Anda di halaman 1dari 20

Emerging Markets Review 7 (2006) 300 – 319

www.elsevier.com/locate/emr

Foreign direct investment in the financial sector and


economic growth in Central and Eastern Europe:
The crucial role of the efficiency channel
Markus Eller a,⁎, Peter Haiss b,1 , Katharina Steiner b,2
a
Centre for Economic and Financial Research (CEFIR), Nakhimovsky prospect 47, office 920, 117418 Moscow, Russia
b
EuropaInstitut, Vienna University of Economics and Business Administration, Vienna, Austria
Received 31 May 2006; received in revised form 31 August 2006; accepted 4 September 2006
Available online 7 November 2006

Abstract

We examine the impact of financial sector FDI (FSFDI) on economic growth via the efficiency channel. We
estimate a panel data model for 11 Central and Eastern European countries in a cross-country growth accounting
framework over 1996 to 2003. We find a hump-shaped impact of FSFDI on economic growth. Medium FSFDI
supports growth if human capital suffices. Above a certain threshold, crowding-out of local physical capital via
foreign bank entry slows growth. We combine the FDI-growth and the finance-growth-literature and conclude that
the level and quality of foreign investment influences the financial sectors' contribution to growth in emerging
markets.
© 2006 Elsevier B.V. All rights reserved.

JEL classification: C23; F36; G10; O16; P2

Keywords: Financial sector foreign direct investment; Economic growth; Financial economics; Emerging markets; Panel
data analysis

1. Introduction and motivation

In the Central and Eastern European countries (CEECs), the inflow of foreign capital (FDI) was
seen as a key component for the solution of problems of capital scarcity and low productivity
⁎ Corresponding author. Tel.: +7 495 105 5002; fax: +7 495 105 5003.
E-mail addresses: MEller@cefir.ru (M. Eller), peter.haiss@wu-wien.ac.at (P. Haiss), katharina.steiner@wu-wien.ac.at
(K. Steiner).
1
Peter Haiss lectures at the EuropaInstitut, Vienna and is with Bank Austria Creditanstalt, Vienna.
2
Katharina Steiner works as a research assistant at the EuropaInstitut, Vienna.

1566-0141/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.ememar.2006.09.001
M. Eller et al. / Emerging Markets Review 7 (2006) 300–319 301

(Sergi, 2004). Accordingly, economic research has developed two different streams of literature:
the FDI-growth and the finance-growth-literature. On the one hand, various studies analyze the
overall impact of general FDI on the host economy (e.g., Borensztein et al., 1998; Dimelis and
Louri, 2004). Results for CEECs are mixed but argue that FDI can be a major growth trigger. On
the other hand, the finance-growth nexus literature has elaborated links between the financial
sector and economic growth: among others, a supply-leading role of the financial sector in
underdeveloped markets (e.g., Beck et al., 2000); mixed evidence in time-series and regional
studies (e.g., Mehl and Winkler, 2003); and changing direction of the causality link with the level
of economic development (e.g., Rousseau and Wachtel, 2005). Countries usually experience
higher growth rates when they have higher investment to GDP ratios. Among other reasons, (in)
efficient allocation of resources contributes to differences in growth (Wachtel, 2001, p. 339).
The positive view of the finance-led growth hypothesis normally focuses on more open and
liberalized financial systems such as the extreme case of banking markets in the CEECs. Empirical
evidence on the economic impact of sectoral FDI is scarce (Levine, 1996).3 This article is one of
the first attempts to fill this gap. We observe that FSFDI induces a variety of micro-structure
changes in the host countries, e.g., potential efficiency improvements for the whole financial
sector. We thus test for the hypothesis that economic growth is led by FSFDI-induced efficiency
gains. We are going to estimate the impact on economic growth in 11 CEECs4 (henceforth CEE-
11) from 1996 to 2003 and adopt a cross-country growth accounting model. This article progresses
as follows: Section 2 characterizes the efficiency channel, Section 3 presents some stylized facts,
Section 4 develops the theoretical model, Section 5 elaborates the empirical model, Section 6
discusses the estimation results, and Section 7 concludes.

2. Transmission between FSFDI and economic growth

The analysis of possible benefits and drawbacks of foreign bank entry is needed to detect
transmission channels through which foreign ownership in banking may affect economic de-
velopment in the CEECs (see Eller et al., 2005, for a review). In the following, we will particularly
focus on the efficiency channel.

2.1. Transmission via the efficiency channel

Change-of-control-driven banking consolidation has contributed to increasing bank efficiency


in the CEECs (Košak and Zajc, 2005). In general, bank efficiency gains at the microeconomic level
are usually attributed to improvements in managerial efficiency (i.e., in input allocation and
technology utilization; also referred to as x-efficiency), and to the realization of economies of scale
and scope. (FS)FDI may increase managerial cost or profit efficiency by transferring superior
managerial skills, bank management systems and technology to the target bank (Amel et al., 2002).
Macroeconomic efficiency gains require better diversification of risks, lower transaction costs and
imply improved pooling and allocation of available financial resources to high-productivity
projects having welfare- and stability-related implications (Fink et al., 2004). An efficiently
operating banking sector can live with a narrower net interest rate margin (lower credit and/or
higher deposit rates) and thereby enhance investment activity and stimulate economic growth

3
For a detailed tabular overview of related studies see Eller et al. (2005).
4
The new EU member states and accession countries, i.e., Bulgaria (BG), Croatia (HR), Czech Republic (CZ), Estonia
(EE), Hungary (HU), Latvia (LV), Lithuania (LT), Poland (PL), Romania (RO), Slovenia (SI), Slovakia (SK).
302 M. Eller et al. / Emerging Markets Review 7 (2006) 300–319

(Holló and Nagy, 2006). As financial markets in the CEECs show extremely high levels of foreign
ownership (Domanski, 2005), it is of crucial interest whether and how the inflow of FSFDI induces
microstructure changes within the affiliate and whether they spill over to efficiency gains for the
financial sector and the overall economy. Fig. 1 depicts the respective line of argumentation (for a
detailed description on the various aspects of foreign bank entry, see Eller et al., 2005).
Arguments of strategic reorientation, the implementation of internal group standards and the
clearing of risky credit portfolios suggest that foreign owned banks can exploit efficiency gains.
However, “foreignness” does not automatically imply higher efficiency. The method and timing
of privatization, the type of investor (e.g., financial vs. strategic) and the competitiveness of the
acquiring banks' home country environment, among others, play a major role. For example,
Bonin et al. (2005b) found that voucher privatized banks were less profit efficient than banks
privatized by other methods. Bonin et al. (2005a) conclude for the 1996–2000 period that in
CEE international institutional investors provide better service and are more cost-efficient than
other banks. However, domestic banks may have a competitive advantage from their previous
interaction with local clients. Bank risk management and the extent to which the domestic
banks' lending decisions are risk/return-driven need to be questioned. While foreign banks do
not have a better credit risk management by definition, it can be assumed that prudent foreign
owners from mature market economies indeed frequently improved operational and
management standards in CEE as most FSFDI into CEE came from the EU-15. However,
cost efficiency does not always depend on ownership (Green et al., 2004). In terms of
economies of scale and scope, foreign owned banks are not more efficient than an average
domestic bank. An explanation for these results might be that foreign owners have upfront costs
in modernizing the acquired banks. Cost efficiency might only occur after some time. In our
empirical analysis, we consider these results and implement time lags and use M&A data to
distinguish between forms of market entry.
Altogether, the ownership question (foreign versus domestic) needs further research. Papi and
Revoltella (2003) provide an explanation for the contradictory findings concerning the
relationship between ownership and bank efficiency on which we will draw in interpreting our
empirical analysis. They argue that a certain threshold for foreign ownership is necessary to
change the efficiency levels of the acquired bank.

Fig. 1. Financial Sector (FS) FDI-induced efficiency-led growth.


M. Eller et al. / Emerging Markets Review 7 (2006) 300–319 303

2.2. Efficiency spillovers on the whole financial sector

What about the impact of FSFDI on the whole financial sector and on domestic banks? A more
efficient and cheaper funding as well as a more diversified business portfolio allows competitive
foreign banks from mature market economies to incorporate lower risk premiums in the interest
margins. As interest margins were and still are higher in the CEECs than in EU-15 where most
foreign entrants come from, they are able to set narrower interest margins in the host markets and
offer their services and products at lower prices. Foreign ownership may thus drive down banks'
interest rate margins, thereby reducing companies' cost of borrowing and facilitating investment.
These institutional reforms within privatized banks and foreign entrants dominating financial
markets in most CEECs push local and regional competition (Drakos, 2003). In general, greenfield
investors represent new competitors per se, whereas acquisitions foster competition due to new
market policies implemented by foreign owners. If domestic banks are able to cope with foreign
contenders, competition with foreign banks will improve the efficiency of the domestic banking
system (Claessens et al., 2001). On the one hand, they might succeed by offering services more
closely tailored to the needs of the local population rather than by trying to compete on price with
foreign competitors that are mostly backed by large financial groups. On the other hand, they might
also be crowded out as foreign owners “cherry pick” and only riskier target groups remain. This
may have a delirious impact on stability if it causes the banks' charter value to drop, thus reducing
the incentives for prudent risk-taking behavior (Mamatzakis et al., 2005). Periods of severe
instability in the financial sector might follow (IADB, 2005). Another negative effect is that while
competition will rise, particularly cross-border mergers and acquisitions contribute to growing
concentration in the financial systems of emerging market economies (BIS, 2004). Mamatzakis
et al. (2005) find evidence for monopolistic competition in South Eastern Europe over the period
1998–2002 which leads to undesirable exercise of market power by banks and hampers further
efficiency gains. External shocks may also have an impact on the activities of foreign owned banks
in the host market, e.g., through the reduction of foreign operations (contagion effect).
Altogether, these issues point out that foreign ownership may have a direct impact on
efficiency gains. In turn, this creates an environment where the entire financial system is forced to
become more efficient, with interest rate margins and, thus, lending rates under severe downward
pressure. Empirical evidence on foreign versus domestic private ownership and efficiency
provides mixed evidence, though competent foreign banks seem to provide positive triggers.
Negative effects, such as the crowding out of domestic banks and systemic risk need to be
considered as well.

2.3. Spillovers to economic growth

How is financial sector efficiency linked to economic development? The efficiency, overall cost
of financial services and depth and breadth of intermediation have an impact on economic growth
(King and Levine, 1993). Privatization and foreign ownership allows for efficient capital
allocation in the CEECs as discussed in Eller et al. (2005). If these gains are being forwarded to the
customer by lower prices for services and products, the cost of borrowing for corporations and
households will decrease and facilitate (real) investment. Following Levine (1997), higher
financial sector efficiency induced by higher financial sector competition should result in an
overall reduction of transaction costs. As a consequence, cost of capital might decline, as interest
margins shrink. Koivu (2004) found evidence that increasing financial sector efficiency measured
by interest margins has growth-enhancing effects on economies in transition. She applied cross-
304 M. Eller et al. / Emerging Markets Review 7 (2006) 300–319

Fig. 2. Economic growth vs. growth of FSFDI in CEE-10, annual observations 1997–2003 (Eller et al., 2005).

country and time-series regressions on nine CEECs over 1995–2002. With cheaper and faster
available capital, corporations might be inclined to invest more which encourages economic
development and growth (Rossi et al., 2004, p. 77). But, heightening competition might also cause
systemic risk, e.g., if banks overly raise their risk appetite in the fight for market share, causing a
credit bubble (Cottarelli et al., 2005). Prudential regulation and tight supervision is necessary.

3. Stylized facts

Foreign penetration in the financial sector has particularly increased since 1999 and reached a
level of more than EUR mn 25,000 or 6.8% of GDP in 2003 in CEE-11 (see Eller et al., 2005).
The financial sector also accounts for a quite high share in total inward FDI stock. The overall
respective average for CEE-105 increased from 15% in 1998 to 19% in 2003. In this context, the
question arises whether one can observe from descriptive data that FSFDI and economic growth
are connected. Fig. 2 shows a simple scatter plot and suggests a slight, positive link between
economic growth and FSFDI. Differences in development among the CEECs are reflected in the
broad dispersion. Particular outliers are Croatia and Latvia, showing a rather high (low) growth in
FSFDI, but at the same time a low (high) growth rate of real GDP per worker.
Financial sector M&A volumes (as a second proxy for FSFDI) vary considerably over time and
across CEECs, leading – at first glance – only to an ambiguous connection with economic growth
(Eller et al., 2005, p. 19). We take this as an indication for the need to include country and time
fixed effects in the empirical setting. Such effects cover the timing of banking crises (e.g., Croatia
1998/1999) and different waves of privatization.
Fig. 3 demonstrates that the annual growth rate and total inward FDI stock per GDP are linked
non-linearly to each other in CEE-11. A kind of “hump-shaped” relationship appears. To a certain
extent, this graph supports the results of studies that have come up with a skeptical view
concerning the growth effects of FDI. Among them, Mencinger (2003) was even able to detect a
negative relationship between growth and FDI for a panel of CEECs. We take this as an indication

5
Because of data limitations, CEE-10 refers for the inward FSFDI indicator to CEE-11 except Romania and for the
financial M&A indicator to CEE-11 except Croatia.
M. Eller et al. / Emerging Markets Review 7 (2006) 300–319 305

Fig. 3. Hump-shaped relationship between economic growth and total FDI in CEE-11 (Eller et al., 2005).

that FDI might not have an unlimited positive impact on growth, but presumably there is a certain
threshold from which on negative effects dominate. This might hold for financial sector FDI as
well. The subsequent empirical investigations of the efficiency channel linking FSFDI and
economic growth will try to render these first stylized facts in a more convincing manner.6

4. Analytical framework

We are conscious about deficiencies of informal growth regressions that do not rely on an
explicit theoretical derivation of the impact of FSFDI, but stick to an MRW-type model (Mankiw
et al., 1992), and introduce FDI as additional explanatory variable (Campos and Kinoshita, 2002;
Mencinger, 2003). Instead, we apply the concept of cross-country growth accounting following
Temple (1999, 124 f.) and Badinger (2003, 180 ff.). This enables the examination of the relative
contributions of growth in inputs and growth in efficiency or technical progress, respectively.
Efficiency is not treated as an unobserved variable as in the MRW-framework and thus is
explicitly considered in our investigation. The change of output is directly regressed on the
changes in factor inputs. Let us start with the following (augmented) standard neoclassical
production function (constant returns to scale and perfect competition are assumed):

Y ¼ AK a H b L1−a−b ; ð1Þ

where Y is the output (GDP), A represents total factor productivity or overall efficiency, respec-
tively, K is the physical capital, H is the human capital and L is the used labor force. Considering
the constant returns to scale, the intensive form of Eq. (1) can be written as follows:

y ¼ Ak a hb ; ð2Þ
where y is the output-labor ratio (Y/L), k is the physical capital-labor ratio (K/L) and h is the
human capital-labor (H/L) ratio. Expanding this equation to the cross-section and time dimension,

6
For a descriptive evidence on financial sector efficiency and foreign ownership see, e.g., Naaborg et al. (2003), Koivu
(2004). For further stylized facts, see Eller et al. (2005).
306 M. Eller et al. / Emerging Markets Review 7 (2006) 300–319

taking logarithms of both sides, and taking first differences, we get the following expression for
the per-worker output growth rate:7
Dlnðyit Þ ¼ DlnðAit Þ þ aDlnðkit Þ þ bDlnðhit Þ; ð3Þ
where i = 1, …, N (country index), t = 1, …, T (time index). Eq. (3) can be used to estimate the
impact of different factor shares on per-worker output growth. The notation allows additionally for
the estimation of the physical capital elasticity (α) and human capital elasticity (β) of the output. In
traditional single country growth accounting, these elasticities are usually imposed (Temple, 1999).
How can FSFDI be included into Eq. (3)? It is a crucial issue to identify the appropriate input
factor through which FSFDI may affect the output. The few theoretical approaches modeling
explicitly economic growth depending on FDI detect three possible ways of influence (Borensztein
et al., 1998; Ruschinski and Sturm, 2004). Firstly, FDI is treated as an inflow of foreign capital
affecting the domestic physical capital stock k (either positively via greenfield investments, or also
negatively via crowding-out of domestic investment). Secondly, FDI influences economic growth
via knowledge spillovers, which contribute to the development of local human capital h. Thirdly,
FDI is expected to spur economic growth through its positive effect on overall efficiency A.
We argue that FSFDI works mainly through the efficiency-channel, since greenfield (de novo)
investments play a minor role compared to privatization-related M&A in the financial sector (e.g.,
Baudino et al., 2004; CGFS, 2005; Hainz and Clayes, 2005). We follow primarily this line
(qualitative arguments presented in Section 2) in modeling the impact of FSFDI. Nevertheless, the
econometric estimations try to consider also the two other effects—inspired primarily by the
findings of Borensztein et al. (1998).
Let us assume that the change in overall efficiency is determined, ceteris paribus, by an
exogenous component (γA0) and the contribution of FSFDI (FSFDI-induced efficiency). Assuming
further that both the change and the level of FSFDI affect the total factor productivity growth rate, we
can specify the subsequent relationship (analogously Badinger, 2003, p. 181):
DlnðAit Þ ¼ gA0 þ gA1 DlnðFSFDIit Þ ð4aÞ
p p
DlnðAit Þ ¼ gA0 þ gA1 FSFDIit : ð4bÞ

Substituting (4a) and (4b) into (3), we can distinguish two equations expressing two different
testable hypotheses. Eq. (5a) represents the temporary FSFDI-induced efficiency-led growth
hypothesis and Eq. (5b) the permanent FSFDI-induced efficiency-led growth hypothesis.
Dlnðyit Þ ¼ gA0 þ gA1 DlnðFSFDIit Þ þ aDlnðkit Þ þ bDlnðhit Þ ð5aÞ
p p
Dlnðyit Þ ¼ gA0 þ gA1 FSFDIit þ aDlnðkit Þ þ bDlnðhit Þ: ð5bÞ

Let us expand this basic setting by the inclusion of control variables that have been found to
have a substantial impact on economic growth in earlier studies. Such variables have the task to
allow for robustness checks of our subsequent estimation results. Out of the manifold pool of
variables that have a potential impact on economic growth (e.g., Sala-i-Martin, 1997), we have
decided to stick to the size of the public sector and to the inflation rate.
Firstly, the size of the public sector is frequently approximated by the government consumption
to GDP ratio (GC). According to Barro and Sala-i-Martin (1995), government consumption

7
The differenced logarithmic series represents the growth rate in continuous time Δlnyit = ∂lnyit/∂t = ∂yit/∂t/yit.
M. Eller et al. / Emerging Markets Review 7 (2006) 300–319 307

proxies for political corruption, bad government, as well as for direct effects of nonproductive
public expenditures. Government consumption covers the expenditures for administration ser-
vices. Wages and salaries are a large component and they have been shown to be unambiguously
associated with lower growth (Barro and Sala-i-Martin, 1995). Various studies show robustly that
government consumption and economic growth are related negatively to each other (e.g., Barro
and Sala-i-Martin, 1995; Romero de Ávila and Strauch, 2003). Nevertheless, Bassanini and
Scarpetta (2001) found that controlling for taxation, both government consumption and in-
vestment seem to have even a positive impact on output per capita. The effect of government
consumption becomes significantly negative when taxes are not included (high mutual correlation
between government consumption and tax variables). Given the different ways for the inter-
pretation of government consumption, one needs to distinguish explicitly between “public sector
size effects” and “expenditure effects”. Bassanini and Scarpetta (2001, p. 30) interpret accordingly:
“The coefficient on consumption when the tax variable is not included should be taken to indicate
the effect on growth of the ‘size’ of government, rather than the true effect on growth of one
specific element of total expenditure.” Since we will not include tax variables as regressors, we rely
on the size interpretation of government consumption.
Secondly, our attempted inclusion of inflation (π) as a conditioning variable has not only a
theoretical reason that becomes manifest in Okun's Law (suggesting a positive relationship
between inflation and economic growth). There are also arguments that transition economies are
characterized by stages of high inflation that may have a negative impact on economic growth
(especially at the beginning of restructuring; Mehl and Winkler, 2003; Valdovinos, 2003). High
inflation will inhibit long-term financial contracting, causing financial intermediaries to keep their
portfolios rather liquid, which in turn retards economic growth (Bordo and Rousseau, 2006;
Rousseau and Wachtel, 2001). Khan and Senhadji (2000) and Rousseau and Wachtel (2002) show
that there is an inflation threshold for the finance-growth relationship. As a consequence, various
empirical investigations of the finance-growth nexus in transition economies have controlled for
inflation (Mamatzakis et al., 2005; Cottarelli et al., 2005). Including these two control variables,
we can rewrite Eqs. (5a) and (5b) as follows:
Dlnðyit Þ ¼ gA0 þ gA1 DlnðFSFDIit Þ þ aDlnðkit Þ þ bDlnðhit Þ þ /1 DlnðGCit Þ þ /2 pit ð6aÞ
p p
Dlnðyit Þ ¼ gA0 þ gA1 FSFDIit þ aDlnðkit Þ þ bDlnðhit Þ þ /1 DlnðGCit Þ þ /2 pit : ð6bÞ

Eqs. (6a) and (6b) form the theoretical baseline. Following the reasoning of above, the γs, α,
and β are expected to exhibit a positive and ϕ1 is expected to have a negative sign, while the sign
for ϕ2 cannot be predicted unambiguously.8
As argued in Section 2, FSFDI needs presumably some time to generate growth-improving
efficiency gains. Accordingly, it is preferable to test not only with contemporaneous right-hand
side values in Eqs. (6a) and (6b), but also allow for time lags. Since the data for our basic
variables are only available from 1996 to 2003, we have a quite small sample size. The inclusion
of time lags would diminish our sample size further and thus deteriorate the power of several
statistical tests. Consequently, we are not able to implement a fully specified dynamic model that
allows for the determination of the optimal lag size. Rather, we will provide estimates for

8
Crespo-Cuaresma and Silgoner (2004) detected a non-linear relationship between inflation and growth in Europe.
They estimated in particular the thresholds where the inflation rate is correlated negatively or positively, respectively,
with economic growth.
308 M. Eller et al. / Emerging Markets Review 7 (2006) 300–319

Eqs. (6a) and (6b) with one lag per each explanatory variable and allow the FSFDI variable to
exhibit even a higher order lag structure (up to lag three).

5. Estimation methodology and procedure

5.1. Fixed-effects panel data model

Our data set comprises the mentioned 11 CEECs with entries from 1996 to 2003. The framework,
which has been responsible for restructuring and opening of the financial sector, is comparable across
the countries under consideration. Thus, we are interested in detecting a considerable behavior across
the different countries and over the relatively short time dimension to evaluate the impact of FSFDI
on economic growth. Panel estimation enables us to take advantage of more observations, “increas-
ing the degrees of freedom and reducing the collinearity among explanatory variables—hence
improving the efficiency of econometric estimates” (Hsiao, 2003, p. 3). The panel data approach
enables to control for omitted variables that are persistent over time. “By utilizing information on
both the inter-temporal dynamics and the individuality of the entities being investigated, one is able to
control in a more natural way for the effects of missing or unobserved variables” (Hsiao, 2003, p. 5).
To make the theoretical Eqs. (6a) and (6b) amenable to our envisaged panel estimation, we
rewrite them as follows:
Dlnðyit Þ ¼ li þ kt þ gDlnðFSFDIit Þ þ aDlnðkit Þ þ bDlnðhit Þ
ð7aÞ
þ /1 DlnðGCit Þ þ /2 pit þ eit

Dlnðyit Þ ¼ li þ kt þ g p FSFDIit þ aDlnðkit Þ þ bDlnðhit Þ


ð7bÞ
þ /1 DlnðGCit Þ þ /2 pit þ eit :

Eqs. (7a) and (7b) represent a static variable-intercept panel data model with country-fixed (μi)
and time-fixed (λt) effects, which take the form of dummy variables. εit is the error term
representing the effects of those unobserved variables that vary over i and t. It is assumed to be an
independently identically distributed random variable with zero mean and variance σε2.9 Let us
briefly discuss the econometric particularities of this model.

5.1.1. Fixed or random effects?


One can think about the possibility that FSFDI may induce cross-country externalities. They are
difficult to measure, remain persistent over time and vary across countries. Country-specific effects
can take such externalities into account. Further, we have to consider that panel data growth
regressions based on annual frequency data are often determined by short-run business cycle
fluctuations (Eller, 2004). As a consequence, it is broadly acknowledged to construct perennial
averages or to use annual data with time-specific effects (Davoodi and Zou, 1998)10 that are treated

9
Given these properties of the noise term, the ordinary least squares (OLS) estimator is the best linear unbiased
estimator (BLUE). Since the μi takes the form of a dummy variable, the OLS estimator is called the least square dummy
variable (LSDV) estimator.
10
The inclusion of these effects is not only an optional tool in order to get more information about individual- or time-
specific characteristics of the examined countries, it is also necessary to inhibit correlation between the regressors purely
because of contemporaneous time or country shocks. Ignoring these effects can lead to parameter heterogeneity in the
model specification, what, in turn, “could lead to inconsistent or meaningless estimates of interesting parameters” (Hsiao,
2003, p. 8).
M. Eller et al. / Emerging Markets Review 7 (2006) 300–319 309

as fixed or random effects. Our examination focuses on differences between specific CEECs not
randomly sampled from a pool of worldwide countries. Therefore, we give exclusive priority to the
fixed-effects estimation procedure.

5.1.2. Dynamic or static panel?


Although we recognize the advantages of a dynamic panel model (e.g., the possibility to control
for omitted variable bias), we would like to stick strictly to the theoretical model, which does not
explicitly call for an inclusion of a lagged dependent variable. Nevertheless, as mentioned above,
also a static panel model allows for the imposition of right-hand side lags.

5.1.3. Homogeneous or heterogeneous coefficients?


Since we are primarily interested in testing whether the behavioral relationship predicting
economic growth is the same across the 11 CEECs and over the 8-year period, the slope coefficients
of the prediction equation are assumed to not vary neither from one country to the other nor from
one year to the other (see Eqs. (7a) and (7b) where the slope coefficients do not possess a country- or
time-specific index). Assuming that the slope coefficients are independently distributed, their
values can be restricted to be constant across countries and over time (similar reasoning in Bottasso
and Sembenelli, 2001, p. 173).

5.2. Estimation procedure

Eqs. (7a) and (7b) are estimated in the following way: we start from a panel data model where all
the variables (inclusive intercept) are homogeneous across the countries and over time. F-tests for
the heterogeneity of the intercept are run11 and deliver the information whether we can include a
country-varying intercept or not. Time dummies are included per assumption in each equation to
control for short-run business cycle fluctuations.
Secondly, we implement the Durbin-Wu-Hausman (DWH) endogeneity test in line with
Davidson and MacKinnon (1993).12 Intuitively, one could argue that (FS)FDI does not only
induce higher growth rates, but that also countries with higher growth rates are more attractive for
foreign investors, and thus, growth affects also (FS)FDI. Hence, we could have an endogeneity
problem in our setting if E{(Δ)FSFDIit·εit} ≠ 0. In this case, ordinary least squares (OLS)
estimation is biased and inconsistent. Instrumental variable (IV) estimation techniques are
required. Under the null of the DWH test, FSFDI is exogenous and both OLS and IVare consistent.
Under the alternative, FSFDI is endogenous and only IV yields consistent estimates.13 Thus, this

11
The null hypothesis of a homogeneous (common) intercept across the countries is tested against the alternative
hypothesis of a heterogeneous (country-specific) intercept conducting simple F-tests with and without restrictions.
12
Procedure: (1) Run the first stage of an IV estimation (i.e. regress the variable with endogeneity suspicion X⁎ on a set
of instruments Z; it is necessary that dim(Z) ≥ dim(X), E{Z·X⁎} ≠ 0, and E{Z·ε} = 0). (2) Take the residuals u from (1)
and run the auxiliary regression y = Xβ + uγ + ε. (3) Employ an F-statistic with K⁎ and NT − K − K⁎ degrees of freedom
that can be used to test the joint significance of the elements in γ. Under the null of exogeneity, γ = 0 and under the
alternative of endogeneity, γ ≠ 0. If K⁎ = 1, a significant t-statistic in the auxiliary regression indicates already
endogeneity since F = t2 (in this case, we compare the t-statistic with the critical values from the standard normal since the
DWH test is an asymptotic test). Therefore, we report in the estimation output tables either the t- or the F-statistic of the
auxiliary regression.
13
Here we could also link our argumentation to the Granger causality issue. Once a regressor is exogenous, it cannot be
that the dependent variable (Granger) causes this regressor. For estimations where the coefficient for the FSFDI variable
is significant and at the same time the DWH test delivers the result that FSFDI is an exogenous regressor for economic
growth, the “direction of causation” runs only from FSFDI to growth but not in the reverse direction.
310 M. Eller et al. / Emerging Markets Review 7 (2006) 300–319

step of the estimation procedure provides the information whether we shall estimate Eqs. (7a) and
(7b) with OLS (i.e., LSDV) or IV. We have to determine already for the DWH test itself proper
instruments, because it requires a first-stage IV estimation step, where the explanatory variable
with endogeneity suspicion is regressed on a set of instruments (see the procedure as described in
footnote 11). Since we will focus our interpretation mainly on the relationship between FSFDI and
economic growth, we test only the FSFDI-related variables in the various equations for
endogeneity.14 Concerning the DWH first stage, we regress the FSFDI variable on the following
set of instruments: the growth rate of the human and physical capital stock per worker, the inflation
rate, one or two lags of the FSFDI variable itself, and the ratio of each CEE country's interest
spread to the aggregate Eurozone interest spread.15 Once the interest spread ratio is bigger than 1,
there is an incentive for a Eurozone bank to invest in the specific CEE country. The correlation of
this ratio with FSFDI is quite high, while its correlation with innovations to economic growth (i.e.,
the residuals of the output growth equations) is relatively low, qualifying it as a suitable instrument.
Thirdly, considerable differences in the standard deviations of the country-specific residuals
indicate group-specific heteroskedasticity. Since heteroskedasticity leads to biased standard
errors, we use White heteroskedasticity-robust standard errors and covariance to allow for
reliable significance interpretations.
Fourthly, all the estimations are accompanied by several sensitivity checks. We check whether
the base regressors remain significant and of the theoretically predicted sign when government
consumption and/or inflation are excluded from the setting. Lags for the explanatory variables are
imposed (one lag for all explanatory variables, higher order lags for FSFDI). Inspired by
Borensztein et al. (1998), we are additionally including interaction terms between FSFDI and the
per-worker human as well as physical capital stock to examine the interaction of FSFDI with the
countries' absorptive capacities. Finally, note that we are implementing two different indicators
for FSFDI: (1) inward FSFDI (Hunya and Stankovsky, 2005; acronym “FSFDI”), (2) completed
financial cross-border M&A (acronym “FINMA”). Both of them are scaled to GDP as well as
measured per employee.

6. Estimation results

The estimation output is summarized in Tables 1–4. Using the inward FSFDI stock per
employee or per GDP (change model in Table 1; level model in Table 2), or using cross-border
financial M&A (Table 3) yields similar results. All the various regressions share two common
particularities. The F-test for heterogeneous intercept rejects always the null of homogeneity and,
thus, calls for the implementation of country-fixed effects taking the form of dummy variables
(estimates available upon request). On the other hand, the DWH test does practically never (only
in Table 1, column (3), at the 10% significance level) reject the null hypothesis that the FSFDI
variable is exogenous. Hence, we are not switching to an instrumental variable estimation.
Additionally, this is an indication that causality in the Granger sense is not very likely to run from
growth to FSFDI. Instead, for FSFDI variables that have a significant coefficient, one could
conclude that causality is only one-directional, namely, from FSFDI to economic growth. This

14
Also, government consumption might be endogenous. Think, for example, about the issue of automatic stabilizers
that could lead to an increase of GC during downswings and to a decrease of GC during upswings.
15
This specification assumes implicitly that the growth rate of the human and physical capital stock per worker and the
inflation rate are exogenous regressors for the output growth equation. A detailed specification and the estimation results
for the DWH first stage regression are available for all the various equations upon request.
M. Eller et al. / Emerging Markets Review 7 (2006) 300–319 311

Table 1
Fixed effects panel data results—impact of FSFDI on economic growth in CEE-10 (cross-country growth accounting,
annual data 1996–2003): temporary effects
Explanatory Dependent variable: Δln(RGDPLit)
variables
Regression (1) (2) (3) (4) (5) (6)
FSFDIit = Δln FSFDIi,t−m = Δln FSFDIit = Δln FSFDIi,t−m = Δln FSFDIit = Δln FSFDIi,t-m =
(FSFDIEMPit) (FSFDIEMPi,t−2) (FSFDIEMPit) (FSFDIEMPi,t−2) (FSFDIGDPit) Δ(FSFDI
HUMPi,t−2)
Constant 0.036⁎⁎⁎ 0.020⁎⁎ 0.068 0.065 0.034⁎⁎⁎ 0.030⁎⁎⁎
(9.320) (2.363) (1.536) (1.026) (12.193) (4.063)
FSFDIit, 0.0004 0.019⁎⁎⁎ 0.002 0.020⁎⁎⁎ − 0.0012 0.002
or FSFDIi,t−m (0.060) (2.750) (0.241) (3.103) (−0.131) (1.216)
Δln(kit) 0.087⁎⁎⁎ 0.090⁎⁎⁎ 0.102⁎⁎⁎ 0.113⁎⁎⁎ 0.103⁎⁎⁎ 0.114⁎⁎⁎
(3.077) (3.227) (5.176) (3.692) (4.675) (3.448)
Δln(hit) 0.037 0.179 0.078 0.251 − 0.008 0.043
(0.139) (0.695) (0.272) (0.865) (−0.007) (0.192)
Δln(GCit) − 0.110⁎ − 0.016 − 0.113⁎ − 0.119⁎
(− 1.658) (−0.180) (− 1.733) (−1.765)
πit − 0.009 0.221 0.211 0.170
(− 1.185) (1.307) (1.188) (1.026)
ln(FSFDIit) × ln(hit) 0.005⁎ 0.006⁎
(1.802) (1.675)
ln(FSFDIit) × ln(kit) − 0.003⁎ −0.003⁎
(− 1.859) (− 1.825)
Number of 63 50 63 50 63 45
total observations
Estimation method LSDV LSDV LSDV LSDV LSDV LSDV
DWH: K⁎; t- 1; 0.796 1; 0.732 3; 2.88⁎ 3; 1.48 1; 0.823 1;−0.931
or F-aux
Adj. R2 0.404 0.272 0.418 0.279 0.409 0.344
F-value (sig. of R2) 3.218⁎⁎⁎ 1.960⁎⁎ 3.220⁎⁎⁎ 1.953⁎⁎ 3.391⁎⁎⁎ 2.442⁎⁎⁎
⁎Static variable-intercept panel data model with country-fixed and time-fixed effects (whose coefficients are not reported
here, but available upon request). t-statistics are in parentheses, basing on heteroskedasticity-robust standard errors (White
diagonal s.e. and covariance; no df correction). Asterisks indicate significance at the 10%(⁎), 5%(⁎⁎), and 1% (⁎⁎⁎) levels,
respectively. DWH (Durbin–Wu–Hausman test): K⁎ represents the number of regressors with endogeneity suspicion; the
t- (for K⁎ = 1) or F-statistic (for K⁎ N 1) represent the significance of the first-stage residuals in the auxiliary regression.
Romania is not included in this setting because of limited FSFDI data. For the variable definitions and sources see the data
appendix.

could be interpreted, for example, via a kind of self-fulfilling prophecy: perceived (but not yet
realized) growth attracts FSFDI, but FSFDI itself realizes growth via raising efficiency (and may
further reinforce growth perceptions).
The standard growth regressors behave as expected: the change of the physical capital stock
per employee is related positively and highly significant to economic growth. The change of
human capital per employee shows in most specifications the expected positive sign, albeit not
significant. Government consumption to GDP shows always the expected negative sign, in
particular highly significant for the estimations employing financial M&A as FSFDI indicator
(Table 3). This confirms the negative impact of the size of the public sector on economic
growth. The inflation rate changes its sign often and is only significant in one regression
(column 4, Table 3). The insignificance in other specifications can be explained by the ex ante
312 M. Eller et al. / Emerging Markets Review 7 (2006) 300–319

Table 2
Fixed effects panel data results—impact of FSFDI on economic growth in CEE-10 (cross-country growth accounting,
annual data 1996–2003): permanent effects
Explanatory Dependent variable: Δln(RGDPLit)
variables
Regression (1) (2) (3) (4) (5)
FSFDIit = ln FSFDIi,t−m = ln FSFDIit = ln FSFDIi,t−m = ln FSFDIi,t-m = ln
(FSFDIEMPit) (FSFDIEMPi,t−3) (FSFDIGDPit) (FSFDIGDPi,t−3) (FSFDIHUMPi,t−2)
Constant 0.027 0.101⁎⁎ 0.035 − 0.065⁎ 0.030⁎⁎⁎
(0.713) (2.226) (1.424) (− 1.998) (3.434)
FSFDIit, resp. FSFDIi,t−m 0.001 − 0.013⁎ 0.0005 − 0.022⁎⁎⁎ 0.007
(0.216) (− 1.792) (0.064) (− 2.782) (1.329)
Δln(kit) 0.087⁎⁎⁎ 0.087⁎⁎⁎ 0.103⁎⁎⁎ 0.100⁎⁎⁎ 0.079⁎⁎⁎
(3.102) (3.249) (4.722) (3.936) (2.718)
Δln(hit) 0.031 − 0.074 0.005 0.064 0.272
(0.119) (− 0.323) (0.019) (0.252) (1.043)
Δln(GCit) −0.112⁎ − 0.102 − 0.123⁎ − 0.097 −0.028
(− 1.717) − 1.225 (− 1.849) (− 1.213) (− 0.406)
πit −0.009 0.135 0.023 −0.072
(− 1.108) (0.694) (0.116) (− 0.953)
ln(FSFDIit) × ln(hit) 0.008⁎⁎⁎
(2.147)
ln(FSFDIit) × ln(kit) − 0.003⁎⁎⁎
(− 2.369)
Number of total observations 64 50 64 50 54
Estimation method LSDV LSDV LSDV LSDV LSDV
DWH: K⁎; t- or 1; 0.965 1;−0.459 1; 0.792 3; 2.30 1; 1.050
F-aux
Adj. R2 0.395 0.245 0.400 0.278 0.323
F-value (sig. of R2) 3.063⁎⁎⁎ 1.837⁎ 3.215⁎⁎⁎ 1.902⁎ 2.406⁎⁎⁎
⁎Static variable-intercept panel data model with country-fixed and time-fixed effects (whose coefficients are not reported
here, but available upon request). t-statistics are in parentheses, basing on heteroskedasticity-robust standard errors (White
diagonal s.e. and covariance; no df correction). Asterisks indicate significance at the 10%(⁎), 5%(⁎⁎), and 1% (⁎⁎⁎) levels,
respectively. DWH (Durbin–Wu–Hausman test): K⁎ represents the number of regressors with endogeneity suspicion; the
t- (for K⁎ = 1) or F-statistic (for K⁎ N 1) represent the significance of the first-stage residuals in the auxiliary regression.
Romania is not included in this setting because of limited FSFDI data. For the variable definitions and sources see the data
appendix.

unclear sign pattern.16 Excluding the inflation rate does not change the results for other
coefficients in a decisive manner. This is an indication of robustness of our specification.
The results for the FSFDI variables show that the temporary and permanent FSFDI-induced
efficiency-led growth hypotheses cannot be confirmed by direct estimates with contempora-
neous values of the two FSFDI indicators. Although the contemporaneous FSFDI variable shows
the expected positive impact on economic growth in most specifications, the coefficient is not
statistically significant. Therefore, we have imposed higher order lags (from one up to three) and
reported the respective significant results in the output tables. Table 1 shows that the growth
of FSFDI per employee is related positively to economic growth with a lag of two periods

16
Apparently, growth-enhancing and growth-decreasing effects of inflation outweigh each other. One could dig deeper
and search for a non-linear impact of inflation on growth (see Crespo-Cuaresma and Silgoner, 2004).
M. Eller et al. / Emerging Markets Review 7 (2006) 300–319 313

Table 3
Fixed effects panel data results—impact of FINMA on economic growth in CEE-10 (cross-country growth accounting,
annual data 1996–2003)
Explanatory Dependent variable: Δln(RGDPLit)
variables
Regression Permanent effects
(1) (2) (3) (4) (5) (6)
FINMAit = ln FINMAit = ln FINMAit = ln FINMAit = ln FINMAi,t−m = ln FINMAi,t−m = ln
(FINMAEMPit) (FINMAEMPit) (FINMAGDPit) (FINMAGDPit) (FINMA (FINMA
HUMPi,t−2) HUMPi,t−3)
Constant 0.036⁎⁎⁎ 0.024⁎⁎⁎ 0.026⁎⁎ 0.048⁎⁎⁎ 0.024⁎⁎⁎ 0.025⁎⁎⁎
(5.008) (3.304) (2.183) (3.684) (5.321) (4.994)
FINMAit, resp. − 0.0009 − 0.001 0.015⁎⁎⁎ 0.011⁎⁎
FINMAi,t−m (−0.489) (− 0.577) (2.752) (2.251)
Δln(kit) 0.111⁎⁎⁎ 0.113⁎⁎⁎ 0.112⁎⁎⁎ 0.077⁎⁎ 0.100⁎⁎⁎ 0.101⁎⁎⁎
(3.717) (4.899) (3.718) (2.607) (4.654) (4.256)
Δln(hit) 0.958 0.292 0.965 0.405 − 0.458 0.177
(1.537) (0.459) (1.550) (0.659) (−1.159) (0.589)
Δln(GCit) − 0.199⁎⁎⁎ − 0.170⁎⁎⁎ − 0.198⁎⁎⁎ − 0.160⁎⁎⁎ − 0.201⁎⁎⁎ −0.175⁎⁎⁎
(−3.512) (−3.076) (− 3.481) (− 3.077) (−3.471) (− 3.089)
πit − 0.008 − 0.008 − 0.016⁎⁎ − 0.005 −0.009
(−0.968) (− 0.943) (− 2.292) (−0.673) (− 0.932)
ln(FINMAit) 0.007⁎⁎⁎ 0.007⁎⁎⁎
× ln(hit) (3.443) (3.433)
ln(FINMAit) − 0.003⁎⁎⁎ − 0.002⁎⁎⁎
× ln(kit) (−3.597) (− 3.560)
Number of total 45 45 45 45 70 60
observations
Estimation LSDV LSDV LSDV LSDV LSDV LSDV
method
DWH: K⁎; t- 1;−0.644 2; 2.27 1;−0.812 2; 1.45 1; 0.118 1; 1.264
or F-aux
Adj. R2 0.416 0.418 0.417 0.491 0.483 0.454
F-value (sig. 2.655⁎⁎ 3.352⁎⁎⁎ 2.660⁎⁎ 3.128⁎⁎⁎ 4.235⁎⁎⁎ 3.873⁎⁎⁎
of R2)
⁎Static variable-intercept panel data model with country-fixed and time-fixed effects (whose coefficients are not reported
here, but available upon request). t-statistics are in parentheses, basing on heteroskedasticity-robust standard errors (White
diagonal s.e. and covariance; no df correction). Asterisks indicate significance at the 10%(⁎), 5%(⁎⁎), and 1% (⁎⁎⁎) levels,
respectively. DWH (Durbin–Wu–Hausman test): K⁎ represents the number of regressors with endogeneity suspicion; the
t- (for K⁎ = 1) or F-statistic (for K⁎ N 1) represent the significance of the first-stage residuals in the auxiliary regression.
Croatia is not included in this setting because of limited FINMA data. For the variable definitions and sources see the data
appendix.

(columns 2 and 4). In contrast, Table 2 shows that the level of FSFDI–both per employee or
scaled to GDP–has a slightly negative impact on economic growth with a lag of 3 years. These
results rather underline the “skeptical” view concerning FSFDI and support the hump-shaped
hypothesis. Note that a lag of 3 not only reduces the number of observations, but shifts the sample
to the period 1999–2003. In these years in most new EU member states, privatization and, thus,
the accumulation of FSFDI have already slowed down, which could lead to the interpretation that
evidence differs among subperiods. Note that for the financial M&A indicator, higher order
lags are not significant. Therefore, we report only the contemporaneous specification results
(columns 1 to 4, Table 3).
314 M. Eller et al. / Emerging Markets Review 7 (2006) 300–319

Table 4
Fixed effects panel data results—impact of FSFDI and FINMA on economic growth in CEE-10 (cross-country growth
accounting, annual data 1996–2003)
Explanatory variables Dependent variable: Δln(RGDPLit)
Regression Temporary Permanent
(1) (2)
FSFDIi,t−1 = Δln(FSFDIEMPi,t−1) FSFDIi,t-m = ln(FINMAHUMPi,t-3)
Constant 0.050⁎⁎⁎ 0.034⁎⁎⁎
(10.668) (7.705)
FSFDIi,t−1, resp. FSFDIi,t−m − 0.004 0.022⁎⁎⁎
(−0.751) (3.770)
Δln(ki,t−1) − 0.034 − 0.037⁎
(−1.294) (−1.686)
Δln(hi,t−1) − 0.083 − 0.189
(−0.225) (−0.389)
Δln(GCi,t−1) 0.076 0.055
(1.303) (1.137)
πi,t−1 − 0.011 − 0.010
(−1.445) (−1.351)
Number of total observations 54 60
Estimation method LSDV LSDV
DWH: K⁎; t- or F-aux 1; 0.506 1;−0.691
Adj. R2 0.236 0.237
F-Value (sig. of R2) 1.911⁎⁎ 1.965⁎⁎
⁎Static variable-intercept panel data model with country-fixed and time-fixed effects (whose coefficients are not reported
here, but available upon request). t-statistics are in parentheses, basing on heteroskedasticity-robust standard errors (White
diagonal s.e. and covariance; no df correction). Asterisks indicate significance at the 10%(⁎), 5%(⁎⁎), and 1% (⁎⁎⁎) levels,
respectively. DWH (Durbin–Wu–Hausman test): K⁎ represents the number of regressors with endogeneity suspicion; the
t- (for K⁎ = 1) or F-statistic (for K⁎ N 1) represent the significance of the first-stage residuals in the auxiliary regression.
Romania resp. Croatia are not included in this setting because of limited data. For the variable definitions and sources see
the data appendix.

In order to investigate further dynamics in our system, we have lagged all the explanatory
variables of the baseline regression by one period (Table 4). This specification turns out to be
very unstable. We have only reported those cases that are somehow acceptable, like the signi-
ficant and positive lagged effects of FSFDI in the permanent specification. We have also
experimented with lags of order two and three. The instability of the specification is still prevalent
and it turns out that these lags are practically never significant—especially in the case of the
physical capital stock.
The theoretical pros and cons of FSFDI for the host economy (Section 2) suggest that there are
limits for economic gains from FSFDI. Thus, the optimal degree of FSFDI lies somewhere in
between an extremely high and an extremely low level. One can think about a hump-shaped
relationship between economic growth and FSFDI, which is also suggested by the stylized facts
presented in Section 3.17
As a consequence, we constructed a transformed index representing a hump-shaped impact of
FSFDI on economic growth. This index is related positively to economic growth with a lag of two
or three periods (see column 6 in Table 1, column 5 in Table 2, and columns 5 and 6 in Table 3).

17
For further theoretical foundation of this “hump-shaped hypothesis” see Eller et al. (2005).
M. Eller et al. / Emerging Markets Review 7 (2006) 300–319 315

While it is highly significant for the estimation employing the financial M&A indicator, it is not
significant for the estimations with the inward FSFDI stock indicator (although it shows a higher
t-statistic than the contemporaneous estimates for this indicator). The effect of the hump-shaped
index for financial M&A (lagged by three periods) remains positive and significant also with the
other base regressors lagged by one period (see column 2, Table 4).
Borensztein et al. (1998) detected a positive and significant interaction between the stock of
human capital and FDI. They interpret this finding with the observation that “the flow of advanced
technology brought along by FDI can increase the growth rate of the host economy only by
interacting with that country's absorptive capability.” Evans et al. (2002) find human capital and
bank-sector size to be complements and suggest that the presence of a developed banking system
is necessary to exploit the productivity-enhancing potential of human capital. Following this line
of research, we implement as a second improvement interaction terms between the stock of FSFDI
and the stock of human and physical capital. We enclose the products of FSFDI and human and
physical capital simultaneously in five different regressions. While the interaction of the FSFDI
stock with the index of employees' education has a positive impact on economic growth, the
interaction of the FSFDI stock with the stock of physical capital is associated negatively to growth.
Both effects together can explain the insignificant impact of contemporaneous FSFDI values in
other equations. The specifications in columns (2) and (4) in Table 3 replace the financial M&A
variable by the mentioned interaction terms and yield coefficients that are highly significant which
may be the effect of the omission of other relevant factors, in particular, the FSFDI variable by
itself. Therefore, we include FSFDI, human capital, and physical capital individually alongside
their product in the respective regressions of Table 1 (columns 3 and 4: change of inward FSFDI
stock per employee) and Table 2 (column 4: level of inward FSFDI stock per GDP). In this way,
we can test jointly whether these variables affect growth by themselves or through the interaction
term (Borensztein et al., 1998). Compared to other equations of Table 1 with the same amount of
total observations, regression (3) delivers the highest adjusted R2. In Table 1, the two interaction
terms do not change their sign and are still significant, albeit only at the 10% level. FSFDI by itself
enters the equation positively but is still statistically significant only with a lag of 2 years.
Let us try to interpret these findings more accurately. Firstly – considering the positive human
capital-related interaction term – we can detect complementary effects between FSFDI and human
capital on economic growth. FSFDI seems to spur economic growth depending on a higher human
capital stock, which is in line with the finding of Borensztein et al. (1998). Knowledge spillovers
to domestic banks associated with the inflow of FSFDI (e.g., in the form of new technology
introduced by foreign banks, Zajc, 2004) can be an explanation. Secondly – considering the
negative physical capital-related interaction term – substitutive effects between FSFDI and
domestic physical capital on economic growth are indicated. De Mello (1999) similarly finds that
FDI among OECD economies is growth enhancing only for countries in which domestic and
foreign capital are complements. This can be interpreted by the crowding-out of local physical
capital caused by the entry of a foreign bank. Schumpeterian effects of creative destruction seem to
be at work. Analogously to Carkovic and Levine (2002), and Campos and Kinoshita (2002), we
could argue that FSFDI is only growth enhancing in countries with low physical capital stocks. In
any case, respective analysis deserves more attention. In general, market microstructure features
like the mode of entry, the main line of business of foreign owners and their target groups and have
to be considered with regard to the impact on local physical capital formation.
As mentioned before, the panel data estimations have also been encompassed by overall cross-
section regressions. The positive impact of FSFDI on economic growth can be broadly confirmed
by the cross-section results, where the trend growth rate of FSFDI per employee between 1996
316 M. Eller et al. / Emerging Markets Review 7 (2006) 300–319

and 2003 is related positively and highly significant to economic growth. Nevertheless, the small
sample of 11 CEECs could lead to sample biases.

7. Conclusion

We analyze the impact of FSFDI on economic growth via the efficiency channel for 11 CEECs
over 1996–2003 with a fixed-effects static panel data model. We find a hump-shaped relationship
between FSFDI and economic growth. Approaching a medium degree of financial M&A is
rewarded by higher economic growth after 2 years. Beyond it, FSFDI seems to spur economic
growth depending on a higher human capital stock. Above a certain threshold, the crowding-out
of local physical capital caused by the entry of a foreign bank seems to hamper economic growth.
The finding of non-linear effects of FSFDI on economic growth is important and has clear policy
implications. While sizable involvement of prudent foreign banks can enhance financial sector
efficiency and spur economic growth, a country's financial sector is more efficient if domestic
banks spice up competition. Looking at the Hungarian case might confirm that position with early
and massive FSFDI and the strong local OTP bank. Concerning privatization, the acquirers' core
banking competency and commitment for know-how transfer and raising efficiency rather than
sales proceeds is important.
Comparing our findings to the arguments brought forward by Rousseau and Wachtel (2005)
provides interesting points for further research. They point out that the effect of finance on growth
possibly varies with the level of economic development. During the 1996–2003 period all CEECs
covered made major progress in the development into market-driven economies. In turn, our
hump-shaped finding can be interpreted as strong initial gains from FSFDI in the catch-up phase
from a lower level of development, with a shrinking impact of FSFDI in the CEECs coming closer
to mature EU economies' standards. This interpretation would imply that FSFDI is most
important in earlier stages of emerging economies. Our finding of substitutive effects between
FSFDI and the domestic physical capital fits that notion. However, this finding should not be
misinterpreted in the sense that once countries rose in economic development, they should try to
reduce foreign bank involvement on the grounds of the hump-shaped relationship to growth.
Another explanation for the non-linearity might rest in investor behavior. Investors usually go into
the most promising markets first and then look further into neighboring markets. More FSFDI
may follow initial FSFDI into related markets even though upgrading banking operations there
may prove more difficult, costly or time consuming because the market environment is (not) yet
ready (copycat strategies, mimicry effects, Košak and Zajc, 2005). In aggregate terms and across
countries, later stage FSFDI may thus look less beneficial to economic development than earlier
stage FSFDI. If correct, this might be a point of concern for home country supervisors of banks
investing abroad. While we concentrate on the efficiency channel between FSFDI and growth,
further channels and regional broadening need still to be investigated. Future analysis should also
strengthen the estimation approach in order to detect potential non-linearity between (FS)FDI and
growth.

Acknowledgement

We gratefully acknowledge the support by the Jubiläumsfonds of the Oesterreichische


Nationalbank (project no. 8868) and are indebted to comments by Harald Badinger, Xiaoqiang
Cheng, Gerhard Fink, Emilia Jurzyk, Robert Kunst, Dušan Mramor, Márton Nagy, Paul Wachtel,
and Martin Wagner. We received constructive comments at the 20th Workshop of the Austrian
M. Eller et al. / Emerging Markets Review 7 (2006) 300–319 317

Working Group on Banking and Finance, Nov. 2005, Graz; the XIV ‘Tor Vergata’ Conference,
Dec. 2005, Rome; the Int'l Conference on Transition Economies at the Academy of the National
Economy of the Russian Federation, March 2006, Moscow; the 14th Conference of the Austrian
Economic Society, May 2006, Vienna; the 2006 Marmara Finance Symposium, May 2006,
Istanbul; the EIASM Workshop, May 2006, Prague, the EFMA's Annual Meeting, June 2006,
Madrid and the 5th GCBE Congress, July 2006, Cambridge. Special thanks to Ettore Dorrucci,
ECB, for data support. The opinions expressed are the authors' personal views. For further
research of the finance-growth nexus group see http://fgr.wu-wien.ac.at/institut/ef/nexus.html.

Appendix A. Data Appendix

For details see Eller et al. (2005).


Growth rates: following Temple (1999, 119).
RGDPL (y): real GDP at 1995 domestic market prices divided by the number of employed persons
of the total economy. Source: AMECO (annual macro-economic) database of the European
Commission's Directorate General for Economic and Financial Affairs (DG ECFIN), April 2005.
Physical capital stock (k): real physical capital stock per employee at 1995 domestic market
prices. Source: WIIW Research Reports 314, March 2005; International Financial Statistics (IFS)
of the IMF; AMECO database.
Human capital stock (h): constructed index using reported education levels of employees 1996–
2003. Source: EUROSTAT.
Government consumption (GC): Real final consumption expenditure of the general government
to real GDP at 1995 domestic market prices, representing the size of the public sector. Source:
AMECO database.
Inflation (π): Growth rate of the GDP price deflator at market prices (national currency). Source:
AMECO database.
Total financial cross border M&A (FINMA): flow data relating to completed M&A (mn EUR) in
the financial sector. The data exclude corporate transactions involving less than 5% of ownership
of banks and nonbank financial institutions or less than 3% if the transaction value is greater than
1 million US$. Source: European Central Bank (Baudino et al., 2004).
Inward financial sector FDI stock (FSFDI): financial intermediation (mn EUR) according to
NACE code J including from 1996 to 2003. Source: Hunya and Stankovsky (2005).
FDI: total inward FDI stock (mn EUR) including equity capital, reinvested earnings and loans
from 1996 to 2003. Source: Hunya and Stankovsky (2005).
Hump-shaped index for FSFDI (FSFDIHUMP): is constructed analogously to Eller (2004) .
Spread: ratio of each CEE country's interest spread to the aggregate Eurozone interest spread
(lending rate minus deposit rate, in percent per annum, end-year). Source: IMF, IFS statistics.

References

Amel, D., Barnes, C., Panetta, F., Salleo, C., 2002. Consolidation and Efficiency in the Financial Sector: A Review of the
Internatitional Evidence. Federal Reserve Board FEDS Series, vol. 2002-47. http://ww.federalreserve.gov/pubs/feds/
2002/200247pap.pdf.
Badinger, H., 2003. Wachstumseffekte der Europäischen Integration. Schriftenreihe des Europainstituts 21. Springer.
Vienna University of Economics and Business Administration, Wien, New York.
Barro, R., Sala-i-Martin, X., 1995. Economic Growth. McGraw-Hill, New York.
Bassanini, A., Scarpetta, S., 2001. The driving forces of economic growth: panel data evidence for the OECD countries.
OECD Economic Studies 33 (2001/II).
318 M. Eller et al. / Emerging Markets Review 7 (2006) 300–319

Baudino, P., Caviglia, G., Dorrucci, E., Pineau, G., 2004. Financial FDI to the EU accession countries. European Central
Bank Paper Submitted to the BIS Committee on the Global Financial System. http://www.bis.org/publ/cgfs22ecb.pdf.
Beck, T., Levine, R., Loayza, N., 2000. Financial intermediation and growth: causality and causes. Journal of Monetary
Economics 46, 31–77.
BIS, 2004. Foreign direct investment in the financial sector of emerging market economies. Report Submitted by the
Committee on the Global Financial System. Bank for International Settlement. http://www.bis.org/publ/cgfs22.pdf.
Bonin, J., Hasan, I., Wachtel, P., 2005a. Bank performance, efficiency and ownership in transition countries. Journal of
Banking and Finance 29 (1), 31–53.
Bonin, J., Hasan, I., Wachtel, P., 2005b. Privatization matters: bank efficiency in transition countries. Journal of Banking
and Finance 29 (8–9), 2155–2178.
Bordo, P., Rousseau, P., 2006. Legal–Political Factors and the Historical Evolution of the Finance-Growth-Link.
Oesterreichische Nationalbank Working Paper, vol. 107.
Borensztein, E., De Gregorio, J., Lee, J.-W., 1998. How does foreign direct investment affect economic growth? Journal of
International Economics 45, 115–135.
Bottasso, A., Sembenelli, A., 2001. Market power, productivity and the EU single market program: evidence from a panel
of Italian firms. European Economic Review 45, 167–186.
Campos, N.F., Kinoshita, Y., 2002. Foreign direct investment as technology transferred: some panel evidence from the
transition economies. The Manchester School 70 (3), 398–419.
Carkovic, M., Levine, R., 2002. Does foreign direct investment accelerate economic growth. Unpublished Working Paper.
University of Minnesota.
CGFS, 2005. Foreign Direct Investment in the Financial Sector—Experiences in Asia, Central and Eastern Europe and
Latin America. CGFS Paper, vol. 25. Committee on the Global Financial System. http://www.bis.org/publ/cgfs25.htm.
Claessens, S., Demirgüç-Kunt, A., Huizinga, H., 2001. How does foreign entry affect domestic banking markets? Journal
of Banking and Finance 25 (5), 891–991.
Cottarelli, C., Dell' Ariccia, G., Vladkova-Hollar, I., 2005. Early birds, late risers, and sleeping beauties: bank credit growth
to the private sector in Central and Eastern Europe and in the Balkans. Journal of Banking and Finance 29 (1), 83–104.
Crespo-Cuaresma, J., Silgoner, M.A., 2004. Growth Effects of Inflation in Europe: How Low is Too Low, How High is
Too High? Vienna Economics Paper, vol. 0411. University of Vienna, Department of Economics.
Davidson, R., MacKinnon, J.G., 1993. Estimation and Inference in Econometrics. Oxford University Press, New York, Oxford.
Davoodi, H., Zou, H., 1998. Fiscal decentralization and economic growth—a cross-country study. Journal of Urban
Economics 43, 244–257.
De Mello, L.R., 1999. Foreign direct investment-led growth: evidence from time series and panel data. Oxford Economic
Paper 51, 131–151.
Dimelis, S., Louri, H., 2004. Foreign direct investment and technology spillovers: which firms really benefit? Review of
World Economics 140 (2), 230–253.
Domanski, D., 2005. Foreign banks in emerging market economies: changing players, changing issues. BIS Quarterly
Review 69–81 (Dec.).
Drakos, K., 2003. Assessing the success of reform in transition banking 10 years later: an interest margins analysis. Journal
of Policy Modeling 25, 309–317.
Eller, M., 2004. The determinants of fiscal decentralization and its impact on economic growth: empirical evidence from a
panel of OECD countries. Master's thesis. Vienna University of Economics and Business Administration.
Eller, M., Haiss, P., Steiner, K., 2005. Foreign Direct Investment in the Financial Sector: The Engine of Growth for Central and
Eastern Europe? EI Working Paper, vol. 69. Europainstitut, Vienna University of Economics and Business Administration.
http://fgr.wu-wien.ac.at/institut/ef/wp/WP69.pdf.
Evans, A., Green, C., Murinde, V., 2002. Human capital and financial development in economic growth: new evidence
using the translog production function. International Journal of Finance and Economics 7 (2), 123–140.
Fink, G., Haiss, P., Mantler, H.C., 2004. Financial-sector macro-efficiency. In: Balling, M., Lierman, F., Mullineux, A.
(Eds.), Financial Markets in Central and Eastern Europe. Routledge, London, pp. 61–98.
Green, C., Murinde, V., Nikolov, I., 2004. The efficiency of foreign and domestic banks in Central and Eastern Europe:
evidence on economies of scale and scope. Journal of Emerging Market Finance 3 (2), 175–205.
Hainz, C., Clayes, C., 2005. Modes of foreign bank entry and the effects on bank interest rates: theory and evidence. Paper
Presented at the ECB/OeNB/CFS-Conference on European Economic Integration. OeNB (Oesterreichische
Nationalbank), Vienna (November).
Holló, D., Nagy, M., 2006. Bank efficiency in the enlarged European Union. Magyar Nemzeti Bank Working Papers,
vol. 2006/3. http://english.mnb.hu.
Hsiao, C., 2003. Analysis of Panel Data. Cambridge University Press, Cambridge.
M. Eller et al. / Emerging Markets Review 7 (2006) 300–319 319

Hunya, G., Stankovsky, J., 2005. Foreign direct investment in Central and Eastern Europe with special attention to
Austrian FDI activities in this region. WIIW-WIFO Database.
IADB, 2005. Unlocking credit: the quest for deep and stable bank lending. Inter American Development Bank Report.
http://www.iadb.org/res/ipes/2005/index.efm.
Khan, M., Senhadji, A., 2000. Threshold Effects in the Relationship between Inflation and Growth. IMF Working Paper,
vol. 110. June.
King, R., Levine, R., 1993. Finance and growth: Schumpeter might be right. Quarterly Journal of Economics 108 (8),
717–737.
Koivu, T., 2004. Banking sector development and economic growth in transition countries. In: Balling, M., Lierman, F.,
Mullineux, A. (Eds.), Financial Markets in Central and Eastern Europe. Routledge, London, pp. 47–60.
Košak, M., Zajc, P., 2005. Bank Consolidation and Bank Efficiency in Europe. University of Ljubljana Faculty of
Economics Working Paper, vol. 182. http://www.ef.uni-lj.si/en/units/rcef/workingPapers.asp.
Levine, R., 1996. Foreign banks, financial development and economic growth. In: Barfield (Ed.), International Financial
Markets: Harmonization Versus Competition. AEI Press, pp. 224–254.
Levine, R., 1997. Financial development and economic growth: views and agenda. Journal of Economic Literature 35 (2),
688–726.
Mamatzakis, E., Staikouras, C., Koutsomanoli-Fillipaki, N., 2005. Competition and concentration in the banking sector of
the south eastern European region. Emerging Markets Review 6 (2), 192–209.
Mankiw, N.G., Romer, D., Weil, D.N., 1992. A contribution to the empirics of economic growth. Quarterly Journal of
Economics 107 (2), 407–437.
Mehl, A., Winkler, A., 2003. The finance-growth nexus and financial sector environment: new evidence from southeast
Europe. Paper Presented at The Ninth Dubrovnik Economic Conference on “Banking and the Financial Sector in
Transition and Emerging Market Economies”. Croatian National Bank. http://www.hnb.hr/dub-konf/9-konferencija-
radovi/mehl-winkler.pdf.
Mencinger, J., 2003. Does foreign direct investment always enhance economic growth? Kyklos 56 (4), 493–510.
Naaborg, I., Scholtens, B., Bol, H., De Haan, J., De Haas, R., 2003. How Important are Foreign Banks in the Financial
Development of European Transition Countries? CESifo Working Paper Series, vol. 1100.
Papi, L., Revoltella, D., 2003. Foreign direct investment in the banking sector: experiences and lessons from CEECs. In:
Colombe, E., Driffill, J. (Eds.), The Role of Financial Markets in the Transition Process. Physica, pp. 155–178.
Romero de Ávila, D., Strauch, R., 2003. Public Finances and Long-Term Growth in Europe: Evidence from a Panel Data
Analysis. Working Paper, vol. 246. European Central Bank, Frankfurt. July 2003.
Rossi, S., Schwaiger, M., Winkler, G., 2004. Banking Efficiency in Central and Eastern Europe. OeNB Financial Stability
Report, vol. 8 (12).
Rousseau, P., Wachtel, P., 2001. Inflation, financial development and growth. In: Negishi, T., Ramachandran, R., Mino, K.
(Eds.), Economic Theory, Dynamics and Markets: Essays in Honor of Ryuzo Sato. Kluwer, Boston, pp. 309–324.
Rousseau, P., Wachtel, P., 2002. Inflation threshold and the finance-growth nexus. Journal of International Money and
Finance 21, 777–793.
Rousseau, P., Wachtel, P., 2005. Economic growth and financial depth: is the relationship extinct already? Paper Presented
at the UNU/WIDER Conference on Financial Sector Development for Growth and Poverty Reduction. Helsinki. July.
Ruschinski, M., Sturm, J.-E., 2004. Foreign direct investment and growth: panel data evidence for 22 OECD countries.
In: Dreger, C., Hansen, G. (Eds.), Advances in Macroeconometric Modeling, Papers and Proceedings of the 3rd IWH
Workshop in Macroeconometrics. Nomos Verlagsgesellschaft, Baden-Baden, pp. 13–24.
Sala-i-Martin, X., 1997. I just ran two million regressions. American Economic Review 87 (2), 178–183.
Sergi, B., 2004. Understanding the “EU-factor”: the Balkans regions as recipients of FDI and industries. South-East
Europe Review 7 (4), 7–20.
Temple, J.R.W., 1999. The new growth evidence. Journal of Economic Literature XXXVII, 112–156.
Valdovinos, C., 2003. Inflation and economic growth in the long run. Economic Letters 80 (2), 167–173.
Wachtel, P., 2001. Growth and finance, what do we know and how do we know it? International Finance 4 (3), 335–362.
Zajc, P., 2004. The effect of foreign bank entry on domestic banks in Central and Eastern Europe. In: Balling, M., Lierman, F.,
Mullineux, A. (Eds.), Financial Markets in Central and Eastern Europe. Routledge, London, pp. 189–205.

Anda mungkin juga menyukai