural logarithms
lnDistance = Distance between home country i and host country j transformed into
natural logarithms
Time = Time trend variable (2000 = 1, …, 2008 = 9)
DEU = EU accession dummy K transformed into natural logarithms (K = 0 if 20002003 period, K = 1 if 2004-2008 period)
DBSR = BSR country dummy K transformed into natural logarithms
(K = 0 if non-BSR country, K = 1 if BSR country)
ln ε = Error term
i = Home country i
j = Host country j
t = time period t.
As seen, the variables chosen for our FDI gravity model (1) follow the standard pattern for
most analysis applications in trade and FDI research (e.g., Africano and Magalhães, 2005;
Ledyaeva and Linden, 2006; Paas, 2000; Paas and Tafenau, 2005; Petri, 1994; Santos Silva
and Tenreyro, 2003; Stone and Jeon, 1999). As the dependent variable, total FDI stocks are
chosen as a measure for FDI due to better data availability for all countries and years in the
sample. The independent variables are the gross domestic product (GDP) of countries i and
j, GDP per capita (GDPC) of countries i and j, geographic distance between countries i and j
(Distance) and total value of trade flows between countries i and j (Trade).
Since we are using cross-section time series data, there are obvious risks for various factors
inherited in the data that might result in spurious estimations, making regression results unreliable. In order to detect the presence of unit roots and non-stationarity in the data, Dickey-Fuller tests were executed on the estimated residuals and all dependent and independent
variables (except for Distance). The test results2 allowed us to reject the null hypothesis of the
existence of a unit root as observed values exceeded critical DF values at the 5% level. Thus,
no tests for variable cointegration were performed due to the indicated absence of unit roots
in Dickey-Fuller tests (e.g., Studenmund 2011).
In addition to this set of basic regressors, dummy variables were introduced. The purpose
of these dummy variables was to estimate 1) the statistical significance of changes in the
FDI pattern from the 2004 accession of the four Eastern European countries Estonia, Latvia,
Lithuania and Poland to the EU, and 2) the propensity for FDI in the BSR to be a “regional”
issue (i.e., that BSR FDI occurs between countries located around the Baltic Sea). Some notes
on the interpretation of the dummy parameters have to be made here, since we have transformed the dummy variable K into natural logarithms in order to maintain the consistency
2
DF tests are not reported here due to space constraints.
FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?
95
of the estimation model. Obviously, this makes a traditional interpretation of the dummy
parameters misleading and the parameter estimation has to be antilogged in order to obtain
an estimated magnitude of the influence of the dummy parameters on the dependent variable
(Giles, 2011a, 2011b). Therefore, the estimations for the dummy variables were antilogged in
the final analysis in order to obtain correct interpretations.
Table 1. Summary of the variables of the main gravity model.
Variable
FDIij
Variable type
Dependent
Description
Total bilateral FDI stocks of home country i and host country j transformed into
natural logarithms
Expected sign
Data source
OECD
WIIW
Eurostat
Home GDPi
Independent
GDP of the home country i transformed
into natural logarithms
+
OECD
Eurostat
Host GDPj
Independent
GDP of the host country j transformed
into natural logarithms
+
OECD
Eurostat
Home GDPCi
Independent
GDP per capita of the home country i
transformed into natural logarithms
+
OECD
Eurostat
Host GDPCj
Independent
GDP per capita of the host country j
transformed into natural logarithms
+
OECD
Eurostat
Tradeij
Independent
Total bilateral trade between home
country i and host country j transformed
into natural logarithms
+
IMF DOTS
EcoWin
Distanceij
Independent
-
Google Earth
Time
Independent
DEU
Dummy
DBSR
Dummy
Total distance (in km) between home
country i and host country j transformed
into natural logarithms
Time trend variable (2000 = 1, …, 2008
=9
EU accession dummy K transformed
into natural logarithms (K = 0 if 20002003 period, K = 1 if 2004-2008 period)
Home and host BSR country dummy K
transformed into natural logarithms (K
= 0 if non-BSR country, K = 1 if BSR
country)
Note: For currency conversion purposes between WIIW, Eurostat, IMF DOTS and OECD data sets, annual average
exchange rates for EUR, GBP and USD were obtained from the
Furthermore, some important points regarding estimation problems with the gravity model
require discussion. In the past, mostly ordinary least square (OLS) estimations have been
used to estimate gravity models. Recent contributions (e.g., Santos Silva and Tenreyro, 2006;
Flam and Nordström, 2010) point to bias problems in the presence of heteroskedasticity.
This makes estimations less reliable or, in the worst case, even misleading due to the different weights given to the observations in a sample. Santos Silva and Tenreyro (2006) point
to the fact that “the expected value of the logarithm of a random variable is different from
Baltic Journal of Economics 12(2) (2012) 89-108
96
the logarithm of its expected value” (Santos Silva and Tenreyro, 2006: 641), which has been
ignored in past gravity analyses in economics. Instead, the authors suggest employing the
Poisson Pseudo-Maximum Likelihood (PPML) estimation method, which gives equal weight
to the observations in a sample. Through Monte Carlo simulations, Santos Silva and Tenreyro (2006) show that estimations yielded from the PPML method were robust to different
patterns of heteroskedasticity. Being aware of the bias risk from heteroskedasticity in OLS
estimations of gravity models, we will follow Santos Silva and Tenreyro (2006) and use the
PPML method for this study. All estimations and test statistics were obtained with a Stata 11
statistical package.
4. Data sources and descriptive statistics
We define the BSR as the EU member countries surrounding the Baltic Sea, i.e., Denmark,
Estonia, Finland, Germany, Latvia, Lithuania, Poland and Sweden. Furthermore, in order to
control for major FDI countries outside this region, the main world FDI countries that are
OECD members (i.e., US, UK, France and Japan) are also included in the sample. In total,
this sample makes 12 countries, covering FDI (defined as the FDI stock of each sample country), GDP, GDP per capita and trade between 2000 and 2008 (i.e., 4 years before and 4 years
after EU accession by the three Baltic states and Poland). Finally, distance, defined as the
geographic distance between the capitals of each bilateral relationship, was used as a proxy
for the physical distance between the sample countries.
Table 2. Descriptive statistics (after conversion of variables to natural logarithms except for
the time trend variable)
Variable
year
country
lnFDI_stock
lnhome_GDP
lnhost_GDP
lnhome_GDPC
lnhost_GDPC
lntrade
lndistance
EU_acc_dummy
BSR_dummy
time
Obs
1080
1080
1041
1080
1080
1080
1080
1080
1080
1080
1080
1080
Mean
2004
6.1
7.142286
26.409601
26.409601
9.887204
9.879565
7.782435
7.281833
0.7222222
0.5333333
5
Std. Dev.
2.583185
3.351676
2.713552
2.204863
2.204863
0.8487946
0.8550506
1.846522
1.143067
0.4481107
0.4991188
2.583185
Min
2000
1
-2.67
22.45947
22.45947
8.09
8.09
2.56
4.43
0
0
1
Max
2008
12
12.8
30.27672
30.27672
11.04
11.04
12.36
9.1
1
1
9
Since the quality and availability of macro data for the sample countries (especially for the
three Baltic states) vary, the data used in this study were obtained and compiled from the
following sources: OECD, The Vienna Institute for International Economic Studies (WIIW),
IMF Direction of Trade Statistics (IMF DOTS), Eurostat, and EcoWin. For currency conversion purposes when transforming EcoWin and WIIW data, denoted in Euro, into US Dollars
(current prices) in order to make all data comparable, annual average exchange rates for
Euro, British Pounds and US Dollars were obtained from the ECB. Distances between sample
country capitals were obtained through the Google Earth distance measurement tool.
FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?
97
Table 2 presents the descriptive statistics for the sample, which contains 1080 observations.
Of these, 39 observations for FDI stocks were negative, and deleted during conversion to
natural logarithms. This left us a total of 1041 observations. Since the PPML method only allows for zero or positive values of natural logarithms, another 14 observations were omitted,
which left us with a total of 1027 observations to be estimated through the PPML method.
5. Results
Table 3 reports the cross-section estimation results of the basic version of the main gravity
model (i.e., without dummies for EU accession and home-host country location) for each of
the years 2000 to 2008. As stated earlier, due to bias problems in using OLS in gravity model
estimations (see Santos Silva and Tenreyro, 2006) and detection of error term heteroskedasticity in our OLS estimation3, we used the PPML method for this analysis.
The first item to note is the overall significance of the estimated models. Secondly, all variables except for the GDP variable have the theoretically expected signs. For all individual
years, trade tends to be the most influential variable when estimating FDI dynamics. This is
also confirmed in a correlation analysis, where the partial correlation between the Trade variable and the dependent variable was 0.796 at the 1% level4. This follows the FDI theory outlined and discussed in UNCTAD (1996), Blomström and Kokko (2003) and Lipsey (2006),
and the findings of, e.g., Petri (1994). What is seemingly more theoretically puzzling is the
negative sign of the GDP variable. Mainstream theory would suggest such a variable would
yield a positive sign. We interpret this unexpected sign by the very nature of the BSR FDI
pattern; when compared to the largest foreign direct investor countries in the world, the main
share of FDI activity in the BSR is made by firms from small open economies investing in
other small open economies of the BSR. Therefore, as one of two control variables for size effects (i.e., GDP and GDPC) included in the classic form of gravity models, the negative sign
of GDP has to be considered together with GDPC estimations. Turning our attention to the
size of the home and host economies measured as GDP per capita also gives some indication
of the importance of this factor for the build-up of BSR FDI stocks. This result is rather intuitive, since the increasing wealth of a country, all other things being equal, tends to increase
the average income level of the population, which in turn makes host markets interesting
for foreign investors. However, we should not forget that FDI is global in scope, directed by
business and profit opportunities identified by investing firms no matter where the location,
which means that what we have observed here is not unique for the BSR; we will return to
this observation later in our continued discussion.
Finally, regarding geographic distance as a theoretically impeding factor for trade and FDI
location (e.g., Hummels, 1999; Krugman, 1991), this study supports the findings of, e.g.,
Stone and Jeon (1999) that distance is a statistically insignificant factor when determining the
location propensity of FDI. This is also a highly interesting result, suggesting that distance is
a secondary factor to FDIs in the BSR.
3
4
Tests are not reported here due to space constraints.
Partial correlations are not reported here due to space constraints.
Baltic Journal of Economics 12(2) (2012) 89-108
98
Table 3. PPML estimation results of basic FDI gravity equations from 2000 through 2008.
2000
2001
2002
0.643
0.952
1.000
(0. 917)
(0.908)
(0.900)
ln(Home GDP)
-0.045
-0.061**
-0.066**
(0.030)
(0.031)
(0.031)
ln(Host GDP)
-0.046
-0.062**
-0.065**
(0.030)
(0.030)
(0.030)
ln(Home GDPC)
0.095***
0.110***
0.119***
(0.031)
(0.034)
(0.034)
ln(Host GDPC)
0.098***
0.113***
0.119***
(0.032)
(0.036)
(0.036)
ln(Trade)
0.201
0.223***
0.229
(0.035)
(0.033)
(0.035)
ln(Distance)
-0.028
-0.035
-0.0367
(0.041)
(0.040)
(0.039)
Pseudo R2
0.181
0.196
0.187
Log Pseudolikelihood
-219.750
-233.501
-230.822
No. of observations (= n)
110
116
114
Coefficients with robust standard errors within parenthesis. *, ** and
10%, 5% and 1% levels respectively.
Constant
2003
2004
2005
0.225
-0.560
-1.056
(0.785)
(0.602)
(0.729)
-0.020
-0.006
0.000
(0.027)
(0.018)
(0.020)
-0.019
-0.003
0.002
(0.024)
(0.019)
(0.021)
0.075***
0.097***
0.114***
(0.023)
(0.025)
(0.032)
0.074***
0.080***
0.105***
(0.024)
(0.025)
(0.031)
0.174***
0.140***
0.126***
(0.030)
(0.020)
(0.020)
-0.012
-0.025
-0.035
(0.038)
(0.026)
(0.030)
0.160
0.144
0.138
-230.087
-228.301
-233.606
114
114
116
*** represents statistical significance at the
Table 3. Continued.
Constant
ln(Home GDP)
ln(Host GDP)
ln(Home
GDPC)
ln(Host GDPC)
ln(Trade)
ln(Distance)
2006
-0.594
(0.552)
0.001
(0.017)
0.003
(0.018)
0.084***
2007
-0.261
(0.434)
-0.009
(0.015)
-0.007
(0.017)
0.084***
2008
-0.098
(0.406)
-0.021
(0.013)
-0.019
(0.014)
0.092***
(0.028)
0.078***
(0.029)
0.131***
(0.017)
-0.031
(0.026)
0.126
-233.182
(0.028)
0.080***
(0.029)
0.138***
(0.014)
-0.014
(0.024)
0.115
-234.044
(0.025)
0.081***
(0.024)
0.161***
(0.015)
0.008
(0.021)
0.107
-222.386
Pseudo R2
Log Pseudolikelihood
No. of observa116
116
111
tions (= n)
Coefficients with standard errors within parenthesis. *, ** and *** represents statistical significance at the 10%, 5%
and 1% levels respectively.
Turning our attention to the cross-section time series version of the same data set, we can
make more interesting observations. In Table 4, we can see that besides the fact that OLS estimators systematically overestimate the coefficients for the independent variables compared
FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?
99
to PPML estimators5, the PPML parameters of all independent variables are also statistically
significant except for the Distance variable. The signs of the parameters follow the pattern
from the cross-section estimations reported in Table 3.
Table 4. OLS and PPML estimation results of the FDI gravity model for the entire period,
controlled for EU accession and geographic proximity.
Model 1: OLS
estimation
of the main
gravity model
for the period
2000-2008
Constant
-6.674***
(1.423)
ln(Home GDP)
-0.172***
(0.047)
ln(Host GDP)
-0.156***
(0.046)
ln(Home GDPC)
0.653***
(0.068)
ln(Host GDPC)
0.597***
(0.068)
ln(Trade)
1.301***
(0.051)
ln(Distance)
0.023
(0.075)
Time
-0.062***
(0.016)
DEU
DBSR
Model 2: PPML
estimation of
the main gravity
model for the
period
2000-2008
-0.063
(0.221)
-0.023***
(0.007)
-0.021***
(0.007)
0.100***
(0.010)
0.094***
(0.010)
0.166***
(0.008)
-0.003
(0.010)
-0.009***
(0.002)
Model 3:
PPML Gravity
model controlling for 2004
EU accession
-0.211
(0.233)
-0.021***
(0.007)
-0.022***
(0.007)
0.110***
(0.011)
0.101***
(0.010)
0.167***
(0.008)
-0.000
(0.010)
-0.015***
(0.002)
-0.061***
(0.016)
Model 4: PPML
Gravity model
controlling for
home and host
country being a
BSR country
-0.235
(0.241)
-0.020***
(0.008)
-0.018**
(0.008)
0.100***
(0.010)
0.094***
(0.010)
0.164***
(0.008)
0.003
(0.010)
-0.010***
(0.002)
-0.033***
(0.013)
Model 5: PPML
Gravity model controlling for 2004 EU
accession and home
and host country being
a BSR country
-0.317
(0.246)
-0.019**
(0.008)
-0.020***
(0.008)
0.109***
(0.011)
0.100***
(0.010)
0.166***
(0.008)
0.004
(0.010)
-0.015***
(0.003)
-0.057***
(0.016)
-0.022*
(0.013)
Adj R2
0.828
Pseudo R2
0.156
0.156
0.156
0.156
Log Pseudolike-2070.025
-2068.620
-2069.649
-2068.457
lihood
No. of observa1041
1027
1027
1027
1027
tions (= n)
Coefficients with standard errors within parenthesis. *, ** and *** represent statistical significance at the 10%, 5%
and 1% levels respectively.
However, the estimation results of the two dummy variables DEU and DBSR tell us an interesting story. As expected, EU accession per se did have a statistically significant influence on
changes in FDI stocks. In the PPML estimation models 3 and 5 (see Table 4), the EU accession dummy yielded statistically significant results suggesting that, holding all other variables
constant, post-accession FDI stock levels were on average 6% higher than prior to EU accession. The location dummy for BSR countries’ FDI stocks (models 4 and 5 of Table 4) yield a
Probably due to the presence of heteroskedastic errors in the sample; the magnitude of this bias can be easily controlled for by comparing the elasticities (i.e., the coefficients) of the OLS and PPML estimations, respectively. The
formula to compute the percentage effects is (eβi – 1) x 100, where βi is the coefficient estimated (Santos Silva &
Tenreyro, 2006: 651).
5
100
Baltic Journal of Economics 12(2) (2012) 89-108
statistically significant coefficient, indicating that, holding all other variables constant, intraBSR FDI stocks are on average between 2% and 3% higher than FDI stocks held by firms
from the major industrialized countries outside the BSR.
All in all, the PPML estimations of the FDI gravity model by and large confirm the “common
wisdom” of the theoretical and empirical literature on FDI cited in this paper, with two interesting exceptions: the negative sign of the GDP variable, and the statistically insignificant
parameter estimation for the Distance variable in combination with the statistically significant BSR location dummy variable. We will now move on to interpret the PPML estimation
results in the next section.
6. Discussion and conclusions
In this study, we used a gravity model to analyze recent FDI trends in the BSR measured by
changes in FDI stocks in bilateral relationships between countries in and outside the BSR.
First of all, we would like to forward the methodological contribution of this paper to the
FDI literature, namely the econometric analysis on macro data using the Poisson PseudoMaximum Likelihood estimation method, which has been demonstrated by Santos Silva and
Tenreyro (2006) to be unbiased and robust to different patterns of heteroskedasticity. This
would imply the appropriateness of using PPML estimations for our sample here, by being
the most consistent estimation method for gravity analyses of the type presented in this article. This has, to the best of the authors’ knowledge, not yet been done on BSR FDI data, and
leads us to forward our next contribution to the literature on BSR FDI, which is to show that
the close connection between trade and FDI discussed in the theoretical literature also has
its validity for the BSR setting. Earlier gravity model analyses have stopped at studying the
relationship between trade and distance, and done so by using OLS for their estimation (e.g.,
Paas, 2000; Paas and Tafenau, 2005). Here, we present a variant of the gravity model that also
includes FDI. The findings of our study are expected in line with Petri’s (1994) and Hummels’ (1999) discussions and earlier, similar, studies on FDI and trade (e.g., Africano and
Magalhães, 2005; Bergstrand, 1985; Broadman, 2005; Christie, 2003; Ledyaeva and Linden,
2006; Santos Silva and Tenreyro, 2003; Santos Silva and Tenreyro, 2006), i.e., the estimation
results suggest trade volume, the size of home and host economies and the location of investing countries as important factors for understanding foreign direct investment activity in the
Baltic Sea Region.
The assumption of investing firms’ general preference for countries and cultures close to their
home countries – at least in the early stages of an internationalization process – is something
that we recognize from mainstream theory, such as the early theories on FDI and firm internationalization (see, for example, Johanson and Vahlne, 1977; Dunning, 1977, 1979). This
“basic paradigm” of early FDI theories has since been challenged by the current development
pattern of global FDI, where we can observe firms such as the so-called “born globals” and
where physical distance seems to play a lesser role in investment decisions (e.g., UNCTAD,
2011) for the benefit of other investment incentives such as market size, costs of obtaining
resources and low production costs that might be decisive for investment decisions. This
could be the explanatory factor behind the statistically insignificant results for the Distance
parameters in combination with the other, statistically significant, parameters. Large physical
distances per se might have become less of an effective deterrent for foreign direct investors.
FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?
101
In the midst of all this, our results also suggest that we should not abandon or forget that the
“basic paradigm” is still valid, and that the “traditional” FDI pattern lingers on at least in the
BSR in the sense that foreign direct investments in the BSR tend to be between countries
within the same region. FDIs follow the trade pattern very closely in the BSR, which is associated with the location of the principal trade partners of BSR countries. The answer to the
title of this paper is therefore: “For the BSR, FDI is indeed a regional issue rather than a
global concern” – with the addition “but this pattern is undergoing changes”.
Thus, we can conclude that the FDIs taking place in the BSR are still characterized as being
an “internal affair” of the region, which is also correlated to the overall trade pattern of the
area. The imports of eastern BSR countries are dominated by products and services produced
in western BSR countries or produced by firms owned by western BSR MNCs, and that onethird of all exports of the eastern BSR go to western BSR countries (Olsson et al., 2010).
It is also interesting, and highly relevant, to reflect our results with what is said in EU political circles in connection with the EU accession of the eastern BSR countries. The political
rhetoric of hope and intentions prior to 2004, and manifested in the so-called EU Baltic Sea
Region strategy in 2009 (Baltic Development Forum, 2009), envisioned significant effects
and dynamism in economic integration, trade and FDI from the EU enlargement project. This
is also what we would expect from the FDI literature (e.g., Lipsey, 2006). We might well use
our results to discuss BSR economic integration also from a somewhat different viewpoint
than those held by, e.g., policymakers in the BSR, who were hailing the EU accession of
the Baltic states and Poland as historic for BSR economic development. We are not arguing
against this fact as such. However, we would like to point to the statistically insignificant result of the Distance variable, which normally yields a negative parameter in gravity models.
In combination with the statistically significant dummy parameter result for EU accession,
we would suggest that FDI decisions are more dependent on, e.g., FDI regimes, trade regimes
and production cost advantages rather than geographical closeness (i.e., the distance) of markets and production locations.
Blomström and Kokko (2003) argue in their overview of the FDI literature that the existence
of other basic economic foundations concurring with regulatory reforms and political rhetoric are crucial to make firms carry out FDI and host economies benefit from positive FDI
spillovers. The eastern BSR countries are still in their “catching-up” transition process, albeit
growing at a slower pace now than during the 1990s, and their market potential, defined as
the overall purchasing power of the consumers in these markets, continues to attract FDIs into
the region. Furthermore, the economic growth and development of the eastern BSR countries
also spills over to the flow of outward FDIs of eastern BSR firms, which have increased
steadily during the last 15 years.
In conclusion, policymakers in countries struggling to attract inward FDIs should be aware
of the non-perpetual and “liquid” nature of FDI, especially in the manufacturing sector. As
inward FDIs in the BSR are dominated by investments in the manufacturing sector, this particular nature of FDI in the region should encourage governments to formulate realistic FDI
policies attractive for foreign investors, aiming for sustained long-term retention of foreign
investment in, e.g., non-manufacturing sectors. Contemporary global investor groups are not
necessarily of EU origin, and BSR policymakers staying comfortable with inward FDI pro-
102
Baltic Journal of Economics 12(2) (2012) 89-108
motion programs based on freedom of labor and capital within the EU might miss out factors
which foreign investors consider more important than geographic home country closeness
and free factor movements.
Suggestions for Future Research
This paper is intended as a study of longitudinal dynamics in the FDI pattern of the BSR,
explained by selected macro and spatial factors and estimated by using the Poisson PseudoMaximum Likelihood method. By doing this study, we have opened up a number of possible developments for this research. One is, as suggested by, e.g., Nakamura (2005), that
cross-border investment in the form of M&A tends to follow the overall M&A trend in the
particular economy, that is, a tendency for behavioral isomorphism in the organizational field
that the companies belong to (i.e., that firms are exhibiting a “follow-the-herd” mentality).
The question is whether such behavior is also observable in the BSR. Another phenomenon
that should be investigated in a BSR setting is institutional reasons for foreign companies to
invest in a particular BSR economy that is not geographically close to the home country of
the investor. Furthermore, in this study we only observed a small temporal “window” in the
history of BSR FDI; to study the influence of firm-specific historical, path dependent, factors
is probably still an alternative way to study historical FDI trends in the BSR.
FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?
References
103
Africano, Ana Paula and Magalhães, Manuela (2005). “FDI and trade in Portugal: A gravity
analysis”, FEP Working Papers no. 174, Universidade do Porto, Oporto.
Barney, Jay B. (1991). “Firm Resources and Sustained Competitive Advantage”, Journal of
Management 17(1), 99-120.
Baltic Development Forum (2009). State of the Region Report 2009: Boosting the Top of
Europe. Copenhagen/Helsinki: Baltic Development Forum.
Bergstrand, Jeffrey H. (1989). “The gravity equation in international trade: Some microeconomic foundations and empirical evidence”, Review of Economics and Statistics 67,
474-481.
Blomström, Magnus and Kokko, Ari (2003). ”The economics of foreign direct investment
incentives”, NBER Working Paper 9489.
Broadman, Harry G. (2005). From disintegration to integration: Eastern Europe and the former Soviet Union in international trade. Washington DC: The World Bank.
Buckley, P. J., Clegg, L. J., Cross, A. R., Liu, X., Voss, H. and Zheng, P. (2007). “The Determinants of Chinese Outward Foreign Direct Investment”, Journal of International Business Studies, 38(4), 499-518.
Christie, Edward (2003). “Foreign direct investment in Southeast Europe”, WIIW Working
Papers no. 24.
Dixit, Avinash K. and Norman, Victor (1980). Theory of International Trade. Cambridge:
Cambridge University Press.
Dunning, John H. (1977). “Trade, Location of Economic Activity and the MNE: A Search
for an Eclectic Approach”, in Bertil Ohlin, Per-Ove Hesselborn and Per Magnus Wijkman
(eds), The International Allocation of Economic Activity: Proceedings of a Nobel Symposium. London: Macmillan, pp. 395-418.
Dunning, John H. (1979). “Explaining Changing Patterns of International Production: In Defense of the Eclectic Theory”, Oxford Bulletin of Economics and Statistics 41(4), 269295.
Dunning, John H. (1992). Multinational Enterprises and the Global Economy. Wokingham:
Addison-Wesley.
Eicher, Teo S., Mutti, John H. and Turnovsky, Michelle H. (2009). International Economics,
7th ed. Oxon: Routledge.
Flam, Harry and Nordström, Håkan (2010). “Gravity estimation of the intensive and extensive margin: An alternative procedure and alternative data”, Manuscript, Conference
paper at the SNEE 12th annual conference on European Economic Integration in Swedish
Research, Mölle, Sweden.
Giles, David E. (2011a). “Interpreting dummy variables in semi-logarithmic regression models: exact distributional results”, Econometrics Working Paper EWP1101, Department of
Economics, University of Victoria.
Giles, David E. (2011b). “On the inconsistency of instrumental variables estimators for the
coefficients of certain dummy variables”, Econometrics Working Paper EWP1106, Department of Economics, University of Victoria.
Görg, Holger and Strobl, Eric (2003). “Multinational Companies, Technology Spillovers and
Plant Survival”, Scandinavian Journal of Economics 105(4), 581-595.
Hummels, David (1999). “Have International Transportation Costs Declined?”, Manuscript,
Graduate School of Business, University of Chicago.
104
Baltic Journal of Economics 12(2) (2012) 89-108
Hymer, Stephen H. (1960 [1976]). The International Operations of National Firms: A Study
of Direct Investment. Cambridge, MA: MIT Press.
Johanson, Jan and Vahlne, Jan-Erik (1977). “The internationalization process of the firm – A
model of knowledge development and increasing foreign market commitments”, Journal
of International Business Studies 8(1), 23-32.
Jovanović, Miroslav N. (2006). The Economics of International Integration. Cheltenham:
Edward Elgar.
Krugman, Paul (1991). Geography and Trade. Cambridge, MA: The MIT Press.
Lipsey, Robert E. (2006). “Measuring the Impacts of FDI in Central and Eastern Europe”,
NBER Working Paper 12808.
Marchant, Mary A. and Peng, Xuehua (2004). “Prospects for China’s agricultural FDI inflows: A gravity model approach”, Manuscrpt, Conference paper at the Southern Agricultural Economics Association Annual Meeting, Oklahoma, USA.
Nakamura, H. Richard (2005). Motives, Partner Selection and Productivity Effects of M&As:
The Pattern of Japanese Mergers and Acquisitions. Ph.D. Thesis. Stockholm: IIB.
Olsson, Mikael, Nakamura, H. Richard and Lönnborg, Mikael (2010). “Is there a Baltic Sea
Region? A review of two decades of economic developments in trade and investment
flows in the light of the European Union Strategy for the Baltic Sea Region”, Manuscript,
Conference paper at the SNEE 12th annual conference on European Economic Integration in Swedish Research, Mölle, Sweden.
Orbis (2011). Orbis European company database. Brussels: Bureau van Dijk.
Paas, Tiiu (2000). “The gravity approach for modelling international trade patterns for economies in transition”, International Advances in Economic Research 6(4), 633-648.
Paas, Tiiu and Tafenau, Elge (2005). “Regional trade clusters in promoting eastward enlargement of European Union”, Transition Studies Review 12(1), 77-90.
Penrose, Edith T. (1959). The Theory of the Growth of the Firm. Oxford: Basil Blackwell.
Petri, Peter A. (1994). “The regional clustering of foreign direct investment and trade”, Transnational Corporations 3(3), 1-24.
Pradhan, Jaya Prakash (2004). “The Determinants of Outward Foreign Direct Investment: A
Firm-level analysis of Indian manufacturing”, Oxford Development Studies 32(4), 619639.
Santos Silva, João M. C. and Tenreyro, Silvana (2003). “Gravity-defying trade”, Federal
Reserve Bank of Boston Working Paper No. 03-1.
Santos Silva, João M. C. and Tenreyro, Silvana (2006). “The log of gravity”, The Review of
Economics and Statistics 88(4), 641-658.
Sauvant, Karl P., Maschek, Wolfgang A. and McAllister, Geraldine (2009). “Foreign Direct
Investment by Emerging Market Multinational Enterprises, the Impact of the Financial
Crisis and Recession and Challenges Ahead”, Global Forum on International Investment
VIII, 7-9 December 2009, OECD.
Shapiro, D. and S. Globerman (2001). “National Infrastructure and Foreign Direct Investment”, Manuscript, Simon Fraser University.
Stone, Susan F., and Jeon, Bang Nam (1999). “Gravity-model specification for foreign direct
investment: A case of the Asia-Pacific economies”, Journal of Business and Economic
Studies 5(1), 33-42.
Tinbergen, Jan (1962). The World Economy: Suggestions for an International Economic Policy. New York: Twentieth Century Fund.
FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?
105
UNCTAD (1996). World Investment Report: Investment, Trade and International Policy Arrangements. New York and Geneva: UNCTAD.
UNCTAD (2011). World Investment Report: Non-Equity Modes of International Production
and Development. New York and Geneva: UNCTAD.
Zephyr (2011). Zephyr Global M&A database. Brussels: Bureau van Dijk.
Source: Orbit/Zephyr M&A database. *Asterisks denote deal value estimations only. aIPOs and JVs not included. bCountry codes: EE=Estonia, FI=Finland, SE=Sweden.
Table A2. Cross-border M&A deals in Estonia worth over USD 1 billion 1998-2010a, b.
Source: Orbit/Zephyr M&A database. *Asterisks denote deal value estimations only. aInitial Public Offers (IPOs) and Joint Ventures (JVs) not included. bCountry codes: AT=Austria,
BH=Bahrain, DK=Denmark, FR=France, GB=Great Britain, IS=Iceland, LU=Luxemburg, NL=Netherlands, NO=Norway, SE=Sweden, US=USA.
Table A1. Cross-border M&A deals in Denmark worth over USD 1 billion 1998-2010a, b.
106
Baltic Journal of Economics 12(2) (2012) 89-108
Britain, IT=Italy, PL=Poland.
Source: Orbit/Zephyr M&A database. *Asterisks denote deal value estimations only. aIPOs and JVs not included. bCountry codes: AT=Austria, DE=Germany, FR=France, GB=Great
Table A5. Cross-border M&A deals in Poland worth over USD 1 billion 1998-2010a, b.
Source: Orbit/Zephyr M&A database. *Asterisks denote deal value estimations only. aIPOs and JVs not included. bCountry codes: LT=Lithuania, PL=Poland.
Table A4. Cross-border M&A deals in Lithuania worth over USD 1 billion 1998-2010a, b.
Source: Orbit/Zephyr M&A database. *Asterisks denote deal value estimations only. aIPOs and JVs not included. bCountry codes: DK=Denmark, FI=Finland, GB=Great Britain,
SE=Sweden, US=USA, ZA=South Africa.
Table A3. Cross-border M&A deals in Finland worth over USD 1 billion 1998-2010a, b.
FDI in the post-EU accession Baltic Sea Region: A global or a regional concern?
107
Source: Orbit/Zephyr M&A database. *Asterisks denote deal value estimations only. aIPOs and JVs not included. bCountry codes: AE=United Arab Emirates, AT=Austria,
BH=Bahrain, CN=China, DE=Germany, DK=Denmark, FI=Finland, FR=France, GB=Great Britain, IS=Iceland, IT=Italy, NL=Netherlands, NO=Norway, SE=Sweden, US=USA.
Table A6. Cross-border M&A deals in Sweden worth over USD 1 billion 1998-2010a, b.
108
Baltic Journal of Economics 12(2) (2012) 89-108