Resource rents,balance of power,and political stability

Abstract

We study the association between natural resource rents and internal political stability, highlighting the importance of the distribution of political power as a mediating factor. We present a simple theoretical model demonstrating that increased rents are likely to be positively associated with the internal stability of a powerful incumbent while destabilizing a less powerful incumbent. Our empirical analysis confirms this prediction. Employing panel data for more than 120 countries from the period 1984–2009, our estimation results demonstrate that resource rents can promote political stability but only when political power is sufficiently concentrated. Indeed, if the incumbent is sufficiently weak, rents fuel instability. Our main results hold when we control for the effects of income, quality of institutions (rule of law, democratic accountability and corruption), persistence of political stability, time-varying common shocks, country fixed effects, possible endogeneity of rents and power balance to political stability, and various additional covariates. Our analysis departs from the existing literature by emphasizing not (only) the type of government, but rather the strength of government, as a key determinant of the impact of resource rents on political stability. This analysis sheds light on current political transformations and reconfigurations in the resource rich Middle East and North Africa.

Keywords

conflict political power political stability resource rent

Introduction

Several studies suggest that higher resource rents increase political instability and intensify conflicts by funding rebel groups, weakening state institutions, and making separatism financially attractive in resource-rich regions.¹ At the same time, we observe that some of the most resource rich countries in the world are very stable: even as the Arab Spring brought down regimes all over the Middle East, the Persian Gulf monarchies did not falter.

Clearly, one source of stability of resource rich countries is their ability to buy peace.² Shortly after the collapse of former Egyptian president Hosni Mubarak’s regime in February 2011, the Saudi Arabian government announced a social welfare program worth $10.7 billion to spend on new employment opportunities and loan forgiveness, a program that reached $93 billion in March 2011. Similar initiatives were introduced in the UAE, Qatar, Oman, and Bahrain.³ The cost of co-opting opposition groups, however, depends on the power of the regime relative to the opposition; the price tag of a political stability program is likely to be lower for a powerful incumbent than a less powerful one. As higher resource rents may also make it more attractive to challenge the incumbent, whether rents promote political stability may crucially depend on the power balance in society.

Andersen & Aslaksen (2013) analyze the impact of resource rents on political stability, measured by leadership duration, and demonstrate that the impact is contingent on the type of resource and the type of government. In particular, oil is found to prolong the life span of regimes in autocracies, whereas minerals are shown to have the opposite effect. Political stability appears to be less sensitive to resource rents in democratic polities. We depart from their study by emphasizing not the type of government but rather the strength of government as a key determinant of the impact of resource rents on political stability and demonstrate that resource rents promote political stability when the incumbent is sufficiently powerful, whereas they have a destabilizing effect when opposition groups are strong.

To set the scene, the next section presents a simple theoretical model that demonstrates how political dominance may shape the impact of increased resource rents. The following section discusses our empirical strategy and the data. We then proceed to present and discuss the empirical evidence and some robustness analyses. The final section concludes the article.

The model

There are two power groups, a and b, in society. Group a is the incumbent; b is the opposition. The group in power controls rents r and also enjoys a non-pecuniary utility v, which we can think of as derived from the ability to determine policy. Moreover, the incumbent has the ability to make a take-it-or-leave-it peace offer to the opposition (in the form of a monetary transfer).⁴

The sequence of moves is as follows. At the first stage of the game, the incumbent decides how much transfer t to offer. At the second stage, the opposition decides whether to accept the offer or to challenge the incumbent. If the incumbent is not challenged, the transfer is paid, and the game ends. If challenged, the game moves on to the next stage, where there is a power contest, with the two groups simultaneously deciding on fighting efforts and where the Nash equilibrium defines the winning probability of each group. The winner collects the rents, leaving the loser with nothing.

We assume that the incumbent can credibly commit to a transfer and that the opposition is able to commit to not challenging the incumbent once the transfer has been received.⁵ We can think of the transfers coming in the form of government jobs, where the government can make binding job offers and where the acceptance of such jobs by the opposition also serves as a credible commitment on their part not to challenge as it allows the government to keep a close eye on the activities of the opposition members.

The alternative to a peaceful solution is conflict. We employ a standard rent-seeking model to describe the conflict equilibrium (Tullock, 1980). Let p_i be the relative power of group i, where $p_{i} \in (0, 1)$ and $p_{a} + p_{b} = 1$ . The closer p is to one-half, the more balanced is the power.

Employing the logic of backward induction, we begin by defining the conflict equilibrium. The winning probability is defined by the relative fighting force of each group, where the fighting force is the result of both power (p) and effort (f). The objective function of group i is given by Equation (1):

π_{i} = \frac{p_{i} f_{i}}{p_{a} f_{a} + p_{b} f_{b}} (v + r) - f_{i}

In equilibrium, both groups exert the same effort, given by Equation (2):

f_{a}^{*} = f_{b}^{*} = p_{a} (1 - p_{a}) (v + r) \equiv f^{*}

In a conflict equilibrium, the expected profit of group i is given by Equation (3):

π_{i}^{*} = \frac{p_{i} f^{*}}{p_{i} f^{*} + (1 - p_{i}) f^{*}} (v + r) - f^{*} = p_{i}^{2} (v + r)

To pacify the opposition, the incumbent needs to make a transfer t such that the opposition prefers to work for the government rather than against it. The minimal transfer that ensures this is given by $t^{*} = π_{b}^{*}$ . Now, for the incumbent to be able to make this transfer, it must have the necessary resources, that is, $r \geq π_{b}^{*}$ . It can be shown that it is always profitable for the incumbent to pacify the opposition, when able to do so. The situation where this condition binds can be expressed in terms of the power balance as in Equation (4):

p_{a} = 1 - \frac{\sqrt{r (v + r)}}{v + r} \equiv p_{a}^{*}

Figure 1 illustrates the $p_{a}^{*}$ -curve for a given level of v, focusing on $p_{a} > 0.5$ , that is, where the incumbent has a power advantage over the opposition.

Figure 1.

Conflict and stability

To the right of the $p_{a}^{*}$ -curve, there is peace, and to the left, there is conflict. For levels of $p_{a}$ close to 0.5 (high degree of power balance), an increase in r means intensified conflict because $f^{*}$ increases. For somewhat higher levels of $p_{a}$ (a ‘moderately’ dominant incumbent), an increase in r can change the equilibrium from conflict to peace, as we move from a point to the left of the $p_{a}^{*}$ -curve to a point to the right of this curve. For high levels of $p_{a}$ (high degree of power dominance), an increase in r simply means even stronger control by the incumbent. An increase in v tilts the curve upward around the point of intersection with the vertical axis, making conflict a more likely outcome.

We observe that the impact of an increase in r on the political economy of the country depends on p_a , that is, the strength of the incumbent. We can make the following two observations:

Observation 1: In the conflict equilibrium, an increase in rents has a smaller impact on fighting efforts the higher is p_a .

Observation 2: The probability that an increase in rents changes the equilibrium from conflict to peace is higher for a higher p_a .

Observation 1 is based on Equation (2), while Observation 2 is based on Equation (4), the latter due to the fact that $p_{a}^{*}$ is a falling function of r. Note that both observations point to a stronger incumbent reducing the potentially negative impact of increased rents on the economy, at the intensive margin (Observation 1) by a weakening of the fighting effort and at the extensive margin (Observation 2) by changing the equilibrium from conflict to peace. We shall use the term ‘stabilizing’ to describe the reduction in conflict on both margins. Similarly, the term ‘destabilizing’ refers to both an increase in fighting effort and a change in equilibrium from stability to conflict.

Data and empirical specification

Our main prediction from the theoretical model is that the distribution of power matters for whether resource rents are a politically stabilizing or destabilizing force. In particular, such rents are more likely to have a stabilizing effect if the incumbent is powerful, whereas adding rents in a situation with a less dominant incumbent may lead to political instability.

We test this hypothesis utilizing panel data for more than 120 countries from 1984–2009. To estimate whether the relationship between rents and internal stability varies systematically with the degree of dominant political power, we employ the following model:

\begin{aligned} S t a b_{i t} & = β_{1} \cdot R e n t_{i t} + β_{2} \cdot L a c k_P o w e r_{i t} \\ + β_{3} \cdot (R e n t_{i t} \times L a c k_P o w e r_{i t}) + β_{4} \cdot Z_{i t} \\ + μ_{i} + δ_{t} + ∊_{i t} \end{aligned}

with country i and time t. Stab is the political stability, Rent is the share of total resource rents in GDP,⁶ Lack_Power is a measure of (lack of) dominance of political power, Rent × Lack_Power is the interaction of rents and power imbalance, and Z stands for the vector of control variables. According to our theoretical prediction, the sign of the interaction term should be negative (β ₃ < 0); the higher the degree of political factionalism, the lower should be the stability effect of rents. The marginal stability effect of rents can be calculated by examining the following partial derivative on the basis of Equation (5):

\frac{\partial (S t a b_{i t})}{\partial (R e n t_{i t})} = β_{1} + β_{3} \cdot (L a c k_P o w e r_{i t})

Political stability

Political stability (Stab) is an assessment of political violence in the country and its actual or potential impact on governance. It is based on the internal conflicts index of the International Country Risk Guide (ICRG, 2011) published by the Political Risk Services (PRS) group. The Stab scores vary from 0 (least stable system) to 12 (most stable system). The Stab index is the sum of three subcomponents, each with a maximum score of 4 points and a minimum score of 0 points. A score of 4 means a very low risk, and a score of 0 refers to a very high risk. These three subcomponents are civil war/coup threat, terrorism/political violence, and civil disorder. This can be seen both as change in equilibrium (coup) and a change in the intensity of conflict in the conflict equilibrium (that is, both on the extensive and intensive margin).

The advantages of employing the Stab measure of political stability are threefold. First, it captures our notion of political instability, reflecting the struggle between power groups in society. Second, it covers the time period 1984–2009, and compared to other available datasets such as World Governance Indicators, it has the largest number of observations, enabling us to take advantage of the panel data method. Third, the Stab index is widely used in the literature (see, for example, Jinjarak, 2009; Farzanegan, Lessmann & Markwardt, 2013; Bjorvatn & Farzanegan, 2013). The ICRG indicators, including political stability, are based on the assessments of experts. This raises the question of the extent to which these subjective evaluations match reality. We address the potential subjective bias by employing fixed effects in which we focus on within country variation of data including the political stability index (see Bezemer & Jong-A-Pin, 2013 for a similar approach).

Political dominance

The lack of power dominance index (Lack_Power), which goes from 0 to 1, is defined as the probability that two randomly picked members of parliament from governing parties belong to different parties (Beck et al., 2001). We employ the Govfrac variable in the Database of Political Institutions to operationalize the concept of power (Keefer, 2010). Higher values of this index demonstrate that the government consists of a large number of small parties, lacking a dominant strong party. In other words, higher scores for Lack_Power is an indicator of weaker government, whereas a lower score demonstrates that the government consists of a small number of strong parties.

In terms of the theoretical model presented in the previous section, a high degree of power concentration corresponds to a p close to 1 (which in the empirical specification is Lack_Power is close to 0). The idea is that a dominant political party in parliament will also be a dominant player in a conflict because it can mobilize more political and military support and will thus be in a better position to appease the opposition through transfers than would a less dominant party.

Riker (1964) provides a theoretical basis for the positive role of strong governments for the provision of public goods and economic growth. Several studies have examined the empirical implications of Riker’s theory. For example, Poteete (2009) argues that power dominance has been important in enabling Botswana to make productive use of its natural resources. Enikolopov & Zhuravskaya (2007) also demonstrate that a lack of dominant power measured by government fractionalization has a negative effect on public goods provision. Bjorvatn, Farzanegan & Schneider (2012, 2013) emphasize the key role of the distribution of political power for economic success in a sample of oil rich economies.

Resource rents

The resource rents data are obtained from the World Bank (2012). Total natural resources rents are the sum of oil, natural gas, coal, mineral, and forest rents. Rents are defined as revenues above the cost of extracting the resources. Natural resources give rise to economic rents because they are not produced. For produced goods and services, competitive forces expand supply until economic profits are driven to zero, but natural resources in fixed supply often command returns well in excess of their cost of production. The estimates of natural resource rents are calculated as the difference between the price of a commodity and the average cost of producing the resources. We are interested in the degree of the resource dependency of the economy. Thus, we utilize the share of resource rents in GDP rather than a resource abundance indicator such as underground reserves. The balance of political power and fractionalization has more significant linkages with tangible resource rents in the economy. The share of resource rents in GDP has been used to explain the corruption and conflict effects of rents in sub-Saharan Africa and the moderating role of democracy by Arezki & Gylfason (2013). Table I reports the descriptive statistics of the major variables in our empirical analysis.

Table I.

Descriptive statistics of key variables (1984–2010)

Variable	Obs.	Mean	Standard deviation	Max.
Stab	3,529	8.76	2.62	12
Rent	4,813	8.14	16.15	214.4
Lack_Power	3,936	0.20	0.27	1
Rent × Lack_Power	3,687	1.33	5.10	75.6

Stab, Rent, and Lack_Power refer to the ICRG internal conflict index, share of total natural resource rents in GDP, and the lack of dominant political power, respectively. There are a few countries whose total resource rents as a share of GDP in some years exceeds 100%: for example, Turkmenistan and Iraq.

Other control variables

The remaining variables are from the World Bank (2012). Control variables include the logarithm of real income per capita, consumer price inflation rate, secondary school enrollment rate, and population growth. We also control for the quality of institutions as measured by a simple average of the rule of law, democracy, and corruption, which are taken from the ICRG (2011). The rule of law score is from 0 (worst score) to 6 (best score) and measures the strength and impartiality of the legal system in addition to popular observance of the law. The democracy index shows the responsiveness of the government to the citizens. The index examines whether there are free and fair elections, whether there is an independent judiciary, and whether there are constitutional or legal guarantees of personal liberties. It ranges from 0 to 6, with higher values indicating more democracy. The corruption index is scored from 0 (most corrupt) to 6 (least corrupt). Thus, we call it ‘the lack of corruption’ index. It measures corruption in the political system such as excessive patronage, nepotism, job reservations, ‘favor-for-favors’, secret party funding, and suspiciously close ties between politics and business. This index of corruption is frequently employed in the literature (e.g. Knack & Keefer, 1995; Bhattacharyya & Hodler, 2010; Fredriksson & Svensson, 2003; Biswas, Farzanegan & Thum, 2012). In our robustness checks, we add the democracy index to our main key independent and interaction terms (Rent, Lack_Power, and Rent × Lack_Power).⁷ These controls will clarify that the theoretical concept of power dominance does not capture the effect of democracy or other institutional factors.

Fixed effects

We allow for country (µ _i) and time (δ _t) specific effects, controlling for the unobservable time-invariant country characteristics and shocks, which are common to all countries. There are several time-invariant country characteristics that affect the political stability of a country, increasing the risk of omitted variable bias. Such country specific factors are, for example, geography, history, and ethnicity, among others. The fixed time effects control for shocks common to all countries such as the collapse of the Soviet Union in 1991 and the end of the Cold War or events related to the post-9/11 terrorist attacks. We address the arbitrary heteroskedasticity and serial correlation by utilizing cluster-robust standard errors at the country level (Wooldridge, 2002). We employ the White period coefficient covariance matrix estimator, which is robust to serial correlation within countries and changing variances over time in all estimations (Arellano, 1987). The appendix summarizes the description and sources of the data.

Results

Main results

Table II reports the estimates of Equation (5) utilizing data covering the period from 1984 to 2009. Our specific model includes our three main variables of interest, namely, Rent, Lack_Power, and in particular their interaction term, Rent × Lack_Power, which we shall refer to as the ‘RP-interaction’ for short. We add other control variables in subsequent models. Employing this approach enables us to examine the robustness of the interaction term between Rent and Lack_Power across different specifications (Models 2.1 to 2.11). All of our regressions are based on panels and include a full set of country and time fixed effects. The interaction term between rents and lack of dominant power is negative and statistically significant at the conventional levels in all models (except for Model 2.9). In all models, the lower dominance of political power is reducing the stability effects of resource rents. In other words, the stabilizing impact of resource rents indeed depends on the level of political dominance in the economy. To test the robustness of our main hypothesis, we add other control variables such as Income (log of real GDP per capita), Inflation (consumer price index growth rate), Education (gross secondary school enrollment rate), and Population growth rate in Models 2.2–2.5, one by one. In Model 2.6, we include all of these control variables and the size and significance of the RP interaction remains robust. Furthermore, we examine the role of Corruption, Rule of law, and their interaction terms with oil rents in Models 2.7 and 2.8. The lack of corruption and higher rule of law buy internal stability. However, their interactions with resource rents have no moderating impact on the resource rents–stability nexus. The RP interaction term remains statistically relevant in these specifications. Does the RP interaction depend on the rule of law and lack of corruption? To test this case, we include a triple interaction (Rent × Lack_Power × Lack_Corruption and Rent × Lack_Power × Law) in Models 2.9 and 2.10. These triple interactions are statistically insignificant, which imply that our main interaction term, namely, RP (Rent × Lack_Power), is independent of law and corruption quality within countries.

Table II.

Stability effects of total resource rents and lack of dominant political power

Dependent variable: Stab (ICRG internal conflict index)
Models	2.1	2.2	2.3	2.4	2.5	2.6	2.7	2.8	2.9	2.10	2.11
Rent	0.012	0.013	0.017	0.018	0.012	0.013	0.013	0.015	0.013	0.029	0.044**
	(0.96)	(0.85)	(1.16)	(1.28)	(1.02)	(0.77)	(0.63)	(0.63)	(0.57)	(1.09)	(3.01)
Lack_Power	–0.098	–0.105	0.134	0.104	–0.097	0.252	0.248	0.268	0.181	1.05	0.661†
	(–0.30)	(–0.32)	(0.40)	(0.31)	(–0.30)	(0.77)	(0.78)	(0.93)	(0.23)	(1.45)	(1.92)
Rent × Lack_Power	–0.030†	–0.029†	–0.038*	–0.046†	–0.031*	–0.048*	–0.045†	–0.036†	–0.040	–0.093*	–0.041*
	(–1.87)	(–1.85)	(–2.40)	(–1.73)	(–1.99)	(–1.99)	(–1.94)	(–1.66)	(–1.06)	(–2.46)	(–2.34)
log(Income)		0.329				0.915	1.08†	–0.021	1.08†	–0.073	–0.041
		(0.65)				(1.38)	(1.68)	(–0.03)	(1.68)	(–0.12)	(–0.22)
Inflation			–0.010*			–0.031**	–0.031**	–0.018**	–0.031**	–0.018**	–0.010
			(–1.98)			(–4.04)	(–5.00)	(–2.82)	(–5.10)	(–2.85)	(–1.18)
Educ				0.009		0.003	0.004	0.006	0.004	0.006	0.000
				(1.22)		(0.39)	(0.57)	(0.87)	(0.55)	(0.84)	(0.26)
Popg					0.096	–0.248*	–0.232*	–0.102	–0.231*	–0.100	0.048
					(0.98)	(–2.36)	(–2.34)	(–1.07)	(–2.35)	(–1.03)	(0.92)
Lack_Corruption							0.402**		0.398**		–0.002
							(3.53)		(3.45)		(–0.05)
Rent × Lack_Corruption							0.000		0.000		0.002
							(0.05)		(0.08)		(0.73)
Law								0.912**		0.969**	0.300**
								(10.0)		(9.16)	(5.68)
Rent × Law								0.000		–0.003	–0.008**
								(0.04)		(–0.48)	(–3.06)
Rent × Lack_Power × Lack_Corruption									–0.002		0.003
									(–0.11)		(0.40)
Lack_Power × Lack_ Corruption									0.023		0.076
									(0.10)		(0.76)
Rent × Lack_Power × Law										0.018	0.007
										(1.58)	(1.16)
Lack_Power × Law										–0.221	–0.198*
										(–1.39)	(–2.35)
Democracy											0.044
											(1.13)
Rent × Democracy											–0.002
											(–1.06)
Stab (–1)											0.732**
											(32.09)
Observations	2,949	2,919	2,721	2,275	2,946	2,110	2,110	2,110	2,110	2,110	2,023
Number of countries	134	133	130	133	134	127	127	127	127	127	126
Adj. R-sq	0.67	0.67	0.67	0.70	0.67	0.70	0.71	0.77	0.71	0.77	0.90

Effective sample period: 1984–2009. The method of estimation is panel OLS (country and time fixed effects). The constant term is included (not reported). t statistics shown in parentheses are based on robust standard errors, which are clustered at the country level (the White period method). †p < 0.1, *p < 0.05, **p < 0.01.

Model 2.11 is a general specification in which we control for a full set of other drivers of internal stability and introduce democracy and its interaction term with rents and lag of internal stability to the previous controls in Models 2.2–2.10. Inclusion of the lag of stability as a predicator of current stability converts the static fixed effects regressions into a dynamic panel model. Inclusion of the lag of the dependent variable in the RHS (right hand side) of the model may raise the problem of Nickell bias (Nickell, 1981) in short samples (large N and small T). Our sample in Table II covers more than 20 years, and the Nickell bias problem is thus less important in terms of inconsistency.⁸ Controlling for the dynamic aspect of stability makes the direct effect of resource rents on stability stronger in size for the first time, and makes the direct effect statistically significant at the 1% level. The estimated coefficient of the negative interaction term between rents and lack of dominant political power is similar, on average, to those of other Models (2.1–2.9) and is statistically significant at the 5% level.

Stability effects of control variables

Among the control variables, real income per capita (log(Income)), significant at the 10% level in Models 2.7 and 2.9, has a positive association with internal stability. A higher inflation rate (Inflation) destabilizes the political system. Inflation has a more robust negative contemporaneous association with internal stability. The negative relationship between inflation and internal stability of the political system is statistically significant at the 1% level in most specifications. This is in line with the findings of Paldam (1987) in his analysis of Latin American countries, demonstrating that inflation and political instability are interconnected. Education (Edu), measured as secondary school enrollment rate, has a positive association with stability but is far from statistically significant. The population growth rate (Popg) as a measure of demographic burden increases the pressure on economic resources. It can have serious implications for provision of employment, investment in public goods and infrastructures, and the cost of subsidies. Meeting the economic and social needs of an increasing population is a major challenge for fragile states in resource rich countries (Bjorvatn & Farzanegan, 2013).⁹ Lack of corruption (Lack_Corruption) is directly and positively associated with higher internal stability. A one score increase in control of corruption is accompanied by an approximately 0.4 score increase in the internal stability index. This is in line with Fjelde (2009). Again, in agreement with Fjelde (2009), the interaction of rents and the (lack of) corruption is not statistically significant after controlling for country and time fixed effects. Apart from the role of corruption, we also examine the direct association of the rule of law (Law) with political stability and its moderating role in shaping the oil rents, power balance, and stability nexus. Raising the rule of law index by one score is associated with an increase of more than 0.9 score in the internal stability. On average, 70% of the variation in the political stability variable is explained by the independent variables, country and year fixed effects. This increases to 90% in the general specification of Model 2.11 in which we also control for the history of internal stability in estimation.

Do the main results change if we control for regime type?

A possible analytical concern in understanding the stability–rent nexus and balance of political power is the role of democratic institutions. The relationship we detect between independent variables (rents and lack of dominant political power) and the dependent variable (political stability) may be caused by a third variable that causes changes in both variables at the same time. We have controlled for a set of important drivers of internal stability and institutional factors such as corruption and rule of law. Regime type (democracy vs. autocracy), however, may also affect both stability and balance of political power. To clarify the mechanism behind the empirical regularity in our observational design, we control directly for democratic accountability of the state in regressions.¹⁰ In Table II, Model 2.11, we have included democracy in the estimations. However, we examine the estimations of the main variables of interest (rents, lack of power, and their interaction) by including and keeping democracy in all specifications in Table III. We also check how oil rents as a part of total resource rents affect stability. Total resource rents include other types of resources such as coal, minerals, and forests, in addition to oil. As we see in Models 3.1–3.16, the direct positive association between oil rents and internal stability is statistically more significant than the direct association between total rents and stability. Nevertheless, the negative interaction term between rents and the lack of dominant power remains robust and statistically significant across different specifications in Table III, controlling for the democratic quality of institutions. In Models 3.3–3.10, we add other control variables (income per capita, inflation, education, and population growth), one by one. Our main interaction term is not sensitive to these controls. In Models 3.11 and 3.12, we have a full set of controls in the estimation. Although democracy is strongly and positively associated with higher internal stability, its inclusion does not undermine the RP-interaction channel. In other words, regime type is not an underlying factor for our empirical observation. In Models 3.13–3.16, we employ an average of lack of corruption, democracy, and rule of law under a single variable Inst in the estimations. The Inst variable is a strong, positive driver of internal stability. However, its inclusion does not change our observation about the moderating role of a lack of dominant power in the stability–rent nexus.

Table III.

Stability effects of oil and total resource rents and lack of dominant political power, controlling for democracy and average of institutional quality

Dependent variable: Stab (ICRG internal conflict index)
Models	3.1	3.2	3.3	3.4	3.5	3.6	3.7	3.8	3.9	3.10	3.11	3.12	3.13	3.14	3.15	3.16
Oil	0.039**		0.045*		0.040*		0.040*		0.039*		0.032		0.032	0.039*
	(2.58)		(2.30)		(2.04)		(2.38)		(2.56)		(1.35)		(1.51)	(2.13)
Rent		0.021†		0.020		0.023		0.024		0.021†		0.014			0.017	0.022†
		(1.65)		(1.35)		(1.51)		(1.61)		(1.67)		(0.81)			(1.13)	(1.66)
Lack_Power	–0.174	–0.048	–0.144	–0.024	–0.017	0.145	–0.043	0.112	–0.175	–0.048	0.126	0.262	0.148	0.040	0.264	0.171
	(–0.62)	(–0.17)	(–0.51)	(–0.08)	(–0.05)	(0.51)	(–0.14)	(0.36)	(–0.62)	(–0.17)	(0.42)	(0.86)	(0.55)	(0.15)	(0.94)	(0.65)
Oil × Lack_Power	–0.041*		–0.040*		–0.046**		–0.046		–0.041*		–0.051*		–0.042†	–0.041**
	(–2.25)		(–2.56)		(–2.83)		(–1.51)		(–2.29)		(–2.08)		(–1.69)	(–2.90)
Rent* Lack_Power		–0.036*		–0.035*		–0.044**		–0.047†		–0.036*		–0.048*			–0.040†	–0.038*
		(–2.12)		(–2.12)		(–2.74)		(–1.75)		(–2.16)		(–2.00)			(–1.72)	(–2.50)
log(Income)			0.502	0.465							0.947	0.909	0.670	0.560	0.637	0.518
			(1.07)	(0.99)							(1.53)	(1.47)	(1.14)	(1.09)	(1.08)	(1.01)
Inflation					–0.008	–0.009					–0.030**	–0.031**	–0.023**	–0.004	–0.025**	–0.004
					(–1.43)	(–1.52)					(–3.87)	(–3.94)	(–3.06)	(–0.92)	(–3.25)	(–1.03)
Education							0.012	0.011			0.006	0.005	0.008		0.008
							(1.41)	(1.37)			(0.67)	(0.65)	(1.00)		(1.00)
Popg									0.025	0.034	–0.252*	–0.250*	–0.177†	–0.127†	–0.173†	–0.122
									(0.24)	(0.34)	(–2.36)	(–2.37)	(–1.82)	(–1.65)	(–1.81)	(–1.60)
Democracy	0.431**	0.426**	0.439**	0.436**	0.438**	0.435**	0.357**	0.349**	0.428**	0.422**	0.355**	0.352**
	(5.22)	(5.12)	(5.20)	(5.13)	(5.17)	(5.16)	(3.05)	(3.97)	(5.10)	(5.00)	(3.94)	(3.92)
Inst													1.19**	1.27**	1.19**	1.27**
													(8.64)	(9.18)	(8.62)	(9.17)
Observations	2,926	2,926	2,896	2,896	2,698	2,698	2,258	2,258	2,923	2,923	2,093	2,093	2,093	2,679	2,093	2,679
Number of countries	133	133	132	132	129	129	132	132	133	133	126	126	126	129	126	129
Adj. R-sq	0.69	0.69	0.69	0.69	0.69	0.69	0.72	0.71	0.69	0.69	0.71	0.71	0.75	0.74	0.75	0.74

Do our main results change if we use the lag of right-hand side (RHS) variables?

In Tables II and III, we present the contemporaneous effects of the variables on internal stability. However, it may take time for the effects of rents and the balance of political power to manifest themselves on the internal stability of the political system. To check this possibility, we employ a one-year lag of all independent variables and re-estimate all specifications of Table III. Employing lagged RHS variables further reduces the problem of simultaneity bias due to reverse feedback from dependent to independent variables. The results are shown in Table IV. Generally, we observe that even after controlling for this time lag, the negative interaction term of RP remains robust. Its statistical significance is somewhat less than the results in Table III (contemporaneous effects).

Table IV.

Stability effects of oil and total resource rents and a lack of dominant political power, controlling for democracy and average of institutional quality

Dependent variable: Stab (ICRG internal conflict index)
Models	4.1	4.2	4.3	4.4	4.5	4.6	4.7	4.8	4.9	4.10	4.11	4.12	4.13	4.14	4.15	4.16
L_Oil	0.041**		0.044*		0.041*		0.040*		0.040**		0.027		0.027	0.038*
	(2.72)		(2.28)		(2.12)		(2.43)		(2.72)		(1.24)		(1.36)	(2.07)
L_Rent		0.023†		0.020		0.026†		0.027*		0.023†		0.018			0.020	0.024†
		(1.83)		(1.42)		(1.81)		(1.93)		(1.89)		(1.18)			(1.50)	(1.88)
L_Lack_Power	–0.052	0.021	–0.034	0.041	0.052	0.167	0.070	0.0181	–0.058	0.021	0.211	0.305	0.227	0.096	0.306	0.187
	(–0.19)	(0.07)	(–0.12)	(0.15)	(0.18)	(0.59)	(0.23)	(0.58)	(–0.22)	(0.07)	(0.71)	(0.99)	(0.85)	(0.38)	(1.08)	(0.71)
L_Oil × Lack_Power	–0.028†		–0.029*		–0.030†		–0.030		–0.030†		–0.033		–0.026	–0.027*
	(–1.70)		(–1.97)		(–1.93)		(–1.08)		(–1.87)		(–1.48)		(–1.18)	(–1.98)
L_Rent × Lack_Power		–0.022		–0.023		–0.029*		–0.031		–0.024†		–0.032			–0.026	–0.024†
		(–1.48)		(–1.60)		(–2.03)		(–1.36)		(–1.64)		(–1.59)			(–1.34)	(–1.83)
L_log(Income)			0.238	0.214							0.797	0.770	0.584	0.371	0.560	0.342
			(0.52)	(0.46)							(1.34)	(1.30)	(1.05)	(0.74)	(1.01)	(0.69)
L_Inflation					–0.007	–0.007					–0.028**	–0.030**	–0.024**	–0.003	–0.025**	–0.004
					(–1.42)	(1–51)					(–3.21)	(–3.35)	(–3.12)	(–0.84)	(–3.27)	(–0.93)
L_Educ							0.009	0.009			0.004	0.004	0.006		0.006
							(1.23)	(1.20)			(0.60)	(0.62)	(0.88)		(0.92)
L_Popg									0.118	0.128	–0.102	–0.097	–0.045	–0.020	–0.038	–0.015
									(1.23)	(1.33)	(–0.87)	(–0.83)	(–0.39)	(–0.24)	(–0.33)	(–0.18)
L_Democracy	0.354**	0.349**	0.360**	0.358**	0.351**	0.349**	0.279**	0.274**	0.341**	0.336**	0.276**	0.274**
	(3.56)	(3.47)	(3.58)	(3.52)	(3.36)	(3.36)	(3.37)	(3.30)	(3.36)	(3.26)	(3.30)	(3.29)
L_Inst													0.917**	0.993**	0.919**	0.996**
													(7.50)	(7.83)	(7.52)	(7.84)
Observations	2,926	2,926	2,896	2,896	2,698	2,698	2,258	2,258	2,923	2,923	2,093	2,093	2,093	2,679	2,093	2,679
No. of countries	133	133	132	132	129	129	132	132	133	133	126	126	126	129	126	129
Adj. R-sq	0.68	0.68	0.68	0.67	0.68	0.68	0.71	0.71	0.68	0.68	0.71	0.71	0.73	0.71	0.73	0.71

Effective sample period: 1984–2009 (one-year lag of all RHS variables).The method of estimation is panel OLS (country and time fixed effects). The constant term is included (not reported). t statistics shown in parentheses are based on robust standard errors, which are clustered at the country level (the White period method). Inst is the average of democracy, rule of law, and lack of corruption indicators from the ICRG. L represents a one-year lag. †p < 0.1, *p < 0.05, **p < 0.01.

Endogeneity issue

The exogeneity assumption of a lack of a dominant political power (Lack_Power) index with respect to resource rents is another possible analytical concern in our study.¹¹ A portion of the literature (e.g. Robinson, Torvik & Verdier, 2006) suggests that natural resource rents can be utilized to change political institutions, which ultimately may also shape the balance of power inside a system. We have examined this possibility and our implicit exogeneity assumption of balance of power with respect to resource rents. First, the correlation between the lack of dominant power index and total rents (% GDP) is only –0.08, which is quite low. In another exercise, we regress the Lack_Power index on total resource rents (% GDP), controlling for country and year fixed effects. The estimated coefficient for resource rents is 0.0008 with a robust t statistic of 0.83, which is far from any conventional level of statistical significance. Even if we consider the possible time lag to transfer the effects of rents on the domestic balance of power by employing a one-year lag of rents, the estimated effect is 0.001 with a t statistic of 1.11, which shows it is not significantly different from zero. Even controlling for longer time lags (for example, a 5-year lag of rents) lends support to the notion that resource rents is not a determinant of the political power balance, considering the self-explanatory power of the balance of power, country, and time fixed effects. Furthermore, simultaneity between resource rents and stability as well as balance of power and stability may increase the risk of endogeneity in our regressions. The current flow of rents may affect the current situation of stability. However, developments in internal stability and conflict may also have significant consequences for the extraction and export of natural resources and their rents in the economy. Similarly, although changes in the balance of political power may have implications (directly and/or indirectly) for the stability of the system, changes in internal stability due to other reasons may affect the balance of internal political power. As mentioned earlier, one of our strategies is to employ lagged values of rents, balance of power, and other right-hand side variables while controlling for fixed effects. This reduces the reverse feedback from stability (see Table IV). In the next analysis, while utilizing dynamic panel data and the Generalized Method of Moments (GMM), we also restrict our sample period to 1991–2002. This sample period looks at the time period after the collapse of the Soviet Union and the end of the Cold War (1990/91) and before the events related to the military intervention by the USA and its allies in a resource-rich country, namely, Iraq in 2003. Our sample also does not include the special political turmoil during the Arab Spring. By this period restriction, we reduce the effect of extreme events in our empirical investigation. We compare the results from the full sample period with those from the restricted period, and we find no major change in our initial findings on the moderating role of a lack of dominant power in the stability–rents nexus.

To maximize the sample size and to identify the parameters of interest more precisely, we employ annual data. For this reason and to account for the possible partial adjustment to the steady state position for internal stability, we need to allow for dynamics in the estimations. This has been done in the fixed effects Model 2.11 in Table II. However, as we already discussed, there are uncertainties regarding the application of the fixed effects OLS method with the presence of the lagged dependent variable in the right-hand side of the model. There is a correlation between the independent variables and the error term because lagged internal stability (Stab (–1)) depends on ∊_it ₋₁, which is a function of µ_i (the country specific effect in Equation (5)). Considering the fixed effects, the dynamic model is biased by the order of 1/T (increasing sample period reduces this bias). We also re-estimate the dynamic model by restricting the sample to 1991–2002, for which the application of the dynamic fixed effects would have a higher bias. The most common method in such a case is the dynamic GMM developed by Arellano & Bond (1991). This method differences the model to eliminate country specific time-constant factors (e.g. geographical, historical, cultural, traditional, ethnic, and religious properties). We also control for time dummies in this method to account for common time specific shocks. Additionally, the difference GMM ensures that all variables in our model are stationary. This method, therefore, addresses one source of endogeneity because of the correlation of country specific and time-constant factors and the right-hand side independent variables. In addition, we treat all the right-hand side regressors as endogenous and employ two of their lags and the lagged dependent variable as instruments in the GMM method. The instruments need to be valid, meaning that they should not be correlated with the error term. We test the validity condition with the Sargan test. In addition, we look at the AR (2) test for serial correlation in the first-differenced residuals, under the null of no serial correlation. The p-value of the AR (2) statistic should be insignificant (Arellano & Bond, 1991). Table V presents the dynamic GMM estimations. The difference GMM estimations for the total sample period (1984–2010), controlling for additional drivers of stability (e.g. income, inflation, education, and population growth, and lag of stability) and country and period fixed effects, support our earlier results. The Sargan test p-value is 21%, indicating the validity of the internal double lags of the RHS variables and lagged dependent variable as instruments. The AR (2) test also exhibits a satisfactory result. Interestingly, the estimated coefficient for Stab (–1) in Diff-GMM is close to the estimated coefficient in the dynamic fixed effects Model 2.11 in Table II. Looking at our restricted sample period – post-Soviet era and pre-Iraq war – does not change the sign or statistical significance of the negative interaction term between Rent and Lack_Power (–0.036 in full sample period vs. –0.27 in restricted period, both significant at the 1% level). The Sargan test in the restricted period again is higher than 10%, providing further evidence on the validity of instruments. In the restricted period specification, the direct positive effect of resource rents (Rent) on stability demonstrate a significant increase in size (from 0.008 to 0.115), and it becomes statistically significant at the 1% level. For comparison, we also report the estimations based on the system GMM method suggested by Blundell & Bond (1998). Monte Carlo simulations by Blundell & Bond (1998) demonstrate that the system GMM estimator performs better than the first difference GMM estimator when some of the variables are highly persistent. However, as we can compare the Diff-GMM and Sys-GMM results in Table V, the estimated interaction term remains negative and statistically robust at the conventional level, and its size is almost identical in both methods (–0.027 in Diff-GMM vs. –0.024 in Sys-GMM). Controlling for democracy and a lack of corruption and their interactions with resource rents does not eliminate the significance or change the negative sign of the RP. The lag of stability is positively and significantly correlated with the current stability, suggesting considerable persistence. It is, however, significantly different from 1. According to Roodman (2007), to have dynamic stability in GMM, the estimated coefficient of the lagged dependent variable (in our case, Stab) should be lower than 1. The estimated coefficient of the lagged dependent variable (Stab in Table V) is approximately 0.75, which means that the steady-state assumption holds. We employ robust two-step standard errors in the GMM, which are shown by Roodman (2009) to be superior to the one-step standard error.¹²

Table V.

Dynamic panel data analysis

Dependent variable: Stab (ICRG internal conflict index)
Models	Diff-GMM 1984–2010	Diff-GMM 1991–2002	Sys-GMM 1991–2002	Diff-GMM 1991–2002	Diff-GMM 1991–2002
Stab (–1)	0.743**	0.754**	0.81**	0.74**	0.708**
	(40.12)	(51.57)	(92.29)	(51.98)	(28.85)
Rent	0.008	0.115**	0.070**	0.104**	0.051†
	(1.51)	(23.94)	(19.99)	(13.17)	(1.79)
Lack_Power	0.496**	–0.150	0.313**	–0.208	0.002
	(3.03)	(–1.26)	(3.40)	(–1.25)	(0.009)
Rent × Lack_Power	–0.036**	–0.027**	–0.024*	–0.037*	–0.059*
	(–2.73)	(–2.65)	(–2.38)	(–2.40)	(–2.49)
Log(Income)	0.24	1.16**	0.029	0.441	1.24**
	(0.90)	(3.79)	(0.24)	(1.40)	(3.24)
Inflation	0.013	–0.015**	–0.011**	–0.012**	–0.020**
	(1.43)	(–5.94)	(–15.82)	(–5.69)	(–3.65)
Educ	–0.015**	–0.062**	–0.005**	–0.049**	–0.062**
	(–3.49)	(–26.51)	(–3.95)	(–13.39)	(–9.00)
Popg	–0.32**	0.079	–0.106*	0.101†	0.098
	(–3.55)	(1.29)	(–2.25)	(1.73)	(0.87)
Democracy				0.017
				(0.77)
Rent × Democracy				–0.000
				(–0.48)
Lack_Corruption					–0.328**
					(–7.13)
Rent × Lack_Corruption					0.028**
					(3.43)
Observations	1,599	719	719	707	719
Number of countries	121	108	108	107	108
Adj. R-sq
Sargan (p-value)	0.21	0.25	0.19	0.33	0.42
AR(2)- p-value	0.99	0.99	na	0.99	0.99

Diff-GMM and Sys-GMM refer to First Differences and Orthogonal Deviations Panel Generalized Method of Moments. Cross-section fixed (first differences and orthogonal deviations) and period fixed (dummy variables) are included in all models. For GMM weight, White period (AB n-step) is employed to compute Arellano-Bond 2-step. White period standard errors and covariance are utilized to report robust t statistics in parentheses. Two lags of all RHS and lagged dependent variables are employed as instruments. Sargan (p-value) null hypothesis tests that overidentifying restrictions are valid. AR (2) null hypothesis tests that there is no serial correlation in the first-differenced errors at order 2. †p < 0.1, *p < 0.05, **p < 0.01.

Magnitude of stability effects of rents

What is the magnitude of the marginal impact of resource rents on stability at different levels of (lack of) dominant political power? When does an increase in rents lead to lower/higher political stability? To answer such questions, we calculate the marginal impacts of a 1% increase in the size of rents in the economy on the internal stability index at different levels of lack of dominant political power over the entire period of study (1984–2010). We utilize the results of the difference GMM in Table V:

\frac{\partial (S t a b_{i t})}{\partial (R e n t_{i t})} = 0.008 - 0.036 \times (L a c k_P o w e r_{i t})

In countries with a Lack_Power above 0.22 (from 0 to 1 scale) (= 0.008/0.036), an increase in rents leads to lower political stability. The average of the Lack_Power index in our sample is 0.204. In other specifications, overall, the final effect of rents on stability is positive, but this is true to a lesser extent in politically fractionalized countries. In the majority of specifications, the power must be highly fractionalized to change the stability effects of rents from the positive zone to the negative zone.

In our sample, there are 16 resource rich countries (for example, the share of total rents in GDP is larger than 30%) with Lack_Power of more than 0.22 in at least one year of the analysis.¹³

Conclusion

We studied how the impact of increases in resource rents on political stability may be contingent on the distribution of political power. Employing a simple theoretic model, we demonstrate that resource rents are likely to weaken a weak incumbent and strengthen a strong incumbent.

To test this prediction, we employ panel data covering the period 1984–2009 and more than 120 countries. Our theoretical prediction is supported by the data. In particular, the positive stability effects of rents decrease with higher levels of political fractionalization. Our main results hold when we control for the effects of income, inflation, education, population growth, time-varying common shocks, country fixed effects, and quality of institutions (corruption, democracy, and rule of law).

We believe our results can shed light on the current political changes in the Middle East. For instance, the homogenous rentier states at the southern borders of the Persian Gulf were able to manage the contagious effects of the Arab Spring through a combination of increased welfare and security spending. Clearly, in these societies, resource wealth has been an important stabilizing factor. Our results demonstrate, however, that in resource rich countries with a less dominating elite, the ability of resource rents to buy peace and stability is weaker.

Footnotes

Replication data

All statistical analyses were done using EViews 8.1. The EViews workfile for replicating the presented results is available at .

Acknowledgements

The authors wish to thank the JPR Associate Editor (Scott Gates), two anonymous reviewers, Richard Jong-A-Pin, Christian Neugebauer, and participants in the 2014 meeting of the European Public Choice Society (Cambridge) for helpful comments and suggestions.

Notes

Appendix

References

Andersen

Jørgen Juel

Aslaksen

Silje

(2013) Oil and political survival. Journal of Development Economics 100(1): 89–106.

Arellano

Manuel

(1987) Computing robust standard errors for within-groups estimators. Oxford Bulletin of Economics & Statistics 49(4): 431–433.

Arellano

Manuel

Bond

Stephen

(1991) Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies 58(2): 277–297.

Arezki

Rabah

Gylfason

Thorvaldur

(2013) Resource rents, democracy and corruption: Evidence from sub-Saharan Africa. Journal of African Economies 22(4): 552–569.

Basedau

Matthias

Lay

Jann

(2009) Resource curse or rentier peace? The ambiguous effects of oil wealth and oil dependence on violent conflict. Journal of Peace Research 46(6): 757–776.

Beck

Thorsten

Clarke

George

Groff

Alberto

Keefer

Philip

Walsh

Patrick

(2001) New tools in comparative political economy: The database of political institutions. World Bank Economic Review 15(1): 165–176.

Bezemer

Dirk

Jong-A-Pin

Richard

(2013) Democracy, globalization and ethnic violence. Journal of Comparative Economics 41(1): 108–125.

Bhattacharyya

Sambit

Hodler

Roland

(2010) Natural resources, democracy and corruption. European Economic Review 54(4): 608–621.

Biswas

Amit

Farzanegan

Mohammad Reza

Thum

Marcel

(2012) Pollution, shadow economy and corruption: Theory and evidence. Ecological Economics 75(C): 114–125.

10.

Bjorvatn

Kjetil

Farzanegan

Mohammad Reza

(2013) Demographic transition in resource rich countries: A blessing or a curse? World Development 45(C): 337–351.

11.

Bjorvatn

Kjetil

Naghavi

Alireza

(2011) Rent-seeking and regime stability in rentier states. European Journal of Political Economy 27(4): 740–748.

12.

Bjorvatn

Kjetil

Farzanegan

Mohammad Reza

Schneider

Friedrich

(2012) Resource curse and power balance: Evidence from oil-rich countries. World Development 40(7): 1308–1316.

13.

Bjorvatn

Kjetil

Farzanegan

Mohammad Reza

Schneider

Friedrich

(2013) Resource curse and power balance: Evidence from Iran. Review of Middle East Economics and Finance 9(2): 133–158.

14.

Blundell

Richard

Bond

Stephen Roy

(1998) Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics 87(1): 115–143.

15.

Collier

Paul

Hoeffler

Anke

(2004) Greed and grievance in civil war. Oxford Economic Papers 56(4): 563–595.

16.

Dunning

Thad

(2005) Resource dependence, economic performance, and political stability. Journal of Conflict Resolution 49(4): 451–482.

17.

Enikolopov

Ruben

Zhuravskaya

Ekaterina

(2007) Decentralization and political institutions. Journal of Public Economics 91(11–12): 2261–2290.

18.

Farzanegan

Mohammad Reza

Witthuhn

Stefan

(2014) Demographic transition and political stability: Does corruption matter? CESifo Working Paper No. 5133, Munich (http://ssrn.com/abstract=2550129).

19.

Farzanegan

Mohammad Reza

Lessmann

Christian

Markwardt

Gunther

(2013) Natural-resource rents and internal conflicts: Can decentralization lift the curse? CESifo Working Paper No. 4180, Munich (http://ssrn.com/abstract=2244830).

20.

Fearon

James D

(2005) Primary commodity exports and civil war. Journal of Conflict Resolution 49(4): 483–507.

21.

Fearon

James D

Laitin

David

(2003) Ethnicity, insurgency, and civil war. American Political Science Review 97(1): 75–90.

22.

Fjelde

Hanne

(2009) Buying peace? Oil wealth, corruption and civil war, 1985–99. Journal of Peace Research 46(2): 199–218.

23.

Fredriksson

Per G

Svensson

Jakob

(2003) Political instability, corruption and policy formation: The case of environmental policy. Journal of Public Economics 87(7–8): 1383–1405.

24.

Gaibulloev

Khusrav

Sandler

Todd

Sul

Donggyu

(2014) Dynamic panel analysis under cross-sectional dependence. Political Analysis 22(2): 258–273.

25.

ICRG (2011) International Country Risk Guide (https://www.prsgroup.com/about-us/our-two-methodologies/icrg).

26.

Jinjarak

Yothin

(2009) Trade variety and political conflict: Some international evidence. Economics Letters 103(1): 26–28.

27.

Keefer

Philip

(2010) DPI2010 Database of political institutions: Changes and variable definitions. Development Research Group, World Bank (http://www.nsd.uib.no/macrodataguide/set.html?id=11&sub=1).

28.

Knack

Stephen

Keefer

Philip

(1995) Institutions and economic performance: Cross-country tests using alternative institutional measures. Economics and Politics 7(3): 207–227.

29.

Le Billon

Philippe

(2003) Buying peace or fuelling war: The role of corruption in armed conflicts. Journal of International Development 15(4): 413–426.

30.

Lensink

Robert

Scholtens

Bert

(2006) Financial development, inflation uncertainty and growth volatility. In: Bagella

Michele

Becchetti

Leonardo

Hasan

Iftekhar

(eds) Transparency, Governance and Markets. Amsterdam: Elsevier, 95–112.

31.

Mehlum

Halvor

Moene

Karl

(2006) Fighting against the odds. Economics of Governance 7(1): 75–87.

32.

Nickell

Stephen John

(1981) Biases in dynamic models with fixed effects. Econometrics 49(6): 1417–1426.

33.

Omgba

Luc Désiré

(2009) On the duration of political power in Africa: The role of oil rents. Comparative Political Studies 42(3): 416–436.

34.

Paldam

Martin

(1987) Inflation and political instability in eight Latin American countries 1946–83. Public Choice 52(2): 143–168.

35.

Poteete

Amy R

(2009) Is development path dependent or political? A reinterpretation of mineral-dependent development in Botswana. Journal of Development Studies 45(4): 544–571.

36.

Riker

William H

(1964) Federalism: Origins, Operation, Significance. Boston, MA: Little, Brown.

37.

Robinson

James A

Torvik

Ragnar

Verdier

Thierry

(2006) Political foundations of the resource curse. Journal of Development Economics 79(2): 447–468.

38.

Roodman

David

(2007) A short note on the theme of too many instruments. Center for Global Development Working Paper No. 125 (http://www.cgdev.org/content/publications/detail/14256/).

39.

Roodman

David

(2009) How to do xtabond2: An introduction to difference and system GMM in Stata. Stata Journal 9(1): 86–136.

40.

Sørli

Mirjam E

Gleditsch

Nils Petter

Strand

Håvard

(2005) Why is there so much conflict in the Middle East? Journal of Conflict Resolution 49(1): 141–165.

41.

Tullock

Gordon

(1980) Efficient rent-seeking. In: Buchanan

James

Tullock

Gordon

Tollison

Robert

(eds) Toward a Theory of the Rent-Seeking Society. College Station, TX: Texas A&M University Press, 97–112.

42.

Wooldridge

Jeffrey M

(2002) Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: MIT Press.

43.

World Bank (2012) World Development Indicators (http://data.worldbank.org/data-catalog/world-development-indicators).

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB