Resources Policy 53 (2017) 147–163

Contents lists available at ScienceDirect

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

Oil dependence, quality of political institutions and economic growth: A MARK

panel VAR approach☆
⁎
Nikolaos Antonakakisa,b, , Juncal Cunadoc, George Filisd, Fernando Perez de Graciab
a
Department of Business and Management, Webster Vienna Private University, Praterstrasse 23, 1020 Vienna, Austria
b
Economics and Finance Subject Group, University of Portsmouth, Portsmouth Business School, Portland Street, Portsmouth PO1 3DE, United Kingdom
c
Universidad de Navarra, Facultad de Económicas y Empresariales, Departamento de Economía.Campus Universitario, 31080 Pamplona, Spain
d
Bournemouth University, Department of Accounting, Finance and Economics, Executive Business Centre, 89 Holdenhurst Road, BH8 8EB Bournemouth,
UK

A R T I C L E I N F O A BS T RAC T

JEL codes: This paper examines the resource curse hypothesis both within and between countries of different democratic
C33 footprint, based on a dynamic model that properly accounts for endogeneity issues. To achieve that, we apply a
O47 panel Vector Auto-Regressive (PVAR) approach along with panel impulse response functions to data on oil
Q32 dependence variables, economic growth and several political institutional variables in 76 countries classified by
Q33
different income groupings and level of development, over the period 1980–2012. Our results suggest that
Keywords: controlling for the quality of political institutions, and in particular the constraints to the executives, is
Resource curse important in rendering the resource curse hypothesis significant. Doing so, the resource curse hypothesis is
Oil dependence
documented mainly for developing economies and medium-high income countries. Specifically, when
Economic growth
economies from the aforementioned groups are characterised by weak quality of political institutions, then
Institutions
Panel VAR oil dependence is not growth-enhancing.

1. Introduction example, demonstrate that high endowments of oil have a positive

In their 1995 influential study titled “Natural resource abundance
and economic growth”, Sachs and Warner started a well-known line of
research focusing on natural resources. They obtained a negative
conditional relationship between economic growth and resource de-
pendence using a cross section of international data, in line with the
resource curse hypothesis. More specifically, they report that econo-
mies with abundant natural resources tend to experience lower
economic growth compared to economies with scarce natural re-
sources. Sachs and Warner (1999, 2001), Gylfason et al. (1999) and
Rodriguez and Sachs (1999), among many others, also find a negative
relationship between growth and resource abundance or dependence.1
However, the evidence in favour of the resource curse hypothesis is by
no means conclusive (see, for example, Raddatz, 2007; Brunnschweiler
and Bulte, 2008; Alexeev and Conrad, 2009; van der Ploeg and
Poelhekke, 2010, among others).2 Alexeev and Conrad (2009), for
effect on per capita Gross Domestic Product (GDP), contradicting most
of the empirical literature on the resource curse, while Brunnschweiler
and Bulte (2008) find that resource dependence does not negatively
affect growth and they define the resource curse as a “red herring”.
However, van der Ploeg and Poelhekke (2010) challenge these results
explaining that ignoring the volatility channel may lead to conclude
that there is no resource curse. These authors find that while resource
exports boosts growth in stable countries, they make especially volatile
economies even more volatile, worsening growth opportunities in these
countries.
In this study we re–examine the resource curse hypothesis in an
attempt to shed more light into that field. The resource curse hypoth-
esis literature reveals the following empirical regularities.
First, natural resource abundance is associated with various
negative development outcomes (Sachs and Warner, 1995, 1999,
2001; Brückner, 2010), although the opposite evidence is still present

☆
Juncal Cunado and Fernando Perez de Gracia acknowledge financial support from the Spanish Ministry of Economics and Competitiveness through project ECO-2014-55496R.
George Filis acknowledges the financial support from the European Union's Horizon 2020 Research and Innovation Programme, which has funded them under the Marie Sklodowska-
Curie Grant agreement no 658494.
⁎
Correspondence to: Department of Business and Management, Webster Vienna Private University, Praterstrasse 23, 1020 Vienna, Austria.
E-mail addresses: nikolaos.antonakakis@webster.ac.at (N. Antonakakis), jcunado@unav.es (J. Cunado), gfilis@bournemouth.ac.uk (G. Filis), fgracia@unav.es (F.P.d. Gracia).
1
Previously, Gelb (1988) and Auty (2002) also documented this relationship.
2
See, for example, Frankel (2010) and van der Ploeg (2011) for recent surveys.

http://dx.doi.org/10.1016/j.resourpol.2017.06.005
Received 18 July 2016; Received in revised form 28 October 2016; Accepted 6 June 2017
Available online 23 June 2017
0301-4207/ © 2017 Elsevier Ltd. All rights reserved.
N. Antonakakis et al.
(Alexeev and Conrad, 2009).
Second, existing explanations for the resource curse do not
adequately account for the role of social forces or external political
and economic environments in shaping development outcomes in
resource abundant countries, nor for the fact that, while most resource
abundant countries have performed poorly in developmental terms
(i.e., the cases of Angola and Congo, rich in oil, or the group of OPEC
countries) a few have done quite well (i.e., Norway).
Third, recommendations for overcoming the resource curse have
not generally taken into account the issue of political feasibility. More
generally, it is argued that the basic problem with the literature is that
researchers have been too reductionist since they have tended to
explain development performance solely in terms of the size and
nature of countries' natural resource endowments. Nevertheless, a
consensus is emerging that various political and social variables
mediate the relationship between natural resource wealth and devel-
opment outcomes (i.e., Isham et al., 2005; Mehlum et al., 2006a,
2006b; Andersen and Aslaksen, 2008; Bhattacharyya and Hodler,
2010; Bjorvatn et al., 2012; Collier and Goderis, 2012; El Anshasy
and Katsaiti, 2013). Even more, most of the studies have not fully
addressed the issue of endogeneity and reverse causality between the
variables of interest (Collier and Goderis, 2012). In this paper, we
address all the above issues when analysing the resource curse
hypothesis.3
Thus, more specifically, the objective of this paper is to re–examine
the dynamic links of the resource curse hypothesis both within and
between countries of different democratic footprint. To achieve this, we
apply a panel Vector Auto-Regressive (PVAR) approach along with
panel impulse response functions to data on oil dependence (approxi-
mated by oil rents as a percentage of GDP, oil share as a percentage of
GDP and oil revenue per capita),4,5 economic growth and several
political institutional variables (i.e., polity IV index and its sub-indices
and the political rights index), together with additional control vari-
ables. We consider 76 countries classified by different income group-
ings, level of development, as well as, their level of democracy over the
period 1980–2012.
Three are the main contributions of the paper to previous existing
economic literature. First, as far as the methodology is concerned,
instead of using previous methodological approaches such as cross-
section (Sachs and Warner, 1995, and many others), panel data
(Bhattacharyya and Hodler, 2010; Boyce and Emery, 2011;
Cavalcanti et al., 2011; Bjorvatn et al., 2012), panel error correction
models (Collier and Goderis, 2012) or time-varying cointegration
(Apergis and Payne, 2014) models, in this paper we estimate different
panel VAR models. To our knowledge, this is the first paper that adopts
a panel VAR approach and panel impulse response analysis to study the
dynamic impact among oil dependence, the quality of political institu-
tions, and economic growth by taking into account the endogeneity of
these variables, as well as controlling for commonly used variables in
the endogenous economic growth theory.
The advantages of using a panel VAR methodology relative to
methods previously discussed so as to examine the oil-based resource
curse hypothesis are several. First, and in contrast to cross-country,
panel data models allow us to control for unobservable time-invariant
country characteristics, reducing concerns of omitted variable bias.
3
In addition, related literature also considers the called “dutch disease’, where the
resource exports increase exchange rates and reduce the competitiveness of other
domestic exporting sectors (i.e., Sachs and Warner, 1995; Gylfason et al., 1999; Sala-i
Martin and Subramanian, 2013), nevertheless this falls outside the scope of this paper.
4
Brückner (2010) shows that the negative relationship between resource dependence
and economic growth is larger when resource dependence is calculated using a real
measure of the share of natural resource exports in GNP adjusted for cross-country
differences in the prices on non-tradables.
5
We concentrate on oil dependence, given that oil is one of the most consumed and
tradable natural resources IEA (2015). In addition, as suggested by Ross (2001), oil
resources have antidemocratic properties.
148
Resources Policy 53 (2017) 147–163
Second, time fixed effects can also be added to account for any global
(macroeconomic) shocks that may affect all countries in the same way.
Third, the inclusion of lags of the variables helps to analyse the
dynamic relationship between the different variables. Thus, impulse
response functions based on PVARs can account for any delayed effects
on and of the variables under consideration and thus determine
whether the effects between the variables of interest are short-lived,
long-lived or both. Such dynamic effects would not have been captured
by panel regressions. Fourth, and most importantly, PVARs are
explicitly designed to address the endogeneity problem, which is one
of the most serious challenges of the empirical research on the resource
curse hypothesis, by treating all variables as potentially endogenous.6
Last but not least, PVARs can be effectively employed with relative
short-time series due to the efficiency gained from the cross-sectional
dimension.
Our second contribution concerns the variables that are employed
in this study. More specifically, we include three key variables, namely,
oil dependence (proxied by three alternative indicators discussed in
detail in the Section 2.1.), economic growth and institutional quality,
together with other commonly used control variables that can poten-
tially affect economic growth (i.e., labour force participation, gross
fixed capital formation, foreign direct investment and openness). The
inclusion of all these variables, together with their interactions (please
see Section 2.1. for details), will allow us to account for the inter-
dependencies among the quality of political institutions, economic
growth and oil dependence. In order to better characterize the
relationship between these variables, we also estimate the PVAR for
different sub-groups of countries based on different characteristics,
such as, income level or developing stage, so as to check whether the
impact of institutional quality and oil dependence variables on
economic growth potentially differs among each of these sub-groups
of countries.
Finally, our third contribution is the use of a panel database for 76
countries over the period 1980–2012 to conduct a PVAR to explore for
the resource curse hypothesis. According to the literature, the results
on the resource curse hypothesis largely depend on the time frame
analyzed (Freudenburg, 1992), which calls for the need of longitudinal
research studies (Freudenburg, 1992; Easterly, 1999; Ross, 2007). In a
recent paper, Davis and Cordano (2013) also support the use of panel
data and use longitudinal data in 57 countries from 1962 to 1997 in
order to investigate whether economic growth in countries with
substantial mineral or energy extraction has a greater or lesser
tendency to be pro-poor than in countries with less extractive activity.
We believe that the use of this panel database will provide valuable new
insights regarding the resource curse hypothesis from a dynamic
perspective.
Of course, the impact of the quality of the institutions and
democracy on economic growth has also been documented in many
papers (Barro, 1999; Acemoglu et al., 2001, 2008, 2002; Papyrakis
and Gerlagh, 2004; Epstein et al., 2006; Glaeser et al., 2007;
Papaioannou and Siourounis, 2008). Furthermore, the interaction
between natural resources and economic growth, taking into
account the role of institutions has been also previously studied
by Isham et al. (2005), Mehlum et al. (2006a, 2006b), Hodler
(2006), Andersen and Aslaksen (2008), Bhattacharyya and Hodler
(2010), Bjorvatn et al. (2012), Brückner et al. (2012), Collier and
6
The endogeneity problem in cross-country and panel data models has been
previously addressed by the inclusion of different instrumental variables (Alexeev and
Conrad, 2009; Cotet and Tsui, 2013), and by estimating the model using 2 or 3 Step Least
Squares models (Brunnschweiler and Bulte, 2008; Busse and Gröning, 2013),
Generalised Method of Moments (Lederman and Maloney, 2003; Maloney and
Lederman, 2008) or Arellano-Bond Generalised Method of Moments (Yaduma et al.,
2013). The difficulty in measuring good instruments of the variables included in these
types of studies, such as oil dependence and quality of institutions, better justifies the use
of panel VAR models, which help to alleviate the endogeneity problem by treating all
variables as potentially endogenous.
N. Antonakakis et al. Resources Policy 53 (2017) 147–163

Goderis (2012) and El Anshasy and Katsaiti (2013), among many Table 1
others. Countries included in the sample.
Isham et al. (2005), for example, find that not only institutional
Panel A: Income Groups
quality has a significant effect on economic growth, but it is also
determined by the resource abundance of each of the countries. Low and Medium-Low Income
Papyrakis and Gerlagh (2004) examine the impact of natural resource
Bangladesh Bolivia Cameroon Congo Brazzaville
abundance on economic growth considering alternative transmission
Congo (Dem Rep) Egypt Ghana Guatemala
channels (corruption, investment, openness, terms of trade and school- India Indonesia Nigeria Pakistan
ing). Despite the negative relationship between natural resource Paraguay Philippines Syria Vietnam
abundance and economic growth, when these transmission channels
are included, they obtain a positive relationship between natural Yemen

resources and economic growth. Hodler (2006), on the other hand,

develop a model in which natural resources cause fighting activities Medium-High Income
between rivalling groups, while fighting reduces productive activities
and weakens property rights, and thus, production activities. According Albania Algeria Angola Argentina
Bulgaria China Colombia Cuba
to this author, apart from the natural resources’ direct positive income
Dominican Rep Ecuador Gabon Hungary
effect, natural resources have an indirect effect on income through Iran Iraq Jordan Libya
property rights, which depends on how fractionalized a country is. In Malaysia Mexico Peru Romania
addition, Mehlum et al. (2006a, 2006b) use the same dataset as Sachs Thailand Trinidad and Tunisia Turkey
and Warner (1995), including an interaction effect between quality of Tobago

institutions and resource dependence, and obtain that institutional Venezuela

quality is the key to understand the resource curse. They maintain that
when institutions are bad, resource dependence is a curse, while it is a
blessing when institutions are good. Furthermore, Andersen and High Income

Aslaksen (2008) analyze how public income shocks from natural

Australia Austria Bahrain Belgium
resources have different long run economic effects dependent on Brazil Canada Chile Denmark
constitutional designs. Using data from 90 economies divided into Finland France Germany Greece
democratic and nondemocratic countries, they find that the form of Ireland Israel Italy Japan
Kuwait New Zealand Norway Netherlands
government matters more than the democratic rule. In a recent study,
Oman Poland Portugal Qatar
Bhattacharyya and Hodler (2010) also analyse both theoretically and Korea South Russia Saudi Arabia Singapore
empirically whether and how the quality of the democratic institutions Spain Sweden Switzerland United Arab Emirates
affects the relationship between natural resources and corruption. They United Kingdom United States
confirm that the relationship between resource rents and corruption
also depends on the quality of institutions. Panel B: Level of development
Despite that there are some studies that have considered the quality
Developed
of institutions in the link between resource dependence or abundance
and economic growth, none of these studies have considered the effects Australia Austria Bahrain Belgium
of the constraints to the executives. This is rather important as there Canada Chile Denmark Finland
are cases where countries are autocratic, yet with strong constraints to France Germany Greece Ireland
the executive, which reduces the powers of the autocrat and thus, these Israel Italy Japan Kuwait
New Zealand Norway Netherlands Oman
economies may be closer to be democracies. An example of such Poland Portugal Qatar Korea South
country is Indonesia, where during the mid-60s Suharto overruled Russia Saudi Arabia Singapore Spain
Sukarno with coups d'état, yet he was committed to maintain the Sweden Switzerland Trinidad and United Arab
property rights and investments of the business sector. During Tobago Emirates
United Kingdom United States
Suharto's era the country experienced significant growth with heavy
investments in public goods and numerous reforms in the banking
sector, as well as, in import trade monopolies (Hadiz and Robison, Developing
2005). This is one of the key innovations of the paper, as it is the first
Albania Algeria Angola Argentina
study to examine the effects of these constraints in the context of the
Bangladesh Bolivia Brazil Bulgaria
natural resource hypothesis. Cameroon China Colombia Congo Brazzaville
The results of our empirical analysis, which remain sound to several Congo (Dem Rep) Cuba Dominican Rep Ecuador
robustness checks, reveal the following empirical regularities. A Egypt Gabon Ghana Guatemala
positive relationship between resource dependence and economic Hungary India Indonesia Iran
Iraq Jordan Libya Malaysia
growth is documented for the overall sample. Put differently, the
Mexico Nigeria Pakistan Paraguay
resource curse hypothesis is not present in the above case. However, Peru Philippines Romania Syria
controlling for the quality of political institutions, and more impor- Thailand Tunisia Turkey Venezuela
tantly the constraints to the executives, is important in rendering the Vietnam Yemen
resource curse hypothesis significant. Doing so, we find evidence of the
resource curse hypothesis, mainly for developing economies and Panel C: Level of democracy
medium-high income countries. Specifically, when economies from
Democracies
the aforementioned groups are characterised by weak quality of
political institutions, then oil dependence is not growth-enhancing. Albania Algeria Argentina Australia
The remainder of the paper is structured as follows. Section 2 Austria Belgium Bangladesh Bolivia
presents the PVAR methodology and the data set. Empirical results Brazil Bulgaria Canada Chile
Colombia Congo (Dem Rep) Denmark Dominican Rep
based on alternative estimations are presented in Section 3. Finally, (continued on next page)
Section 4 concludes.

149
N. Antonakakis et al. Resources Policy 53 (2017) 147–163

Table 1 (continued) seeking measure, when examining the said curse. Oil dependence has
the capability to measure this rent-seeking behaviour, given that the
Panel C: Level of democracy
higher the dependency of an economy on its resources, the higher the
Ecuador Finland France Gabon probability of the political elites to engage in a rent-seeking behaviour
Germany Ghana Greece Guatemala (see, for instance Sachs and Warner, 2001; Collier and Hoeffler, 2005).
Hungary India Indonesia Ireland A further argument in favour of using the oil dependence is that since it
Iraq Israel Italy Japan
is a ratio, it is not independent of economic policies and the institutions
Malaysia Mexico New Zealand Nigeria
Norway Netherlands Pakistan Paraguay that produce them. This particular virtue of our measure is valuable in
Peru Philippines Poland Portugal disentangling the resource curse hypothesis. Finally, studies, such as
Korea South Romania Russia Spain this by Alexeev and Conrad (2009) that have used oil abundance, as
Sweden Switzerland Thailand Trinidad and Tobago opposed to oil dependence, have also used per capita GDP. Since they
Turkey United Kingdom United States
use level series as their dependent variable, it is reasonable for them to
Non-Democracies use a resource measure that is not expressed in relative terms. In our
case, though, we use per capital GDP growth rate and thus, a relative
Angola Bahrain Cameroon China measure is capable of showing better the effects of resources on
Congo Brazzaville Cuba Egypt Iran
economic growth. We should highlight of course that Wright and
Jordan Kuwait Libya Oman
Qatar Saudi Arabia Singapore Syria Czelusta (2004) claim that the resource dependence (as a percentage of
Tunisia United Arab Venezuela Vietnam GDP) suffers from endogeneity problems and thus it should not be used
Emirates as an exogenous variable. In our case, though, we do not encounter
such problem as our PVAR model treats all variables as endogenous. To
Yemen
achieve stationarity we convert the proxies of oil dependence variables
Note: The income-country groupings, as well as, the grouping between developed and into their growth rates.
developing are based on the World Bank and United Nations classifications, respectively. c. Quality of political institutions. Again, for robustness purposes,
we use two alternative measures of political institutional quality: (i)
Polity IV index from the Polity IV project (Marshall Monty et al., 2009)
2. Data set and methodology and (ii) Political Rights index (from the Freedom House). The Polity IV
index is a commonly used proxy for institutional quality in several
2.1. Data set studies (see, for example Bhattacharyya and Hodler, 2010; Arezki and
Brückner, 2011; Bjorvatn et al., 2012; Brückner et al., 2012; El
We consider an unbalanced panel of annual data from 76 countries Anshasy and Katsaiti, 2013; Boschini et al., 2013; Caselli and Tesei,
that covers the period 1980–2012. In total we have 1471 country-year 2016). The index allocates annual scores to each country, ranging from
observations. The countries included in our dataset are listed in −10 to +10. According to the Polity IV index, values that range between
Table 1. Table 1 also divides our sample countries into the following +6 and +10 reflect democracies, with those countries scoring between
subgroups that we also examine below: developed and developing and +6 to +9 evaluated as flawed democracies, and those countries with a
different income groups. The variables used in this paper are obtained score of +10 are regarded as full democracies. On the other hand, a
from the World Bank, International Monetary Fund (IMF), US Energy score between −5 and +5 is allocated to these countries which are
Information Administration (EIA), Polity IV project and Freedom regarded as anocracies, whereas a score between −10 and −6 is given to
House (see Table 2 for a detailed description of our dataset and their autocratic regimes. The Political Rights index also approximates the
sources). quality of institutions, although it is constructed based on the
Following previous empirical related studies on natural resources responses to different questions related to the electoral process,
that also use panel models (see, for example, Bhattacharyya and political pluralism and participation and functioning of government,
Hodler, 2010; Boyce and Emery, 2011; Cavalcanti et al., 2011; and it has also been used in the literature (see, e.g., Arezki and
Bjorvatn et al., 2012, among others), we propose different specifica- Brückner, 2011). More specifically, the Political Rights index assesses
tions of PVAR models. We collect the following data: the degree of political and civil freedoms of each country. The Freedom
a. Economic growth. Following most of the papers, we use the House which reports the index allocates a score between 1 to 7,
annual real growth of per capita GDP as one of our endogenous signifying the most free to the least free countries, respectively.
variables in the analysis, which approximates the degree of the d. Interaction term. Economic and political science literature tend
countries' economic development. to include an interactive term between the quality of institutions and
b. Oil dependence. We use the following three alternative endogen- natural resource abundance or share. In our case, we use an interaction
ous variables (for robustness purposes) as proxies of oil dependence: (i) term between the oil dependence and the constraints to the executives,
oil share as a percentage of GDP, (ii) oil rents as a percentage of GDP so as to account for the interdependencies among the quality of
and (iii) oil revenues per capita. More specifically, oil share (as % of political institutions, economic growth and oil dependence. More
GDP) reports the value of crude oil exports as % of GDP, while oil rents specifically, we use an interaction term between oil dependence and
(as % of GDP) depict the difference between the value of crude oil the ‘Regulation of Chief Executive Recruitment’ (INTER_1; as defined
production at world prices and total costs of production. Finally, oil in Table 2). The Polity IV project also provides a score for the extent to
revenues reports the value of crude oil exports per capita. The choice of which institutionalised procedures are put in place for transferring
these proxies are in line with Arezki and Brückner (2011), Bjorvatn executive power (i.e. degree of the constraints to the executives) and
et al. (2012) and Apergis and Payne (2014). We interpret these which is abbreviated as XRREG. The score ranges between 1 to 3, with
variables as a measure of the oil dependence, rather than as a measure 1 denoting no constraint to the executives (i.e. self-selection by seizure
of oil abundance, which could be better proxied by a stock measure of power), 2 denoting a transition stage (i.e. informal competition with
(see, for instance, Brunnschweiler and Bulte, 2008, for a discussion of an elite group or restricted elections) and 3 denoting constraints to the
the differences between resource abundance, resource rents and executives (i.e. via birthright, competitive election, dual executives
resource dependence). We maintain that for the examination of the where ascriptive and designated rulers coexist). This interaction term is
resource curse hypothesis we need a variable that can capture the the third dependent/endogenous variable that we use in the extended
extent to which political elites exhibit a rent-seeking behaviour. As PVAR model.
such, there is a body in the literature that argues in favour of a rent- e. Exogenous control variables. In order to avoid any potential

150
N. Antonakakis et al. Resources Policy 53 (2017) 147–163

Table 2
Variable description and sources.

Name Description Source Notes

Economic Growth rate of real per IMF Log difference of per capita GDP (in PPP,
Growth capita GDP (GDPPCGR) constant 2005 intnl $)

Oil rents Oil rents (as % of GDP) World Bank Difference between the value of crude oil
production at world prices and total costs
of production

Oil share Oil share (as % of GDP) IMF, EIA Value of crude oil exports as % of GDP.

Oil revenue Oil revenue per capita IMF, EIA Value of crude oil exports per capita

Polity IV Rating based on a +10 Polity IV Substracting the AUTOC score from the
(strongly democratic) to project DEMOC score in the Polity IV database
−10 (strongly autocratic)
scale

Xrreg Rating based on a 1–3 Polity IV It is a component of the Polity IV index, and
scale project measures the “Regulation of Chief Executive
Recruitment” mechanism

Political Rating based on a 1–7 Freedom The ratings process is based on a checklist of
rights scale House 10political rights questions related to the
electoral process, political pluralism and
participation and functioning of government

Democracy Dummy variable Polity IV Countries are classified according to the

status project Polity IV index in democracies (Polity IV
scores between 6 and 10), and anocracies/
autocracies (Polity IV scores between −10
and 5)

Xrred Interactive term IMF, EIA, Calculated as the product of Xrreg

× (INTER_1) Polity IV and Oil share

Oil share

Developing Category World Bank Countries are classified according to their degree
of development, based on World Bank data

GFCF Gross fixed capital World Bank Expressed as percentage of GDP

formation

Trade openness Trade openness World Bank The sum of exports and imports as a percentage of GDP

LPFR Labour force World Bank Expressed as a percentage of total population of ages
participation rate 15+

Note: Annual data from 76 countries for the period 1980–2012.

omitted variable bias, we also control for several exogenous variables
typically used in the endogenous growth theory, namely, labour force
participation, gross fixed capital formation and openness. Labour force
participation and gross fixed capital formation approximate the labour
and capital productivity factors, as in a Cobb–Douglas production
function, which have been established to affect economic growth.
Finally, trade openness captures the openness of the economy to
international trade and thus the effects of international competitive-
ness on economic growth.
2.2. PVAR
This paper uses data from 76 economies for the period 1980–2012.
The PVAR methodology we employ, originally developed by Holtz-
Eakin et al. (1988), extends the traditional VAR model introduced by
Sims (1980), which treats all the variables in the system as endogen-
ous, with the panel-data approach, which allows for unobserved
individual heterogeneity. In its general form, the PVAR model can be
expressed as follows:
Yit = A0 + A1Yit −1 + A2 Yit −2 + ⋯ + Aj Yit − j + BXit + μi + λt + εit
where Yit is a vector of our dependent/endogenous variables, namely real
per capita economic growth and oil dependence (proxied by either oil share
as a % of GDP, oil rents or oil revenue per capita). The autoregressive
structure allows all endogenous variables to enter the model with a number
of j lags. Xit is a vector of the exogenous variables (commonly used in
endogenous growth models) comprising: (i) gross fixed capital formation as
a % of GDP, measuring capital input, (ii) imports plus exports as a % of
GDP, capturing the degree of openness, and (iii) labour force participation,
capturing human capital. μi accounts for the unobservable country
characteristics (country fixed-effects) and λt accounts for any global shocks
that may affect all countries in the same way (time fixed-effects). Finally, εit
denotes the error term.
As indicated above, our benchmark specification is a PVAR that
contains the real per capita GDP growth rate and a proxy of oil
dependence, as well as exogenous variables and country- and time-
fixed-effects, for the full sample. However, we also estimate the same
model using only autocratic countries, so as to discover if the quality of
political institutions impacts on the oil dependence-economic growth
relationship. Furthermore, we extend the benchmark model to a PVAR
with the inclusion of an additional dependent/endogenous variable
(1) that captures the effects of the constraints to the executives on the

151
N. Antonakakis et al.
Table 3
Block exogeneity/Granger-causality tests.
Dependent variable
Economic growth rate Oil share INTER_1
Economic growth 11.55** 3.40
rate (excluded)
Oil share (excluded) 16.38*** 13.55***
INTER_1 (excluded) 10.04** 6.19
All variables 28.34*** 17.44** 15.17*
Note: The numbers in the table are the Chi-square block exogeneity Wald tests. Under
the null hypothesis, the excluded variables do not Granger-cause the dependent variable.
* **
, and *** denotes significance at the 10%, 5% and 1% level.
resource curse hypothesis (see, interaction term INTER_1). Thus, we
allow for all these variables to be endogenous, addressing one of the
main empirical problems of the related literature.
In fact, as a first step, and in order to justify the methodology used
in this paper, we pursued Block exogeneity tests, as a test for the
endogeneity/exogeneity of the key variables in the study. A variable is
said to Granger cause another variable if there is enough evidence to
reject the null hypothesis that the coefficients on the lags of the vector
of variables Ykt−j in the PVAR equation of Yit, where i ≠ k , are all equal
to zero. The results of this test reported in Table 3, provide evidence of
causality among the three variables (i.e., economic growth, oil depen-
dence and quality of institutions), suggesting that these variables
should be treated as endogenous.7 This is the approach that we follow
in this study.
In order to get a more complete picture of the dynamic interactions
among oil dependence, economic growth and political institutions, we
perform a panel generalised impulse-response function (PGIRF)
analysis, in order to assess the speed of adjustments to shocks
originating in our aforementioned variables. The panel generalised
impulse response function analysis employed, which is based on Koop
et al. (1996) and Pesaran and Shin (1998), provides a natural solution
when theory does not provide a clear cut guidance on the ordering of
the aforementioned endogenous variables, as in our case. Moreover,
the PGIRFs are also decomposed into the responses of shocks to
specific variables by taking out from the PGIRFs the effects of shocks to
all other variables (Koop et al., 1996), which gives us further insights
into the mechanisms at work.
3. Empirical results
3.1. Descriptive statistics and causality tests
In Table 4, we present the descriptive statistics of our main
variables for the full data sample (i.e., 76 countries between 1980
and 2012). It is evident from this table that, real GDP per capita growth
averaged at 1.67% and the oil dependence variables averaged between
8.40% and 10.09%. Compared to real GDP per capita growth, the oil
dependence variables are more volatile. On average, the countries in
the sample are characterised by high degree of openness (72.04%),
abundant human capital (59.87%) and moderate capital input
(21.91%). According to the panel unit root test, all series are stationary,
indicating the appropriateness of using them in the PVAR analysis.8
7
The Granger-causality results for the subsample groups, which are qualitatively
similar, are available from the authors upon request.
8
The results for the subgroups of countries and proxies of oil dependence and
institutional quality, point towards similar conclusions. Thus, for the sake of brevity,
these are not presented but are available upon request from the authors.
152
Resources Policy 53 (2017) 147–163
3.2. Panel generalised impulse response functions: full sample
analysis
Based on the estimation of Eq. (1), with a lag order of (up to) 4,
determined by the Akaike Information Criterion (AIC) and Schwarz
Bayesian Criterion (SBC), we first calculate the generalised panel
impulse response functions tracing out the reaction of real per capita
GDP growth to a shock on oil share and vice versa.9
For brevity, we report only the PGIRFs, where a positive (negative)
response of one variable to a positive shock of another variable is
depicted by a response line above (below) zero.10 The two dashed lines
(above and below the solid response line) show the 95% confidence
intervals.
Fig. 1 (Panel A) depicts the dynamic path of adjustment to a shock
on oil share in year 1 and in subsequent periods (up to 15 years) based
on a PVAR model with only these two endogenous variables, as well as
the exogenous variables (i.e., labour force participation, openness and
gross fixed capital formation).
Our results indicate that oil share tends to have a positive effect on
per capita real GDP growth in the long-run (up to 15 years).
Furthermore, we observe that a positive shock to the per capita real
GDP growth triggers a positive response from the oil share, yet only
short-lived (up to 2 years), as it becomes insignificant thereafter. This is
suggestive of the fact that, based on the full sample estimation, higher
levels of oil share lead to higher economic growth, contrary to the
empirical evidence of the resource curse hypothesis (as in Raddatz
(2007), Alexeev and Conrad (2009), Brunnschweiler and Bulte (2008)).
In addition, we report for the first time that there is a feedback
mechanism from economic growth to the oil share, which might
suggest that economic growth could lead to better exploitation of oil
resources and thus increase the oil share for a country, which points
again to the endogeneity of the oil dependence variables in this type of
studies (Collier and Goderis, 2012).
In order to analyse the role of the institutions, we estimate the
previous PVAR model distinguishing between democratic and auto-
cratic countries and the results are displayed in Panel B of Fig. 1.
Interestingly enough, the response of per capita real GDP growth to a
positive shock to oil share, considering only the autocracies, is still
positive, however, only in the short-run. In addition, the magnitude of
the response is lower compared to the PGRIFS when all countries are
considered (Panel A). This fact suggests that oil share has a higher
positive effect on economic growth in democratic rather than in
autocratic countries. Similarly, a positive shock to economic growth
is not translated into a positive response from the oil share of the
autocracies, as evident by the insignificant response of INTER_1 to a
per capita real GDP growth shock. Even though we do not report a
negative relationship between oil share and economic growth for the
non-democratic countries (as in Sachs and Warner (1999, 2001),
Gylfason et al. (1999), Rodriguez and Sachs (1999)), this finding
allows us to confirm the resource curse hypothesis and the role of
the institutions in explaining the oil dependence and economic growth
relationship (as in Isham et al. (2005), Mehlum et al. (2006a, 2006b)).
Although the findings for the resource curse hypothesis have been
previously reported, we also show for the first time that the reverse
causality is still evident, yet only for democracies. Thus, the role of
institutions in explaining the positive significant relationship from
economic growth to oil rents adds a new channel through which
institutions should be considered when analysing the resource curse
hypothesis.
This result suggests that the institutional channel is key to under-
stand the relationship between resource dependence and economic
9
Post-PVARs estimation of misspecification tests reveal no evidence of residual
autocorrelation and heteroscedasticity. In addition, the stability conditions are satisfied.
These results are available upon request from the authors.
10
Full estimation results of the various PVAR models are included in the Appendix.
N. Antonakakis et al. Resources Policy 53 (2017) 147–163

Table 4
Descriptive statistics.

Mean Minimum Maximum Std. Dev. Skewness Kurtosis J-B LLC

GDPPCGR 1.6688 −43.0161 40.5673 4.5809 −1.2640 14.4295 13194.21* −20.93*

OIL_SHARE 8.4982 0.0000 98.8086 14.4599 3.0219 14.0225 9685.52* −28.07*
OIL_RENT 10.0863 0.0000 80.2375 14.9963 1.8375 5.8979 1852.66* −24.53*
OIL_REVENUE 8.4171 0.0001 245.0232 20.1332 5.8563 50.3884 154090.8* −27.17*
GFCF 21.9134 2.1000 59.7324 6.1869 0.5755 5.2960 618.13* −4.80*
OPENNESS 72.0425 6.3203 439.6567 49.4087 3.1378 18.3525 26877.76* −39.86*
LPFR 59.8670 15 86.7 9.5022 −0.2659 3.3369 41.4322* −5.61*

Note: * denotes significance at the 1% level. J-B denotes the Jarque-Bera test for normality. LLC is the panel unit root test (with just a constant) of Levin, Lin and Chu (2002), which test
the null hypothesis of a unit root, against the alternative that the panel is stationary.

Fig. 1. Cumulative generalised impulse response functions: Full sample.

growth, and it adds new and valuable evidence to an outcome that had
already been found in the literature, although using different sample of
countries, data span and methodologies (see for example Bulte et al.,
2005; Mehlum et al., 2006a, 2006b; Brunnschweiler and Bulte, 2008;
Alexeev and Conrad, 2009; Davis and Cordano, 2013).
Next, we assess whether the results are different when we take into
consideration the degree of constraints on the executive (as approxi-
mated by the xrreg variable; as defined in Table 2). As aforementioned,
this is a rather important innovation of this study, as constraints to the
executives could “force” an autocratic regime to exhibit democratic

153
N. Antonakakis et al.
Fig. 2. Cumulative generalised impulse response functions: Developing.
traits.
To capture the effects of these constraints we employ our
interaction term (INTER_1; as defined in Table 2). The results are
reported in Panel C of Fig. 1. We notice, that the response of the
economic growth to a positive shock to oil share, given an autocracy
with high constraints on the executive (INTER_1), is of higher
magnitude compared to the response in Panel B. However, the effect
is of lower magnitude (and limited to a shorter time period) than the
obtained in Panel A, suggesting again that the positive relationship
between oil share and economic growth is higher for democratic
countries and countries with constraints to the executive. This result
justifies again the inclusion of quality of institutions as a channel
154
Resources Policy 53 (2017) 147–163
through which oil rents may influence economic growth (Mehlum
et al., 2006a, 2006b). In addition, the oil share for those autocracies
with high constraints to the executive responds positively (although
only in the short-run) to an economic shock, as evident by the right
PGIRF of Panel C.
3.3. Panel generalised impulse response functions: subgroup analysis
In this section we analyse the robustness of our results by means of
estimating previous specifications of the PVAR for different subgroups
of countries (as classified in Table 1).
First, we estimate the PVAR model for developing and developed
N. Antonakakis et al.
Fig. 3. Cumulative generalised impulse response functions: Developed.
countries, and display the results in Figs. 2 and 3, respectively. As
shown in the Panels A of Figs. 2 and 3, oil share leads to higher growth
in both developing and developed countries, although this effect is of a
lower magnitude for the developed countries (see left PGIRFs of Panels
A of Figs. 2 and 3). This is rather expected given that the oil sector in
the developed countries may not be a key sector for the economic,
whereas the reverse is true for the developing economics. Indicatively,
the oil revenues in Venezuela account for about 25% of the country's
GDP and 95% of its exports, whereas for the UK, the same ratios are
about 1.2.
Again, when we focus only on the autocratic countries we find that oil
share does not have any effect on the economic growth of autocracies in
developing countries (see left PGIRF in Panel B of Fig. 2), however; a
155
Resources Policy 53 (2017) 147–163
short-run positive response is evident for the autocratic developed
countries (see left PGIRF in Panel B of Fig. 3). In addition, we observe
no significant response from the oil share in the autocratic developing
countries to positive economic shocks, whereas a marginal and extremely
short-run positive response is observed for the autocratic developed
economies (see right PGIRFS in Panels C of Figs. 2 and 3). Once again
when we consider the constraints to the executives (see Panels C of Figs. 2
and 3) we observe for the case of the autocratic developing economies that
there is indeed a positive response of the economic growth to a positive
shock of the oil share. Thus, we show again that in the case developing
economies with autocratic regimes and low constraints to the executives
the resource curve is evident. We should highlight here again that it is
important to consider the constraints to the executives in this line of
N. Antonakakis et al.
Fig. 4. Cumulative generalised impulse response functions: Low and Medium-Low Income group.
research, as simply distinguishing between autocratic and democratic
regimes, does not reveal the full dynamics of the relationship between oil
dependence and economic growth.
Finally, we control for the income group (low/medium-low, med-
ium-high and high income countries) when analysing the resource
curse hypothesis. The results are shown in Figs. 4, 5 and 6.
We find evidence of heterogeneous responses among the different
income groups. In particular, there is no significant relationship
between oil share and economic growth in the low/medium-low
income group and this finding is robust even when we consider only
the autocratic countries or the interaction term (see Fig. 4).
Turning to the medium-high income group of countries we
156
Resources Policy 53 (2017) 147–163
document that the positive long run effects of oil share are mainly
driven by democratic countries (left PGIRFs in Panels A and B in
Fig. 5) or by autocracies that have in place significant constraints to
the executive (see left PGIRF in Panel C of Fig. 5). The reverse
causality (i.e. from the economic growth to oil share) does not exist
for this income group.
More importantly, we find bidirectional relationship between oil
share and economic growth for the high-income countries group,
although these effects are short-lived. This finding holds true for all
specifications (see PGIRFs of Fig. 6).
Overall, our results suggest that controlling for the quality of
political institutions, and more importantly the constraints to the
N. Antonakakis et al.
Fig. 5. Cumulative generalised impulse response functions: Medium-High Income group.
executives is important in rendering the resource curse hypothesis
significant. Doing so, the resource curse hypothesis is documented
mainly for developing economies and medium-high income countries.
Different arguments could explain this result. Specifically, when
economies from the aforementioned groups are characterised by weak
quality of political institutions (autocracies with limited constraints to
the executive), oil dependence is not growth-enhancing. This might
suggest that these autocrats or the political elite exploit the benefits of
the country's oil resources to accommodate their own rent-seeking
behaviour, without considering the potential positive long-run benefits
to the wider economy.11 Gylfason (2001), for example, argues that
11
Last but not least, our main results are robust to different proxies of oil dependence,
economic growth (growth rates, GDP per capita growth, 5-year period growth rates) and
157
Resources Policy 53 (2017) 147–163
nations that consider their natural resources to be their most important
asset may neglect the development of other resources, such as
education, due to the rent-seeking behaviour. Furthermore, this
relationship between resource dependence and economic growth in
countries with low quality institutions could be explained by the drag
effect (Alexeev and Conrad, 2009; Davis, 2011). Davis (2011), for
example, shows that when the resource drag is modelled (by including
the variable “change in mineral production” as an explanatory variable
to the Sachs and Warner studies), the evidence for the resource curse
hypothesis is weaker, while the results do not reject the drag effect
(footnote continued)
quality of political institutions (polity index and the freedom house political rights index).
For the sake of brevity we do not report these results, which are, however, available upon
request from the authors.
N. Antonakakis et al.
Fig. 6. Cumulative generalised impulse response functions: High Income group.
hypothesis. In any case, the existence of high-quality institutions (those
able to create positive incentives for entrepreneurial growth) is crucial
to translate the benefits from oil to productive activities (Mehlum et al.,
2006a, 2006b).
4. Conclusions
In this paper we shed more light to the contested literature on the
resource curse hypothesis, by estimating a PVAR approach along
with PGIRFs to data on oil dependence, economic growth and
several political institutional variables for 76 countries grouped
under different income groupings and level of development, over
the period 1980–2012. To our knowledge, this is the first paper that
adopts a PVAR and PGIRFs analyses, to study the impact of oil
158
Resources Policy 53 (2017) 147–163
dependence on economic growth taking into account the endogene-
ity of institutional quality, as well as controlling for commonly used
indicators in the growth literature in order to shed more light into
the natural resource curse hypothesis. The use of this methodology
allow us to control for cross-country unobservable heterogeneity,
account for time fixed-effects, analyse the dynamic relationship
between the different variables, and most importantly, to address
the endogeneity problem often found in these type of studies.
Furthermore, it is the first study to consider how the constraints to
the executives could influence the relationship between oil depen-
dence and economic growth.
The results of our empirical analysis reveal the following regula-
rities. First, we document the need of considering per capita real GDP
growth, oil dependence and quality institutions as endogenous vari-
N. Antonakakis et al. Resources Policy 53 (2017) 147–163

ables, which justifies the use of panel VAR models in analysing the
relationship between these variables. Second, we find significant
evidence that positive oil share shocks are growth-enhancing, when
we do not account for institutional quality, suggesting thus, evidence
against the resource curse hypothesis in that case. Third, controlling for
the quality of political institutions seems important in rendering the
resource curse hypothesis significant. Doing so, the resource curse
hypothesis is documented mainly for developing economies and
medium-high income countries. Specifically, when economies from
the aforementioned groups are characterised by weak quality of
political institutions, then oil dependence is not growth-enhancing.
These findings are important to support the new consensus appearing
in the literature that resource dependence is neither a curse or a blessing:
all depends on how a country is prepared to take advantage of its natural
resources, and quality of institutions is a key determinant of how each of
the countries will use its resources to promote economic growth. That is,
while in countries with high quality institutions, having resources can be a
blessing, in those countries with low quality institutions, resources can
become a curse for economic growth.
These results are robust to different proxies of oil dependence,
economic growth (overall GDP growth, 5-year period growth rates) and
quality of political institutions (polity index and the freedom house
political rights index).
Overall, our findings, based on the suggested dynamic approach
that deals with a number of issues in the estimation process, provide
new insights in the resource curse hypothesis. Moreover, our analysis
shows that the resource curse hypothesis is mainly driven by the
quality of political institutions, but more importantly, the constraints
imposed to the executives. This suggests that the natural resource
hypothesis hold true for autocracies with limited constraints to the
executive.
Acknowledgements
We would like to thank the editor (Gary A. Campbell) and one
anonymous reviewer for their very insightful comments that helped to
improve a previous version of this paper. We also like to thank the
participants of the Kingston University London Economic Staff
Seminar Series for their valuable comments. The usual disclaimer
applies.

Appendix

Table A.1–Table A.6.

Table A.1
PVAR estimates: Full Sample.

All Countries Only Autocracies Only Autocracies with the interaction terms

Endogenous variables

GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE INTER_1

GDPPCGR(−1) 0.386*** −0.051 0.384*** −0.125 0.414*** 0.069 −0.353

GDPPCGR(−2) 0.117*** −0.081 0.122*** 0.019 0.052 −0.174 0.241
GDPPCGR(−3) 0.017 −0.055 0.016 −0.076 0.050 0.088 0.073
GDPPCGR(−4) 0.024 0.311 0.022 −0.046 0.006 0.208 −0.196
OIL_SHARE(−1) 0.004* −0.217*** 0.003 −0.312*** 0.001 −0.180*** −0.031
OIL_SHARE(−2) −0.001 −0.144*** −0.006 −0.224*** 0.001 −0.095*** −0.022
OIL_SHARE(−3) −0.002 −0.022 −0.015*** −0.155*** 0.003 0.052 −0.039
OIL_SHARE(−4) 0.001 0.031 −0.012*** 0.153*** 0.004 −0.032 −0.007
INTER_1(−1) 0.002 −0.073*** −0.275***
INTER_1(−2) −0.004 −0.066*** −0.214***
INTER_1(−3) −0.006*** −0.102*** −0.115***
INTER_1(−4) −0.004* 0.078*** 0.151***
Constant −0.009 −0.012 −0.011 0.077 −0.014 0.007 0.189

Exogenous variables

GFCF 0.001*** −0.008*** 0.001*** −0.005*** 0.001*** −0.011*** −0.015***

OPENNESS 0.001 0.001*** 0.001 0.001*** 0.001 0.002*** 0.002***
LPFR 0.001 0.003* 0.001 0.001 0.001 0.003** 0.001

Adj. R-squared 0.257 0.061 0.263 0.181 0.253 0.087 0.158

F-statistic 32.991 6.974 25.263 16.042 22.530 7.051 12.969
Akaike information criterion −2.556 −3.077 −1.477
Schwarz criterion −2.441 −2.844 −1.232

Note: In each country-grouping specification, the lag-length of the VAR was determined by the the AIC and BIC criteria. *, ** ***
, denotes significance at the 10%, 5% and 1% level.

159
N. Antonakakis et al. Resources Policy 53 (2017) 147–163

Table A.2
PVAR estimates: Developing Countries.

All Countries Only Autocracies Only Autocracies with the interaction terms

Endogenous variables

GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE INTER_1

***
GDPPCGR(−1) 0.365 −0.102 0.365 ***
−0.203 0.386 ***
−0.134 −0.524
GDPPCGR(−2) 0.111*** −0.191 0.113*** −0.051 0.015 −0.414 −0.043
GDPPCGR(−3) 0.009 −0.111 0.008 −0.183 0.041 0.129 −0.388
GDPPCGR(−4) 0.038 0.324 0.034 0.081 0.023 0.271 0.244
OIL_SHARE(−1) 0.010*** −0.146*** −0.003 −0.186*** 0.007 −0.119*** −0.037
OIL_SHARE(−2) 0.006 −0.063 0.008 −0.147*** 0.005 0.034 −0.010
OIL_SHARE(−3) 0.001 0.142*** −0.015 0.173*** 0.008 0.109*** −0.005
OIL_SHARE(−4) 0.001 0.000 −0.018** −0.088** 0.004 0.035 −0.013
INTER_1(−1) 0.002 −0.044 −0.172***
INTER_1(−2) 0.003 −0.095*** −0.165***
INTER_1(−3) −0.005 0.027 0.175***
INTER_1(−4) −0.005 −0.066 −0.088**
Constant −0.019* 0.059 −0.020* 0.069 −0.028 0.069*** 0.124

Exogenous variables

GFCF 0.001*** −0.007*** 0.001*** −0.005*** 0.001*** −0.008*** −0.010***

OPENNESS 0.000 0.001*** 0.000 0.001*** 0.000 0.001*** 0.002***
LPFR 0.000 0.001 0.000 0.000 0.001** 0.001 0.000

Adj. R-squared 0.248 0.059 0.251 0.125 0.233 0.075 0.121

F-statistic 19.752 4.600 15.012 6.974 12.363 4.016 6.139
Akaike information criterion −2.751 −3.497 −2.008
Schwarz criterion −2.581 −3.157 −1.638

Note: In each country-grouping specification, the lag-length of the VAR was determined by the the AIC and BIC criteria. *, ** ***
, denotes significance at the 10%, 5% and 1% level.

Table A.3
PVAR estimates: Developed Countries.

All Countries Only Autocracies Only Autocracies with the interaction terms

Endogenous variables

GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE INTER_1

GDPPCGR(−1) 0.493*** −0.252 0.465*** −0.419 0.461*** −0.131 −1.074

GDPPCGR(−2) 0.123*** 0.591 0.136*** 0.589 0.134*** 0.787 1.608
GDPPCGR(−3) 0.038 0.371 0.042 0.124 1.022
OIL_SHARE(−1) 0.003 −0.174*** 0.018*** −0.335*** −0.002 −0.287*** −0.036
OIL_SHARE(−2) −0.005*** −0.140*** −0.009** −0.391*** −0.002 −0.190*** −0.047
OIL_SHARE(−3) −0.011*** −0.249*** 0.001 −0.031 −0.060
INTER_1(−1) 0.006*** −0.026 −0.295***
INTER_1(−2) −0.004** −0.102*** −0.368***
INTER_1(−3) −0.005*** −0.103*** −0.209***
Constant 0.019 −0.239 0.021 0.158 0.022 −0.334 0.358

Exogenous variables

GFCF 0.000 −0.020*** 0.000 −0.018*** 0.000 −0.016*** −0.042***

OPENNESS 0.000 0.000 0.000 0.001*** 0.000 0.001 0.002***
LPFR 0.000 0.012*** 0.000 0.003 0.000 0.011*** 0.007

Adj. R-squared 0.349 0.061 0.355 0.344 0.347 0.167 0.302

F-statistic 34.493 5.070 19.939 19.017 19.291 7.899 15.900
Akaike information criterion −2.302 −2.562 −0.937
Schwarz criterion −2.153 −2.182 −0.558

Note: In each country-grouping specification, the lag-length of the VAR was determined by the the AIC and BIC criteria. *, ** ***
, denotes significance at the 10%, 5% and 1% level.

160
N. Antonakakis et al. Resources Policy 53 (2017) 147–163

Table A.4
PVAR estimates: Low and Medium-Low Income Group.

All Countries Only Autocracies Only Autocracies with the interaction terms

Endogenous variables

GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE INTER_1

*** *** ***

GDPPCGR(−1) 0.504 0.030 0.309 0.001 0.429 0.478 0.218
GDPPCGR(−2) 0.048 −0.465 0.002 0.000 0.029 −0.342 −0.011
GDPPCGR(−3) 0.156*** −0.260 0.168*** −0.004 0.151*** 0.039 −0.156
GDPPCGR(−4) −0.006 0.776
OIL_SHARE(−1) −0.004 −0.229*** 0.548*** 0.838*** 0.005 −0.207** −0.073
OIL_SHARE(−2) −0.009 −0.184*** −0.038 −0.026 −0.016 −0.212** −0.094
OIL_SHARE(−3) −0.010 0.129** −0.172 0.025 −0.003 −0.130 −0.039
OIL_SHARE(−4) 0.002 −0.076
INTER_1(−1) −0.005 −0.068 −0.258***
INTER_1(−2) 0.009 0.029 −0.048
INTER_1(−3) −0.001 0.195*** 0.303***
Constant 0.002 −0.117 −0.001 0.000 0.001 −0.035 −0.117

Exogenous variables

GFCF 0.000 −0.010*** 0.000 0.000 0.000 −0.017*** −0.011*

OPENNESS 0.000 0.002*** 0.000 0.001** 0.000 0.003*** 0.003**
LPFR 0.000 0.003 0.000 0.000 0.000 0.003 0.003

Adj. R-squared 0.372 0.109 0.422 0.726 0.279 0.124 0.150

F-statistic 12.784 3.430 14.991 51.743 7.929 3.527 4.158
Akaike information criterion −3.002 −8.379 −1.930
Schwarz criterion −2.632 −7.798 −1.320

Note: In each country-grouping specification, the lag-length of the VAR was determined by the the AIC and BIC criteria. *, ** ***
, denotes significance at the 10%, 5% and 1% level.

Table A.5
PVAR estimates: Medium-high Income Group.

All Countries Only Autocracies Only Autocracies with the interaction terms

Endogenous variables

GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE INTER_1

GDPPCGR(−1) 0.312*** −0.130 0.311*** −0.214 0.339*** −0.279 −0.701

GDPPCGR(−2) 0.096*** −0.029 0.100*** 0.015 0.011 −0.192 0.158
GDPPCGR(−3) −0.016 −0.002 −0.012 −0.180 0.012 0.434 −0.450
GDPPCGR(−4) 0.032 0.300 0.031 0.080 0.019 0.193 0.437
OIL_SHARE(−1) 0.017*** −0.116*** 0.012 −0.129*** 0.011 −0.102* −0.021
OIL_SHARE(−2) 0.012* −0.018 0.017 −0.191*** 0.010 0.057 −0.001
OIL_SHARE(−3) 0.006 0.107*** −0.016 0.104** 0.012 0.089 −0.012
OIL_SHARE(−4) 0.000 0.022 −0.019 −0.062 0.001 0.053 0.001
INTER_1(−1) 0.008 −0.021 −0.113**
INTER_1(−2) 0.004 −0.109*** −0.222***
INTER_1(−3) −0.002 0.003 0.127***
INTER_1(−4) −0.002 −0.057 −0.051
Constant −0.043*** 0.168 −0.043*** 0.149*** −0.059*** 0.192 0.330***

Exogenous variables

GFCF 0.001*** −0.009*** 0.001*** −0.006*** 0.001*** −0.010*** −0.012***

OPENNESS 0.000 0.001*** 0.000 0.001*** 0.000 0.001*** 0.001***
LPFR 0.001*** 0.000 0.001*** −0.001 0.001*** 0.000 −0.002

Adj. R-squared 0.231 0.043 0.233 0.101 0.251 0.047 0.100

F-statistic 11.956 2.639 9.164 4.003 9.316 2.229 3.767
Akaike information criterion −2.635 −3.540 −1.940
Schwarz criterion −2.397 −3.064 −1.436

Note: In each country-grouping specification, the lag-length of the VAR was determined by the the AIC and BIC criteria. *, ** ***
, denotes significance at the 10%, 5% and 1% level.

161
N. Antonakakis et al. Resources Policy 53 (2017) 147–163

Table A.6
PVAR estimates: High Income Group.

All Countries Only Autocracies Only Autocracies with the interaction terms

Endogenous variables

GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE GDPPCGR OIL_SHARE INTER_1

***
GDPPCGR(−1) 0.426 −0.216 0.444 ***
−0.374 0.440 ***
−0.157 −0.965
GDPPCGR(−2) 0.108** 0.037 0.145*** 0.675 0.143*** 0.699 1.826
GDPPCGR(−3) 0.056 0.145 0.014 0.425 0.019 0.077 1.171
GDPPCGR(−4) −0.058 0.723
OIL_SHARE(−1) 0.000 −0.308*** 0.019*** −0.340*** −0.002 −0.288*** −0.035
OIL_SHARE(−2) −0.006*** −0.241*** −0.008* −0.392*** −0.002 −0.189*** −0.052
OIL_SHARE(−3) −0.004 −0.135*** −0.011*** −0.254*** 0.001 −0.031 −0.061
OIL_SHARE(−4) 0.001 −0.001
INTER_1(−1) 0.006*** −0.028 −0.302***
INTER_1(−2) −0.004** −0.103*** −0.370***
INTER_1(−3) −0.005*** −0.104*** −0.215***
Constant 0.032*** −0.492 0.026 0.224 0.027 −0.271 0.514

Exogenous variables

GFCF 0.000 −0.012 0.000 −0.022*** 0.000 −0.018*** −0.052***

OPENNESS 0.000 0.000 0.000 0.001*** 0.000 0.001 0.003***
LPFR 0.000 0.013*** 0.000 0.003 0.000 0.011*** 0.007

Adj. R-squared 0.259 0.101 0.318 0.351 0.309 0.167 0.312

F-statistic 12.597 4.739 16.082 18.506 15.452 7.461 15.671
Akaike information criterion −2.618 −2.484 −0.861
Schwarz criterion −2.362 −2.086 −0.464

Note: In each country-grouping specification, the lag-length of the VAR was determined by the the AIC and BIC criteria. *, ** ***
, denotes significance at the 10%, 5% and 1% level.

Appendix A. Supplementary data

Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.resourpol.2017.06.005.

References
Acemoglu, D., Johnson, S., Robinson, J.A., 2001. The colonial origins of comparative
development: an empirical investigation. Am. Econ. Rev. 91 (5), 1369–1401.
Acemoglu, D., Johnson, S., Robinson, J.A., 2002. Reversal of fortune: geography and
institutions in the making of the modern world income distribution. Q. J. Econ. 117
(4), 1231–1294.
Acemoglu, D., Johnson, S., Robinson, J.A., Yared, P., 2008. Income and democracy. Am.
Econ. Rev., 808–842.
Alexeev, M., Conrad, R., 2009. The elusive curse of oil. Rev. Econ. Stat. 91 (3), 586–598.
Andersen, J.J., Aslaksen, S., 2008. Constitutions and the resource curse. J. Dev. Econ. 87
(2), 227–246.
Apergis, N., Payne, J.E., 2014. The oil curse, institutional quality, and growth in MENA
countries: evidence from time-varying cointegration. Energy Econ. 46, 1–9.
Arezki, R., Brückner, M., 2011. Oil rents, corruption, and state stability: evidence from
panel data regressions. Eur. Econ. Rev. 55 (7), 955–963.
Auty, R., 2002. Sustaining Development in Mineral Economies: The Resource Curse
Thesis. Routledge.
Barro, R.J., 1999. Determinants of democracy. J. Political Econ. 107 (S6), S158–S183.
Bhattacharyya, S., Hodler, R., 2010. Natural resources, democracy and corruption. Eur.
Econ. Rev. 54 (4), 608–621.
Bjorvatn, K., Farzanegan, M.R., Schneider, F., 2012. Resource curse and power balance:
evidence from oil-rich countries. World Dev. 40 (7), 1308–1316.
Boschini, A., Pettersson, J., Roine, J., 2013. The resource curse and its potential reversal.
World Dev. 43, 19–41.
Boyce, J.R., Emery, J.H., 2011. Is a negative correlation between resource abundance
and growth sufficient evidence that there is a resource curse? Resour. Policy 36 (1),
1–13.
Brückner, M., 2010. Natural resource dependence, non-tradables, and economic growth.
J. Comp. Econ. 38 (4), 461–471.
Brückner, M., Ciccone, A., Tesei, A., 2012. Oil price shocks, income, and democracy. Rev.
Econ. Stat. 94 (2), 389–399.
Brunnschweiler, C.N., Bulte, E.H., 2008. The resource curse revisited and revised: a tale
of paradoxes and red herrings. J. Environ. Econ. Manag. 55 (3), 248–264.
Bulte, E.H., Damania, R., Deacon, R.T., 2005. Resource intensity, institutions, and
development. World Dev. 33 (7), 1029–1044.
Busse, M., Gröning, S., 2013. The resource curse revisited: governance and natural
resources. Public Choice 154 (1–2), 1–20.
Caselli, F., Tesei, A., 2016. Resource windfalls, political regimes, and political stability.
Review of Economics and Statistics, forthcoming.
Cavalcanti, T.V.d.V., Mohaddes, K., Raissi, M., 2011. Growth, development and natural
resources: new evidence using a heterogeneous panel analysis. Q. Rev. Econ. Financ.
51 (4), 305–318.
Collier, P., Goderis, B., 2012. Commodity prices and growth: an empirical investigation.
Eur. Econ. Rev. 56 (6), 1241–1260.
Collier, P., Hoeffler, A., 2005. Resource rents, governance, and conflict. J. Confl. Resolut.
49 (4), 625–633.
Cotet, A.M., Tsui, K.K., 2013. Oil and conflict: what does the cross country evidence really
show? Am. Econ. J.: Macroecon. 5 (1), 49–80.
Davis, G.A., 2011. The resource drag. Int. Econ. Econ. Policy 8 (2), 155–176.
Davis, G.A., Cordano, A.L.V., 2013. The fate of the poor in growing mineral and energy
economies. Resour. Policy 38 (2), 138–151.
Easterly, W., 1999. Life during growth. J. Econ. Growth 4 (3), 239–276.
El Anshasy, A.A., Katsaiti, M.-S., 2013. Natural resources and fiscal performance: does
good governance matter? J. Macroecon. 37, 285–298.
Epstein, D.L., Bates, R., Goldstone, J., Kristensen, I., O'Halloran, S., 2006. Democratic
transitions. Am. J. Political Sci. 50 (3), 551–569.
Frankel, J.A., 2010. The natural resource curse: A survey. Tech. rep., National Bureau of
Economic Research.
Freudenburg, W.R., 1992. Addictive economies: extractive industries and vulnerable
localities in a changing world economy. Rural Sociol. 57 (3), 305–332.
Gelb, A.H., 1988. Oil Windfalls: Blessing or Curse?. Oxford University Press.
Glaeser, E.L., Ponzetto, G.A., Shleifer, A., 2007. Why does democracy need education? J.
Econ. Growth 12 (2), 77–99.
Gylfason, T., 2001. Natural resources, education, and economic development. Eur. Econ.
Rev. 45 (4), 847–859.
Gylfason, T., Herbertsson, T.T., Zoega, G., 1999. A mixed blessing. Macroecon. Dyn. 3
(02), 204–225.
Hadiz, V., Robison, R., 2005. Neo-liberal reforms and illiberal consolidations: the
Indonesian paradox. J. Dev. Stud. 41 (2), 220–241.
Hodler, R., 2006. The curse of natural resources in fractionalized countries. Eur. Econ.
Rev. 50 (6), 1367–1386.
Holtz-Eakin, D., Newey, W., Rosen, H.S., 1988. Estimating vector autoregressions with
panel data. Econometrica, 1371–1395.
IEA, 2015. 2015 Key World Energy Statistics. International Energy Agency.
Isham, J., Woolcock, M., Pritchett, L., Busby, G., 2005. The varieties of resource
experience: natural resource export structures and the political economy of economic
growth. World Bank Econ. Rev. 19 (2), 141–174.
Koop, G., Pesaran, M.H., Potter, S.M., 1996. Impulse response analysis in nonlinear
multivariate models. J. Econ. 74 (1), 119–147.
Lederman, D., Maloney, W.F., 2003. Trade structure and growth. World Bank Policy
Research Working Paper (3025).

162
N. Antonakakis et al.
Maloney, W., Lederman, D., 2008. In search of the missing resource curse. J. Lat. Am.
Caribb. Econ. Assoc. 29 (1), 1–20.
Marshall Monty, G., Jaggers, K., Gurr, T.R., 2009. Polity IV Project: Political regime
characteristics and transitions, 1800–2009. Dataset Users Manual. Center for
International Development and Conflict Management, University of Maryland.
Mehlum, H., Moene, K., Torvik, R., 2006a. Cursed by resources or institutions? World
Econ. 29 (8), 1117–1131.
Mehlum, H., Moene, K., Torvik, R., 2006b. Institutions and the resource curse*. Econ. J.
116 (508), 1–20.
Papaioannou, E., Siourounis, G., 2008. Democratisation and growth. The Economic
Journal 118 (532), 1520–1551.
Papyrakis, E., Gerlagh, R., 2004. The resource curse hypothesis and its transmission
channels. J. Comp. Econ. 32 (1), 181–193.
Pesaran, H.H., Shin, Y., 1998. Generalized impulse response analysis in linear
multivariate models. Econ. Lett. 58 (1), 17–29.
Raddatz, C., 2007. Are external shocks responsible for the instability of output in low-
income countries? J. Dev. Econ. 84 (1), 155–187.
Rodriguez, F., Sachs, J.D., 1999. Why do resource-abundant economies grow more
slowly? J. Econ. Growth 4 (3), 277–303.
Ross, M., 2007. How mineral-rich states can reduce inequality in Humphreys, Sachs,
Stiglitz (eds.), Escaping the Resource Curse. Columbia University Press and Revenue
163
Resources Policy 53 (2017) 147–163
Watch Institute, Ch. 9, pp. 237–255.
Ross, M.L., 2001. Does oil hinder democracy? World Polit. 53 (3), 325–361.
Sachs, J.D., Warner, A.M., 1995. Natural resource abundance and economic growth.
NBER Working Papers 5398, National Bureau of Economic Research, Inc.,
Sachs, J.D., Warner, A.M., 1999. The big push, natural resource booms and growth. J.
Dev. Econ. 59 (1), 43–76.
Sachs, J.D., Warner, A.M., 2001. The curse of natural resources. Eur. Econ. Rev. 45 (4),
827–838.
Sala-i Martin, X., Subramanian, A., 2013. Addressing the natural resource curse: an
illustration from Nigeria. J. Afr. Econ. 22 (4), 570–615.
Sims, C.A., 1980. Macroeconomics and reality. Econometrica 48 (1), 1–48.
van der Ploeg, F., 2011. Natural resources: curse or blessing? J. Econ. Lit. 49 (1),
366–420.
van der Ploeg, F., Poelhekke, S., 2010. The pungent smell of red herrings: subsoil assets,
rents, volatility and the resource curse. J. Environ. Econ. Manag. 60 (1), 44–55.
Wright, G., Czelusta, J., 2004. Why economies slow: the myth of the resource curse.
Challenge 47 (2), 6–38.
Yaduma, N., Kortelainen, M., Wossink, A., et al., 2013. An investigation of oil curse in
OECD and Non-OECD oil exporting economies using green measures of income.
Economics Discussion Paper Series EDP-1321, The Manchester University.