European Journal of Political Economy

Vol. 21 (2005) 932 – 952

www.elsevier.com/locate/ejpe

Evidence on corruption as an incentive for foreign

direct investment
Peter Eggera, Hannes Winnerb,T
a
Ludwig-Maximilans-University of Munich, CESifo, and ifo-Institute for Economic Research,
Poschingerstr. 5, 81631 Munich, Germany
b
Department of Economics and Statistics, University of Innsbruck, Universitaetsstr. 15, 6020 Innsbruck, Austria
Received 8 September 2004; received in revised form 21 January 2005; accepted 27 January 2005
Available online 20 April 2005

Abstract

This paper assesses the relationship between corruption and inward foreign direct investment
(FDI). Previous research has presumed that corruption directly enters the cost function of
multinationals, suggesting a negative relationship between corruption and FDI. For a sample of
73 developed and less developed countries and the time period 1995–1999, we find a clear positive
relationship between corruption and FDI. Corruption is thus a stimulus for FDI.
D 2005 Elsevier B.V. All rights reserved.

JEL classification: F23; R38; C23

Keywords: Corruption; Foreign direct investment; Panel econometrics

1. Introduction

Corruption, defined as the misuse of power by public officials for private gains (see
Bardhan, 1997), affects economic development and is a characteristic of low-income
countries (see, e.g., Easterly, 2001; Abed and Gupta, 2002). Recent research has
emphasized the relation between corruption and economic growth (see Shleifer and
Vishny, 1993; Mauro, 1995; Ehrlich and Lui, 1999), inequality and poverty (see Gupta et

T Corresponding author. Tel.: +43 512 507 7156; fax: +43 512 507 2788.
E-mail address: Hannes.Winner@uibk.ac.at (H. Winner).

0176-2680/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.ejpoleco.2005.01.002
 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952 933

al., 2002), inflation (see Cukierman et al., 1992), real exchange rates (see Bahmani-
Oskooee and Nasir, 2002), and the provision of public goods (see Mauro, 1998; Tanzi and
Davoodi, 2000).1
Prior research also suggests that corruption affects flows of foreign direct investment
(FDI). Corruption is determined by a country’s institutional environment, which, in turn,
has been found to be important for the locational attractiveness. Harms and Ursprung
(2002) provide evidence that FDI inflows per capita positively depend on political rights
and civil liberties (see Schneider and Frey, 1985; Wheeler and Mody, 1992, for earlier
support on a positive nexus between political freedom and FDI). Harms and Ursprung
(2002) further illustrate that FDI increases as a country’s degree of political risk decreases.
In Egger and Winner (2003), we find a positive impact of the viability of contracts on FDI
inflows. A high level of corruption is usually associated with an unfavorable institutional
environment. Accordingly, one would expect FDI to decline with corruption: Corruption is
expressed as a bgrabbing handQ that increases costs of multinational firms, so lowering the
incentives to invest abroad. On the other hand, in the presence of regulations and other
administrative controls, corruption can act as a bhelping handQ to foster FDI, as proposed
by Leff (1964).2
Prior empirical research has mainly analyzed cross-section data, thereby implicitly
focusing on the long run effects of corruption. The studies tend to find a negative long run
impact of corruption on FDI. Hines (1995) lends support to the view that US FDI locates
in less corrupt countries. Wei (2000), in a cross-section of bilateral FDI, also finds a
negative relationship between corruption and FDI. See also Smarzynska and Wei (2000)
for related firm level evidence, and our 2004 paper for evidence within country-pairs.
Habib and Zurawicki (2002) analyze the relationship between corruption and FDI in a
cross-section of 89 developed and less developed countries and find that corruption tends
to impede FDI. In contrast, Akcay (2001) does not find a significant relationship between
corruption and FDI to less developed countries.
We propose that a country’s degree of perceived corruption is correlated with
determinants of the institutional environment that are difficult to measure and are typically
not or only partly controlled for in these studies. The cross-country variation in corruption
is consequently likely to be correlated with omitted variables, inducing a bias of the long
run (i.e., cross-sectional) corruption parameter. Moreover, the literature on the
determinants of corruption suggests that corruption should be treated as an endogenous
variable (see Bliss and Di Tella, 1997; Treisman, 2000; Paldam, 2002; Clarke and Xu,
2004; Mocan, 2004).
This paper extends the previous empirical evidence on the impact of corruption on FDI
in several directions. First, we assess the short and long run impact of corruption on inward
FDI stocks. Empirically, (i) we utilize a large panel of 73 (developed and less developed)
host countries between 1995 and 1999, covering more than 90% of the world’s inward
FDI. (ii) We estimate a Hausman–Taylor model, which allows us to disentangle the short

1
For surveys on the causes and economic consequences of corruption, see Bardhan (1997), Tanzi (1998), Jain
(2001) and Aidt (2003).
2
Glass and Wu (2002) demonstrate that corruption may have positive effects on FDI in a general equilibrium
model.
934 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

and additional long run impact of corruption on FDI, thereby accounting for the potential
endogeneity of the latter one. (iii) We use information from three different corruption data
sources to account for the robustness of our findings. We find a positive short run impact
of corruption on FDI, which supports a bhelping handQ interpretation. In accordance with
Glass and Wu (2002), we also identify a positive long run impact of corruption on a host
country’s attractiveness for foreign investors, which is at odds with previous hypotheses
and results (Hines, 1995; Wei, 2000; Habib and Zurawicki, 2002).
Second, in a counterfactual simulation exercise, we assess how corruption influences
the international distribution of FDI. In particular, we ask how the observed change in
corruption over the period 1995–1999 contributed to both the growth and the distribution
of FDI. Our findings suggest that the long run contribution of the perceived corruption
may be up to 40% of the observed overall FDI growth in our sample of countries. Further,
the observed change in corruption has accounted for an equalization effect on the
international distribution of real inward FDI shares, which was mainly in favor of shares in
the EU and the South and Central American areas and mostly at the expense of FDI stock
shares in the rest of NAFTA and the Asian nations.
The paper is organized as follows. Section 2 presents the theoretical background
focusing on the distinction between short and long run effects of corruption on FDI.
Section 3 motivates our econometric specification and discusses the variables of interest as
well as other controls. This section also elaborates the empirical model and discusses the
key issues of estimation. Section 4 presents the empirical results and the findings of the
simulation analysis. Finally, Section 5 concludes.

2. Theoretical background

Theoretically, one has to distinguish between dgrabbing handT and dhelping handT
influences of corruption on inward FDI.3 In the short run, corruption raises the cost of a
firm’s foreign investment, since (i) firms have to pay bribes (similar to taxes),4 (ii) they are
engaged in resource-wasting rent seeking activities (see Applebaum and Katz, 1987;
Murphy et al., 1991; Shleifer and Vishny, 1993),5 and (iii) they have to bear additional
contract-related risks, because corruption contracts are not enforceable in courts (see
Boycko et al., 1995). Corruption in the host country thus acts as a bgrabbing handQ,
reducing the profits of firms and therefore lowering a firm’s incentives to invest abroad.
Further, corruption reduces the productivity of public inputs (e.g., infrastructure) which, in
turn, decreases a country’s locational attractiveness (see Bardhan, 1997; Rose-Ackermann,
1999; Lambsdorff, 2003). On the other hand, multinational firms might be willing to

3
See also Olson (1993) who distinguishes rolling and stationary bandits. Dalgic and van Long (in press) follow
Olson (1993, 2000) to develop a dynamic framework of corruption. They distinguish between the central and the
local government level, focussing on the trade-off between grabbing hand and helping hand corruption in local
governments.
4
b. . . corruption is associated with an extra fee or bribe paid to a government official by a private entrepreneur
for obtaining an economic profitQ (Glass and Wu, 2002, p. 3).
5
b. . . corruption causes people to give up productive activities to engage in rent-seeking behavior . . .Q (Glass
and Wu, 2002, pp. 3–4).
 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952 935

accept paying bribes in order (i) to speed up the bureaucratic processes to obtain the legal
permissions for setting up a foreign plant (see Lui, 1985, for the related dspeed moneyT
argument), and (ii) to gain access to publicly funded projects (see Tanzi and Davoodi,
2000). In this case, corruption acts as a bhelping handQ, increasing profits of multinational
firms. If the revenue effects outweigh the cost effects, corruption is expected to increase
FDI. In the presence of pre-existing government and/or bureaucratic failures, however,
corruption may be efficiency-enhancing (see Leff, 1964, and the surveys by Bardhan,
1997; Aidt, 2003).6 Corruption then may also be rent-creating, with the rents from controls
over foreign investment shared by corrupt officials and foreign investors.
Glass and Wu (2002) focus on the impact of corruption on FDI in a general equilibrium
model with Northern innovation and Southern imitation (see also Grossman and Helpman,
1991).7 Multinational firms are vertically organized and shift their production to the low-
cost country. The firms bear the risk of illegal imitation of their innovations, and of the
requirement to pay bribes to public officials. Analyzing four types of bribes (i.e., bribes on
sales or profits as well as repeated and one-time bribes, e.g., to obtain a permission to sell
in the foreign market), Glass and Wu (2002) demonstrate that the general equilibrium
effects of corruption on FDI are in principle ambiguous, and conclude (ibid., p. 19) that
bcorruption need not be bad for FDI...Q, but rather corruption may foster inward FDI.8

3. The empirical model

To isolate the impact of corruption on FDI, we follow the related literature and control
for other determinants of the locational choice of horizontally (i.e., market seeking) and
vertically (i.e., low-cost seeking) organized multinational enterprises.

3.1. FDI

General equilibrium models of a country’s outward or inward FDI analyze the role of
(physical or human) capital endowments for the choice of going multinational (see
Markusen, 2002; Markusen and Venables, 2000; Glass and Wu, 2002). Previous empirical
work employs either foreign affiliate sales or FDI stocks as the dependent variable (see
Markusen and Maskus, 2002; Blonigen et al., 2003; Blonigen and Davies, 2004; Egger
and Pfaffermayr, 2004). Real stocks of FDI ( F) are approximated in the following way.

6
The investment could itself in principle be inefficient or welfare-reducing, if protectionist policies encourage
the investment. See Bhagwati (1968) on bimmiserizingQ growth.
7
Empirically, this effect can be associated with the long run impact, since general equilibrium outcomes are
likely to not take place immediately.
8
Similar to the bgrabbing handQ, corruption increases a firm’s costs by the payment of bribes and the waste of
resources through rent seeking activities. This effect underlies a disincentive to invest in the low-cost country.
Second, and in contrast to the cost effect, the payment of bribes (i) increases the relative wages of the North and
makes it more profitable to shift production to the low cost South (it reduces the benefits of innovating by raising
costs while producing in the North), and (ii) it raises aggregate expenditure of households and therefore the
volume of a firm’s sales. Glass and Wu (2002) demonstrate that this positive effect dominates the negative cost
effect in general equilibrium, irrespective of which type of bribe is applied.
936 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

Nominal stocks in US$ from UNCTAD’s World Investment Indicators are depreciated by
investment deflators to come up with real stocks in the initial period. We choose 1995 as
the base year. Further, we assume a constant depreciation rate of 100d d=8% throughout
(see Keller, 2000, p. 41, for a justification of this assumption). Real FDI stocks in the
following periods are then defined as F t = F t
1(1
d) + FI t , where F t denotes inward FDI
stocks in year t and FI t is real foreign direct investment in t. FI t is based on the annual
difference in nominal FDI stock data deflated by the investment deflators available from
the World Bank (World Development Indicators).

3.2. Corruption

We apply rating based data as collected by Transparency International (TI, 1995–2001).

The TI index summarizes the degree to which corruption is perceived to exist among public
officials. It is a composite index, drawing on 14 different polls and surveys from seven
independent institutions carried out among business people and country analysts (see
Transparency International, 2001, for further details). We use data from 1995 and later, since
Transparency International does not publish historical data at an annual basis. Additionally,
no information on the variance of the corresponding point estimates of the corruption index
scores are reported. The TI index takes values in between 0 and 10, where 0 corresponds to a
very high level of corruption and 10 to a low one. Hence, a negative correlation between the
index value and FDI indicates a positive relationship between corruption and FDI.
To check the robustness of our results, we additionally apply the corresponding index of
the International Country Risk Guide (ICRG, see http://www.icrgonline.com/) and the
World Bank’s 1997/98 index (Kaufmann et al., 1999, 2002). The former is ranging from 0 to
6, where 0 corresponds to a high level of corruption. The latter is the point estimate of
governance from Kaufmann et al. (2002, pp. 22–24), which does not exhibit any time
variation.

3.3. Other controls

First, the theoretical literature on horizontally organized multinationals stresses the

importance of the proximity to the market (see Brainard, 1993). Accordingly, multina-
tional activity abroad is driven by the market size of destination countries. We follow the
related empirical literature and use country size in terms of real GNP as a proxy of foreign
market size (see Brainard, 1997; Markusen and Maskus, 2002).9 Caves (1996, p. 29–30)
emphasizes that banother consequence of scale economies is that where the MNE locates
its production depends ... also on the sizes of ... national marketsQ. We expect a positive
impact of an increase in market size on inward FDI. Second, vertical multinational activity
and FDI is predominantly determined by differences in relative factor endowments rather
than size. Since the developed economies (i.e., the OECD countries) are the main sending
countries of FDI, we would expect more vertical FDI to take place in countries with a low
high-skilled to low-skilled labor ratio (proxied by secondary school enrolment figures),

9
Caves (1996, p. 2930) emphasizes that another consequence of scale economies is that where the MNE locates
its production depends ... also on the sizes of ... national markets.
 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952 937

and more horizontal FDI in the opposite case (see also Markusen and Maskus, 2002;
Wheeler and Mody, 1992). Additionally, Markusen and Maskus (2002) suggest
accounting for the importance of vertical FDI by the inclusion of an interaction term
between country size and the skilled to unskilled labor endowment ratio.10 As illustrated in
Markusen and Maskus (2002), especially the small, skilled labor abundant economies are
expected to conduct outward FDI. Hence, especially large, skilled-labor scarce economies
are expected to receive FDI. A negative co-efficient estimate of this interaction term would
support this hypothesis. GNP and secondary school enrolment come from the World
Bank’s World Development Indicators. The literature on the determinants of bilateral FDI
underpins the negative impact of distance as a measure of transaction costs on FDI (see
Markusen and Maskus, 2002; Blonigen et al., 2003; Egger and Pfaffermayr, 2004). We
implicitly account for this and any other time-invariant determinant by country-specific
effects, see below.
Finally, we account for the quality of the legal system as an additional determinant of
contract risk and FDI. Legal quality is measured by a rating based index as collected by the
Economic Freedom Network (see Gwartney et al., 1996, 2000, and recently Gwartney and
Lawson, 2003). This index is reported annually in Area II of the Economic Freedom
Network and includes ingredients such as the independence of the judicial and court
system, the protection of property rights (including the enforcement of contracts, and the
mutually agreeable settlement of disputes), military interference in rule of law and the
political process, and the integrity of the legal system (see Gwartney et al., 2000; The
Economic Freedom Network, 2001).11 Also this index takes values in between 0 and 10,
where the latter indicates a high level of quality of a country’s legal system. Since a higher
legal quality reduces contract risk, we expect a positive impact on inward FDI.

3.4. Specification

We estimate the impact of corruption on inward FDI at the aggregate level in an error
components framework (see Baltagi, 2001, for further details). To isolate the corruption
effect, we control for the most important determinants of FDI as outlined before:
Fit ¼ b0 þ b1 Giðt
1Þ þ b2 Siðt
1Þ þ b3 Giðt
1Þ d Siðt
1Þ þ b4 Lit þ b5 Cit þ b6 Cid þ uit ;
ð1Þ
where F denotes country i’s real inward stocks of FDI in year t. G is real GNP in terms of
1995 US$, S is the secondary school enrolment share, and Gd S is the interaction effect as

10
In studies of bilateral FDI, the absolute difference between two countries’ school enrolment ratios may be
employed. In our unilateral context, this boils down to an interaction term between GNP and secondary school
enrolment ratio. The various measures of endowments with skilled labor are only available until 1999 so far.
11
An alternative measure is assembled by the Heritage Foundation (see Johnson et al., 1998). According to de
Haan and Sturm (2000, pp. 223–226) both data collections are similar with respect to their country ratings. In our
study, we refer to the Economic Freedom Network, which allows for a distinction between corruption and legal
quality (for details see Kaufmann et al., 2002). Further, we only apply the institutional components of the freedom
index. Thus, other areas mainly covering macroeconomic aspects (for instance, money supply, inflation rate, or
capital controls) are not used here (for details see Gwartney et al., 2000).
938 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

discussed above. We take the log of all continuous variables ( F, G, S). Lagged values of
G, S and the interaction term are used as predetermined variables, to guard against the
potential problem of endogeneity (see Hayashi, 2001, p. 109). L denotes legal quality, and
C is the corruption index scoring. The data sources are referred to above. Noteworthy, b 5
captures only the short run impact of corruption, whereas b 5 +b 6 is the overall long run
effect. Thereby, C i. is the country-specific, time-averaged, between impact of corruption.

3.5. Estimation

First, we estimate Eq. (1) using fixed country effects. Although this provides consistent
(but not necessarily efficient) parameter estimates, it wipes out any time-invariant
information. In our case, the additional long run effect of corruption associated with the
between effect is of interest but captured by the fixed effects. The fixed effects parameters
represent only short-term influences of the variables of interest (see Pirotte, 1999), and the
short run impact of corruption may only account for part of the long run influence. Also,
we run a between regression, which exploits information from the cross-country
dimension of the data. However, the index of perceived corruption is likely highly
correlated with hardly measurable institutional or cultural factors being part of the error
term. Then, the between parameter of the corruption variable cannot be consistently
estimated in a cross-sectional between regression. Therefore, we have to apply an
approach, which allows us to consistently infer the short run (within) and the long run
(between) impact simultaneously.
Hausman and Taylor (1981) propose to estimate the impact of time invariant variables
in panels by generalized least squares (GLS), applying an instrumental variables estimator
to guard against the possible endogeneity of some of the regressors. In our case, this
approach is eager to overcome the potential endogeneity of the corruption variable. In
addition to the short run effect, the Hausman and Taylor technique allows us to
consistently estimate the additional long run (i.e., the additional between) impact of
corruption (C i. ), which cannot be estimated by dynamic panel data methods due to the
short time coverage and the strong unbalancedness of our data. The advantage of Hausman
and Taylor’s approach is that no instruments from outside the model are required. But
rather, instruments are constructed from inside the model. This is possible, since each
variable is split into two components: the within and the between component. Following
Cornwell et al. (1992), we split the set of variables into the doubly exogenous and the
singly exogenous ones. The latter are correlated with the error term, and the former are not.
Which group a variable belongs to has to be decided on theoretical grounds and according
to sensitivity analysis. In our case, it seems adequate to treat all size-related variables like
GDP and the interaction term but also corruption as singly exogenous. We assume a
country’s school enrolment ratio and legal quality as doubly exogenous.
In the estimation, we proceed as follows (see Baltagi, 2001). First, we estimate the fixed
effects model, which forms the consistent benchmark case, where the additional long run
impact of corruption (b 6) cannot be estimated. This model obtains an estimate of the
within error component (r v2) from its remainder error term. Second, we subtract the
estimated within model impact of the time-variant variables plus the constant from the left-
hand-side variable to obtain the so-called Amemiya-type residuals. We average them over
 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952 939

time and regress them on the time-averaged corruption (C i. ), using the doubly exogenous
variables as instruments. The time-averages of the corresponding residuals serve to
compute an estimate of the second overall variance component (r 12). Third, we transform
each variable in the following way:
ỹy it ¼ yit
hyi: ; ð2Þ

r2v
h¼1
; ð3Þ
r21
where y i. denotes the average over time. Additionally, we need the within ( y it
y i. ) and the
between part ( y i. ) of each variable to construct the set of instruments. Finally, we run an
instrumental variable regression on the transformed model, where all within and between
parts of the doubly exogenous determinants and only the within part of the singly
exogenous variables are used as instruments.
The appropriateness of the choice of instruments and the decision of whether a variable
is treated as doubly exogenous or singly exogenous can be tested by the Hausman and
Taylor (1981) over-identification test. This is a variant of the Hausman test, distributed as
v 2 with the number of doubly exogenous minus the number of singly exogenous variables
as degrees of freedom. Instead of the simple random effects model, the Hausman and
Taylor model is tested against its fixed effects counterpart. If the test statistic is
insignificant, the model is consistent and more efficient than its fixed effects counterpart.

4. Estimation results

4.1. Descriptive statistics

Table 1 summarizes the descriptive statistics. The EU and the NAFTA together account
for more than 65% of real stocks of FDI in the whole sample of 73 countries covering the
period 1995–1999 (see Appendix A for the country sample). The EFTA and the EU
countries exhibit the lowest perceived corruption (highest scoring). Perceived corruption is
highest, i.e., the index is lowest, for the African, South and Central American as well as
the Asian economies. With respect to the quality of the legal system it turns out that the
EFTA countries exhibit the highest values, whereas the lowest index values are observed
in the African as well as the South and Central American countries.

4.2. Regression results

The data set is strongly unbalanced. The average time series length is 4.7 years, varying
considerably across nations. In sum, we come up with about 320 observations. We run
fixed effects, Hausman–Taylor and between regressions on (1) using all non-missing FDI
data in our sample. Table 2 presents the results for the whole country sample and a second
one, which excludes outliers (all observations with a remainder error in the upper or lower
end 1 percentile range, i.e., 8 observations or about 2.5% of the sample).
940 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

Table 1
Descriptive statistics
Country bloc Real FDI stocks Legal qualitya Corruption indexa
Share in Average Average Average Average Average
final period annual score in annual score in annual
in % growth in % final period growth in % final period growth in %
EU 40.4 16.3 8.5 0.6 7.5 0.2
EFTA 2.7 10.4 9.1 0.2 8.9
0.1
CEEC 1.3 71.3 6.5 0.3 4.1
2.9
NAFTA 25.0 13.4 7.6 0.0 6.8 0.5
OECD-Rest 6.3 2.6 7.6 0.6 6.4 0.0
Africa 1.8
4.8 5.9 1.5 4.0 0.5
South and 6.3 31.2 4.8 0.8 3.9 1.9
Central America
Asia 16.1 4.4 6.4 0.4 4.7
0.8
Total 100.0 18.8 7.1 0.4 5.8 0.0
a
Unweighted figures.

In general, the model fit is well and the short run estimates of the corruption coefficient
are fairly robust with respect to the correction for outliers (Models C to D; see below for
more details). In fact, the significant Hausman test in any estimated model indicates that the
random effects model is inconsistent, and that the long run impact of corruption cannot be
consistently estimated from the between regression. This becomes obvious from the
between estimates reported in Model E. In fact, the estimated corruption parameter in
Model E even exhibits the opposite sign. This also leads to positive or heavily biased
negative random effects model estimates of the additional long run coefficient of corruption
(not reported). According to the Ramsey RESET tests, the panel data model results seem
neither prone to an omitted variables bias nor to misspecified functional form of the
estimated relationship. This is not the case for the between regression in Model E, which
casts further doubt on these results. The reported Hausman and Taylor over-identification
tests do not reject the validity of our choice of instruments. Although the coefficient of G is
lower in the Hausman and Taylor models than in their fixed effects counterparts, the
difference in the parameters is small as compared to the efficiency gain. Hence, the
Hausman and Taylor specifications are not rejected at traditional significance levels.
In Model B, only the long run impact of corruption is significant. However, when
excluding outliers in Models C and D, both the short run and the additional long run
parameter estimates of corruption are significant. They amount to
0.05 and
0.42,
respectively. Hence, an additional score in the corruption index (i.e., less corruption)
implies an overall long run impact of about 100d (exp(
0.05
0.42)
1)=
37.5% in the
average country’s inward FDI stocks.12 The biased between regression results in Model E
erroneously support a negative nexus between corruption and FDI, being in line with
previous evidence (see Hines, 1995; Wei, 2000). Clearly, the Hausman and Taylor
estimates are preferred here. However, they point to a positive nexus between corruption
12
Although corruption exhibits a sizeable effect, we have to bear in mind two points: (i) the estimated effect is
predicted to hold ceteris paribus (i.e., holding everything else constant), and (ii) it represents a long run impact
rather than a short run effect.
 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952 941

Table 2
The impact of corruption on FDI stocks (dependent variable is log of inward FDI stocks)
Explanatory variablea Model A Model Bb Model C Model Db Model Ec
Fixed effects Hausman–Taylor FE-no HT-no Between
(FE) (HT) outliers outliers
Log lagged real GDP 2.395TT 1.129# 2.215TTT 1.021T 1.491TT
(0.962) (0.707) (0.690) (0.571) (0.737)
Log secondary
0.114
0.974 2.882 0.242TTT 3.282
school enrolment share (4.888) (4.177) (3.851) (0.052) (3.904)
Log lagged real 0.011 0.053
0.076 0.031
0.143
GDP  log secondary (0.197) (0.163) (0.150) (0.130) (0.172)
school enrolment
Legal quality 0.108# 0.182TTT 0.185TTT 0.369
0.177T
(0.073) (0.068) (0.056) (3.316) (0.100)
Corruption (short run impact)
0.028
0.026
0.053TT
0.052TT –
(0.030) (0.029) (0.023) (0.022)
Corruption (additional –
0.315T –
0.423TTT 0.285TTT
long run impact) – (0.179) – (0.144) (0.092)
Observations 326 326 318 318 68
Cross-sections 68 68 68 68 68
R2 0.978 0.742 0.999 0.823 0.879
Fixed country effectsd 26.94TTT – 48.67TTT – –
Hausman test: v 2 (5) 48.38TTT – 111.81TTT – –
Over-identification test: v 2(2) – 1.74 – 0.89 –
Average h – 0.95 – 0.95 –
Harmonic mean of – 0.61 – 0.62 –
canonical correlations
RESET test for omitted 0.18 – 0.52 – 1.88#
variable biase
White (1980) robust standard errors in parentheses.
a
Constant and country effects not reported for the sake of brevity.
b
Observations with errors within the 99% and the 1% percentile are excluded.
c
This corresponds to a cross-sectional regression on time-averaged data.
d
F(67,253) in (A) and F(67,245) in (C).
e
F(3,250) in (A) and F(3,242) in (C).
T Significant at 10%.
TT Significant at 5%.
TTT Significant at 1%.
#
Significant at 15%.

and inward FDI. This is at odds with prior empirical evidence, but it is in line with both the
dhelping handT argument and the theoretical hypotheses of Glass and Wu (2002) as
summarized in Section 2.13

13
This does not necessarily imply that we should also expect a positive long run impact of corruption for FDI at
the firm level. It can well be the case that more corruption is detrimental for a firm’s foreign investment at a
location given that the firm runs already a plant there. Nevertheless, it can lead to the entry of new investors that
would not have entered the market in the absence of helping hand type corruption ceteris paribus. The results in
Smarzynska and Wei (2000) support this view. Egger and Winner (2004) indicate that this is even the case within
country-pairs. However, the findings in Egger and Winner (2004) point to a declining short run impact of
corruption on FDI.
942 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

Country size (GNP) exhibits a positive overall impact, as expected (see Caves, 1996).
Further and in line with Markusen and Maskus (2002), we find a positive and significant
impact of secondary school enrolment on inward FDI in Model D, which supports the
model of horizontal FDI (see also Markusen and Venables, 2000). In contrast to Markusen
and Maskus (2002), we cannot identify any significant impact of the interaction term
between country size and the secondary school enrolment ratio. Hence, there is no
evidence for a dominance of vertical FDI in the sample at hand. Vertically organized
multinationals do their high-skilled labor intensive production and research in the
developed countries and locate their low-skilled labor intensive production stages in
countries, where low-skilled labor costs are low (see Caves, 1996). In all models but
Model D, the estimated (short run) effect of a change in the legal quality is significant and
amounts to about 0.15. This supports the hypothesis of a positive relationship between
legal quality and inward FDI.

4.3. Robustness

We undertake several robustness checks, always excluding outliers as in Model D. First,

performing a jackknife analysis (see Efron and Tibshirani, 1993) on Model D, we identify
the most influential countries with respect to maximum changes in the corruption coefficient
(see Table 3). The minimum (maximum) of the short run parameter is
0.064 (
0.043),
where the most influential country is Japan (Ireland). The additional long run parameter
takes a minimum (maximum) value of
0.478 (
0.390), and is caused by the exclusion of
Singapore (Ireland). Nevertheless, neither the short run nor the long run corruption
parameter changes its sign and in both cases the estimated average reported parameter in the
middle column of Table 3 lies in the center of the two identified extreme bounds.
Second, we investigate whether the positive long run effect of corruption on inward
FDI in the long run is sensitive to the definition of the dependent variable, the use of
alternative factor endowment measures, the inclusion of additional controls, and the choice
of the country sample. In the first block of experiments, we use inward FDI stocks per
capita (#1) and inward FDI stocks to GDP ratios (#2) instead of inward FDI stocks in
levels. Since inward FDI evolves generally much more dynamically than GDP and
population, it is not very surprising that the results are not sensitive to this choice. In the
second block of results, we use alternatively three different measures for a country’s

Table 3
Jackknife analysis
Explanatory variable Minimum coefficient Point estimate Maximum coefficient
Parameter Country Parameter Parameter Country
Log lagged real GDP 0.610 Australia 1.021 1.436 Malawi
Log secondary school enrolment share
2.112 Australia 0.369 2.850 Malawi
Log lagged real GDP  log
0.065 Malawi 0.031 0.132 Australia
secondary school enrolment
Legal quality 0.211 South Korea 0.242 0.280 Costa Rica
Corruption (short run impact)
0.064 Japan
0.052
0.043 Ireland
Corruption (additional long run impact)
0.478 Singapore
0.423
0.390 Ireland
Table 4
Robustness of long run and short run effects of corruption with respect to specification and sample choice
Explanatory variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Log lagged real GDP 0.236 0.167 7.267TTT 5.831TTT 0.481 0.327 1.513T 1.177T 1.311T 1.905T 0.744 0.896

P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

(1.021) (0.825) (1.646) (1.714) (1.155) (1.140) (0.892) (0.735) (0.800) (1.116) (0.690) (0.800)
Log secondary school
4.652 1.485 – – –
4.104
0.423
1.947
4.371
1.149
5.232
7.725T
enrolment share (6.030) (4.920) – – – (6.770) (5.297) (5.071) (4.702) (6.489) (3.901) (4.681)
Log lagged real 0.216
0.025 – – – 0.195 0.053 0.106 0.190 0.063 0.273T 0.308T
GDP  log secondary (0.234) (0.193) – – – (0.262) (0.206) (0.198) (0.183) (0.251) (0.152) (0.181)
school enrolment
Log real GDP per capita – –
22.451TTT – – – – – – – – –
– – (4.338) – – – – – – – – –
Log lagged real GDP  – – 0.938 TTT – – – – – – – – –
log real GDP per capita – – (0.171) – – – – – – – – –
Log human capital per capita – – –
3.810 – – – – – – – –
– – – (2.887) – – – – – – – –
Log lagged real – – – 0.222T – – – – – – – –
GDP  log human – – – (0.117) – – – – – – – –
capital per worker
Log capital–labor ratios
4.467 – – – – – – –
– – – – (2.935) – – – – – – –
Log lagged real – – – – 0.187T – – – – – – –
GDP  log – – – – (0114) – – – – – – –
capital–labor ratios
Legal quality 0.278TTT 0.202TT 0.006
0.022 0.086T 0.280TTT 0.251TTT 0.094 0.225TTT 0.274TTT 0.237TTT 0.209TTT
(0.091) (0.088) (0.050) (0.061) (0.052) (0.092) (0.084) (0.089) (0.070) (0.078) (0.059) (0.067)
Corruption (short
0.012
0.030
0.015
0.015
0.018
0.012
0.016 0.006
0.005 0.012
0.050T
0.004
run impact) (0.032) (0.028) (0.027) (0.026) (0.024) (0.032) (0.026) (0.041) (0.033) (0.034) (0.029) (0.032)
Corruption (additional long
0.458T
0.274T
0.455TT
0.691TTT
0.729TT
0.439T
0.645TTT
0.321T
0.685TTT
0.611TT
0.885TTT
0.648TT
run impact) (0.265) (0.159) (0.210) (0.193) (0.301) (0.261) (0.241)
(0.201) (0.264) (0.244) (0.231) (0.259)
Observations 250 264 285 285 360 323 261 224 296 265 244 302
R2 0.471 0.925 0.700 0.750 0.586 0.640 0.683 0.658 0.635 0.670 0.745 0.637
Over-identification test 1.74 0.76 1.35 0.01 1.81 2.14 1.12 1.67 2.06 2.20 0.56 2.09
(continued on next page)

943
 944
P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952
Table 4 (continued)
Explanatory variable (13) (14) (15) (16) (17) (18) (19) (20) (21)
Log lagged real GDP 0.790 0.461 1.371T 0.808 0.818 0.768 0.793 1.258 0.427
(0.682) (0.775) (0.786) (0.691) (0.685) (0.704) (0.711) (0.783) (0.400)
Log secondary school enrolment share
0.380
0.833
0.070
0.093
0.452
1.003
0.373
0.261 3.330
(3.968) (4.691) (4.196) (4.069) (3.977) (4.139) (4.137) (4.147) (2.280)
Log lagged real GDP  log secondary 0.050 0.077 0.035 0.039 0.052 0.074 0.050 0.045
0.128
school enrolment (0.157) (0.188) (0.166) (0.161) (0.157) (0.164) (0.16) (0.164) (0.091)
Change in log real GDP 0.267 0.168 0.710 0.253 0.293 0.343 0.298 0.624 0.377
(0.662) (0.704) (0.740) (0.666) (0.665) (0.676) (0.675) (0.740) (0.424)
Inflation
0.003
0.003
0.003
0.003
0.003
0.003
0.003
0.003 0.001
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.001)
Trade openness 0.009TTT 0.003 0.008TTT 0.009TTT 0.009TTT 0.009TTT 0.009TTT 0.008TTT 0.001
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.001)
Political accountability 0.000
0.017
0.004 0.000
0.001 0.001 0.000
0.006 0.030
(0.036) (0.047) (0.038) (0.036) (0.036) (0.036) (0.036) (0.038) (0.023)
External conflicts
0.025
0.025
0.022
0.025
0.024
0.023
0.025
0.021
0.009
(0.019) (0.022) (0.020) (0.019) (0.019) (0.019) (0.019) (0.020) (0.012)
Internal conflicts
0.026
0.029
0.020
0.026
0.025
0.026
0.026
0.020
(0.016) (0.018) (0.018) (0.017) (0.017) (0.017) (0.017) (0.018) (0.011)
Bureaucratic quality 0.005 0.034 0.000 0.005 0.006 0.009 0.008 0.006
0.021
(0.047) (0.055) (0.050) (0.047) (0.047) (0.050) (0.051) (0.050) (0.030)
Legal quality 0.217TTT 0.150TT 0.186TTT 0.216TTT 0.216TTT 0.217TTT 0.217TTT 0.098 0.067T
(0.064) (0.074) (0.070) (0.064) (0.064) (0.068) (0.068) (0.115) (0.041)
Share of literate adults – 0.008 – – – – – – –
– (0.012) – – – – – – –
Reduction in transboundary emission pressure – – 2.374T – – – – 2.153T
0.028
 – – (1.290) – – – – (1.274) (0.191)
Environmental governance – – – 0.562 – – – – –
– – – (0.937) – – – – –
Environmental sustainability – – – – 0.038 – – – –
– – – – (0.042) – – – –

P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

Membership in international environmental agreements – – – – –
0.318 – – –
– – – – – (0.375) – – –
Protected area in percent of total area – – – – – – 0.187 – –
– – – – – – (0.241) – –
Corruption (short run impact)
0.040
0.032
0.041
0.041
0.040
0.040
0.040
0.181
0.013
(0.025) (0.037) (0.027) (0.026) (0.026) (0.026) (0.026) (0.147) (0.016)
Corruption (additional long run impact)
0.338TT
0.322T
0.500TTT
0.458T
0.416TT
0.273T
0.356TT
0.620TTT
0.118T
(0.141) (0.194) (0.173) (0.269) (0.174) (0.158) (0.151) (0.215) (0.071)
Corruption  legal quality – – – – – – – 0.018 –
– – – – – – – (0.019) –
Observations 264 158 264 264 264 258 258 264 259
R2 0.880 0.948 0.866 0.880 0.880 0.883 0.883 0.876 0.091
Over-identification test 0.63 0.85 0.53 0.64 0.60 0.63 0.67 1.44 0.000
We use log FDI/capita in (1), log FDI/GDP in (2). We exclude the following country blocks from the sample: the ones in the lowest decile of the GDP distribution in (6),
the ones in the highest decile of the GDP distribution in (7), the EU in (8), NAFTA in (9), Africa in (10), Asia in (11), oil exporting countries according to the IMF
definition in (12).
Constant and country effects not reported for the sake of brevity. In (21), the dependent variable is the inward FDI flows to GDP ratio.
T Significant at 10%.
TT Significant at 5%.
TTT Significant at 1%.

945
946 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

relative endowments: GDP per capita (#3), a measure of the human capital stock used in
Baier et al. (2001) (#4), and capital–labor ratios (#5). It should be noted that the correlation
coefficients between these three measures are extremely high. Accordingly, it is not
surprising that the results are also robust in this regard. In the next block of exercises, we
consider the robustness of the corruption effect on inward FDI with respect to the
exclusion of different subsamples of countries: the ones in the lowest and highest deciles
of the GDP distribution in (#6) and (#7), respectively; the EU in (#8), NAFTA in (#9),
Africa in (#10), Asia in (#11), and the oil exporting countries according to the IMF
definition in (#12). The results are fairly robust in these regards. Hence, our main findings
are not driven by the Asian crisis or the presence of oil exporting economies in the sample.
In the second part of Table 4, we consider the relevance of including other control
variables for our findings. Before discussing the robustness of the long run corruption
effect, we should mention that trade openness and legal quality seem the most important
determinants in these specifications. Both of them tend to enter significantly. Further, in
the augmented models the standard errors are likely to be larger than in the parsimonious
ones (for instance, the correlation between lagged real GDP as a measure of market size
and trade openness renders the former insignificant in most specifications). In (#13) and
(#14) we augment our specification by other controls that were used by related
contributions (see Wei, 2000; Harms and Ursprung, 2002). Again, the negative long run
impact of corruption proves robust in this regard. In the remaining specifications, we
control for the impact of environmental regulation on FDI (see List and Co, 2000;
Damania et al., 2003; Fredriksson et al., 2003). We alternatively employ several proxies as
published in the World Economic Forum’s Global Competitiveness Report 2001 (see
World Economic Forum, 2001). Also in this case, we find a robust positive impact of
corruption on FDI.
Third, in specification (#20) we include an interaction effect between corruption and
legal quality. One might argue that the positive effect of corruption on FDI declines as
governance improves. Indeed, the positive parameter estimate of this variable points in this
direction. However, the parameter cannot be estimated significantly and we should be
careful with the interpretation. Finally, in (#21) we estimate the augmented model (#15)
using inward FDI flow as a share of GDP rather than log inward FDI stocks as the
dependent variable. Of course, the associated parameter is much smaller in absolute terms.
However, we again find support in favor of a positive long run impact of corruption in this
flow specification.
Fourth, we assess the sensitivity of our results with respect to the alternative
International Country Risk Guide (ICRG) and the World Bank’s corruption indices, using
Model D from Table 2 again. As Table 5 indicates, both indices suggest a pronounced
effect of corruption on inward FDI, whereas the other parameter estimates are almost
unchanged. For the ICRG index, both the short run and the additional long run impact of
corruption are positive and significant. We may conclude that the finding of a heavily
biased between parameter of corruption in aggregate FDI regressions with a tendency to
exhibit the wrong sign is invariant to alternative choices of the corruption measure
(compare columns 2 and 5 with columns 3 and 6 in Table 5). Again, these results confirm
the theoretical hypothesis of Lui (1985) and Glass and Wu (2002) and of corruption acting
as a dhelping handT for aggregate inward FDI.
 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952 947

Table 5
The impact of contract risk on FDI stocks. Alternative corruption indices (dependent variable is log of inward FDI
stocks)
Explanatory variablea ICRG corruption index World Bank corruption index (1997)
(1) (2) (3) (4) (5) (6)
FE Between HT FE Between HT
Log lagged real GDP 2.310TTT 1.239# 1.615TTT 2.408TTT 1.271T 1.855TTT
(0.396) (0.842) (0.377) (0.322) (0.742) (0.321)
Log secondary school enrolment share 3.861TT 2.335 3.500T 2.774# 2.259 2.579#
(2.065) (4.418) (1.977) (1.695) (3.942) (1.686)
Log lagged real GDP  log secondary
0.123#
0.092
0.098
0.082
0.097
0.065
school enrolment (0.085) (0.196) (0.082) (0.070) (0.174) (0.070)
Legal quality 0.030 0.064 0.047 0.044T
0.108 0.065TT
(0.033) (0.075) (0.032) (0.026) (0.098) (0.026)
Corruption (short run impact)
0.059TTT –
0.092TTT – – –
(0.022) – (0.021) – – –
Corruption (additional long run impact) – 0.061
0.704TT – 0.523T
1.163TTT
– (0.130) (0.357) – (0.280) (0.381)
Observations 315 70 315 393 72 393
Cross-sections 70 70 70 72 72 72
R2 0.994 0.852 0.823 0.993 0.871 0.766
Fixed country effectsb 80.29TTT – – 78.14TTT – –
Hausman test: v 2 (5) in (1) 44.77TTT – – 377.66TTT – –
and v 2 (4) in (3)
Over-identification test: v 2 (2) – – 1.28 – – 1.07
Average h – – 0.96 – – 0.97
Harmonic mean of canonical correlations – – 0.79 – – 0.95
RESET test for omitted variable biasc 0.75 1.98# – 0.17 1.80# –
White (1980) robust standard errors in parentheses.
a
Constant and country effects not reported for the sake of brevity. Observations with errors within the 99%
and the 1% percentile are excluded in all models.
b
F(67,253) in (1) and F(71,317) in (4).
c
F(3,237) in (1) and F(3,314) in (4).
T Significant at 10%.
TT Significant at 5%.*
TTT Significant at 1%.
#
Significant at 15%.

Finally, we undertake a Monte Carlo exercise on the impact of corruption. To take the
imprecision of the reported point estimates of the corruption index into account (see
Kaufmann et al., 1999, p. 10, on this issue), we randomly draw 1000 observations from
each country’s distribution of the corruption index, which is only available for the indices
published by Transparency International (1995–2001) and the World Bank (Kaufmann et
al., 2002).14 Since higher moments (skewness, kurtosis) are not available, we have to
assume that the underlying distribution the reported standard errors refer to is normal
(Table 6).

14
We use the dgraftT variable, which measures the perception of corruption defined as the use of public power
for private gain (see Kaufmann et al., 1999, p. 8).
948 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

Table 6
Monte Carlo results on the impact of corruption (using the reported standard errors of Transparency International
and World Bank corruption indices; 1000 replications)
Explanatory variablea Transparency International World Bank index (1997)
HT-no outliersa HT-no outliersa
Short run
Mean
0.013 –
Standard deviation 0.012 –
Minimum
0.057 –
Maximum 0.030 –
Additional run
Mean
0.419
1.260
Standard deviation 0.020 0.028
Minimum
0.505
1.369
Maximum
0.344
1.161
a
Obervations with errors within the 99% and the 1% percentile are excluded.

4.4. Simulation analysis

With our regression results of Model D in Table 2 at hand, we are able to investigate the
long run impact of the observed change in the corruption index scoring on both the growth
and the cross-country distribution of real stocks of FDI.
We can assess the long run impact of the observed change in the perceived
corruption level on real stocks of inward FDI from a growth-accounting-like exercise.
We compute the model prediction for FDI growth given the observed changes in the
explanatory variables (dbaselineT). Further, we compute the growth of FDI in a
situation where all variables except corruption are allowed to change as observed (i.e.,
assuming that the perceived corruption had not changed since the initial period;
dcounterfactualT). The difference in FDI growth between the dbaselineT and the

Asia
South and Central
America
Africa

OECD-Rest

NAFTA

CEEC

EFTA

EU

-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

Fig. 1. Long run impact of an observed annual change in corruption on inward FDI shares.
 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952 949

dcounterfactualT scenario can be viewed as the pure impact of corruption on real FDI
stocks. According to Model D, the overall long run impact of the change accounts for
about 40% of the overall predicted growth of FDI in the whole sample. The short run
effect accounts for about 5%.
Further, we assess the impact on the change in the country structure of real FDI stocks
on the basis of this difference between predicted and counterfactual FDI stocks. Fig. 1
summarizes the results.
Obviously, the change in corruption mainly has favored FDI shares in the EU and South
and Central American areas, mostly at the expense of FDI shares in the Asian and NAFTA
economies. In sum, the observed change in the perceived corruption had an equalization
effect on the international distribution of real inward FDI stocks. The change in corruption
alone accounts for a 1% long run increase in the entropy index15 of the distribution of FDI
stocks across country blocs.

5. Conclusions

Using data set of 73 developed and less developed countries, we find that corruption is
a stimulus for FDI, which confirms the position of Leff (1964) that corruption can be
beneficial in circumventing regulatory and administrative restrictions. The finding of a
positive short run and long run impact of corruption lends empirical support to existence of
the bhelping handQ type of corruption with regard to foreign investment. The short run
impact is considerably smaller than the long run counterpart, which we take to indicate a
short run bgrabbing handQ effect.
In a thought experiment, we considered the long run effects of the observed change
in corruption between 1995 and 1999 on inward FDI stocks. The contribution of the
change in perceived corruption in the long run may account for up to 40% of the
observed overall FDI growth between 1995 and 1999. Further, the observed change in
this variable has accounted for an equalization effect on the international distribution
of real inward FDI stocks, which was mainly in favor of FDI shares in the EU and
South and Central American areas and mostly at the expense of FDI shares in the
Asian and NAFTA nations. Our evidence suggests that the change in corruption is not
only able to explain part of the growth of FDI but also the change of its worldwide
distribution.
Finally, while larger perceived corruption appears associated with more direct
investment to low-income economies, our results, of course, should not be interpreted
as support for corrupt regimes. As Aidt (2003) and others point out, the socially most
beneficial policy is eliminating rather than circumventing corruption. The positive relation
between FDI and corruption suggests that administrative controls and bureaucratic
discretion is used to allow government officials to share in the profits from foreign
investment.

15 P
Defined as E=
ni si log si , with si as country bloc i’s share in overall stocks of inward FDI. E exhibits its
maximum with equally distributed shares.
950 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

Acknowledgements

We are grateful to two anonymous referees and the editors in charge (Heinrich
Ursprung and Arye Hillman) for their detailed and constructive comments. Financial
support from the Austrian Fonds zur Foerderung der wissenschaftlichen Forschung (FWF,
project no. P17028-G05) is gratefully acknowledged.

Appendix A. Country sample

EU Austria, Belgium, Denmark, Finland, France, Germany, Great

Britain, Greece, Ireland, Italy, Netherlands, Portugal, Spain, Sweden.
EFTA Iceland, Norway, Switzerland.
CEEC Albania, Bulgaria, Czech Republic, Georgia, Hungary, Poland, Romania,
Slovak Republic, Slovenia.
NAFTA Canada, Mexico, USA.
Rest of OECD Australia, Japan, New Zealand, South Korea, Turkey.
Africa Botswana, Cote d’Ivoire, Egypt, Kenya, Madagascar, Malawi, Mauritius,
Morocco, Namibia, Senegal, South Africa, Tanzania, Tunisia, Uganda.
Asia China, Hong Kong, India, Israel, Jordan, Malaysia, Philippines, Singapore,
Sri Lanka, Thailand.
South and Central America Argentina, Brazil, Chile, Colombia, Costa Rica, Dominican Republic,
El Salvador, Guatemala, Nicaragua, Panama, Paraguay, Peru, Trinidad
and Tobago, Uruguay, Venezuela.

References

Abed, G.T., Gupta, S. (Eds.), 2002. Governance, Corruption, and Economic Performance. International Monetary
Fund, Washington, DC.
Aidt T.S., 2003. Economic analysis of corruption: a survey. Economic Journal 113, F632 – F652.
Akcay S., 2001. Is corruption an obstacle for foreign investors in developing countries? Cross-country evidence.
Yap Kredi Economic Review 12, 27 – 34.
Applebaum E., Katz E., 1987. Seeking rents by setting rents: the political economy of rent seeking. Economic
Journal 97, 685 – 699.
Bahmani-Oskooee M., Nasir A., 2002. Corruption, law and order, bureaucracy, and the real exchange rate.
Economic Development and Cultural Change 50, 1021 – 1028.
Baier, S.L., Dwyer, G.P., Tamura, R., 2001. How important are capital and total factor productivity for economic
growth. Unpublished working paper. Clemson University.
Baltagi B.H., 2001. Econometric Analysis of Panel Data, 2nd ed. Wiley & Sons, Chichester.
Bardhan P., 1997. Corruption and development: a review of issues. Journal of Economic Literature 35,
1320 – 1346.
Bhagwati J., 1968. Distortions and immiserizing growth: a generalization. Review of Economic Studies 35,
481 – 485.
Bliss C., Di Tella R., 1997. Does competition kill corruption. Journal of Political Economy 105, 1001 – 1023.
Blonigen B.A., Davies R.B., 2004. The effects of bilateral tax treaties on U.S. FDI activity. International Tax and
Public Finance 11, 601 – 622.
Blonigen B.A., Davies R.B., Head K., 2003. Estimating the knowledge capital model of FDI: a comment.
American Economic Review 93, 980 – 994.
 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952 951

Boycko M., Shleifer A., Vishny R.W., 1995. Privatizing Russia. MIT Press, Cambridge MA.
Brainard L.S., 1993. A simple theory of multinational corporations and trade with a trade-off between proximity
and concentration. NBER Working Paper, vol. 4269. National Bureau of Economic Research, Cambridge, MA.
Brainard L.S., 1997. An empirical assessment of the proximity-concentration trade-off between multinational
sales and trade and concentration. American Economic Review 87, 520 – 544.
Caves R.E., 1996. Multinational Enterprise and Economic Analysis, 2nd ed. Cambridge University Press,
Cambridge.
Clarke G.R.G., Xu L.C., 2004. Privatization, competition, and corruption: how characteristics of bribe takers and
payers affect bribes to utilities. Journal of Public Economics 88, 2067 – 2097.
Cornwell C., Schmidt P., Wyhowski D., 1992. Simultaneous equations and panel data. Journal of Econometrics
51, 151 – 181.
Cukierman A., Edwards S., Tabellini G., 1992. Seignorage and political instability. American Economic Review
82, 537 – 555.
Dalgic, E., van Long, N., in press. Corrupt local government as resource farmers: the helping hand and the
grabbing hand. European Journal of Political Economy.
Damania R., Fredriksson P.G., List J.A., 2003. Trade liberalization, corruption, and environmental policy
formation: theory and evidence. Journal of Environmental Economics and Management 46, 490 – 512.
de Haan J., Sturm J.E., 2000. On the relationship between economic freedom and economic growth. European
Journal of Political Economy 16, 215 – 241.
Easterly W., 2001. The Elusive Quest for Growth: Economists Advantages and Misadvantages in the Topics. MIT
Press, Cambridge MA.
Efron B., Tibshirani R.J., 1993. An Introduction to the Bootstrap. Chapman & Hall, New York.
Egger P., Pfaffermayr M., 2004. Distance, trade and FDI: a Hausman–Taylor SUR approach. Journal of Applied
Econometrics 19, 227 – 246.
Egger P., Winner H., 2003. Does contract risk impede foreign direct investment? Swiss Journal of Economics and
Statistics 139, 155 – 172.
Egger, P., Winner, H., 2004. How corruption influences FDI: a panel data study, Unpublished working paper.
University of Innsbruck.
Ehrlich I., Lui F.T., 1999. Bureaucratic corruption and endogenous economic growth. Journal of Political
Economy 107, 270 – 293.
Fredriksson P.G., List J.A., Millimet D.L., 2003. Bureaucratic corruption, environmental policy and inbound US
FDI: theory and evidence. Journal of Public Economics 87, 1407 – 1430.
Glass, A.J., Wu, X., 2002. Does corruption discourage foreign direct investment and innovation? Unpublished
working paper. Texas A&M University.
Grossman G.M., Helpman E., 1991. Quality ladders and product cycles. Quarterly Journal of Economics 106,
556 – 586.
Gupta S., Davoodi H., Alonso-Terme R., 2002. Does corruption affect income inequality and poverty? Economics
of Governance 3, 23 – 45.
Gwartney J.D., Lawson R.A., 2003. The concept and measurement of economic freedom. European Journal of
Political Economy 19, 405 – 430.
Gwartney J.D., Lawson R.A., Block W., 1996. Economic Freedom of the World 1975–1995. The Fraser Insitute,
Vancouver.
Gwartney, J.D., Lawson, R.A., Samida, D., 2000. Economic Freedom of the World: 2000 Annual Report. The
Fraser Institute, Vancouver.
Habib M., Zurawicki L., 2002. Corruption and foreign direct investment. Journal of International Business
Studies 33, 291 – 307.
Harms P., Ursprung H.W., 2002. Do civil and political repression really boost foreign direct investments?
Economic Inquiry 40, 651 – 663.
Hausman J.A., Taylor W.E., 1981. Panel data and unobservable individual effects. Econometrica 49, 1377 – 1398.
Hayashi F., 2001. Econometrics. Princeton University Press, Princeton.
Hines J.R., 1995. Forbidden payment: foreign bribery and American business after 1977. NBER Working Paper
vol. 5266. National Bureau of Economic Research, Cambridge, MA.
Jain A.K., 2001. Corruption: a review. Journal of Economic Surveys 15, 71 – 121.
952 P. Egger, H. Winner / European Journal of Political Economy 21 (2005) 932–952

Johnson B.T., Holmes K.R., Kirkpatrick M., 1998. The Index of Economic Freedom. The Heritage Foundation,
Washington, DC.
Kaufmann D., Kraay A., Zoido-Labatón P., 1999. Governance matters. Policy Research Working Paper, vol.
2196. World Bank, Washington, DC.
Kaufmann D., Kraay A., Zoido-Labatón P., 2002. Governance matters: II. Updated Indicators for 2000/01. Policy
Research Working Paper vol. 2772. World Bank, Washington, DC.
Keller W., 2000. Do trade patterns and technology flows affect productivity growth? The World Bank Economic
Review 14, 17 – 47.
Lambsdorff J., 2003. How corruption affects productivity. Kyklos 56, 457 – 474.
Leff N., 1964. Economic development through bureaucratic corruption. American Behavioral Scientist 8, 8 – 14.
List J.A., Co C.Y., 2000. The effects of environmental regulations on foreign direct investment. Journal of
Environmental Economics and Management 40, 1 – 20.
Lui F.T., 1985. An equilibrium queuing model of bribery. Journal of Political Economy 93, 760 – 781.
Markusen J.R., 2002. Multinational Firms and the Theory of International Trade. MIT Press, Boston.
Markusen J.R., Maskus K.E., 2002. Discriminating among alternative theories of the multinational enterprise.
Review of International Economics 10, 694 – 707.
Markusen J.R., Venables A.J., 2000. The theory of endowment, intra-industry and multi-national trade. Journal of
International Economics 52, 209 – 234.
Mauro P., 1995. Corruption and growth. Quarterly Journal of Economics 110, 681 – 712.
Mauro P., 1998. Corruption and the composition of government expenditure. Journal of Public Economics 69,
263 – 279.
Mocan N., 2004. What determines corruption? International evidence from microdata. NBER Working Paper, vol.
10460. National Bureau of Economic Research, Cambridge, MA.
Murphy K.M., Shleifer A., Vishny R.W., 1991. The allocation of talent: implications for growth. Quarterly
Journal of Economics 106, 503 – 530.
Olson M., 1993. Dictatorship, democracy, and development. American Political Science Review 87, 567 – 576.
Olson M., 2000. Power and Prosperity: Outgrowing Communist and Capitalist Dictatorships. Basic Books, New
York.
Paldam M., 2002. The cross-country pattern of corruption: economics, culture and the seesaw dynamics.
European Journal of Political Economy 18, 215 – 240.
Pirotte A., 1999. Convergence of the static estimation toward the long run effects of dynamic panel data models.
Economics Letters 63, 151 – 158.
Rose-Ackermann S., 1999. Corruption and Government. Causes, Consequences, and Reform. Cambridge
University Press, Cambridge UK.
Schneider F., Frey B.S., 1985. Economic and political determinants of foreign direct investment. World
Development 13, 161 – 175.
Shleifer A., Vishny R.W., 1993. Corruption. Quarterly Journal of Economics 108, 599–617.
Smarzynska B., Wei S.-J., 2000. Corruption and composition of foreign direct investment. Discussion Paper, vol.
2360. World Bank, Washington, DC.
Tanzi V., 1998. Corruption around the world. IMF Staff Papers 45, 559 – 594.
Tanzi, V. Davoodi, H., 2000. Corruption, growth, and public finances. Working Paper No. 00/182. International
Monetary Fund, Washington, DC.
The Economic Freedom Network, 2001. Economic Freedom of the World: 2001 Annual Report. http://
www.freetheworld.com/.
Transparency International, 2001. Global Corruption Report 2001. http://www.globalcorruptionreport.org/.
Treisman D., 2000. The causes of corruption: a cross-national study. Journal of Public Economics 76, 399 – 457.
Wei S.-J., 2000. How taxing is corruption on international investors? Review of Economics and Statistics 82,
1 – 11.
Wheeler D., Mody A., 1992. International investment location decisions: the case of U.S. firms. Journal of
International Economics 33, 57 – 76.
White H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for hetero-
skedasticity. Econometrica 48, 817 – 838.
World Economic Forum, 2001. 2001 Environmental Sustainability Index. World Economic Forum, Davos.