Corrup SD Threshold Non-Linear Analysis EAP2023

Economic Analysis and Policy 78 (2023) 505–523
Contents lists available at ScienceDirect
Economic Analysis and Policy

journal homepage: www.elsevier.com/locate/eap
Analyses of topical policy issues
How does corruption affect sustainable development?

A threshold non-linear analysis
∗
Fredj Fhima a , Ridha Nouira b , Khalid Sekkat c ,
a
University of Sousse, IHEC, LaREMFiQ, Route de la Ceinture Sahloul 3 BP40, 4054 Sousse, Tunisia
b
University Sousse, ISFF, LAMIDED, Avenue 18 janvier 1952, BP436, 4000 Sousse, Tunisia
c
University of Brussels, CP 114/03, 50 Avenue F.D. Roosevelt, 1050 Brussels, Belgium
article info a b s t r a c t
Article history: This paper investigates the impact of corruption on sustainable development using the
Received 13 August 2022 Seo and Shin (2016) threshold model for a panel data set of 96 to 103 developed and
Received in revised form 20 March 2023 developing countries from 1996 to 2019. The paper contributes to the ongoing de-
Accepted 26 March 2023
bate about whether corruption greases or sands the wheels of sustainable development.
Available online 31 March 2023
All results support the hypothesis of the existence of thresholds. Examining the full
JEL classification: sample, the impact of corruption on sustainable development is regime-specific and
D73 dependent on the quality of governance. In developed countries, corruption always hin-
G3 ders sustainable development, while in developing countries the impact of corruption on
O47 sustainable development is regime-specific and dependent on the quality of governance.
Q01
When the quality of governance is low, highly corrupt countries achieve higher levels of
C24
sustainable development, which is consistent with the ‘‘grease the wheels’’ hypothesis.
Keywords: © 2023 Economic Society of Australia, Queensland. Published by Elsevier B.V. All rights
Corruption reserved.
Governance
Sustainable development
Threshold models
1. Introduction
Corruption, commonly defined as the misuse of public office for private gains (Rose-Ackerman, 2006), is recognized as
immoral and unethical.1 It is seen, moreover, as a hindrance to social and economic well-being (e.g. Mauro, 1995; Wei,
2000; Lambsdorff, 2003; Reinikka and Svensson, 2005; Aidt, 2009; Uberti, 2022). National and international organizations
such as the OECD, the United Nations, the International Court of Justice, the World Bank and the International Monetary
Fund, recommend fighting corruption. This view is supported by the proponents of the ‘‘sand in the wheels’’ consequences
of corruption. This detrimental effect of corruption is, however, not unanimously accepted. There are proponents of a
positive effect of corruption; their ‘‘grease the wheels’’ theory argues that corruption may be beneficial in a ‘‘second best
world’’ by alleviating the distortions caused by ill-functioning institutions (e.g. Leff, 1964; Huntington, 1968; Lui, 1985;
Acemoglu and Verdier, 1998; Méon and Weill, 2010; Stojanović et al., 2016). The outcomes from the empirical literature
are varied and, in some instances, contradictory.
∗ Corresponding author.
E-mail addresses: [email protected] (F. Fhima), [email protected], [email protected] (R. Nouira),
[email protected] (K. Sekkat).
1 For a detailed discussion of the various definitions, see Sekkat (2018).
https://doi.org/10.1016/j.eap.2023.03.020
0313-5926/© 2023 Economic Society of Australia, Queensland. Published by Elsevier B.V. All rights reserved.
F. Fhima, R. Nouira and K. Sekkat Economic Analysis and Policy 78 (2023) 505–523
This variation in outcomes relates in part to differences in methodologies, definitions and interpretations. For example,
various studies in the existing literature use time series data, others employ panel data, and yet others rely on cross-
national data. In addition, these works examine different countries over different periods of time.2 Furthermore, the
relationship between corruption and growth is not clear-cut and involves complex and sometimes conflicting mechanisms
(Cariolle, 2018, p. 8), which induces a non-linearity of the effect of corruption (Swaleheen, 2011, pp. 23–24). For instance,
Shleifer and Vishny (1993) argue that the challenge may arise from the organization of a given corruption market. This
organization depends, among others, on the number of actors, their potential coalitions and the complementarity of the
tasks each of the actors is in charge of. In particular, weak governments may not be able to efficiently control corruption,
which allows the phenomenon to prosper. In the same vein, Olson (2000) shows that when corrupt officials belong
to a small, well organized and privileged group, the detrimental effect of corruption is greater than when the corrupt
officials belong to large, latent groups that face difficulties organizing toward successful collective action. Rock and Bonnett
(2004) explain the so called ‘‘East Asian paradox’’ of high corruption and high growth in terms of stable and mutually
beneficial exchanges of government promotional privileges for bribes and kickbacks. Thus, with an alternate organization
of corruption, the impact of corruption on growth will be different as we move along a corruption index, which may
lead to a non-linearity in the corruption-growth nexus. Therefore, the non-linear effect of corruption-growth in a multi-
country framework should not be ruled out. In this paper, we investigate whether the quality of governance may lead to
a non-linearity in the effect of corruption on sustainable development. Depending on a country’s quality of governance,
the effect of corruption on sustainable development differs. In other words, there may exist a threshold of governance
quality that determines whether corruption has a positive or a negative effect on sustainable development.
This paper contributes to a relatively recent and still limited empirical literature on nonlinearities in the effect of
corruption. At the economic level, this paper widens the scope of study to examine the effect of corruption on three
dimensions that are important for human well-being. While the preceding studies focused mainly on growth, here we
also examine the impact of corruption on ‘‘human development’’ and ‘‘environmental performance’’. At the methodological
level, we improve on existing studies by using the threshold model proposed by Seo and Shin (2016). Méon and Sekkat
(2005) use an interaction term between corruption and the quality of governance to highlight such non-linearity. In
contrast, we employ a threshold methodology that allows all parameters of the model to change across regimes, while
Méon and Sekkat (2005)’s methodology pre-supposes that only the corruption coefficient is a function of the index used
to measure the quality of governance. Aidt et al. (2008) use the method proposed by Caner and Hansen (2004) to examine
the role of political accountability as a determinant of non-linearity. In contrast to Aidt et al. (2008) who exogenously set
the level of the threshold, we let the data determine the threshold endogenously using Seo and Shin (2016)’s approach.
This approach allows us to conduct the sample-splitting and threshold estimation in a panel data set-up, while the method
of threshold estimation of Caner and Hansen (2004) is only available for cross-sectional analysis (see Aidt et al., 2008, p.
207).
The results using the whole sample of countries show that thresholds exist and depend on the levels of development
and of the quality of governance. For developed countries, the impact of corruption on sustainable development is always
harmful. For developing countries, the results indicate that the impact of corruption on sustainable development is regime-
specific and dependent on the quality of governance. When the quality of governance is low, highly corrupt countries
have higher levels of sustainable development. Comparison of our findings with of the international literature can only
concern growth because the effects of corruption on ‘‘human development’’ and ‘‘environmental performance’’ has not
been examined enough. Our findings are in line with Méon and Weill (2010) who report that corruption has a detrimental
effect on productivity in economies with effective institutions but this effect may be positive in economies with ineffective
institutions. In the same vein, Aidt et al. (2008) show that the relationship between corruption and growth is regime-
specific. Swaleheen (2011) also establishes that the effect of corruption on growth is non-linear. It is negative but not at
all levels of corruption. In contrast, Méon and Sekkat (2005) find that impact of corruption on growth tends to worsen
when the quality of governance deteriorates.
The rest of the paper is organized as follows. Section 2 presents an overview of the literature on corruption’s effects.
Section 3 describes the empirical strategy. Section 4 presents and discusses the empirical results. Section 5 concludes and
offers recommendations and directions for future research.
2. Literature review
Following the United Nations, sustainable development can be approximated by ‘‘human development’’ (Absalyamova
et al., 2016; Murshed and Mredula, 2018) and ‘‘environmental performance’’ (Emerson et al., 2012). The former is defined
as ‘‘expanding the choices people have to lead lives that they value’’ (Human Development Report, 2001, p. 9). It is often
measured by the Human Development Index (HDI) which published by the United Nations Development Program and is
a summary of the average achievement in three dimensions: (1) living a long and healthy life, (2) attaining significant
education and (3) having a decent standard of living and access to resources. The second piece of sustainable development,
environmental performance, is defined as ‘‘the commitment of firms to protect the environment and to demonstrate
2 Time-invariant heterogeneity among countries in terms of culture, history, religion, colonial past, institutions, development and income per
capita exerts an important effect that explains cross-country differences in the prevalence of corruption (Andvig, 2006).
506
measurable operational parameters that are within the prescribed limits of environmental care’’ (Olayeni et al., 2021, p.
3). It is often measured by the Environmental Performance Index (EPI) published by the Center for Environmental Law and
Policy of Yale University, in collaboration with Columbia University’s Center for International Earth Science Information
Network (CIESIN) and the World Economic Forum (WEF).3 The EPI ranks 180 countries on 24 performance indicators,
including reduction in environmental stresses to human health and the protection of ecosystems and natural resources
(Emerson et al., 2012).
In the rest of this section, we will discuss the empirical evidence linking corruption and human development before
turning to corruption’s impact on environmental performance. We end with evidence regarding the issue of non-linearity.
2.1. The effect of corruption on the Human Development Index (HDI)
Corruption directly affects the HDI by influencing its dimensions (poverty, education and health) and indirectly by
impacting the determinants of these dimensions (e.g., investment and economic growth) (Akçay, 2006). The pioneering
work in the field is Mauro (1998), which reveals a negative, significant and robust relationship between corruption and
government spending on education and health. In line with Mauro (1998), Mo (2001) finds a negative effect of corruption
on human capital, growth and private investment. Gupta et al. (2002) focus on the ways corruption could negatively
affect income distribution and poverty. Using several measures of corruption and the Gini Index as a measure of income
inequality, they find that corruption not only directly reduces income growth of the poor, but it also indirectly impacts
income growth through social spending on health and education. Regardless of level of development, highly corrupt
countries tend to have lower levels of social spending, higher education inequality, lower secondary schooling attainment
and unequal distribution of land. The tests in Gupta et al. (2002) show that a decrease in a country’s corruption index
by 2.5 points on a scale of 0 to 10 is associated with the same increase in the Gini coefficient (indicating worsening
inequality) and a reduction in average secondary schooling by 2.3 years. Corruption increases the cost of education for
students and burdens their families, which in turn leads to lower enrollment rates. Justesen and Bjørnskov (2014) show
that corruption perpetuates poverty, and that compared to rich citizens, poor individuals are 2.5 times more likely to be
the victims of bureaucratic corruption.
In a new study focusing on China (one of the most debated cases in the literature), Han et al. (2022) use five waves of
the China Family Panel Studies (CFPS) data to examine the effects of the latest anti-corruption campaign led by Chinese
authorities on poverty alleviation. They find that in the counties most affected by the anti-corruption campaign, poor
households are associated with a lower probability of remaining in poverty after the campaign. More access to transfer
payments and less expropriation of government funds are the key factors that explain the reduction in poverty following
the anti-corruption campaign. In a study on Indonesia, Iskandar and Saragih (2018) find that corruption tends to be
associated with less social spending because corruption reduces available government resources for the provision of social
services, including education. According to Iskandar and Saragih (2018), corruption leads to inappropriate tax exemptions
and to tax evasion, thus deteriorating tax administration and decreasing tax revenues.
Studies focusing on the health sector show that besides the issue of state capacity, specific features such as uncertainty,
asymmetric information and large numbers of dispersed actors make the sector particularly susceptible to corruption.
Gupta et al. (2002)’s empirical analyses reveal that these factors both increase the cost of health care and worsen its
quality. Furthermore, corruption adversely affects a country’s child and infant mortality rates and the percentage of low-
birth weight babies in relation to total births, and it increases dropout rates in primary school. Hanf et al. (2011, p. 5)
point out that in developing countries, about 80% of health problems can be linked to substandard water and sanitation
services, with corruption likely being a main contributor to these issues. Ortega et al. (2014) find that the negative effect of
corruption on human development seems to play mainly through the health component of the HDI. Siverson and Johnson
(2014) also find that corruption is detrimental to health outcomes and associated with decreased life expectancy. This
conclusion is confirmed by Lio and Lee (2016), who write ‘‘more than the waste of money, corruption costs lives!’’ More
recently, Achim et al. (2019) find that corruption negatively and significantly affects not only physical but also mental
health.
The empirical literature strongly supports the argument that corruption hampers human development, and this is
verified directly in works using the HDI and indirectly in works examining the specific indicators of poverty, education
and health.
2.2. The effect of corruption on environmental performance
Existing literature distinguishes between the direct effect of corruption on environmental performance and an
indirect effect through corruption’s impact on per capita income and the resulting impact of income on environmental
pollution. The direct effect stems from the type of governmental regulation related to environmental performance. Desai
(1998) argues that corruption is a major source of environmental degradation. In some countries (e.g., India, Thailand
and Indonesia), the power of vested interests by officials heavily impacts both the drafting and the enforcement of
3 See also the World Bank Climate Change Knowledge Portal; World Bank: Washington, DC, USA, 2021; Available online: https://climateknowledge
portal.worldbank.org/download-data
507
environmental legislation. Damania et al. (2003) focus on lead content in gasoline and find that corruption weakens
environmental policies. Fredriksson et al. (2004) show that the stringency of energy policy is negatively related to
government corruptibility and positively related to lobbying costs.
The ‘‘Pollution Haven Hypothesis (PHH)’’ (Copeland and Taylor, 1994) claims that as a result of trade and investment
liberalizing, pollution-intensive industries move from developed countries to developing countries with lax environmental
regulations. In this regard, Candau and Dienesch (2017) find evidence in support of the PHH, especially for developing and
emerging countries with intermediate levels of corruption. Corruption lowers environmental regulations, which strongly
attracts ‘‘dirty’’ firms. However, in highly corrupt countries and countries with very little corruption, the PHH does not
apply.
Other papers examine the link between corruption and CO2 emissions. Akhbari and Nejati (2019) investigate this issue
using a threshold model based on the HDI. They show that in countries situated below the HDI threshold, a decrease in
corruption decreases carbon emission, while in countries above the threshold, corruption no longer influences carbon
emission levels. Khan et al. (2021) establish that institutional quality, including corruption, has a significant effect on
carbon emissions. Good (poor) governance systems enhance (deteriorate) environmental quality. Similar findings are
reported by Zhang et al. (2016) for Asia-Pacific Economic Cooperation (APEC) economies, Sinha et al. (2019) for BRICS
and Ganda (2020) for the Southern African region. Lv and Gao (2021) use spatial econometrics to assess the role of spatial
dependence in the effect of corruption on environmental performance. They find highly statistically significant results
showing that the direct, indirect and total effects of corruption on environmental performance are negative and that
there is a significant negative spatial spillover effect between corruption and environmental performance. Their results
also show that being surrounded by highly corrupt countries worsens the local country’s environmental performance.
Country-specific studies tend to confirm the above findings. For instance, Dincer and Fredriksson (2018) investigate
the relationship between environmental policy, corruption and trust in the 48 states of the continental USA and find
that higher corruption reduces the stringency of environmental policies when the level of trust is low, but the effect
declines and even becomes positive at higher levels of trust. Chen et al. (2018) examine the impact of environmental
regulation, the shadow economy and corruption on environment quality in 30 Chinese provinces and show that there is
a close relationship between corruption, environmental regulation and pollution. Moreover, an increase in the number
of corrupt officials results in weakened environmental regulations, which therefore generates higher illegal production
along with higher total pollutant emissions. Wang et al. (2020) use panel data covering 29 Chinese provinces and show
that government corruption directly reduces ecological efficiency.
2.3. Sand or grease the wheels?
The above studies suggest that the effect of corruption on sustainable development is likely to be negative. However,
this conclusion is not unanimous and is in fact contradicted by the ‘‘grease the wheels’’ theory of corruption. First proposed
by Leff (1964), the theory argues that corruption may be beneficial in a ‘‘second best world’’. In countries with a low quality
of governance, corruption can reduce the inconvenience of this low quality. Hence, the core of the argument in favor of
the ‘‘grease the wheels’’ hypothesis lies in the combination of corruption with a low quality of governance, which may
be marked by sluggish or tedious bureaucratic regulations. For instance, Leys (1965) and Lui (1985) show that corruption
could efficiently lessen the time spent in queues and could enhance efficiency by giving bureaucrats an incentive to speed
up the process. Huntington (1968) argues that in the USA during the 1870s and 1880s, corruption by railroad, utility and
industrial corporations resulted in faster growth. He suggests that corruption may improve the likelihood that officials
and governments make the right choice. In cases where these organizational bodies do not have enough information or
are not competent to make some of these decisions, corruption can replicate the outcome of a competitive auction (Beck
and Mahler, 1986; Lien, 1986). Hence, using these arguments, one can argue that the relationship between corruption,
economic growth and sustainable development may be non-linear, depending on the quality of governance. The purpose
of this paper is to rigorously test this hypothesis. The methods used are discussed below, and the merits of our approach
are discussed in the introduction.
Méon and Sekkat (2005) use various proxies for both corruption and governance to explicitly test for non-linearity (or,
alternatively, the ‘‘grease the wheels’’ versus ‘‘sand in the wheels’’ hypotheses) in the impact of corruption on growth.
They find that impact tends to worsen when indicators of the quality of governance deteriorate. In a similar study, Méon
and Weill (2010) analyze the interaction between aggregate efficiency, corruption and various dimensions of governance.
Contrary to Méon and Sekkat (2005), they report that corruption has a detrimental effect on efficiency in economies
with effective institutions but may be positively associated with efficiency in economies with ineffective institutions.
This finding supports the ‘‘grease the wheels’’ hypothesis and suggests a non-linear relationship between corruption and
economic growth. Aidt et al. (2008) construct a composite index of the quality of institutions, which they then use in
a threshold model where corruption is potentially endogenous. They define two governance regimes and show that the
relationship between corruption and growth is regime-specific. Swaleheen (2011) tests a non-linear model where squared
corruption is introduced as an explanatory variable. He finds that the effect of corruption is non-linear. There is a negative
effect on growth but not at all levels of corruption. In countries with high levels of corruption, the effect on growth is
actually significantly negative. Finally, Bose et al. (2008) focus on the quality of public infrastructure and allow for the
existence of a threshold effect in their analysis. Their results suggest two distinct regimes following corruption level. In the
first regime, the incidence of corruption is high, and its effect on the quality of public infrastructure is strongly negative.
508
In the second regime with relatively low levels of corruption, the effect of corruption is neutral or perhaps even slightly
positive.
3. The empirical strategy
3.1. Framework
In this paper, we test the hypothesis that the impact of corruption on sustainable development as measured by the
three key variables of growth, human development and environmental performance is non-linear depending on the
quality of governance. In other words, we examine whether different levels of governance quality affect the impact of
corruption on sustainable development. To this end, we employ the fixed-effect panel threshold model proposed by Seo
and Shin (2016). Under the null hypothesis of the absence of the threshold effect, the impact of corruption on sustainable
development does not depend on the quality of governance. To check the validity of the null hypothesis, we report the
test results for the null hypothesis in Tables 2 to 7.
The model advanced by Seo and Shin (2016) has the following advantages over alternative approaches. First, it provides
insight into the importance of threshold regression methods for dynamic panels with individual-specific fixed effects,
thus allowing the analysis of the effect of a variable (corruption in our study) under different regimes of another variable
(governance in our study). The threshold model can, for instance, show that the effect of corruption varies among countries
of different development levels or among countries with different levels of quality of governance. Second, this approach
does not exogenously set the level of the threshold; rather, it lets the data endogenously determine its level and standard
deviation (Aidt et al., 2008). Third, this approach allows more than one coefficient in the relationship to change with the
threshold. Fourth, the study builds on previous analyses, which almost exclusively focus on growth, by considering other
indicators of human well-being: human development and environmental performance. We think that such an approach
is more suitable to revealing valuable insights into the impact of corruption on human well-being. Finally, the method
improves on the widely used panel model with threshold effect proposed by Hansen (1999). Hansen (1999)’s model is
static and requires the covariates to be strongly exogenous for the estimator to be consistent. Seo and Shin (2016) propose
a GMM estimator in a dynamic panel threshold model with a potentially endogenous threshold variable.
To the best of our knowledge, our study is the first to employ the threshold approach to examine the existence of
contingency on the effect of corruption. We think that such an approach is more suitable to revealing valuable insights
about the impact of corruption on human well-being. In formal terms, the model empirically examines if the relationship
between variables x and y depends on whether a variable z is below or above a threshold level z. To illustrate, consider
the following basic regression:
yit = µi + β ′ (z ) xit + eit (1)
where β is the estimated coefficient and xit is a vector of explanatory variables. If the relationship between yit and xit
is linear, z can be dropped, and a traditional estimation of (1) gives the best fit of the data. However, if the true model
is believed to be non-linear, the econometric analysis should adopt a method that allows testing for non-linearity. If the
test confirms the non-linearity, an estimation of the threshold and an assessment of its significance should be conducted.
Seo and Shin (2016) propose the following dynamic panel threshold model:
yit = Xit′ β + 1, Xit′ δ l {qit > γ } + µi + εit i= 1, . . ..N ; t = 1, . . .T

( )
(2)
where Xit may include lagged dependent variables and qit is the threshold variable. γ is the threshold parameter, l (.)
is an indicator function, and δ and β are the slope parameters. We remove the parameter µi with the first-difference
transformation and estimate the unknown parameters through two types of estimators: first-differenced two-step least
squares (FD-2SLS) and first-differenced GMM (FD-GMM). In this paper, we adopt the first-differenced generalized method
of moments (GMM) estimators. The distinguishing feature of the proposed FD-GMM and FD-2SLS approaches, as compared
to the approach used by Hansen (1999), is that they allow both threshold variable and regressors to be endogenous. The
null hypothesis of this test is:
H0 : δ0 = 0 for any γ ∈ Γ
for any where Γ denotes the parameter space for γ .

Against the alternative:
H1 : δ0 ̸ = 0 for some γ ∈ Γ
for some
Testing whether the model is non-linear is accomplished using the statistic for the null hypothesis, H0 :
supW = SupWn (γ ) , (3)

γ ∈Γ
509
Wn (γ ) is the standard Wald statistic for each fixed γ . According to Seo and Shin (2016), the limiting distribution of this
statistic is not asymptotically pivotal, and critical values cannot be tabulated. We bootstrap critical values or p-values.
3.2. Model
Our model considers three key dependent variables. In addition to human development and environmental perfor-
mance, we add a growth equation for comparison with the literature.
In order to examine whether there is a threshold effect between corruption and sustainable development, we follow the
literature (e.g. Zhu et al., 2020; Asimakopoulos and Karavias, 2016) and focus only on the coefficient of interest, although
many thresholds may exist. In our case, we are interested in the change in the effect of corruption with the threshold
while the other coefficients remain unchanged.
According to Seo and Shin (2016), the model can be written as:
Growthit = αi0 + α1 LogGDPpcit −1 + α2 LogSchoolit + α3 LogInvit + α4 Popit +

α5 LogOpenit + α6L Corruit l {Govit ≤ γ } + α6U Corruit l {Govit > γ } + εit (4)
LogHDIit = βi0 + β1 Economicfreedomit + β2 Democracyit + β3 LogUrbit +
β4 LogGDPpcit −1 + β5 LogOpenit + β6L Corruit l {Govit ≤ γ } + β6U Corruit l {Govit > γ } + µit (5)
LogEPIit = γi0 + γ1 Economicfreedomit + γ2 Democracyit + γ3 LogUrbit +
γ4 LogGDPpcit −1 + γ5 LogOpenit + γ6L Corruit l {Govit ≤ γ } + γ6U Corruit l {Govit > γ } + µit (6)
where γ is the threshold parameter, i is the country index, and t is the time. GDPpc it −1 is per capita GDP of country i
in the year t-1. The rest of the variables are taken at time t. School is primary school enrollment and a proxy for human
capital. Inv is gross fixed capital formation in % of GDP. Pop is the growth rate of the population. Open measures openness
to trade using the ratio of exports and imports to GDP. Urb is the degree of urbanization measured as the percentage of
urban population among the total population. Democracy is an indicator of country’s degree democracy, while Economic
freedom is an index of individuals’ freedom to do business, including individuals’ property rights and financial freedom.
The growth Eq. (4) has been commonly used in the literature since Barro (1991); the HDI Eq. (5) is inspired by Akçay
(2006); and the EPI Eq. (6) draws on Aidt (2011). Growth is the growth rate of real GDP per capita. The HDI is a summary
measure of average achievement in key dimensions of human development: health, education and standard of living. It
is the geometric mean of normalized indices for each of the three dimensions. The health dimension is assessed by life
expectancy at birth. Education is measured by mean of years of schooling for adults aged 25 years and older and expected
years of schooling for children of school-entering age. Standard of living is measured by gross national income per capita.
The EPI provides a national scale of how close countries are to meeting established environmental policy goals.
The expected signs of the different coefficients are based on the above review of the literature.4 In Eq. (4), the purpose
of including the country’s past per capita GDP is to take the convergence effect highlighted in the neo-classical growth
model into account (Barro, 1991; Islam, 1995). If convergence holds, α1 < 0. Similarly, population growth allows us to
take into account the negative effect of demographic growth on the growth rate of per capita income, and α4 < 0. We
use the enrollment ratio in primary school in the initial year as a proxy of human capital (Mankiw et al., 1992). Since an
improvement in human capital boosts growth, α2 >0. Openness is used as a proxy for the exposure of the economy to
foreign markets. Its impact on growth is expected to be positive (Frankel and Romer, 1999), or α5 > 0. Openness is defined
as the ratio of exports plus imports to GDP. Levine and Renelt (1992) have found that investment is always robust and
positively correlated with growth. We expect, therefore, α3 >0.
Our expectations for Eq. (5) are the following. Economic freedom is associated in the literature with good performance
across all components of the index (Grubel, 1998), so β1 > 0. To the extent that democratic governance protects human
rights, promotes wider participation in institutions and the rules that affect people’s lives, and achieves more equitable
economic and social outcomes, we expect β2 > 0. Residing in urban areas not only provides more opportunities to make
a higher income, but it also offers better access to schooling, health care and other social services (Akçay, 2006). As a
result, the expected sign of the coefficient of urbanization is positive, or β3 > 0. By construction, GDP per capita is related
to HDI. However, here we introduce the lagged GDP per capita to examine whether richer countries have an advantage
in improving future human development. Hence, β4 could be positive or negative. Davies and Quinlivan (2006) examine
the impact of trade on countries’ human development levels as measured by the HDI and find that increases in trade are
positively associated with future increases in HDI, or β5 > 0.
Finally, Eq. (6) concerns environmental development. Bjørnskov (2020) finds that economic freedom reduces overall
CO2 and greenhouse gas emissions; we expect γ1 >0. Midlarsky (1998), focusing on the relationship between democracy
and the environment, detects no specific significant sign between the two variables; hence, γ2 could be positive or
negative. Maneejuk et al. (2020) examine the relationship between economic development and environmental degradation
4 Except for α , which should be less than 1 in absolute terms, the literature provides no specific expectations about the level of the coefficient.
1
510
and find that urbanization raises environmental degradation; hence, γ3 < 0. Grossman and Krueger (1995) find no evidence
that environmental quality deteriorates steadily with economic growth. Rather, for most indicators, economic growth
brings an initial phase of deterioration followed by a subsequent phase of improvement; as a result, γ4 can be positive
or negative. Finally, the impact of openness on environmental quality is, following Managi et al. (2009), dependent on
various factors, and γ5 could be positive or negative.
In this paper, the parameters of interest are α6L , α6U , β6L , β6U, γ6L and γ6U . In interpreting the results, keep in mind the
definition of Economic Growth, HDI, EPI, Corruption and Governance. An increase of any of them indicates an improved
situation. For instance, a higher level in the corruption index means less corruption. Moreover, the values of estimates
may differ depending on the governance quality. The level of potential thresholds may also be different across dependent
variables. If the test Wn (γ ) does not reject the null hypothesis (i.e., there is no threshold), the estimates of the parameters
are α6L = α6U , β6L = β6U a, and γ6L = γ6U . Following literature pointing to the harmful effect of corruption, these estimated
parameters are all positive. In contrast, if the test Wn (γ ) rejects the null hypothesis (i.e., there is a threshold), we have
different regimes and different estimated signs and values of the parameters across regimes, which means α6L < 0, α6U >
0, β6L < 0, β6U > 0, γ6L < 0, and γ6U > 0.
To examine whether corruption greases or puts sands in the wheels, we use the supW statistic defined in Eq. (3), which
tests the validity of the null hypothesis of no threshold effect.
3.3. Sources of the data
All variables, except HDI, EPI, Democracy and Economic freedom, are drawn from the World Development Indicators
(WDI) of the World Bank (WB). Democracy is the score of democracy drawn from Polity iv, while Economic freedom is an
index of 12 freedoms that allow individuals to do business, including individuals’ property rights and level of financial
freedom. This measure comes from the Heritage Foundation.5 The scores for the three HDI dimensions are aggregated
into a composite index using the geometric mean.6 The EPI ranks 180 countries on 24 performance indicators across 10
categories covering environmental health and ecosystem vitality, among other categories.7
Corruption and governance are two key explanatory variables for our analysis. Therefore, consistency of their measure-
ment is important. A consistent measure is intended to reveal the nature and impact of corruption, and it is necessary
for developing anti-corruption responses. Thus, any measure must be able to quantify the frequency and the intensity
of corruption and be comparable across countries and over time. Scholars investigating the corruption-growth nexus
commonly use the International Country Risk Guide (ICRG), the World Governance Indicators (WGI) and Transparency
International (TI) as sources of data. Since TI does not provide other measures of governance apart from corruption, we
focus on two sources: WGI and ICRG. From the WGI, we use Control of Corruption, which captures perceptions of the
extent to which the state controls and punishes the use of public power for private gain. It includes both petty and
grand corruption, as well as ‘‘capture’’ of the state by elites and private interests. For governance, we use Government
Effectiveness, which captures perceptions of the quality of public services, the quality of civil servants and the degree of
their independence from political pressures. The indicators are converted to vary between 1 and 7, and the higher the
score, the better the situation. Fromthe ICRG, we draw the indicators Corruption and Democratic Accountability. Corruption is
an assessment of corruption within the whole political system, while Democratic Accountability measures how responsive
a government is to its people. The ICRG indicators are measured along a scale ranging from 0 to 6, with a higher score
once again indicating a better situation.
The descriptive statistics are presented in Table 1.
4. Results
4.1. Preliminary analysis
Based on a panel data of 96 to 103 countries over the period 1996 to 2019, a preliminary investigation of the ‘‘grease
the wheels’’ and the ‘‘sand in the wheels’’ claims examines how the impact of control of corruption varies depending on
the quality of governance. We start with an informal analysis of the data.
To focus on a single idea, let us take growth. We construct different sub-samples, which are obtained as follows. We
take the set of observed couples ‘‘growth, corruption’’ which we call ‘‘Set-1’’. We sort the couples in Set-1 according to
the corresponding quality of governance (from the lowest to the highest) to obtain a new set called ‘‘Set-2’’. In Set-2
the first couple corresponds to the lowest level of governance, the second couple corresponds to the second level of
governance and so forth. Set-2 is, then, used to construct sub-samples of 40 couples of ‘‘growth, corruption’’. The sub-
samples are numbered 1 (for the first 40 couples in Set-2), 2 (for the couples from 41 to 80 in Set-2) and so on. Over
each sub-sample, growth is regressed on a constant and corruption using Ordinary Least Squares. This gives one estimated
5 https://www.heritage.org/index/about
6 https://hdr.undp.org/data-center/human-development-index#/indicies/HDI
7 https://epi.yale.edu/about-epi
511
Table 1
Descriptive statistics.
Mean Min Max St-dev Jarque–Bera
Growth 2.26 −36.5 33.03 3.97 92.86
GDP per capita t−1 15765.48 113.56 92556.32 19064.97 1085.00
Openness 78.39 15.63 239.21 38.90 468.90
Investment to GDP ratio 22.13 2.00 59.72 6.09 625.80
Primary schooling 88.97 23.68 105.49 12.95 333.80
Growth rate of POP 1.23 −3.84 15.70 1.41 193.00
Corruption 3.67 1.98 5.96 1.04 185.00
Governance 3.75 1.65 5.85 0.97 114.30
Freedom 61.79 21.40 84.40 10.45 138.30
Urbanization 61.95 15.41 98.04 20.12 129.30
HDI 0.71 0.24 0.95 0.17 203.90
EPI 53.59 22.39 90.58 17.46 149.60
Fig. 1.a. Coefficients of control of corruption by sub-sample numbers: The Growth equation.
coefficient for each sub-sample. The coefficient is an estimate of the impact of control of corruption on growth. Putting the
sample number in an increasing order on the x-axis and the corresponding coefficient on the y-axis gives Fig. 1.a. Similar
approaches are adopted with the HDI (Fig. 1.b) and the EPI (Fig. 1.c). If the ‘‘grease the wheels’’ hypothesis is valid, the
coefficients corresponding to the low numbered sub-samples should be negative, while the coefficients corresponding to
the high numbered sub-samples should be positive.
We conduct the investigation using the WB indicators. The results are presented in Figure 1. For each indicator, the
curve switches from negative to positive after a certain level of sub-sample numbers. This supports the ‘‘grease the wheels’’
hypothesis. However, in order to be confident in these conclusions, we conduct more rigorous econometric tests, as shown
in Tables 2 to 7.
4.2. Econometric analysis
Tables 2 and 3 concern growth. The first table uses the WB measure of corruption, while the second use the ICRG
measure. Each table contains three columns. The first column is based on the whole sample, the second is based on
developed countries, and the third is based on developing countries. Tables 4 and 5 show HDI, and Tables 6 and 7 show
EPI. Both use a similar presentation to Tables 2 and 3. All the tables give the P-value of the test of non-linearity of the model
(Sup-Wald statistics) as described in the end of Section 3.1. The sample is composed of a panel of 96 to 103 developed
and developing countries over the period 1996 to 2019. The distinction between developed and developing countries is
based on classification in the WDI.
In Table 2, the coefficients of GDP per capita t−1 and investment are all significant at 1% and have the expected signs.
The negative coefficients of initial GDP per capita mean that we observe the familiar convergence effect (Barro, 1991). This
512
Fig. 1.b. Coefficients of control of corruption by sub-sample numbers: The HDI equation.
Fig. 1.c. Coefficients of control of corruption by sub-sample numbers: The EPI equation.
suggests that if a country’s GDP per capita in 1996 is high relative to other countries, its growth rate over subsequent years
will be relatively low. The average investment ratio has positive coefficients implying that investment is a strong catalyst
for economic growth (Aghion et al., 2004; Blackburn et al., 2006). Openness has positive and significant coefficients at
1% in two cases (Developed and Developing countries) and 10% in one case (Whole sample). This indicates that openness
to trade is good for growth (Sachs and Warner, 1995; Kentor, 2001). The coefficient of primary schooling is significant
and correctly signed only in the sample of developed countries. In the sample of developing countries, it is insignificant,
indicating that there is insufficient evidence of a growth-boosting effect of primary schooling. This result is common in
the literature probably due to measurement errors (Kwon, 2009). The population growth effect is always significant and is
positive or negative depending on the sample. The negative effect may imply that more people make more use of available
finite resources, thereby reducing long-term potential economic growth (Linden, 2017). According to Piketty (2014, p. 72),
economic growth ‘‘(. . . ) always includes a purely demographic component and a purely economic component, and only
513
Table 2
The growth equation (WB data).
Whole sample Developed countries Developing countries
GDP per capita t−1 −0.17***(25.10) −0.15***(17.09) −0.07***(5.58)
Openness 0.01*(1.65) 0.18***(12.67) 0.11***(2.97)
Investment to GDP ratio 0.04***(12.38) 0.02***(4.09) 0.08***(6.50)
Primary schooling −0.01(0.65) 0.41***(6.70) 0.06(1.44)
Growth rate of population 0.04***(27.02) −0.01***(9.21) 0.02***(3.69)
Control corruption (below threshold) −0.07***(5.02) 0.01(1.52) −0.05*(1.75)
Control corruption (above threshold) 0.07***(6.58) 0.12***(2.59) −0.03(0.75)
Threshold levels (P-value) 2.94*** (0.00) 5.25***(0.00) 2.75*** (0.00)
95% confidence interval [2.78; 3.11] [5.10; 5.40] [2.46; 3.04]
Number of countries 103 49 54
Number of observations 2472 1176 1296
Period 1996–2019 1996–2019 1996–2019
Sup-Wald statistics (P-value) : 34.39***(0.00) 70.90***(0.00) 18.59***(0.00)
*, ** and *** indicate significance at the 10%, 5% and 1% levels. Unless indicated otherwise, in parentheses there are
the t-statistics.
Table 3
The growth equation (ICRG data).
GDP per capita t−1 −0.01***(2.59) −0.22***(4.00) −0.15***(5.80)
Openness 0.26***(14.69) −0.06(1.32) 0.04*(1.90)
Investment to GDP ratio 0.12***(16.45) 0.15***(4.46) 0.08***(5.17)
Primary schooling −0.29***(9.73) −0.44(1.46) 0.35***(4.12)
Growth rate of population −0.05***(19.87) 0.02***(3.18) 0.07***(5.49)
Control corruption (below threshold) −0.03***(15.03) −0.01(1.01) −0.01*(1.67)
Control corruption (above threshold) 0.06***(22.84) −0.01(0.59) 0.01(1.05)
Threshold level (P-value) 3.15***(0.00) 5.84***(0.00) 4.16**(0.04)
Period 1996–2019 1996–2019 1996–2019
Sup-Wald statistics (P-value) 39.18***(0.00) 31.02***(0.00) 6.22***(0.00)
the t-statistics.
the latter allows for an improvement in the standard of living’’. Overall, the findings are compatible with the theory’s
predictions.
We now turn to our main purpose: the impact of corruption and the existence of a threshold of governance quality.
Our results, based on the indicators of corruption and governance from the World Bank, are consistent between developed
and developing countries. There are thresholds that are significantly within the interval of variation of the governance
indicator between 1 and 7. This supports the idea that the effect of corruption on economic growth depends on quality
of governance.
With the whole sample, the coefficients of control of corruption are significant at 1%. Below the threshold level the
coefficient is negative, and it is positive above the threshold. With low quality of governance, corruption is beneficial to
growth, in line with the ‘‘grease the wheels’’ theory. For a one-unit increase in control of corruption, growth decreases by
0.07 units. For developing countries with low-quality governance (i.e., below 2.75), the coefficient of control of corruption
has a negative sign; for a one-point increase in control of corruption index, growth decreases by 0.05 points. Above 2.75,
the coefficient of control of corruption is insignificant. Looking at developed countries, we observe the reverse situation.
Below the threshold of 5.25, a one-point increase in the index of control of corruption does not affect economic growth.
Above this threshold, similar movement results in 0.12 points more economic growth.
Overall, the findings confirm the ‘‘grease the wheels’’ theory, which postulates that corruption has a positive effect
on growth when the quality of governance is poor (Leff, 1964; Huntington, 1968; Méon and Weill, 2010). In fact, several
countries (e.g., China, Vietnam and Cambodia) experience high growth rates despite having high corruption indices (Beck
and Mahler, 1986; Jiang and Nie, 2014; Stojanović et al., 2016). Interestingly, the estimates suggest another dimension
for the effect of corruption: the level of development. In developing countries, control of corruption has a negative effect
on growth when the quality of governance is low. It has no effect when the quality of governance is high. The opposite
happens in developed countries. Control of corruption has a positive effect on growth when the quality of governance is
514
Table 4
The HDI equation (WB data).
Economic freedom 0.00***(8.46) 0.00***(3.58) −0.00(1.63)
Urbanization 0.74***(42.95) 0.82***(5.01) 0.81***(24.76)
Democracy −0.00*(1.68) 0.01***(3.06) 0.00***(2.65)
Openness 0.00***(3.02) 0.07***(2.30) −0.04***(5.53)
GDP per capita t−1 0.03**(13.83) 0.05***(8.33) 0.07***(9.86)
Control corruption below threshold −0.03**(3.87) −0.02*(1.74) −0.03***(7.06)
Control corruption above threshold 0.01***(11.7) 0.08***(5.32) 0.01(1.07)
Threshold level (P-value) 4.47***(0.00) 4.78***(0.00) 3.29***(0.00)
Period 1996–2019 1996–2019 1996–2019
the t-statistics.
Table 5
The HDI equation (ICRG data).
Economic freedom 0.00***(11.26) 0.01**(2.40) 0.01***(5.18)
Urbanization 0.62***(43.58) 1.09***(7.90) 0.99***(21.00)
Democracy 0.00***(2.36) 0.07*(1.82) −0.00(1.60)
Openness 0.00**(2.34) 0.02(1.08) −0.02***(4.71)
GDP per capita t−1 0.01***(3.74) 0.03**(2.19) −0.01(1.27)
Control corruption below threshold −0.00***(7.67) 0.00(1.04) −0.01***(3.01)
Control corruption above threshold 0.00(1.60) −0.01(0.97) −0.00(0.88)
Period 1996–2019 1996–2019 1996–2019
the t-statistics.
high and no effect when the quality of governance is low (Méndez and Sepúlveda, 2006; Aidt et al., 2008; Dzhumashev,
2014). Although there is a high correlation between the quality of governance and development level, it seems that the
methodology can disentangle their effects.
Table 3 is similar to Table 2, except that the indicators of corruption and governance are drawn from ICRG. The
results are similar to those above. The coefficients of initial GDP per capita and investment are all significant at 1%
and have the expected sign. Openness has positive and significant coefficients at 1% in the whole sample and at 10%
in developing countries. This means that openness to trade is good for growth in developing countries. The coefficient of
primary schooling is significant at the 1% level and correctly signed in the sample of developing countries. In the sample
of developed countries, it is insignificant. The population growth effect is always significant, and it is positive in both
developed and developing countries.
The threshold value is 3.15 for the full sample, 5.84 for developed countries and 4.16 for developing countries. Since the
scale of the ICRG governance is between 0 to 6 and the upper-bound of the confidence interval with developed countries
is above 6 (6.21), the interpretation of the existence of two regimes in this sub-sample should be taken with caution.
The confidence intervals with the whole sample and with the developing countries sample support the existence of two
regimes. The coefficient of control of corruption is significant at 1% in the whole sample. With low quality governance,
corruption is beneficial to growth, in line with the ‘‘grease the wheels’’ theory. For a one-unit increase in control of
corruption, growth decreases by −0.03 units. The results for the developing countries give a coefficient of control of
corruption, significant at 10%. In sum, the results confirm those found with the WB data both with respect to the ‘‘grease
the wheels’’ theory and to the developed–developing countries distinction.
Tables 4 and 5 are focused on the Human Development Index. Although the presentation is similar to our prior tables,
the explanatory variables are different. The results concerning control variables are similar in Tables 3 and 4. Out of 15
couples of comparable coefficients, only 5 are different. The coefficients of economic freedom, democracy and urbaniza-
tion, when significant, are positive. As discussed above, more economic freedom enhances human development insofar as
515
Table 6
The EPI equation (WB data).
Whole Sample Developed countries Developing countries
Economic freedom 0.00***(7.78) −0.00(0.24) −0.00(1.23)
Urbanization 0.82***(22.85) 1.73***(4.85) 0.68***(11.04)
Democracy 0.00(0.22) −0.00(0.57) 0.00(1.51)
Openness −0.02***(6.38) −0.07*(1.71) −0.12***(6.80)
GDP per capita t−1 0.04***(5.12) −0.03(0.47) 0.12***(8.13)
Control corruption (below threshold) −0.07***(4.31) 0.02(0.62) −0.08***(4.38)
Control corruption (above threshold) 0.03***(6.26) 0.13**(2.47) 0.19***(5.26)
Period 1996–2019 1996–2019 1996–2019
the t-statistics.
Table 7
The EPI equation (ICRG data).
Economic freedom 0.00*(1.78) 0.00(0.79) 0.00***(4.78)
Urbanization 0.70***(30.06) 2.30***(6.99) 0.99***(9.08)
Democracy 0.00***(12.18) 0.01(0.34) 0.00***(3.03)
Openness −0.02***(3.02) −0.17***(3.07) −0.01(0.64)
GDP per capita t−1 0.01***(4.99) −0.05(1.29) 0.06***(3.05)
Control corruption (below threshold) −0.01***(7.13) −0.01(0.83) −0.01*(1.85)
Control corruption (above threshold) 0.00(1.01) 0.03(1.56) −0.00(0.59)
Period 1996–2019 1996–2019 1996–2019
the t-statistics.
it protects private property, removes obstacles and barriers that restrict transactions, and encourages entrepreneurship
opportunities for all, thus increasing economic activities that, in turn, increase the standard of living. Economic freedom is
a solution that unleashes the drive and ingenuity of the whole population (Grubel, 1998; Akçay, 2006). Democracy helps
citizens’ preferences be better expressed and weighted in political decisions, provides checks on governmental power and
limits the potential of public officials to accumulate personal wealth (Munck, 2016). Thus, democratic countries protect
human rights, encourage larger contributions in political institutions and the rules that affect people’s lives, and often
experience more equitable economic and social outcomes (Barro, 1999). Urbanization, resulting from a general increase
in population and the amount of industrialization of a settlement, is always considered a central factor of growth and
modernization of society (Jacobs, 1984). Its potential to accelerate economic growth depends on the establishment of
favorable institutions and investments in appropriate public infrastructure (Nguyen and Nguyen, 2018), and its positive
effect on income, health and education is mainly proved in the long run (Mayer-Foulkes, 2011). The coefficients of the
lagged GDP and openness are almost the same for the whole sample but differ between groups of countries.
Regarding the threshold of governance, the tests reject the hypothesis of the absence of a threshold independent
of the data source. However, we uncover a similar phenomenon to the growth analysis. With the scale of the ICRG
governance being between 0 to 6 and the upper-bound of the confidence interval with developed countries is around
7, interpreting the impact in these countries is difficult. The confidence intervals with the whole sample and with the
developing countries sample support the existence of two regimes irrespective of the data source. The effect of corruption
on HDI in developing countries follows two regimes depending on the quality of governance. The threshold is around 3.5
for both measurement sources (WB and ICRG). In developing countries with low-quality governance (below the threshold),
the coefficients of control of corruption are significantly negative. A one-unit increase in control of corruption decreases
the HDI by between 0.03 and 0.01 according to WB and ICRG measurement sources, respectively.
Tables 6 and 7 focus on the EPI and use a similar presentation to the other tables. The estimated coefficients of
urbanization are positive and significant at the 1% level in all samples and both data sources. GDP per capita t−1 has
516
Table 8
Summary of the results.
All countries Developed countries Developing countries
The unbalanced sample 103 countries The growth equation (WB data)
The null hypothesis no threshold effect exists Reject Reject Reject
Control corruption (below threshold) −0.07*** 0.01 −0.05*
Control corruption (above threshold) 0.07*** 0.12*** −0.03
The balanced sample of 87 countries The growth equation (WB data)
Control corruption (below threshold) 0.12*** 0.43*** 0.25***
Control corruption (above threshold) 4.58 4.94 2.99
The unbalanced sample 93countries The growth equation (ICRG data)
Control corruption (below threshold) −0.03*** −0.01 −0.01*
Control corruption (above threshold) 0.06*** −0.01 0.01
The balanced sample of 87 countries The growth equation (ICRG data)
Control corruption (below threshold) −0.03*** 0.00 −0.10***
Control corruption (above threshold) 0.03*** 0.01 0.02***
The unbalanced sample 102countries The HDI equation (WB data)
Control corruption below threshold −0.03** −0.02* −0.03***
Control corruption above threshold 0.01*** 0.08*** 0.01
The balanced sample of 87 countries The HDI equation (WB data)
Control corruption below threshold −0.01*** −0.00 −0.07***
Control corruption above threshold 0.04*** 0.01 0.02*
The unbalanced sample 96 countries The HDI equation (ICRG data)
Control corruption below threshold −0.00*** 0.00 −0.01***
Control corruption above threshold 0.00 −0.01 −0.00
The balanced sample of 87 countries The HDI equation (ICRG data)
Control corruption below threshold −0.01*** −0.01*** −0.03***
Control corruption above threshold 0.01*** 0.01* 0.01***
The unbalanced sample 102 countries The EPI equation (WB data)
Control corruption (below threshold) −0.07*** 0.02 −0.08***
Control corruption (above threshold) 0.03*** 0.13** 0.19***
The balanced sample of 87 countries The EPI equation (WB data)
Control corruption (below threshold) −0.02*** −0.02 −0.11***
Control corruption (above threshold) 0.08*** −0.01 0.18***
The unbalanced sample 96 countries The EPI equation (ICRG data)
Control corruption (below threshold) −0.01*** −0.01 −0.01*
Control corruption (above threshold) 0.00 0.03 −0.00
The balanced sample of 87 countries The EPI equation (ICRG data)
Control corruption (below threshold) −0.01*** −0.02*** 0.01
Control corruption (above threshold) 0.00 0.02*** 0.01*
*, ** and *** indicate significance at the 10%, 5% and 1% levels
517
significant and positive coefficients in the whole sample and in developing countries irrespective of the data source. This
is consistent with the suggestion of Beckerman (1992, p. 482) that ‘‘the best, and probably only, way to attain a decent
environment is to become rich’’. Democracy has significant and positive coefficients in the whole sample and in developing
countries only with the ICRG data. The coefficients of openness are negative and are always significant except in one case.
There is no specific pattern for the coefficients of economic freedom.
The estimated coefficients across dependent variables differ. This is not surprising given the scale and the interval of
variations between growth (−36.5; 33.03), HDI (0.24; 0.95) and EPI (22.39; 90.58).
The results establish the existence of a threshold that is significant at the 1% level in all samples and independent of
the data source. Again, we uncover the phenomenon specific to developed countries with the ICRG data. The confidence
intervals with the whole sample and with the developing countries sample support the existence of two regimes
irrespective of the data source. The effect of corruption on EPI in developing countries follows two regimes depending
on the quality of governance. The threshold is below 3.4 for both measurement sources (WB and ICRG). In developing
countries with low-quality governance (below the threshold), the coefficients of control of corruption are significantly
negative. A one-unit increase in control of corruption decreases the EPI by between 0.01 and 0.08.
The sample of 96 to 103 countries is, however, unbalanced in the sense that because of missing observations, it does
not contain the same countries when estimation is conducted using different dependent variables (i.e., growth, HDI and
EPI). Hence, for robustness, we create a core sample which contains the same countries for all dependent variables. This,
of course, reduces the number of countries to 87. The estimation results, presented in Appendix B, globally validate our
main results. Table 8 summarizes the results with both a different and the same sample of countries.
5. Conclusion
This paper investigates whether quality of governance leads to non-linearity in the effect of corruption on economic
growth (approximated by real GDP per capita), human development (HDI) and environmental performance (EPI).
Based on a panel data of 96 to 103 developed and developing countries over the period 1996 to 2019, we compare
the estimation results of a threshold model using different indicators of corruption and governance from the WB and the
ICRG. For all indicators, samples and sources of the data, the tests reject the hypothesis of the absence of a threshold. In
developed countries that are currently associated with high-quality governance, the coefficients of control of corruption,
when significant, are positive. This means that corruption is always harmful to economic growth, human development
and environmental performance. However, for developing countries and for the whole sample, the results indicate that
the impact of corruption on the three variables is regime-specific, where a regime depends on the quality of governance.
When the quality of governance is low (below the threshold), the coefficient of control of corruption is negative, which
means that the effect of a strong control of corruption is detrimental to the variables in question. The reverse is true when
the quality of governance is high (above the threshold). This non-linear pattern is consistent with the ‘‘grease the wheels’’
theory.
Our study suggests that a methodology that allows the estimation of an endogenous threshold is important for
understanding the effect of corruption. The threshold determines potentially different regimes where the impact of
corruption on growth, human development and environmental performance differs. This approach enables a better
understanding of the effect of corruption on economic growth and sustainable development and could act as an indicator
to help governmental institutions adopt the best policies to fight corruption and improve the quality of governance.
Some promising research avenues might arise from this paper. It would be interesting to investigate the effect of the
legal system origin (civil law, common law, customary law, religious law and mixed legal systems) on the corruption-
growth nexus. In fact, a country’s legal origin shapes its institutions over time (La Porta et al., 1998), thus leading to
different effects of corruption on growth and sustainable development. It would also be interesting to test the causality
between corruption and shadow economies since in corrupt countries, shadow economies are often obstacles to growth
and sustainability (Ruzek, 2015).
Appendix A. List of countries
In Tables A.1 to A.3, countries are classified according to the degree of Development and Governance quality. The
distinction between developed and developing countries is based on the classification in the World Development
Indicators (WDI) of the World Bank (WB). The distinction between low and high Governance quality is based on the
estimated threshold of each indicator.
518
Table A.1
Classification of countries by Governance quality and degree of Development (Based on Table 2).
Developed countries Developing countries
Lower regime Upper regime Lower regime Upper regime
Chile; Croatia; Australia; Austria; Algeria; Azerbaijan; Albania; Armenia;
Czech; Belgium; Canada; Belarus; Burkina Faso; Burundi; Benin; Bolivia;
Estonia; Denmark; Cambodia; Central African; Botswana; Brazil;
Greece; Finland; France; Chad; Ecuador; Gabon; Bulgaria; Colombia;
Hungary; Germany; Gambia; Guatemala; Honduras; Cuba; Dominican; Egypt;
Italy; Japan; Ireland; Madagascar; Mali; El Salvador;
Korea, Rep.; Netherlands; Mozambique; Ghana; India;
Latvia; New Zealand; Niger; Pakistan; Indonesia; Iran;
Lithuania; Norway; Togo; Ukraine; Venezuela; Jordon; Kenya;
Oman; Poland; Sweden; Zimbabwe. Malaysia;
Panama; Portugal Switzerland; UK; Mexico; Morocco;
Romania; USA. Peru; Philippines;
Saudi Arabia; Senegal; Sri Lank;
Slovak Republic; Tunisia; Turkey.
Slovenia; Spain;
United Arab Emirates;
Uruguay.
Table A.2
Classification of countries by Governance quality (Based on Table 4).
Croatia; Australia; Austria; Algeria; Azerbaijan; Albania; Armenia;
Czech; Belgium; Canada; Belarus; Burkina Faso; Benin; Bolivia;
Greece; Chile; Denmark; Burundi; Botswana; Brazil;
Hungary; Estonia; Cambodia; Central; African Bulgaria; Colombia;
Italy; Finland; France; Chad; Ecuador; Gabon; Cuba; Dominican; Egypt;
Latvia; Germany; Gambia; Guatemala; El Salvador;
Lithuania; Ireland; Honduras; Kenya; Ghana; India;
Oman; Poland; Japan; Madagascar; Mali; Indonesia; Iran;
Panama; Korea Rep.; Mozambique; Jordon;
Romania; Netherlands; Niger; Pakistan; Malaysia;
Saudi Arabia; New Zealand; Togo; Ukraine; Venezuela; Mexico; Morocco;
Slovak Republic; Norway; Zimbabwe. Peru; Philippines;
Uruguay. Portugal; Senegal; Sri Lank;
Slovenia; Spain; Tunisia; Turkey.
Sweden;
Switzerland; UK;
United Arab Emirates;
USA.
Table A.3
Classification of countries by Governance quality (Based on Table 6).
Croatia; Australia; Austria; Algeria; Azerbaijan; Albania; Botswana; Brazil;
Czech; Estonia; Belgium; Canada; Armenia; Belarus; Bulgaria;
Greece; Chile; Denmark; Benin; Bolivia; Burkina Faso; Ghana;
Hungary; Finland; France; Burundi; Cuba; Colombia; India;
Italy; Korea, Rep.; Germany; Cambodia; Central African Jordon;
Latvia; Ireland; Chad; Dominican; Egypt; Malaysia;
Lithuania; Japan; El Salvador; Mexico;
Oman; Poland; Netherlands; Ecuador; Gabon; Philippines;
Panama; Portugal; New Zealand; Gambia; Guatemala; Tunisia;
Romania; Norway; Honduras; Indonesia; Turkey.
Saudi Arabia; Slovenia; Spain; Iran; Kenya;
Slovak Republic; Sweden; Madagascar; Mali;
United Arab Emirates; Switzerland; UK; Mozambique; Morocco;
Uruguay. USA. Niger; Pakistan; Peru;
Senegal; Sri Lank;
Togo; Ukraine; Venezuela
Zimbabwe.
519
Appendix B. Results based on the core sample
See Tables B.1–B.6.
Table B.1
The growth equation (WB data).
GDP per capita t−1 −0.15***(21.3) −0.28***(8.68) −0.07***(3.88)
Openness 0.01(1.26) 0.34***(9.4) 0.03(0.61)
Investment to GDP ratio 0.04***(14.33) 0.03(0.69) −0.01(0.34)
Primary schooling −0.13***(4.79) −0.56***(2.24) −0.22***(2.38)
Growth rate of population 0.01***(3.99) −0.03***(6.54) −0.01(1.09)
Control corruption (below threshold) −0.53***(10.97) 0.32***(7.01) −0.22***(5.81)
Control corruption (above threshold) 0.12***(11.42) 0.43***(2.71) 0.25***(5.73)
Threshold levels (P-value) 4.58(0.00) 4.94(0.00) 2.99(0.00)
Period 1996–2019 1996–2019 1996–2019
the t-statistics.
Table B.2
The growth equation (ICRG data).
GDP per capita t−1 −0.14***(12.85) −0.09***(5.28) −0.08***(0.22)
Openness −0.05***(4.15) 0.43***(9.21) 0.03(0.95)
Investment to GDP ratio 0.05***(19.35) −0.07(1.56) 0.00 0.00(0.24)
Primary schooling 0.17***(6.21) 1.19***(5.79) 0.05(0.68)
Growth rate of population −0.02***(15.7) −0.03***(3.73) −0.05***(4.26)
Control corruption (below threshold) −0.03***(2.88) 0.00(0.16) −0.10***(4.32)
Control corruption (above threshold) 0.03***(3.33) 0.01(0.08) 0.02***(4.34)
Threshold levels (P-value) 4.21(0.00) 5.58(0.00) 4.84(0.00)
Period 1996–2019 1996–2019 1996–2019
the t-statistics.
Table B.3
The HDI equation (WB data).
Economic freedom −0.00(0.82) 0.00(0.72) −0.00***(2.73)
Urbanization 0.70***(31.93) 0.88***(13.2) 0.81***(11.93)
Democracy 0.01***(2.6) 0.01***(4.15) 0.00(3.61)
Openness 0.01***(6.36) 0.04***(4.96) 0.06***(4.32)
GDP per capita t−1 0.04***(16.42) 0.02***(3.03) 0.06***(4.91)
Control corruption below threshold −0.01***(15.26) −0.00(0.22) −0.07***(6.19)
Control corruption above threshold 0.04***(4.03) 0.01(1.06) 0.02*(1.76)
Threshold level (P-value) 4.77(0.00) 4.41(0.00) 3.01(0.00)
Period 1996–2019 1996–2019 1996–2019
the t-statistics.
520
Table B.4
The HDI equation (ICRG data).
Economic freedom −0.00(0.35) 0.00(1.07) 0.00***(2.51)
Urbanization 0.71***(32.48) 1.19***(7.57) 0.87***(14.5)
Democracy 0.00***(15.99) 0.00(1.28) 0.00(1.17)
Openness −0.02***(9.44) 0.00(0.56) 0.07***(12.12)
GDP per capita t−1 0.02***(11.14) 0.04***(3.23) 0.03***(3.54)
Control corruption below threshold −0.01***(11.3) −0.01***(2.92) −0.03***(2.26)
Control corruption above threshold 0.01***(8.2) 0.01*(1.94) 0.01***(4.97)
Threshold level (P-value) 4.18(0.00) 5.62(0.00) 5.00
Period 1996–2019 1996–2019 1996–2019
the t-statistics.
Table B.5
The EPI equation (WB data).
Economic freedom −0.00***(3.32) 0.00*(1.94) −0.00***(3.98)
Urbanization 0.88***(14.84) 1.86***(8.48) 0.24**(1.98)
Democracy −0.00(0.21) 0.01***(2.89) −0.01**(1.99)
Openness 0.00(0.83) −0.03**(1.96) 0.04(1.36)
GDP per capita t−1 0.05***(5.39) 0.02(1.13) 0.09***(3.67)
Control corruption (below threshold) −0.02***(5.08) −0.02(0.76) −0.11***(2.88)
Control corruption (above threshold) 0.08***(4.73) −0.01(0.48) 0.18***(2.85)
Period 1996–2019 1996–2019 1996–2019
the t-statistics.
Table B.6
The EPI equation (ICRG data).
Economic freedom 0.00(0.24) 0.00*(1.86) −0.00***(3.68)
Urbanization 1.34***(25.43) 1.82***(5.8) 1.22***(6.04)
Democracy 0.00***(21.45) 0.026***(2.43) 1.22***(6.04)
Openness −0.05(0.81) −0.14***(2.28) 0.12***(6.93)
GDP per capita t−1 −0.08***(10.75) −0.095***(3.56) −0.05(1.05)
Control corruption (below threshold) −0.01***(2.66) −0.02***(2.68) 0.01(0.82)
Control corruption (above threshold) 0.00(0.83) 0.02***(3.34) 0.01*(1.91)
Period 1996–2019 1996–2019 1996–2019
the t-statistics.
References
Absalyamova, S., Absalyamov, T., Khusnullova, A., Mukhametgalieva, C., 2016. The impact of corruption on the sustainable development of human
capital. J. Phys. Conf. Ser. 738 (1), 1–6.
Acemoglu, D., Verdier, T., 1998. Property rights, corruption and the allocation of talent: A general equilibrium approach. Econ. J. 108 (450), 1381–1403.
Achim, M.V., Văidean, V.L., Borlea, S.N., 2019. Corruption and health outcomes within an economic and cultural framework. Eur. J. Health Econ. 21
(1), 195–207. http://dx.doi.org/10.1007/s10198-019-01120-8.
521
Aghion, P., Alesina, A., Trebbi, F., 2004. Endogenous political institutions. Q. J. Econ. 119 (2), 565–611.
Aidt, T.S., 2009. Corruption, institutions, and economic development. Oxf. Rev. Econ. Policy 25 (2), 271–291.
Aidt, T.S., 2011. Corruption and sustainable development. In: International Handbook on the Economics of Corruption, Vol. 2. pp. 3–51.
Aidt, T., Dutta, J., Sena, V., 2008. Governance regimes, corruption and growth: Theory and evidence. J. Comp. Econ. 36 (2), 195–220.
Akçay, S., 2006. Corruption and human development. Cato J. 26 (1), 29–48.
Akhbari, R., Nejati, M., 2019. The effect of corruption on carbon emissions in developed and developing countries: empirical investigation of a claim.
Heliyon 5 (9), e02516. http://dx.doi.org/10.1016/j.heliyon.2019.e02516.
Andvig, J.C., 2006. Corruption and fast change. World Dev. 34 (2), 328–340.
Asimakopoulos, S., Karavias, Y., 2016. The impact of government size on economic growth: A threshold analysis. Econom. Lett. 139, 65–68.
Barro, R.J., 1991. Economic growth in a cross section of countries. Q. J. Econ. 106 (2), 407–443.
Barro, R., 1999. Determinants of democracy. J. Polit. Econ. 107 (6), 158–183.
Beck, P., Mahler, M., 1986. A comparison of bribery and bidding in thin markets. Econom. Lett. 20 (1), 1–5.
Beckerman, W., 1992. Economic growth and the environment: Whose growth? Whose environment? World Dev. 20 (4), 481–496. http://dx.doi.org/
10.1016/0305-750X(92)90038-W.
Bjørnskov, C., 2020. Economic freedom and the CO2 Kuznets Curve. Available at SSRN 3508271.
Blackburn, K., Bose, N., Haque, M.E., 2006. The incidence and persistence of corruption in economic development. J. Econom. Dynam. Control 39 (12),
2447–2467.
Bose, N., Capasso, S., Murshid, A.P., 2008. Threshold effects of corruption: theory and evidence. World Dev. 36 (7), 1173–1191.
Candau, F., Dienesch, E., 2017. Pollution haven and corruption paradise. J. Environ. Econ. Manag. 85, 171–192. http://dx.doi.org/10.1016/j.jeem.2017.
05.005.
Caner, M., Hansen, B.E., 2004. Instrumental variable estimation of a threshold model. Econom. Theory 20 (5), 813–843.
Cariolle, J., 2018. Corruption Determinants in Developing and Transition Economies: Insights from a Multi-Level Analysis. Working Papers P229,
FERDI.
Chen, H., Hao, Y., Li, J., Song, X., 2018. The impact of environmental regulation, shadow economy, and corruption on environmental quality: theory
and empirical evidence from China. J. Clean. Prod. 195, 200–214. http://dx.doi.org/10.1016/j.jclepro.2018.05.206.
Copeland, B.R., Taylor, M.S., 1994. North-South trade and the environment. Q. J. Econ. 109 (3), 755–787. http://dx.doi.org/10.1257/.42.1.7.
Damania, R., Fredriksson, P., List, J., 2003. Trade liberalization, corruption, and environmental policy formation: theory and evidence. J. Environ. Econ.
Manag. 46 (3), 490–512. http://dx.doi.org/10.1016/S0095-0696(03)00025-1.
Davies, A., Quinlivan, G., 2006. A panel data analysis of the impact of trade on human development. J. Socio-Econ. 35 (5), 868–876.
Desai, U., 1998. Ecological Policy and Politics in Developing Countries: Growth, Democracy and Environment. State University of New York Press,
Albany.
Dincer, O.C., Fredriksson, P.G., 2018. Corruption and environmental regulatory policy in the United States: does trust matter? Resour. Energy Econ.
54, 212–225. http://dx.doi.org/10.1016/j.reseneeco.2018.10.001.
Dzhumashev, R., 2014. Corruption and growth: The role of governance, public spending, and economic development. Economic Modelling 37, 202–215.
http://dx.doi.org/10.1016/j.econmod.2013.11.007.
Emerson, J.W., Hsu, A., Levy, M.A., de Sherbinin, A., Mara, V., Esty, D.C., Jaiteh, M., 2012. Environmental performance index and pilot trend
environmental performance index.
Frankel, J.A., Romer, D.H., 1999. Does trade cause growth? Am. Econ. Rev. 89 (3), 379–399.
Fredriksson, P.G., Vollebergh, H.R., Dijkgraaf, E., 2004. Corruption and energy efficiency in OECD countries: theory and evidence. J. Environ. Econ.
Manag. 47 (2), 207–231.
Ganda, F., 2020. The influence of corruption on environmental sustainability in the developing economies of Southern Africa. Heliyon 6 (7), e04387.
Grossman, G.M., Krueger, A.B., 1995. Economic growth and the environment. Q. J. Econ. 110 (2), 353–357. http://dx.doi.org/10.2307/2118443.
Grubel, H.G., 1998. Economic freedom and human welfare: Some empirical findings. Cato J. 18 (2), 287–304.
Gupta, S., Davoodi, H., Alonso-Terme, R., 2002. Does corruption affect income inequality and poverty? Econ. Gov. 3 (1), 23–45. http://dx.doi.org/10.
1007/s101010100039.
Han, L., Li, X., Xu, G., 2022. Anti-corruption and poverty alleviation: Evidence from China. J. Econ. Behav. Organ. 203 (C), 150–172. http://dx.doi.org/
10.1016/j.jebo.2022.09.001.
Hanf, M., Van-Melle, A., Fraisse, F., Roger, A., Carme, B., Nacher, M., 2011. Corruption kills: Estimating the global impact of corruption on children
deaths. PLoS One 6 (11), 1–7.
Hansen, B.E., 1999. Threshold effects in non-dynamic panels: Estimation, testing, and inference. J. Econometrics 93 (2), 345–368.
Huntington, S.P., 1968. Political Order in Changing Societies. Yale Press, New Haven.
Iskandar, A., Saragih, R., 2018. Does corruption affect poverty in indonesia?. Jurnal BPPK: Badan Pendidikan Dan Pelatihan Keuangan 11 (1), 1–14.
Islam, N., 1995. Growth empirics: a panel data approach. Q. J. Econ. 110 (4), 1127–1170.
Jacobs, J., 1984. Cities and the Wealth of Nations. Random House, New York.
Jiang, T., Nie, H., 2014. The stained China miracle: Corruption, regulation, and firm performance. Econ. Lett. 123 (3), 366–369.
Justesen, M.K., Bjørnskov, C., 2014. Exploiting the poor: Bureaucratic corruption and poverty in Africa. World Dev. 58 (C), 106–115.
Kentor, J., 2001. The long term effects of globalization on income inequality, population growth, and economic development. Social Problems 48 (4),
435–455. http://dx.doi.org/10.1525/sp.2001.48.4.435.
Khan, H., Weili, L., Khan, I., Oanh, L.K., 2021. Recent advances in energy usage and environmental degradation: Does quality institutions matter? A
worldwide evidence. Energy Rep. 7, 1091–1103. http://dx.doi.org/10.1016/j.egyr.2021.01.085.
Kwon, D.B., 2009. Human capital and its measurement. In: Paper Presented in the 3rd OECD World Forum on Statistics, Knowledge and Policy
Charting Progress, Building Visions, Improving Life.
La Porta, R., Lopez-de Silanes, F., Shleifer, A., R., Vishny, 1998. Law and finance. J. Polit. Econ. 106 (5), 1113–1155.
Lambsdorff, J.G., 2003. How corruption affects productivity. Kyklos 56 (4), 457–474.
Leff, N., 1964. Economic development through bureaucratic corruption. Am. Behav. Sci. 8 (2), 8–14.
Levine, R., Renelt, D., 1992. A sensitivity analysis of cross-country growth regressions. Am. Econ. Rev. 88 (4), 942–963.
Leys, C., 1965. What is the problem about corruption?. The Journal of Modern African Studies 3 (2), 215–230;
Reprint in, Heidenheimer, A.J., Johnston, M., LeVine, V.T. (Eds.), 1989. Political Corruption: A Handbook. Transaction Books, Oxford, pp. 51–66.
Lien, D.H.D., 1986. A note on competitive bribery games Econom. Lett. 22, 337–341.
Linden, E., 2017. Remember the population bomb? It’s still ticking In: New York Times: Sunday Review, 4.
Lio, M.C., Lee, M.H., 2016. Corruption costs lives: A cross-country study using an IV approach Int. J. Health Plan. Manag. 31 (2), 175–190.
Lui, F.T., 1985. An equilibrium queuing model of Bribery J. Polit. Econ. 93 (4), 760–781.
Lv, Z., Gao, Z., 2021. The effect of corruption on environmental performance: Does spatial dependence play a role? Econ. Syst. 45 (2), 100773.
http://dx.doi.org/10.1016/j.ecosys.2020.100773.
Managi, S., Hibiki, A., Tsurumi, T., 2009. Does trade openness improve environmental quality? J. Environ. Econ. Manag. 58 (3), 346–363.
522
Maneejuk, N., Ratchakom, S., Maneejuk, P., Yamaka, W., 2020. Does the environmental Kuznets Curve exist? An international study Sustainability 12
(21), 9117.
Mankiw, N.G., Romer, D., Weil, D.N., 1992. A contribution to the empirics of economic growth Q. J. Econ. 107 (2), 407–437.
Mauro, P., 1995. Corruption and growth Q. J. Econ. 110 (3), 681–712.
Mauro, P., 1998. Corruption and the composition of government expenditure J. Public Econ. 69 (2), 263–279.
Mayer-Foulkes, D., 2011. Urbanization as a fundamental cause of development In: Economics. the Economics Divison of the centro de investigacion
y Docencia Economicas, Mexico City.
Méndez, F., Sepúlveda, F., 2006. Corruption, growth and political regimes: Cross-country evidence Eur. J. Political Econ. 22 (1), 82–98.
Méon, P.-G., Sekkat, K., 2005. Does corruption grease or sand the wheels of growth? Public Choice 122 (1–2), 69–97.
Méon, P.-G., Weill, L., 2010. Is corruption an efficient grease? World Dev. 36 (3), 244–259.
Midlarsky, M.I., 1998. Democracy and the environment: an empirical assessment J. Peace Res. 35 (3), 341–361.
Mo, P.H., 2001. Corruption and economic growth J. Comp. Econ. 29 (1), 66–79. http://dx.doi.org/10.1006/jcec.2000.1703.
Munck, G.L., 2016. What is democracy? A reconceptualization of the quality of democracy Democratization 23 (1), 1–26.
Murshed, M., Mredula, F.A., 2018. Impacts of corruption on sustainable development: A simultaneous equations model estimation approach J. Account.
Finan. Econ. 8 (1), 109–133.
Nguyen, H.M., Nguyen, L.D., 2018. The relationship between urbanization and economic growth: An empirical study on ASEAN countries Int. J. Social
Econ. 45 (2), 316–339. http://dx.doi.org/10.1108/IJSE-12-2016-0358.
Olayeni, A., Ogbo, A., Okwo, H., Chukwu, B., Ifediora, C., Ezenwakwelu, C., 2021. Green strategy effect on financial and environmental performance:
A mediation analysis of product quality Sustainability 13 (4), 2115. http://dx.doi.org/10.3390/su13042115.
Olson, M., 2000. Power and Prosperity: Outgrowing Communist and Capitalist Dictatorships. Basic Books, New York.
Ortega, B., Casquero, A., Sanjuán, J., 2014. Growth in human development: The role of corruption J. Int. Dev. 26 (7), 974–998.
Piketty, T., 2014. Capital in the Twenty-First Century. Belknap Press of Harvard University Press, Cambridge, MA.
Reinikka, R., Svensson, J., 2005. Fighting corruption to improve schooling: Evidence from a newspaper campaign in Uganda J. Eur. Econom. Assoc. 3
(2–3), 259–267.
Rock, M., Bonnett, H., 2004. The comparative politics of corruption: accounting for the east Asian paradox in empirical studies of corruption, growth
and investment World Dev. 32 (6), 999–1017. http://dx.doi.org/10.1016/j.worlddev.2003.12.002.
Rose-Ackerman, S., 2006. International Handbook on the Economics of Corruption. Edward Elgar, Cheltenham, UK and Northampton, MA, USA.
Ruzek, W., 2015. The informal economy as a catalyst for sustainability Sustainability 7, 23–34.
Sachs, J.D., Warner, A., 1995. Economic reform and the process of global integration In: Brookings Paper-s on Economic Activity. p. 118.
Sekkat, Kh., 2018. Is Corruption Curable?. Palgrave Macmillan.
Seo, M.H., Shin, Y., 2016. Dynamic panels with threshold effect and endogeneity J. Econometrics 195 (2), 169–186.
Shleifer, A., Vishny, R., 1993. Corruption Q. J. Econ. 108 (3), 599–617.
Sinha, A., Gupta, M., Shahbaz, M., Sengupta, T., 2019. Impact of corruption in public sector on environmental quality: implications for sustainability
in BRICS and next 11 countries J. Clean. Prod. 232, 1379–1393. http://dx.doi.org/10.1016/j.jclepro.2019.06.066.
Siverson, R., Johnson, R., 2014. Politics and parasites: The contribution of corruption to human misery Int. Stud. Q. 58 (1), 199–206. http:
//dx.doi.org/10.1111/isqu.12052.
Stojanović, I., Ateljević, J., Stević, R.S., 2016. Good governance as a tool of sustainable development Eur. J. Sustain. Dev. 5 (4), 558–573.
Swaleheen, M., 2011. Economic growth with endogenous corruption: An empirical study Public Choice 146, 23–41.
Uberti, L.J., 2022. Corruption and growth: Historical evidence, 1790–2010 J. Comp. Econ. 50 (2), 321–349. http://dx.doi.org/10.1016/j.jce.2021.10.002.
Wang, S., Zhao, D., Chen, H., 2020. Government corruption, resource misallocation, and ecological efficiency Energy Econ. 85, 104573. http:
//dx.doi.org/10.1016/j.eneco.2019.104573.
Wei, S.-J., 2000. How taxing is corruption on international investors? Rev. Econ. Stat. 82 (1), 1–11.
Zhang, Y.J., Jin, Y.L., Chevallier, J., Shen, B., 2016. The effect of corruption on carbon dioxide emissions in APEC countries: a panel quantile regression
analysis Technol. Forecast. Soc. Chang. 112 (C), 220–227. http://dx.doi.org/10.1016/j.techfore.2016.05.027.
Zhu, X., Asimakopoulos, S., Kim, J., 2020. Financial development and innovation-led growth: Is too much finance better? J. Int. Money Finance 100,
102083.
523

Corrup SD Threshold Non-Linear Analysis EAP2023

Uploaded by

Copyright:

Available Formats

Corrup SD Threshold Non-Linear Analysis EAP2023

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Corrup SD Threshold Non-Linear Analysis EAP2023

Uploaded by

Copyright:

Available Formats

Economic Analysis and Policy 78 (2023) 505–523

Contents lists available at ScienceDirect

Economic Analysis and Policy

Analyses of topical policy issues

How does corruption affect sustainable development?

2.1. The effect of corruption on the Human Development Index (HDI)

2.2. The effect of corruption on environmental performance

2.3. Sand or grease the wheels?

3. The empirical strategy

yit = µi + β ′ (z ) xit + eit (1)

yit = Xit′ β + 1, Xit′ δ l {qit > γ } + µi + εit i= 1, . . ..N ; t = 1, . . .T

for any where Γ denotes the parameter space for γ .

supW = SupWn (γ ) , (3)

Growthit = αi0 + α1 LogGDPpcit −1 + α2 LogSchoolit + α3 LogInvit + α4 Popit +

3.3. Sources of the data

4.1. Preliminary analysis

4.2. Econometric analysis

*, ** and *** indicate significance at the 10%, 5% and 1% levels

Appendix A. List of countries

Appendix B. Results based on the core sample

See Tables B.1–B.6.

You might also like

, and indicate significance at the 10%, 5% and 1% levels