Resources Policy 48 (2016) 117–128

Contents lists available at ScienceDirect

Resources Policy
journal homepage: www.elsevier.com/locate/resourpol

Extractive industries and corruption: Investigating the effectiveness of

EITI as a scrutiny mechanism
Elizabeth Kasekende a, Charles Abuka b, Mare Sarr a,n
a
School of Economics, University of Cape Town, Private Bag X3, Rondebosch 7701, South Africa
b
Bank of Uganda, P.O Box 7120, Kampala, Uganda

art ic l e i nf o a b s t r a c t

Article history: The resource curse literature suggests that many countries have failed to exploit their natural resource
Received 24 August 2015 wealth to finance the growth of their economies. Developing countries appear to be most affected. It is
Received in revised form believed that poor governance, including lack of transparency, poor accountability to citizens and cor-
8 March 2016
ruption, are the main culprits. In 2002, an international initiative sponsored by the UK government and
Accepted 8 March 2016
backed by activist groups launched the Extractive Industries Transparency Initiative (EITI) with a view to
Available online 25 March 2016
mitigating the potential negative effects of resource wealth. The objective of this study is to investigate
JEL: the effectiveness of this initiative that has gained much traction as a scrutiny mechanism for corruption
P4 control. In particular, this paper addresses two key questions: First, what are the observable factors that
Q3
lead a country to voluntarily join the EITI? Second, does EITI membership leads to greater corruption
Keywords: control? Using the Worldwide Governance Indicators (WGI) control of corruption index, we find that EITI
EITI membership has not resulted in reduced corruption scores.
Corruption & 2016 Elsevier Ltd. All rights reserved.
Extractive industries

1. Introduction However, natural resources can be turned into a blessing

rather than a curse through better functioning institutions, en-
A number of developing countries rely heavily on natural hanced transparency and accountability in resource exploitation
resource revenues. However, in many instances, reliance on (Bannon and Collier, 2003). As a result, improved governance
extractive industries has been found to be associated with slow can achieve reduced corruption through greater transparency
growth rates. Although natural resources per se do not ne- and a more effective legal system (Collier and Venables, 2010).
cessarily lead to poor economic performance (Brunnschweiler For these reasons, Collier and Venables (2010) argue that good
and Bulte, 2008; Alexeev and Conrad, 2009), many countries governance is even more important in resource management
have failed to use resource proceeds effectively as a source of than any other area of economic management because of the
sustainable financing for growth. It has been argued, though, enabling role of the state. They maintain that greater public
that dependency on natural resources may lead to slower scrutiny as well as international commitments to transparency
growth through poor institutions (Sala-i-Martin and Sub- initiatives (Collier and Venables, 2010) could influence govern-
ramanian, 2003; Mehlum et al., 2006), and corruption (Leite and ment officials and company representatives to abide by the law
Weidmann, 1999). As a result these countries have experienced and comply with their contractual obligations (Rosenblum and
mismanagement of resource revenues resulting in disappointing Maples, 2009). Public scrutiny by the press and civil society is
economic outcomes (Sala-i-Martin and Subramanian, 2003). therefore believed to be a necessary condition for halting ra-
In a number of developing countries, contracts in the extractive pacious behavior and gross misappropriation of public funds
sector are opaque and often escape proper scrutiny. To corrupt officials, (D'Souza, 2009).
the lack of scrutiny is alluring because they are able to sell mineral or In the past fifteen years, two major scrutiny mechanisms
oil rights under a veil of secrecy. Resource dependence tends to affect have been initiated with a view to reducing corruption and rent
the quality of governance through looting which encourages corrup- seeking through greater transparency. These include the Publish
tion, reduction in accountability and economic mismanagement. What You Pay (PWYP) campaign and more prominently, the
Extractive Industries Transparency Initiative (EITI). EITI aims at
n promoting transparency and accountability in resource-rich
Corresponding author.
E-mail addresses: [email protected] (E. Kasekende), countries by encouraging extractive companies to publish what
[email protected] (C. Abuka), [email protected] (M. Sarr). they pay and governments to disclose their resource revenues so

http://dx.doi.org/10.1016/j.resourpol.2016.03.002
0301-4207/& 2016 Elsevier Ltd. All rights reserved.
118 E. Kasekende et al. / Resources Policy 48 (2016) 117–128

that these countries can fully benefit from their revenues and extractive sector, EITI principles were agreed upon at the 2003
improve living standards (Ölcer, 2009; EITI Source Book, 2005). Lancaster House Conference in London (EITI Source Book,
This paper is concerned with analyzing the effectiveness of 2005).1
Extractive Industries Transparency Initiative (EITI) (a voluntary, Countries joining the initiative expect to derive tangible
“soft law” scrutiny mechanism) in the control of corruption. benefits such as an improvement in their investment prospects,
Despite the growing interest in EITI, there is a dearth of rigorous and potentially more development assistance, as a result of their
studies that have examined its role on governance outcomes commitment to greater transparency. As a recent initiative, EITI
(Corrigan, 2014 is a notable exception). This is partly because has gained recognition and its credibility and popularity have
EITI is still a fairly recent arrangement that was only initiated in increased over the years. It is supported by major international
2002. Our research contributes to understanding the effective- organizations such as the United Nations (UN), the European
ness of EITI as a scrutiny mechanism by introducing a sound Union (EU), the OECD and the African Development Bank, as
methodological approach to the literature. In particular, this well as numerous global investment institutions. The number of
paper addresses two key questions: (i) what observable factors EITI members has increased dramatically. By December 2015,
are associated with EITI membership? (ii) How does EITI there were 49 member countries, of which 31 were listed as
membership impact upon corruption control? In responding to compliant countries. Some of these new members include de-
these questions the paper sheds some light on the determinants veloped countries such as the United States (US) and the United
of EITI membership, and the causal effect of joining the EITI on Kingdom (UK) that joined very recently in 2014. Numerous civil
corruption. Although this paper asks a similar question to Cor- society organizations are also involved with EITI on an inter-
rigan (2014) – albeit focusing entirely on the effect of EITI on national level. These include, inter alia, the Publish What You
corruption control – it differs methodologically to this earlier Pay coalition (PWYP), Global Witness, Oxfam, Transparency In-
contribution in two respects. First, because countries adopting ternational, and Revenue Watch Institute.2 Finally, many of the
EITI are unlikely to be random, any estimation that seeks to world's largest oil and gas and mining companies have shown
understand its effect on governance or economic outcomes support for EITI.
must specifically model this selection process. This implies that Despite EITI's increased popularity, it suffers from numerous
the OLS models estimated by Corrigan (2014) are likely to suffer shortcomings. First, a large number of oil producing countries,
from selection bias. To avoid this pitfall, we jointly estimate the including some of the largest oil producers such as Saudi Arabia,
selection process, i.e., the probability of joining EITI, together Iran and Russia are not members of the initiative. Moreover,
with the corruption outcome since the initiation of EITI in 2002 among industrialized countries, only Norway, and very recently
(using full information maximum likelihood [FIML]) and allow the USA and the UK have so far become members. Second, EITI
for correlation between the error terms of the treatment and members (countries and companies alike) typically disclose
outcome equations. Our second departure from Corrigan (2014) partial or scant revenue information, which allows at best for
lies in the fact that we check that estimations based on short- minimal transparency. In this light, the Revenue Watch Institute
term (yearly) variations in the corruption indices are robust for 2010 report recommends a radical move towards full disclosure
longer term variations in accordance with the recommendations of standardized financial and contractual information (in a de-
by Kaufmann et al. (2010) who constructed the corruption in- tailed and disaggregated manner) in order to achieve greater
dices used in this paper and in Corrigan (2014). Consequently, transparency. Thirdly, it is unclear whether EITI's promotion of
governance is captured here by five-year variation in perceived transparency through independent and rigorous audits brings
corruption between 2002 (year of inception of EITI) and 2012, in about accountability when some audit recommendations are
addition to a year-to-year variation in the corruption score blatantly ignored despite serious irregularities (Keblusek, 2010).
(following Corrigan, 2014). Keblusek (2010:10; 16) discusses the case of Nigeria where in-
The results indicate that there is no evidence of improved depth financial, physical and process audits for the period 1999–
corruption scores as a result of EITI membership. In contrast to 2004 revealed serious irregularities but no one was held to ac-
the EITI goals, member countries of the initiative are, on aver- count for these. Fourthly, by focusing solely on transparency on
age, associated with higher scores of corruption perception the revenue side, EITI ignores the fact that patronage, one of the
compared to non-EITI members, suggesting that EITI has not main avenues for corruption in resource rich countries, relies on
yet been successful in reducing corruption in resource rich the lack of transparency associated with expenditures rather
countries. than revenues (Kolstad and Soreide, 2009; Kolstad and Wiig,
The rest of the paper is structured as follows: the background 2009). Besides, the public reporting of revenues, EITI merely
to the EITI mechanism is discussed in Section 2, followed by the constitutes a step towards improving governance of the ex-
empirical literature, and estimation strategy in Sections 3 and 4, tractive industries value chain. The exploitation of natural re-
respectively. Sections 5 and 6 provide the analysis of the results source wealth for sustainable development requires countries
and conclusion, respectively. indeed to proceed through various critical stages in the value

1
A resource-rich country has to satisfy a number of requirements before it can
2. Background to EITI become a fully compliant member of the EITI. First, it needs to fulfill five sign up
conditions to the satisfaction of the EITI Board to become a candidate. These five
It has often been argued that some officials in resource rich conditions include (1) government issuance of an unequivocal public statement of
developing countries use lack of scrutiny to engage in embez- its intention to implement the EITI; (2) government commitment to work with civil
society and companies on the implementation of the EITI; (3) government ap-
zlement. Transparency is therefore considered a key defense
pointment of a senior individual to lead the implementation of the EITI; (4) es-
against corruption. The EITI, a coalition of governments, ex- tablishment a multi-stakeholder group to oversee the implementation of the EITI;
tractive companies and civil society organizations, was estab- and (5) the multi-stakeholder group, in consultation with key EITI stakeholders,
lished in 2002. This is a voluntary scrutiny mechanism aimed at should agree and publish a work plan, containing measurable targets, and a
reducing the scope for “grand corruption” in the use of re- timetable for implementation and incorporating an assessment of capacity con-
straints (see http://eiti.org/eiti/implementation/signup).
source-based revenues through public reporting of companies’ 2
The Revenue Watch Institute has been renamed the Natural Resource Gov-
payment and revenue earned by resource-rich nations. To pro- ernance Institute (NRGI) in 2014. Because our analysis covers a period prior to the
mote transparency in the payments and revenues in the renaming, we will use the old name.
 E. Kasekende et al. / Resources Policy 48 (2016) 117–128 119

chain such as transparency in contracts and licenses award 4. Data and estimation strategy
procedures, transparency in the system of tax and royalties
collection, effective regulation and monitoring of extractive in- 4.1. Data
dustries projects, and establishment of transparent saving me-
chanisms as well as a macro fiscal framework adapted to volatile In studying EITI effectiveness on reduced corruption scores, we
and finite resources (Alba, 2009). use panel data estimation techniques for the first decade of the
initiative's existence (2002–2012). The choice of countries in the
sample was based on natural resource endowment as listed by the
EITI website and the IMF (2007). We concentrated on countries
3. Empirical literature rich in natural resources because EITI aims at improving trans-
parency in that target group of countries. Our sample includes 76
To better understand the EITI initiative, two empirical ques- countries that are rich in hydrocarbon and mineral resources (see
tions of policy relevance have been recently examined: (i) what Table A1 in the Appendix). Table A1 includes the status of the
are the characteristics of countries that decide to become EITI countries, the year of EITI membership if applicable, the average
members, and (ii) has EITI lived up to the expectation it aroused; corruption score and the average resource rents. It is interesting to
i.e., has EITI been an effective mechanism in mitigating the re- note that some countries such as Iraq, Republic of Congo and
source curse? Equatorial Guinea that had the highest average resource rent as a
In recent years, two important contributions have attempted to percent of GDP registered the highest corruption scores while
shed light on these questions using econometric analysis. The first countries such as the United States, United Kingdom and Norway
article by Pitlik et al. (2010) aims to understand the factors that that had relatively lower shares of resource rents, exhibited lower
determine EITI membership. Typically, Pitlik et al. (2010) estimate corruption levels. This could suggest that some countries that
a probit model to explore the determinants of EITI membership. highly depend on natural resources are more susceptible to
They find that key factors influencing countries’ decision to join corruption.
EITI include the extent of resource dependency, widespread cor- Although there exist several corruption measures used in the
ruption, the level of political freedom (democracy) and freedom of literature, two of the most popular include the Corruption Per-
the media, as well as membership to OPEC. However, the scope of ception Index (CPI) published by Transparency International, and
the Control of Corruption Index (CCI) from the Worldwide Gov-
this study is fairly limited because it overlooks the incidence of
ernance indicators (WGI) developed by Kaufmann's team at the
EITI membership on important outcomes such as governance or
World Bank (Kaufmann et al., 2002, 2006, 2010). With the Trans-
economic performance.
parency International measure, a country is perceived to be most
The second contribution by Corrigan (2014) assesses the ef-
corrupt if it has a CPI of 1, and least corrupt if it has a CPI of 10.
fectiveness of EITI with regard to governance and economic out-
While the index was available for most countries in 2012, only 29
comes. In essence, the question asked is whether joining EITI helps
countries in our sample had available indices for 2002. Therefore,
resource-rich countries improve governance quality and economic we concentrate on the WGI Control of Corruption Index (CCI),
performance, thereby mitigating the resource curse. To address which incidentally is strongly correlated with the CPI index in
this question, Corrigan (2014) uses panel data analysis to establish 2002 (0.92) and 2012 (0.98). The CCI index is a perception-based
the effectiveness (or lack thereof) of EITI across numerous gov- indicator that measures the extent to which public office is ex-
ernance indicators and economic growth. However, given the very ercised for private gain, including petty corruption, grand cor-
brief existence of EITI, these assessments (like ours) are necessarily ruption and state capture. It ranges from
2.5 (most corrupt) to
incomplete and should be taken with some degree of caution. 2.5 (least corrupt) and has a standard normal distribution. For ease
Corrigan (2014)'s overall message is that while the introduction of of interpretation of the results, we rescaled the index between
EITI has failed to result in reduced corruption (perception) or in- 0 and 5 and reversed the order so that 0 indicates lower corruption
creased political stability or democracy, it has nonetheless posi- and 5 higher corruption. For instance, Norway is perceived as the
tively impacted on GDP per capita, government effectiveness and least corrupt country in the sample with an average score of 0.4. To
the rule of law. However, her econometric analysis suffers from exploit the time variation within each country and use panel data
several methodological shortcomings. First, in the context of cross- analysis, this indicator is used first on a yearly basis and subse-
country analysis, a major advantage of using panel data lies in the quently on a 5-yearly basis to ascertain that our findings are not
ability to control for time-invariant characteristics through fixed driven erroneously by year-on-year variation.3 Indeed, both Arndt
effects estimation in order to mitigate the problem of missing and Oman (2006) and Kaufmann et al. (2010) call for caution in
variables (or unobserved heterogeneity). Therefore, any credible comparing these governance indicators over time to avoid po-
tential misuse.
results in this context should involve country fixed effects. How-
Measures of foreign aid, foreign direct investment, real GDP per
ever, in Corrigan (2014), the country fixed-effects specifications
capita and openness Are collected from the World Bank Devel-
(which are few) systematically failed to yield a statistically sig-
opment Indicators while the information on federal states comes
nificant effect of EITI membership on the WGI indicators for re-
from the Forum of Federations (2012). The freedom of the press
source-rich countries. This implies that the claims about the po-
score is obtained from Freedom House. It measures the degree to
sitive effect of EITI on economic performance, government effec- which a country allows free flow of information (Freedom House,
tiveness and rule of law, are not warranted. Secondly, by relying on 2011). On average, EITI members seem to enjoy greater press
ordinary least squares, Corrigan's estimation strategy is likely to freedom, with an average score of 44 compared to 42 for non-
result in biased estimates of the effect of EITI membership on members (see Table 1). Details of the definition and source of
governance scores and economic performance. variables can be found in Table A2 in the Appendix. Detailed
This paper contributes to this literature by providing a meth-
odology that stands on more solid ground to mitigate estimation
3
We thank an anonymous reviewer for suggesting the use of panel data
biases resulting from both omitted variables and self-selection. We
analysis rather than cross-section analysis. A key advantage of panel data analysis is
explain these crucial points in the next section in which we justify that it allows for country fixed effects to capture time-invariant unobserved het-
our proposed analytical framework. erogeneity, thereby mitigating the omitted variables bias.
120 E. Kasekende et al. / Resources Policy 48 (2016) 117–128

Table 1
Mean difference EITI vs. non-EITI.

Variables EITI members Non-EITI members T-tests

Obs. Mean Obs. Mean Mean

difference

Corruption score 167 3.214 669 2.933 0.281nnn

Corruption when 167 3.170 669 2.640 0.530nnn
joining EITI
Aid when joining EITI 163 12.048 546 5.748 6.300nnn
GDP per capita (log) 167 8.051 647 8.888
0.836nnn
Openness (% GDP) 160 34.861 638 41.883
7.022nnn
Resource rent (% 167 22.303 651 22.621
0.317
GDP)
OPEC 167 0.054 669 0.184
0.130nnn
Polity2 score 145 1.807 616 1.580 þ0.227
British Dummy 167 0.21 669 0.326 0.116nnn
Federal States 167 0.054 669 0.167 0.114nnn

nnn
p < 0.01.

summary statistics for this analysis are depicted in Table 1. decision to join EITI and corruption using maximum likelihood.4
This method allows us to estimate the effect of EITI membership
4.2. Analytical framework on perceived corruption while taking into account the possible
endogeneity of EITI through self-selection. Although our primary
The belief that greater transparency in the funds paid and re- interest lies in identifying the effect of EITI membership on cor-
ceived by resource rich countries will result in greater public ruption scores, it is essential that we simultaneously investigate
scrutiny and eventually in reduced corruption is what primarily the determinants of EITI. The point this paper is making is that,
motivates initiatives such as the EITI. Our objective is to analyze contrary to Corrigan (2014), an appropriate estimation strategy,
the effect of joining EITI on the perception of corruption among requires one to simultaneously estimate the selection process to-
countries endowed with natural resources. To do so, the simplest, gether with the corruption effect of EITI (or any outcome resulting
often dubbed naïve approach is to examine the impact of EITI on from EITI, for that matter).
perceived corruption by including in the corruption equation a An important question is how to incorporate the impact of EITI
dummy variable equal to 1 if the country has joined EITI and to membership into the econometric model. Standard treatment ef-
estimate the equation using ordinary least squares (OLS). This fects models typically include a treatment dummy as explanatory
approach followed by Corrigan (2014) suffers, however, from a variable, assuming that the impact on the outcome variable can be
fundamental flaw. It assumes de facto that the EITI dummy variable represented as a simple intercept shift. A treatment effects re-
is exogenous or equivalently that the decision to join EITI is ran- gression consists of two stages. The first stage is a selection
dom. If one takes seriously the work of Pitlik et al. (2010), such an equation, which is based on a dichotomous choice function, e.g.,
assumption is hardly tenable. Indeed, these authors provide evi- the probability of being an EITI member. With regard to expected
dence suggesting that the self-selection process of joining EITI is benefits, the country evaluates whether or not to join EITI based
endogenous rather than random, and depends among others upon on country's observed and unobserved characteristics. A country
a number of observable characteristics as discussed earlier. The decides to join EITI if the expected utility of EITI membership
qualitative literature on EITI also provides ample anecdotal evi- (denoted EITI⁎) is positive, and will abstain from joining EITI
dence of country self-selection based on observable characteristics otherwise. The expected benefit EITI⁎ is not observable, but we
(see our discussion in Section 4.3 that refers to David-Barrett and observe the membership dummy variable EITI which indicates
Okamura, 2013; Ölcer, 2009; Singh and Bourgouin, 2013). In other whether the country has joined or not. The canonical regression
words, EITI-members have systematically different characteristics model used in this literature takes the following form:
from non-members, and may have decided to join based on their
yit = αEITIit + Xit′ β + εit
own expected benefits. Unobservable characteristics of countries
may also affect both the decision to join and corruption percep- EITIit⁎ = Sit′ γ + δZit + υit
tion, resulting in biased estimates of the effect of joining EITI on 1 if EITIit⁎ > 0
corruption. Practically, this implies that the so-called naïve treat- EITIit = {
0 if EITIit⁎ ≤ 0
ment of EITI dummy as an exogenous variable in an OLS estima-
tion as in Corrigan (2014) is likely to be flawed (see Guo and Fraser, where yit is the value of the corruption indicator for country i at
2014 for a non-technical exposition of the sample selection time t; εit and υit are mean-zero error terms; EITIit is a dummy
problem). variable that captures whether a country is an EITI member (i.e.,
Since there are good reasons to suspect the presence of non- has received the “treatment”); α is the effect of the treatment on
random (endogenous) self-selection, the selection process must be yit; Xit is a vector of exogenous covariates and β is the associated
modeled explicitly otherwise the impact of EITI membership is vector of coefficients; Sit is a vector of exogenous variables
likely to be biased. Various methods have been developed to cor-
rect selection bias by modeling explicitly the endogenous treat- 4
Estimation methods that correct for selection bias have been developed over
ment (here the decision of joining EITI) in addition to the effect of
the years since James Heckman's path-breaking work on sample selection in 1979
the treatment on corruption perceptions. In this paper, we use a and have become standard in econometric analysis (see Wooldridge, 2002; Ca-
treatment effects model, which estimates simultaneously the meron and Trivedi, 2005; Moffitt, 2009; Guo and Fraser, 2014).
 E. Kasekende et al. / Resources Policy 48 (2016) 117–128 121

affecting both the probability of joining EITI and the perception of Okamura, 2013). For example, it is alleged that the Republic of
corruption, i.e., Sit is a subset of Xit. For proper identification of the Congo acceded to the World Bank's pressure to have its oil sector
selection equation coefficients, we assume that at least one cov- audited – which paved the way for EITI membership – primarily to
ariate Zit in the EITI equation is excluded from the outcome have its application for a debt relief under the Heavily Indebted
equation, i.e., Zit is not contained in vector Xit. The assumption is Poor Country Initiative (HIPC) approved (Singh and Bourgouin,
called an exclusion restriction and implies that the EITI member- 2013). Ölcer (2009) echoes a similar story for Cameroon. Countries
ship equation will have at least one variable that is excluded from that are heavily dependent on aid could have more incentive to
the corruption equation. Typically, the treatment effects model is join EITI to signal to their donors that they are taking the necessary
estimated with an efficient procedure using full information steps to promote transparency. As the data in Table 1 shows, EITI
maximum likelihood. When the likelihood function fails to con- member countries received more foreign aid compared to non-
verge due to computational difficulties, we use the two-step es- members. Therefore, foreign aid is expected to be positively as-
timator of the treatment effects model, which is similar to Heck- sociated with EITI membership:
man (1979).5 This method allows for correlation between the two H2: Countries with initially greater foreign aid allocation are more
error terms εit and υit – which are assumed to be jointly normally likely to join EITI. Conditional on controlling for factors deemed
distributed. Our contribution here is to improve methodologically important in the literature (especially Pitlik et al., 2010) such as
the current literature on the causal effect of EITI membership on the effect of resource dependence, corruption, OPEC countries and
governance outcomes by taking seriously the ubiquitous problem national income on the likelihood of joining EITI, we will focus on
of selection bias within a panel data framework. testing the two above-mentioned hypotheses. Both variables of
interest (corruption and foreign aid) are evaluated at the year of
4.3. Testable hypotheses EITI membership. However, for non-EITI members such data do
not exist. We therefore have to make assumptions about what
In choosing appropriate covariates, we build on (i) the available corruption and foreign aid levels to select for non-EITI members.
literature on the determinants of EITI membership (Pitlik et al., There is obviously a degree of arbitrariness in making such as-
2010) and the impacts of EITI on governance indicators (Corrigan, sumptions. Cognizant of this, we make the three following as-
2014); (ii) the corruption literature; and (iii) the analysis of the sumptions for the sake of robustness: (1) the non-EITI corruption/
mean difference in key variables for EITI member countries re- aid levels take the value of the average corruption scores and aid
lative to non-members (see Table 1). levels; (2) the non-EITI corruption/aid levels take the 2002 values
(year of EITI establishment); and (3) the non-EITI corruption takes
4.3.1. Determinants of EITI membership the value of a moving average of corruption/aid levels. The former
Whether a country chooses to join such a voluntary initiative is used in our main results table (Table 2) while the other two
hinges on the perceived net benefits it expects to reap. In this measures feature in Table 3 for robustness purpose. Interestingly,
section, we outline two key hypotheses that highlight how ex- all assumptions yield very similar results as we show in the paper.
pected benefits may induce a country to seek EITI membership.6
According to Pitlik et al. (2010), countries that are more dependent 4.3.2. Determinants of corruption
of natural resources, countries that enjoy more political and press Our outcome equation is a standard corruption equation aug-
freedom and countries that are perceived as being more corrupt mented with the effect of EITI membership. One of the most ro-
are more likely to join EITI while OPEC members are less likely to bust findings in this literature is that income level tends to be
do so. Our mean difference analysis of key variables (see Table 1) is negatively associated with high corruption. According to Madani
in tune with these findings. It also indicates that EITI countries and Licetti (2010), wealthier countries are associated with lower
attracted both more foreign aid by the time they joined the in- levels of corruption because of the more efficient and transparent
itiative and tend to be poorer than their counterparts. The ratio- regulations and good institutions in place. Furthermore, Ades and
nale for these key findings and observations is clear (see Pitlik Di Tella (1999) find that an increase in GDP per capita is positively
et al., 2010; David-Barrett and Okamura, 2013; Ölcer, 2009; Singh associated with a decline in corruption. Therefore, we expect GDP
and Bourgouin, 2013). per capita to be statistically significant and negatively associated
Because EITI presents itself as an international standard pro- with corruption.7
moting transparency and good governance (Kolstad and Soreide, In addition, openness to foreign trade was found to help reduce
2009: 528), a number of resource-rich countries perceived as corruption according to Larrain and Tavares (2007). However,
corrupt may decide to join the initiative to improve their reputa- Treisman (2000) found this to be true only with a very small
tion in the international community (David-Barrett and Okamura, magnitude. By contrast, Combes and Saadi-Sedik (2006) argue that
2013). Their reasons can be varied: it may be a strategic move, but openness to trade is more prone to corruption for non-renewable
it may also be that governments that are committed to corruption resources, such as oil and minerals, due to associated resource
reduction or that seek to change the perception about their rents. Since this study only analyses countries rich in oil and mi-
countries will be willing to join the EITI. For example, Singh and nerals (see also Leite and Weidmann, 1999), openness to trade is
Bourgouin (2013) argue that one of the reasons why President expected to be positively associated with corruption.
Obasanjo committed Nigeria to become an EITI member was to Based on empirical literature on corruption, colonial heritage
counter the country's reputation for corruption following Sani tends to influence the level of corruption in a country. In
Abacha's rule. Based on this reasoning, we hypothesize that initial
corruption is positively associated with EITI membership, that is: 7
We thank an anonymous reviewer for pointing to the debates regarding the
H1: Countries perceived to be more corrupt are more likely to join
causality between income and corruption. Early literature on corruption estab-
EITI. Countries also join EITI with the hope of reaping tangible lished that “by far the strongest and most consistent finding is that lower perceived
benefits in the form of aid or debt forgiveness (David-Barrett and corruption correlates closely with higher economic development” (see La Porta at
al., 1999; Ades and Di Tella, 1999; Treisman, 2000), with causality running from
income to corruption. This view has been challenged due to the possible existence
5
Note that this happened for our 5-year panel. of reverse causality. More recent development in the literature, however, suggests
6
Following Pitlik et al. (2010), we will include numerous covariates which will that the causal pathway seems to run from development to reduced perceptions of
serve as mere control variables to mitigate the risk of omitted variables bias. For corruption (Treisman, 2007, Gundlach and Paldam, 2009). This explains our deci-
this reason, we will not outline the hypotheses for those controls. sion to control for GDP per capita in the corruption equation.
122 E. Kasekende et al. / Resources Policy 48 (2016) 117–128

Table 2
Impact of EITI membership on corruption perception: OLS and treatment effects models.

1-Year Panel Data 5-Year Panel Data

OLS Estimator FIML Estimator FIML Estimator Two-step Estimator

Model 1 Model 2 Model 3 Model 4

Panel A: Corruption Equation (Outcome equation)

EITI member
0.0212 0.190nnn 0.171nnn 0.145nn
(0.360) (0.000) (0.000) (0.017)
Corruption when joining EITI 0.481nnn 0.801nnn 0.981nnn
(0.000) (0.000) (0.000)
Aid when joining EITI
0.000432 0.0656nnn
0.0178n 0.00807
(0.962) (0.000) (0.066) (0.367)
GDP per capita (log)
0.196nnn
0.165nn
0.139n
0.237nn
(0.014) (0.037) (0.077) (0.014)
Openness (% GDP) 0.000422 0.0000973 0.000262 0.000581
(0.740) (0.932) (0.820) (0.678)
OPEC
0.558nnn
0.287nn
0.414nnn
1.085nnn
(0.000) (0.018) (0.001) (0.001)
Polity2
0.00223 0.00272 0.00235
0.00816
(0.611) (0.542) (0.595) (0.106)
Press Freedom
0.000994
0.00352nn
0.00468nnn
0.00636nnn
(0.554) (0.044) (0.009) (0.004)
British dummy
0.932nnn
0.776nnn
0.132nnn
0.837nnn
(0.000) (0.000) (0.004) (0.001)
Federal states 0.410n 0.296nnn 0.0755 0.925nnn
(0.059) (0.002) (0.402) (0.001)
Country fixed effect Yes Yes Yes Yes
Time fixed effects Yes Yes Yes Yes

Panel B: EITI Membership Equation (Treatment equation)

Corruption when joining EITI 0.613nnn 0.701n
(0.000) (0.053)
Aid when joining EITI
0.0400nnn
0.0643nnn
0.0322
(0.001) (0.000) (0.343)
GDP per capita (log)
0.513nnn
0.636nnn
0.704nn
(0.000) (0.000) (0.013)
Resource rent (% GDP) 0.0196nnn 0.0183nnn 0.0319nnn
(0.000) (0.000) (0.007)
Polity2
0.00688
0.0126
0.0219
(0.551) (0.295) (0.416)
OPEC
0.564nnn
0.941nnn
0.580
(0.009) (0.000) (0.291)
Press Freedom 0.0166nnn 0.0280nnn 0.0427nnn
(0.000) (0.000) (0.001)
Continents fixed effects Yes Yes Yes Yes
Time fixed effects Yes Yes Yes Yes
Inverse Mills ratio –
0.156nnn
0.148nnn
0.0825nnn
– (0.0130) (0.0137) (0.0384)

Number of countries 623 621 621 169

Log pseudo-likelihood 311.2 157.9 170.9

Dependent variables: Corruption in Panel A (outcome equation) and EITI Membership in Panel B (treatment equation). Control for time fixed effects.
The probit equation controls only for regional dummies because country fixed effects would produce inconsistent estimates in a standard probit model due to the incidental
parameters problem. p-values are in parentheses.
n
p < 0.10.
nn
p < 0.05.
nnn
p < 0.01.

particular, former British colonies have been found to have lower Besides these determinants, we also control for correlates of
levels of corruption than other former colonies (La Porta at al., EITI membership that are known to potentially affect corruption.
1999; Treisman, 2000). Thus, former British colonies in this study We incorporate political factors such as democracy and press
would be expected to have lower corruption levels. freedom which have been found in the literature to potentially
The system of government could also affect the perceived level matter for corruption (Brunetti and Weder, 2003; Shen and Wil-
of corruption. Treisman (2000) finds that once the level of eco- liamson, 2005; Treisman 2007; Ata and Arvas, 2011). Furthermore,
nomic development is controlled for, federal states are perceived we include foreign aid, which is a predictor of corruption ac-
to be more corrupt than unitary states. Similarly, Kunicová and cording to Ali and Isse (2003) and Tavares (2003) though they find
Rose-Ackerman (2005) find that federal states tend to be more opposite effects. Meanwhile, Knack (2001) finds no significant link
corrupt. Therefore, federal states are expected to be associated between aid and corruption. Finally, we include past corruption
with higher corruption levels than unitary states. perception (i.e., corruption when joining EITI), which is assumed
 E. Kasekende et al. / Resources Policy 48 (2016) 117–128 123

Table 3
Robustness: alternative definitions for variables “corruption” and “aid when joining EITI” for non-EITI members.

Counterfactual for non-EITI members Counterfactual for non-EITI members

Corruption, Aid in 2002 Moving average of Corruption, Aid
Model 5 Model 6

Panel A: Corruption Equation (Outcome equation)

EITI member 0.172nnn 0.157nnn
(0.000) (0.000)
Corruption when joining EITI 1.083nnn 0.990nnn
(0.000) (0.000)
Aid when joining EITI
0.0200nn 0.0103
(0.035) (0.220)
GDP per capita (log)
0.138n
0.149nn
(0.081) (0.049)
Openness (% GDP) 0.000239 0.000457
(0.838) (0.691)
OPEC
0.564nnn
1.002nnn
(0.000) (0.000)
Polity2 0.00182 0.00408
(0.688) (0.331)
Press freedom
0.00463nnn
0.00551nnn
(0.009) (0.001)
British dummy
0.296nnn
0.699nnn
(0.000) (0.002)
Federal states 0.541nnn 0.889nnn
(0.000) (0.001)
Country fixed effect Yes Yes
Time fixed effects Yes Yes

Panel B: EITI Membership Equation (Treatment equation)

Corruption when joining EITI 0.571nnn 0.639nnn
(0.000) (0.000)
Aid when joining EITI
0.0657nnn
0.0720nnn
(0.000) (0.000)
GDP per capita (log)
0.643nnn
0.717nnn
(0.000) (0.000)
Resource rent (% GDP) 0.0193nnn 0.0209nnn
(0.000) (0.000)
Polity2
0.0107
0.0181
(0.369) (0.148)
OPEC
0.940nnn
0.974nnn
(0.000) (0.000)
Press freedom 0.0275nnn 0.0331nnn
(0.000) (0.000)
Continents fixed effects Yes Yes
Time fixed effects Yes Yes

Inverse Mills ratio
0.149nnn
0.132nnn

0.0135 0.0164

Number of countries 614 606

Log pseudo-likelihood 166.4 193.3

Dependent variables: Corruption in Panel A (outcome equation) and EITI Membership in Panel B (treatment equation). Control for time fixed effects.
The probit equation controls only for regional dummies because country fixed effects would produce inconsistent estimates in a standard probit
model due to the incidental parameters problem. p-values are in parentheses.
nnn
p < 0.01.
nn
p < 0.05.
n
p < 0.10.

to affect current corruption (see Treisman, 2000 discussion about follows8:

corruption perception and the “rather disheartening but plausible
message about the tenacity of the past”). CCIit = αEITIit + β1CCIi, join + β2 Aidi, join ++ β3 GDPit + β4 OPECit
Conditional on controlling for these factors, the coefficient on
the treatment variable EITI is the parameter of interest. In line with + β5 Democracyit + β6 Pressit + β7 Opennessit + β8 Britishit
the expectation that EITI may help mitigate the resource curse, our + β9 Federalit + εit (1)
major hypothesis is:
H3: Joining EITI contributes to the reduction of corruption perception.

4.4. Econometric model

Based on the discussion of the previous section, our treat-

ment effects model (outcome and treatment equations) reads as 8
See Sarr et al. (2011) for a similar treatment effects model.
124 E. Kasekende et al. / Resources Policy 48 (2016) 117–128

EITIit⁎ = γ 0 + γ 1CCIi, join + γ 2 Aidi, join + treatment effects estimation models (Models 2–4). This result
suggests that EITI membership is associated with greater per-
+ γ 3 GDPit + γ4 OPECit + γ 5 Democracyit + γ 6 Press it + δNR Dependenceit + υ it

ceived corruption. In addition, the fact that the coefficients of the
Controls from Pitlik et al. (2010) maximum likelihood estimation are systematically larger than the
1 if EITIit⁎ > 0 OLS estimates, combined with the negative and significant coef-
EITIit = {
0 if EITIit⁎ ≤ 0 (2) ficients of the inverse Mills ratio indicate that the OLS estimates
are biased downwards (Models 2–4).10 This evidence of selection
Eq. (1) depicts the outcome equation with the Control of Corruption bias justifies our use of treatment effects models. This implies that
Index (CCI) as the dependent variable. The coefficient α on the Corrigan (2014) OLS estimations may well suffer from selection
treatment variable EITI is the parameter of interest. Its sign and sig- bias and therefore may not be interpreted as they have been un-
nificance informs us about the contribution of EITI membership to less self-selection is appropriately corrected for.
changes in perceived corruption levels. EITI is in turn the dependent The inclusion of the variable that captures “corruption when
observable variable in the treatment equation (Eq. (2)). The de- joining EITI” can be somehow interpreted as a lagged dependent
terminants of EITI membership include corruption, aid and FDI when variable (although strictly speaking this is not the case). Using lagged
joining the initiative, real GDP per capita, OPEC membership, de- dependent variables in combination with fixed effects is potentially
mocracy, and press freedom. Finally, for the purpose of identification problematic because the correlation between the fixed effects in the
of the model, resource dependence (NR Dependence ) serves as our error term and the lagged dependent variable would give rise to
exclusion restriction. That is, NR Dependence is included in the se- “dynamic panel bias” (Nickell, 1981). Ideally, we would like to correct
lection equation but excluded from the corruption outcome equation. for this potential dynamic panel bias by using a generalized method of
We interpret this as follows: conditional on all observed character- moments (GMM) estimator à la Arelano–Bond or the ML estimator
istics discussed above, dependence on natural resources affects cor- developed by Allison (2014). However, these two estimators are un-
ruption indirectly through EITI membership or lack thereof.9 To able to accommodate the problem of self-selection. For this reason, we
check that our exclusion restriction is reasonable, we included our decide to address simply the selection bias problem given it is likely to
indicator of resource dependence in the corruption equation, and be the more serious of the biases in our context. We nonetheless verify
found its coefficient to be statistically insignificant, which vindicates that this bias is not a major concern in our analysis. We do so by not
our choice. including our constructed “corruption when joining EITI” variable
We control for country fixed effects in the corruption equation to (Model 2).11 We find that the result is similar for our variables of in-
make sure that any unobserved, time-invariant feature of each in- terest. This could suggest that we should not be overly concerned with
dividual country is taken into account. This enables us to mitigate the this bias.
issue of omitted variables bias. In the EITI membership equation, we The effect of the control variables for the corruption equation
control only for continental dummies. We are unable to control for accords with the main findings of the corruption literature. Indeed,
fixed effects because they produce inconsistent estimates in a stan- the British dummy, press freedom, and GDP per capita are all
dard probit model due to the incidental parameters problem. Finally, significant and negatively related to corruption scores, while fed-
we control for time dummies in all specifications to account for eral states are positively associated with corruption perceptions.
events that affect all countries in a given year. The effect of foreign aid is less clear, just as it is in the literature.
In analyzing the treatment equation, we are particularly interested
in testing how corruption perception and foreign aid affect the deci-
5. Empirical results and discussion sion to join EITI. We find that corruption when joining EITI is statis-
tically significant (at the 1 percent level for the 1-year panel and 10
5.1. Findings percent level for the 5-year panel) and positively related to EITI
membership. This indicates that countries perceived to be more cor-
We start by estimating the effect of EITI membership on cor- rupt are more likely to join the initiative to send a strong signal that
ruption using ordinary least squares (OLS) following Corrigan they are willing to increase transparency and eventually curb cor-
(2014). We subsequently use the treatment effects model pre- ruption. These findings are similar to those presented by David-Bar-
sented in the previous section to account for selection bias. In rett and Okamura (2013) and Keblusek (2010). Surprisingly, countries
doing this, we demonstrate the extent to which the OLS is biased with higher aid at the time of joining the initiative are less likely to
and the importance of appropriately accounting for selection bias. join EITI. As expected, countries that are dependent on natural re-
Table 2 reports the OLS (Model 1) and the treatment effects sources and those with higher levels of press freedom are more likely
(Models 2–4) estimations results. In Model 1, we find EITI to be to join EITI while wealthy countries with higher GDP per capita are
insignificant, suggesting that EITI does not have an effect on cor- less likely to join the initiative.
ruption. However, the effect of most control variables is in line We also find evidence that oil producing countries are less
with the corruption literature. For instance, the British dummy and likely to join the initiative, which accords with the work of Pitlik
GDP per capita negatively related to corruption while initial cor- et al. (2010).12 A close look at the data suggests that Norway and
ruption scores (when joining EITI) and federal states are positively Nigeria are the only major oil producers that have joined EITI.
associated with corruption. North African and Middle Eastern countries, Russia and Venezuela,
In contrast to the OLS estimation and to our expectation, EITI is have showed little interest in being part of this initiative. This
highly significant and positively related to corruption in all
10
The inverse Mills ratio is an important variable that helps correct for se-
9
To address the endogeneity of EITI, treatment effects models resort to an lection bias. The sign and statistical significance of its coefficient (which reflects the
approach that bears resemblance to the instrumental variable approach where the correlation between the error terms of the corruption and the EITI equations) allow
variable of interest must satisfy two conditions: Condition 1, it must be correlated us to test the presence of selection bias.
11
with EITI; and condition 2, it must be uncorrelated with corruption, after con- We thank an anonymous reviewer for this suggestion.
12
trolling for relevant covariates. These conditions imply that the variable of interest According to Pitlik et al. (2010), most OPEC countries are reluctant to join
must feature in the EITI selection equation (condition 1) but must be excluded from EITI because its scrutiny could reveal the suspected free-riding and cheating be-
the corruption equation (condition 2). This is the so-called exclusion restriction havior of many OPEC members in regard to their production commitment (under-
where the variable of interest is excluded from the outcome equation. reporting of the production).
 E. Kasekende et al. / Resources Policy 48 (2016) 117–128 125

might be a concern for the effectiveness of EITI, since major human rights abuses and corruption. This suggests that a country
players which suffered from low or negative growth during the can satisfy the minimum requirements for membership of the EITI
period 1970–2000 prefer to opt out. One might have expected oil while the governance of its resource sector experiences no sub-
producing countries to join the initiative to send a signal to in- stantive improvement.
vestors. However, there might be reasons for the status quo. One
possibility is that oil rich countries are reluctant to invite external,
public scrutiny and would rather continue to operate in a non-
6. Conclusion
transparent environment. Another possibility is that these coun-
tries simply do not need an external agency such as EITI to attract
The main objective of this study is to investigate the effec-
investment in oil extraction. We should however note that the
tiveness of EITI, whose main goal is to strengthen governance
OPEC dummy loses its significance in the 5-year panel despite
through greater transparency and accountability in extractive in-
being negative (Model 4).
Finally, in Table 3, we report robustness checks with regard to dustries. In particular, the paper examines the effect of EITI
our exclusion restriction. We find that the results are very similar membership on corruption perception by taking into account the
to those presented above. possibility of self-selection. By taking seriously the ubiquitous
problem of selection bias that previous studies overlooked, this
5.2. Discussion paper contributes to the EITI literature by introducing a rigorous
methodology to analyze the causal effect of EITI membership on
It has become commonplace in the policy discourse to expect governance outcomes.
that greater transparency will lead to reduced corruption through Our analysis shows that, on average, corrupt countries and
improved governance (Kolstad and Soreide, 2009). However, this countries attracting greater shares of FDI as well as countries with
study suggests that this optimism might be misplaced. Our find- lower per capita GDP are more likely to join the initiative. This is
ings indicate that transparency alone (proxied by increased dis- probably because EITI membership enables them to signal their
closure of resource revenue) may not be sufficient to reduce per- commitment to greater transparency. In this manner, they expect
ceived corruption. Real improvements may require that the public to reap some benefits in the form improved investment and as-
to be able to hold governments accountable for the decisions they sistance. But interestingly, we have found that countries that re-
make (Klitgaard, 1988).13 This paper, like those of Kolstad and Wiig ceived more foreign aid before joining tend to be more reluctant to
(2009), Kolstad and Soreide (2009) and Corrigan (2014), suggests become EITI members. Overall, we find like Pitlik et al. (2010)
that EITI alone may be unable to provide this sort of accountability. countries with more press freedom have incentive to join EITI
In this regard, our study contrasts with David-Barrett and Oka- unlike OPEC countries.
mura (2013) who find evidence that joining EITI reduces perceived Most importantly, the analysis highlights the fact that EITI
levels of corruption. In their view, the implementation of EITI in- membership is positively associated with poorer governance
volves much more than just transparency. In particular, the es- through enhanced perceived corruption. Contrary to conventional
tablishment of a multi-stakeholder group (MSG), according to wisdom we find no evidence, thus far, that EITI has been effective
them, is a critical factor that increases accountability (David-Bar- in reducing corruption. This suggests that the EITI initiative may
rett and Okamura, 2013). well need to implement more effective measures to succeed in
The EITI initiative may fall short of bringing about account- achieving corruption control in extractive industries. Some of
ability in addition to greater transparency. Indeed, it requires a these measures may include more stringent rules on transparency
great deal from civil society organizations which are often weak in and accountability (for both countries and companies) such as the
many resource-rich countries. Transparency is most effective in disclosure of detailed and disaggregated information, or greater
generating accountability where an organized civil society is able emphasis on the expenditure of the proceeds from resource rents
to freely and critically assess complex and technical information to tackle patronage (Kolstad and Wiig, 2009). These issues not-
made available by governments and companies in the extractive
withstanding, EITI has set a new standard in the extractive in-
sector, and hold leaders responsible for decisions made. Although,
dustries, which has the potential, if adequately implemented, to
EITI constitutes an improvement in the regime of opacity that has
contribute to the development of many resource-rich countries.
characterized extractive industries, whether the conditions that
For further research, it would be interesting to undertake case
ensure greater accountability are in place in most resource-rich
studies to explore the legal, regulatory, and administrative climate
nations remains an open question (Kolstad and Wiig, 2009).14
in specific countries in order to assess the degree to which gov-
Finally, inconsistencies in the implementation of EITI high-
ernance relevant to EITI implementation can be made more
lighted by observers may have contributed to the lack of im-
effective.
provement in perceived corruption in EITI countries. International
NGOs such as Human Rights Watch and Global Witness have ap-
proved of the EITI Board expelling serial offenders such as Equa-
Acknowledgements
torial Guinea (Human Rights Watch, 2010). However these NGOs
also claim that the Board sends mixed messages by failing to
We thank audiences at the 2012 AERC Bi-annual Meetings, the
sanction countries (including the Republic of Congo, the Demo-
cratic Republic of Congo, Nigeria, and Peru) that are plagued by 2014 CSAE Conference in Oxford, and the University of Cape Town
for useful comments and discussions. Elizabeth Kasekende and
Charles Abuka gratefully acknowledge financial support from the
13
For better accountability and to build trust between governments, civil so- African Economic Research Consortium (AERC).
ciety and companies, Aguilar and Caspary (2011) argue that EITI could make a
contribution by disclosing disaggregated payments at subnational levels. In addi-
tion, according to Hilson and Maconachie (2009), EITI members should establish
solid local institutions if EITI (by itself) is to become fully effective.
14
Aaronson (2011) attributes EITI shortcomings to lack of common vision,
Appendix
undue restrictions on civil society that impede full participation in government
processes, and lack of knowledge of EITI by the public and legislatures. See Tables A1 and A2.
126 E. Kasekende et al. / Resources Policy 48 (2016) 117–128

Table A1
List of countries and EITI membership.

Country EITI membership Year of Average cor- Average

membership ruption resource
score rent (%)

Afghanistan 1 2010 4.0 1.8

Albania 1 2009 3.2 2.9
Algeria 0 – 3.1 32.3
Angola 0 – 3.8 53.6
Azerbijan 1 2007 3.6 51.2
Bahrain 0 – 2.1 26.3
Bolivia 0 – 3.1 25.8
Botswana 0 – 1.5 4.5
Brunei 0 – 2.0 53.1
Darussalam
Burkina Faso 1 2005 2.8 10.4
Cameroon 1 2007 3.5 11.3
Central African 1 2008 3.5 8.3
Republic
Chad 1 2010 3.8 35.6
Chile 0 – 1.1 17.0
Colombia 0 – 2.7 9.2
Congo, Republic 1 2008 3.5 66.6
of
Dem.Republic of 1 2008 3.9 28.5
Congo
Ecuador 0 – 3.3 19.9
Equatorial 0 – 4.0 72.0
Guinea
Ethiopia 0 – 3.2 19.6
Gabon 1 2007 3.3 48.3
Ghana 1 2007 2.6 12.1
Guatemala 1 2011 3.1 3.7
Guinea 1 2007 3.5 23.4
Honduras 0 – 3.3 4.0
Indonesia 1 2010 3.3 11.4
Iran 0 – 3.1 38.8
Iraq 1 2010 3.9 54.4
Ivory coast 1 2008 3.6 7.6
Jordan 0 – 2.3 1.9
Kazakhstan 1 2009 3.5 40.2
Kuwait 0 – 1.9 53.5
Kyrgyz Republic 1 2009 3.6 7.4
Liberia 1 2007 3.3 33.5
Libya 0 – 3.5 56.8
Madagascar 1 2008 2.7 6.5
Mali 1 2007 3.0 10.8
Mauritania 1 2007 2.9 34.4
Mexico 0 – 2.8 7.5
Mongolia 1 2007 3.0 26.5
Mozambique 1 2009 3.0 14.0
Myanmar 0 – 4.0
Namibia 0 – 2.3 2.3
Niger 1 2007 3.3 9.3
Nigeria 1 2007 3.6 31.2
Norway 1 2009 0.4 16.2
Oman 0 – 2.2 47.4
Papua New 0 – 3.6 42.8
Guinea
Peru 1 2007 2.8 9.9
Philippines 0 2013 3.2 2.9
Qatar 0 – 1.5 39.5
Russia 0 – 3.4 28.0
Sao Tome and 0 – 3.0 2.9
Principe
Saudi Arabia 0 – 2.6 49.1
Senegal 0 2013 2.8 3.3
Seychelles 0 2014 2.3 0.1
Sierra Leone 1 2008 3.4 9.9
Solomon Islands 1 2012 3.1 23.6
South Africa 0 – 2.3 6.3
Sudan 0 – 3.8 18.5
Syria 0 – 3.4 26.8
Tajikistan 0 2013 3.6 1.3
Tanzania 1 2009 3.1 8.9
Timor-Leste 1 2008 3.3 0.7
Togo 1 2010 3.4 8.4
Trinidad and 1 2011 2.7 46.4
 E. Kasekende et al. / Resources Policy 48 (2016) 117–128 127

Table A1 (continued )

Country EITI membership Year of Average cor- Average

membership ruption resource
score rent (%)

Afghanistan 1 2010 4.0 1.8

Tobago
Turkmenistan 0 – 3.9 34.6
USA 0 2014 1.0 1.4
Ukraine 0 2013 3.4 6.8
United Arab 0 – 1.4 24.0
Emirates
United Kingdom 0 2014 0.7 1.7
Uzbekistan 0 – 3.6 54.8
Venezuela 0 – 3.6 33.0
Vietnam 0 – 3.1 13.3
Yemen 1 2007 3.4 28.5
Zambia 0 2009 3.1 19.2

Table A2
Definition of the variables.

Variable Description Source

Corruption score Control of corruption index (CCI). It reflects perception of the extent to which public power is World Bank (2013a)
exercised for private gain, including both petty and grand forms of corruption, as well as capture of
the state by elites and private interests. Ranges from
2.5(weak) to 2.5 (strong) and rescaled to 0
(least corrupt) and 5 (most corrupt)
Corruption when joining EITI Level of control of corruption index for the year of EITI membership World Bank (2013a)
Aid when joining EITI Foreign aid (% GNI) for the year of EITI membership World Bank (2013b)
GDP per capita (log) Log of real GDP per Capita World Bank (2013b)
Openness (% GDP) Total trade as a percent of GDP World Bank (2013b)
Resource rent (% GDP) Natural resource rent as a percent of GDP World Bank (2013b)
Press Freedom Freedom of the Press measures the degree to which a country allows free flow of news and in- Freedom House (2011)
formation Scores range from 0 (best) to 100 (worst) and rescaled to 0 (worst) and 100 (best).
Polity2 score Index of democracy POLITY IV (2013)
British Dummy Dummy variable that takes on a value of 1 for former British Colonies and zero otherwise Treisman (2000) & CIA World Fact-
book (2011)
Federal States Dummy variable that takes on 1 for “states” and 0 otherwise Forum of Federations (2012)
EITI Membership Dummy variable that takes on a value of 1 for member countries and zero otherwise EITI Website (2015)

Cambridge University Press, New York.

CIA, n.d. The World Fact Book. 〈https://www.cia.gov/library/publications/the-
world-factbook/〉 (accessed 2011).
References Collier, Paul, Venables, Antony J., 2010. Natural Resources and State Fragility. Eur-
opean University Institute (EIU) Working Paper.
Aaronson, S.A., 2011. Limited partnership: business, government, civil society and Combes, Jean-Louis, Saadi-Sedik, Tahsin, 2006. How Does Trade Openness Influence
the public in the extractive industries transparency initiative (EITI). Public Adm. Budget Deficits in Developing Countries. IMF Working Paper.
Dev. 31, 50–63. Corrigan, Caitlin, 2014. Breaking the resource curse: transparency in the natural
Ades, Alberto, Di Tella, Rafael, 1999. Rents, competition and corruption. Am. Econ. resource sector and initiative. Resour. Policy 40, 17–30.
David-Barrett, Liz, Okamura, Ken, 2013. The Transparency Paradox: Why do Corrupt
Rev. 89 (4), 982–993.
Countries Join EITI. Working Paper. European Research Centre for Anti-Cor-
Aguilar, J., Caspary, G.V., 2011. Implementing EITI at the Subnational Level: Emer-
ruption and State Building.
ging Experience and Operational Framework. Extractive Industries for Devel-
D'Souza, Errol, 2009. Collusion in government and corruption. J. Interdiscip. Econ.
opment Series 23. World Bank, Washington DC.
21, 17–34.
Alba, Mayorga Eleodoro, 2009. Extractive Industries Value Chain. The World Bank,
EITI. Source Book, 2005, http://eiti.org/document/sourcebook (accessed January
Africa Regional Working Paper.
2015).
Alexeev, M., Conrad, R., 2009. The elusive curse of oil. Rev. Econ. Stat. 91 (3),
EITI. Website, 2015, https://eiti.org/ (accessed January 2015).
586–598. Forum of Federations, 2012. 〈http://www.forumfed.org/en/federalism/by_country/
Ali, Abdiweli M., Isse, Hodan Said, 2003. Determinants of economic corruption: a index.php〉 (accessed 2012).
cross-country comparison. Cato J. 22, 449–466. Freedom House, 2011. Freedom House. 〈https://freedomhouse.org/report/freedom-
Allison, Paul, 2014. Maximum likelihood for dynamic panel models with cross- press/freedom-press-2011#.VPwwtOn9ljo〉.
lagged effects, Mimeo. Gundlach, Erich, Martin, Paldam, 2009. A farewell to critical junctures: Sorting out
Arndt, Christiane, Oman, Charles, 2006. Development Centre Studies Uses and long-run causality of income and democracy. European Journal of Political
Abuses of Governance Indicators. OECD Publishing, Paris. Economy 25 (3), 340–354.
Ata, A. Ylmaz, Arvas, M. Akif, 2011. Determinants of economic corruption: a cross- Guo, Shenyang, Fraser, Mark W., 2014. Propensity Score Analysis: Statistical
country data analysis. Int. J. Bus. Soc. Sci. 2 (13), 161–169. Methods and Applications: Statistical Methods and Applications vol. 11. Sage
Bannon, Ian, Collier, Paul, 2003. Natural Resources and Conflict: What We Can do. Publications.
The International Bank for Reconstruction and Development/The World Bank, Hilson, G., Maconachie, R., 2009. Good governance and the extractive industries in
Washington DC. sub-saharan africa. Miner. Process. Extr. Metall. Rev. 30, 52–100.
Brunetti, Aymo, Weder, Beatrice, 2003. A free press is bad news for corruption. J. Heckman, James, 1979. Sample selection bias as a specification error. Econometrica
Public Econ. 87 (7), 1801–1824. 47 (1), 153–161.
Brunnschweiler, N. Christa, Bulte, Erwin H., 2008. The resource curse revisited and Human Rights Watch, 2010. International: Mixed Message on Resource Revenue
revised: a tale of paradoxes and red herrings. J. Environ. Econ. Manag. 55, Transparency. 〈https://www.hrw.org/news/2010/04/16/international-mixed-
248–264. message-resource-revenue-transparency〉.
Cameron, A., Trivedi, P.K., 2005. Microeconometrics: Methods and Applications. International Monetary Fund, 2007. Guide on Resource Revenue Transparency.
128 E. Kasekende et al. / Resources Policy 48 (2016) 117–128

〈http://www.imf.org/external/np/pp/2007/eng/051507g.pdf〉 (accessed January 116 (508), 1–20.

10, 2011). Moffitt, Robert A., 2009. Issues in the estimation of causal effects in population
Kaufmann, Daniel, Kraay, Aart, Mastruzzi, Massimo, 2010. The Worldwide Governance research, with an application to the effects of teenage childbearing. In: En-
Indicators: Methodology and Analytical Issues. World Bank Working Paper 5430. gelhardt, H., et al. (Eds.), Causal Analysis in Population Studies, The Springer
Kaufmann, D., Kraay, A., Mastruzzi, M., 2006. Measuring governance using cross- Series on Demographic Methods and Population Analysis.
country perceptions data. In: International Handbook on the Economics of Nickell, Stephen, 1981. Biases in dynamic models with fixed effects. Econometrica
Corruption, vol. 52. 49 (6), 1417–1426.
Kaufmann, Daniel, Aart Kraay, Pablo Zoido, 2002. Governance Matters II: Updated Ölcer, Dilan, 2009. Extracting the Maximum from the EITI. OECD Development
Indicators for 2000–01. World Bank Policy Research Working Paper No. 2772. Centre Working Paper 276.
Keblusek, Mary Ella, 2010. Is EITI really helping improve global good governance? Pitlik, Hans, Frank, Björn, Firchow, Mathias, 2010. The demand for transparency: an
In: Examining the Resource Curse, Corruption, and Nigeria's EITI Implementa- empirical note. Rev. Int. Organ. 5 (2), 177–195.
tion Experience, 〈http://nidprodev.org/EITI%20-%20Nigeria%20Analysis.pdf〉. Rosenblum, Peter, Maples, Susan, 2009. Contracts Confidential: Ending the Secret
Klitgaard, Robert, 1988. Controlling Corruption. University of California Press. Deals in the Extractive Industries. Revenue Watch Institute.
Knack, Stephen, 2001. Aid dependence and the quality of governance: cross- Sala-i-Martin, Xavier, Subramanian, Arvind, 2003. Addressing the Natural Resource
country empirical tests. South. Econ. J. 68 (2), 310–329. Curse: An Illustration from Nigeria. NBER, Working Paper No. 9804.
Kolstad, Ivar, Wiig, Arne, 2009. Is Transparency the key to reducing corruption in Sarr, M., Bulte, E., Meissner, C., Swanson, T., 2011. On the looting of nations. Public
resource-rich countries? World Dev. 37 (3), 521–532. Choice 148 (3–4), 353–380.
Kolstad, Ivar, Soreide, Tina, 2009. Corruption in natural resource management: Shen, C., Williamson, J.B., 2005. Corruption, democracy, economic freedom and
implications for policy makers. Resour. Policy 34 (4), 214–226. state strength. Int. J. Comp. Sociol. 46 (4), 327–345.
Kunicová, Jana, Rose-Ackerman, Susan, 2005. Electoral rules and constitutional Singh, Jewellord Nem, Bourgouin, France, 2013. Resource Governance and Devel-
structures as constraints on corruption. Br. J. Polit. Sci. 35 (4), 573–606. opmental States in the Global South: Critical International Political Economy
La, Porta, Rafael, Florencio, Lopez-de-Silanes., Andrei, Shleifer., Robert, Vishny, Perspectives. Palgrave Macmillan, United Kingdom.
2009. "The quality of government.". Journal of Law, Economics, and organiza- Tavares, Jöse, 2003. Does foreign aid corrupt? Econ. Lett. 79 (1), 99–106.
tion 15 (1), 222–279. Treisman, a, Daniel, a, 2007. What have we learned about the causes of corruption
Larrain, B.F., Tavares, J., 2007. Can Openness Deter Corruption? The Role of Foreign from ten years of cross-national empirical research? Annu. Rev. Polit. Sci 10,
Direct Investment. London, Centre for Economic Policy Research CEPR Discus- 211–244.
sion Paper No. 6488. 〈http://www.cepr.org/pubs/dps/DP6488.asp〉. Treisman, Daniel, 2000. The causes of corruption: a cross-national study. J. Public
Leite, Carlos, Weidmann, Jens, 1999. Does Mother Nature Corrupt? Natural Re- Econ. 76 (3), 399–457.
sources, Corruption and Economic Growth. IMF Working Paper. Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. The
Madani, Dorsati, Licetti, Martha, 2010. Business Regulation, Reform and Corruption. MIT Press, London.
The World Bank, Prem Notes Economic Policy. World Bank, 2013a. Worldwide Governance Indicators.
Mehlum, H., Moene, K., Torvik, R., 2006. Institutions and the resource curse. Econ. J. World Bank, 2013b. World Development Indicators.