On the effect of natural resources on interstate war

Abstract

Many analysts expect that a decline in natural resources and a rise in resource demand will increase the risk of interstate war in the coming decades. Other researchers reject this expectation. Empirical examinations are mostly qualitative case studies, and the few statistical models center primarily on waters. We argue changing resource levels can provide either an incentive or a disincentive for countries to go to war. The net effects by resource are empirical issues. We offer statistical models for a large sample, including several types of renewable and non-renewable resources, and other variables. We find that resource changes impact interstate war, and the magnitude of their impacts is on par with that of non-resource forces, and the effects of one time resource changes linger. The paper examines implications for the coming decades in light of the United States’ National Intelligence Council’s Global Trends 2030 outlook.

Keywords

environment fighting formal international model statistical

I Introduction

The United States (US) government’s National Intelligence Council (NIC), the center for forecasting US security threats, argues three baskets of risks could make interstate war more likely by 2030: (1) changing calculations of major countries; (2) rising access to a wider spectrum of tools of war; and (3) more conflict over resources such as energy, food, water, and minerals, due to increasing demand stemming from growing populations and consumption patterns of a larger middle class; and declining supply due to finite stocks and climate-change damages (NIC, 2012).

This paper does not examine the first two baskets of risks since such an analysis requires access to restricted data. Realizing, however, that the NIC is not alone in its risk projection for basket three, we seek to contribute to the dialog about it.¹ In the tradition of social science, we take the approach of using past data to say something about possible developments in the coming decades. Though it is not possible to say for certain whether the future of war will resemble its recent past, we consider that this exercise is useful under the assumption that conditions in the future will generally resemble those when the data were collected.

Considering modern classics in the scientific study of war (e.g. Richardson, 1960; Wright, 1965), studies describe a causal role for resources in specific cases (e.g. Klare, 2002; Maddow, 2014). These studies lack systematic comparison with peaceful cases and consideration of non-resource factors, so it is hard to generalize their results. Meanwhile, only a few statistical models for large samples have been developed (Koubi et al., 2014); herein we offer our contribution.

This paper investigates the effect of a country’s resources on its overall propensity for war in world politics. That is to say, we ask whether countries with more of a given resource are more or less likely to go to war and whether the converse can be said for countries with less of this resource. A country level of analysis will help in answering these questions. The underlying assumption of this well-established research tradition is that national traits or a mixture of traits influence state behavior; states with similar features are expected to act similarly in world politics (e.g. Cashman, 2014; Levy and Thompson, 2011). We examine traits that despite the growing interest of governments and qualitative studies, have received little attention in statistical models: resources.

The NIC and other sources suggest that countries with fewer resources may become more war-prone overall, in line with some existing theories. However, other theories suggest that such countries may be less likely to go to war, or that countries with more resources may be more war-prone. Still other theories offer that states with more resources are less war-prone.

When there are conflicting theoretical effects, the direction of their net impact is an empirical issue. We do not pose an initial hypothesis, but rather seek to gain insight on the net effect by developing multivariate regression models for a large sample of countries and years. This approach enables us to alleviate limitations of the case-based approach by systematically looking at both war and peace and accounting for the potential effects of non-resource variables. Our approach, which asks a question at the country level of analysis, differs from the approach taken by most empirical models in the emerging statistical literature on war and resources, which asks a question at the country-pair level. The issue is not which level is ‘the most correct’, but rather is what can be examined within each level (e.g. Singer, 1969). Levels of analysis, in other words, are imperfect analytical tools to help us study certain questions, not goals in themselves (e.g. Cashman, 2014; Levy and Thompson, 2011).

Since war is probabilistic, we model its likelihood. Our theory suggests inclusion of a broad range of resources in our models. Though we cannot include all the resources people use, we cover key types. For terrestrial renewable resources, we use arable land, permanent cropland, timber, and agricultural output; for non-renewable resources, we examine minerals and fossil fuels; and for water, we study precipitation and freshwater. Since war is multi-causal, we also include non-resource variables. Here, again, we cannot examine all the conceivable inputs; nor can any model. We include those common to the literature and those of interest to our question.

The remainder of this paper is organized as follows. Section II takes stock of prior work. Sections III and IV develop our statistical models. Section V presents estimation results, and section VI places these results in a broader context.

II Prior work

The question of what role resources play in militarized interstate conflict, particularly war, is not new. This section seeks to gain insight from prior work. Since the relevant body of work is too large for us to fully cover it here, we summarize key theories and empirical findings.

1 Theoretical approaches

The extant theories can be classified into five types, depending on their resource-related mechanism. We label these groups Malthusian, Godwinian, Realist, Discontent, and Abundance.

a Malthusian

Malthus (1798) argues that population growth raises pressure on arable land in a country. Ultimately food per capita declines and people fight over land. The problem is classically a growing resource demand facing fixed supply. Many scholars apply this thesis to interstate war. For example, German political geographers before 1945 believed that seizing foreign resources was vital for sustaining Germany’s Lebensraum (e.g. Heske, 1987). Lenin (1917) wrote that in pursuing larger profit while facing fixed resources at home, the capitalist class pushes leaders to seize resources abroad, leading to war. Choucri and North (1975, 1989) say technical advances and population growth raise resource demand above supply in major states. This causes lateral pressure, including predatory trade and investment, acquisition of colonies and spheres of influence, deployment of military forces, and establishment of military bases, all of which raise the risk of war. Russett (1981) argues the resource pressures in the 1970s resemble those before the First World War, warning of coming interstate conflicts. Others say resource depletion may cause wars in the coming decades (section I).

b Godwinian

Godwin (1793), another famous writer of his era, argued that ongoing scientific progress would ease resource pressure due to population growth, rejecting the Malthusian thesis. Many scholars have since taken a Godwinian view: resource scarcity unleashes innovations that alleviate the shortage, thus removing the link to conflict. These innovations include finding more resources, substituting resources by other materials (e.g. Lomborg, 2001; Simon, 1998), conservation motivated by democracy (e.g. Payne, 1995), international trade and resource-use treaties (e.g. Conca, 2001; North, 1995), profit-driven research (e.g. Barro and Sala-i-Martin, 2004), and increased ingenuity due to population growth (e.g. Boserup, 1981; Simon, 1998). Simon (1989) conceptually integrates these measures to argue that resource shortage reduces the impetus to wage war by unleashing mitigations that improve conditions more than if the shortage had not occurred.

c Realist

Political realism assumes states primarily seek security and a sense of insecurity is ultimately the key cause of interstate war. To feel secure, states seek power in the international system, which, in turn, requires resources, among other inputs. Resources are thus implicated in interstate war (e.g. Morgenthau et al., 2005; Waltz, 1979). Theorists then pose that states, especially major states, maximize their power relative to others and in so doing they compete over resources deemed crucial for power, including by using force. Decline in resources at home thus increases the risk of interstate war (Mearsheimer, 2001). Van Evera (2001) argues states compete in particular over cumulative resources, which he defines as those whose control can ease the acquisition and retention of more resources. This competition can lead to war. Lebow (2010) suggests diminishing resources due to limited stocks and climate-change impacts may increase security concerns and, therefore, the risk of wars in which competition over resources becomes the central objective.

d Discontent

The idea that the effect of resources on interstate conflict works through internal discontent links to the diversionary theory of war, or the rally ’round the flag syndrome. Going back to classics such as Wright (1965), many authors suggest that leaders facing local turmoil may attempt to divert public attention and secure their own power by going to war (e.g. Cashman, 2014). Recent studies associate the disorder with resources. One theory links to the Malthusian thesis: resource scarcity at home can cause socio-economic problems, which increase public restlessness and turmoil (Wasson, 2007). Another theory links to the resource-curse literature, which observes that countries relying on resource export revenues as their chief income often perform economically worse and exhibit more political instability and are therefore more discontent than other states.² Schneider (2010) sees a possible link to interstate war, though he does not offer a theory. Soltanov (2011) links this particular discontent to interstate war propensity through the rally ’round the flag syndrome.

e Abundance

In this theory, the effect of resources on interstate war propensity is grounded in their abundance. As Colgan (2013a) discusses, countries with abundant resources may be relatively less war-prone, for war is bad for business. For example, transport costs may grow due to war damages and battle locations, insurance rates may rise due to growing risks, and resource stocks and extraction facilities may be destroyed. But resource income can also be used to buy arms, making states more able to fight, and increase a government’s autonomy in starting war by reducing its reliance on taxes and thus on public opinion. Leaders may also take excessive risk if they assume that facing a defeat, they could use resource revenues to buy enough loyalty to remain in office, and/or, as De Soysa et al. (2011) suggest, that their major resource importers will help them.

2 Empirical studies

Most empirical studies are qualitative. For example, Durham (1979) and Myers (1993) find, respectively, land scarcity partly caused the Soccer War (El Salvador-Honduras) and Ogaden War (Ethiopia-Somalia). Gleick (1993) describes militarized conflict over the Jordan River waters (Israel-Arab states, especially Syria) and tension over the Nile River (especially Egypt-Sudan-Ethiopia). Russett (1981) says scarcity of raw materials partly led to the First World War. Westing (1986) finds resource causes for the two World Wars (land, iron, oil), Chaco war (Paraguay-Bolivia, 1932–1935, oil), Six-Day War (Arab-Israeli, water), and Falkland War (Argentina-UK, fisheries, oil potential). Swatuk (2007) finds the Central African wars in 1998–2003 were partly over timber. Klare (2002, 2005) chronicles fighting over oil, and Colgan (2013a) shows oil enables exporting states to start wars.

The emerging modeling literature typically examines militarized interstate disputes (MIDs). For the country level of analysis, Stalley (2003) finds the risk of any MID in 1980–1992 – measured as a binary variable set to one when a country threatens to use force, displays force, uses force (defined by the MID data as violence killing less than 1000 combatants in a year), or engages in war (≥1000) – rises with soil degradation. Strüver (2010) links non-war MIDs in 1960–2001 to fuel profit, and a binary measure of fuel richness to war. Soltanov (2011) reports the risk of use of force MID in 1970–2002 rises with fuel profits. Colgan (2013b) finds petrostates (oil exporters) exhibit the same or higher risk of any MID (depending on model variant) in 1945–2001, compared to others, while petrostates with revolutionary leaders exhibit a higher risk than their peers. Braithwaite (2006) finds the spread of MIDs in 1993–2001 rises with or is unaffected (across model variants) by presence of oil, gems, or illicit drugs in a country, and falls with or is unaffected by forest size.

For the country-pair level, models primarily study freshwater (Koubi et al., 2014). Gleditsch et al. (2006) find that water scarcity raises the risk of fatal MIDs (at least one battle death) for riparian states in 1980-2001. Furlong et al. (2006) find that countries sharing a river are more likely than others to experience fatal MIDs in 1880–1992. Tir and Stinnett (2012) report similar results for river-sharing countries experiencing MIDs of any type in 1950–2000. As summarized by Koubi et al. (2014), others find that river-sharing countries tend to cooperate, though they examine dependent variables such as water treaties or negotiations over river claims, not MIDs. For non-renewable resources, Wasson (2007) finds iron/steel and oil scarcities raise risk of MID in 1990–2001 for adjacent states or pairs with at least one major power. De Soysa et al. (2011) argue oil exporters in 1946–1999 experience higher risk of MIDs.

To our knowledge, only three models study interstate war, per se. Examining six major European powers in 1870–1914, Choucri and North (1975) find colonial area per country (a proxy for lateral pressure) rises with population (for resource pressure) and income (for technology); war propensity rises with interstate colonial disputes short of war, which, in turn, rise with colonial area. Choucri et al. (1992) find similar results for Japan in 1878–1987. De Soysa et al. (2011) find oil exporters are less or not more likely (across model variants) than other states to fight wars, but are more likely to fight each other.

Interstate war is obviously not the only militarized violence studied in a resource context. Briefly, the literature as a whole seems to imply that scarcity is more often linked to renewable resources in promoting civil war and abundance is more often linked to non-renewable resources (e.g. Couttenier and Soubeyran; 2013; Koubi et al., 2014; Le Billon, 2012). Further investigation of these links, including a meta-analysis, would contribute to the civil war literature and complement our focus, which is whether the resources a country has play a causal role in its propensity to engage in interstate war.

3 Taking stock of the literature

The literature does not address precisely our question. The qualitative case studies show that resources play a causal role in certain interstate wars, but one cannot generalize their results, as they do not pair them with an appropriate counterfactual by examining cases in which war did not occur.

The empirical modeling section of the literature is small and primarily examines a binary variable set to one when MIDs of any type occur, essentially assuming the effect of resources on, say, a threat to use force, resembles their effect on war. Moreover, the all-MIDs measure includes events such as individual rather than state actions, minor border crossing or airspace violations, and fishing boat incidents, which fall short of militarized conflict (Downes and Sechser, 2012).

For war, the six-country sample of Choucri and North (1975) and the Japan focus of Choucri et al. (1992) seem too narrow. Strüver (2010) also tabulates wars and fuel richness, but does not offer a model or include peaceful cases. De Soysa et al. (2011) offer mixed results and use of directed country-pairs raises issues.³

Finally, the case studies find a role for various resources, but other than Stalley (2003) and Wasson (2007), who include three and two resources, respectively, models include only one resource, primarily shared water. Choucri and North (1975) and Choucri et al. (1992) offer a theory for many resources, but use population as a proxy variable for resources. A recent review paper concludes that the number of models for non-renewable resources is surprisingly small (Koubi et al., 2014).

In all, the literature paints a mixed and partial picture, making more wide-ranging statements difficult. While it gives reasons to suspect that resources play a causal role in militarized conflict, it leaves ample room for future research (Koubi et al., 2014). We build on the literature as we attempt to alleviate some of its limitations. Our first step is to develop an empirically oriented theoretical model.

III A decision-making model of interstate war and resources

Section II presents seven theories: Malthusian, Realist, Malthusian Discontent, Godwinian, Resource Curse Discontent, Abundance to War, and Abundance to Peace. We suggest an eighth theory, Malthusian Trap, which flows from Malthus’ prediction that society will end up in a state of poverty and conflict. Countries heading toward this trap may not be able to fight, thus a decline in resources may reduce their war propensity. Table 1 summarizes these theories by mechanism and effect of a resource change stated as they themselves imply. There are four categories: resource reduction leads to (→) smaller war propensity; resource reduction → greater propensity; resource increase → greater propensity; and resource increase → smaller propensity.

Table 1.

Effects of resource changes on interstate war propensity.

Theory	Mechanism	Effect
1. Malthusian	Pursue Lebensraum, profit; Lateral pressure	Resource decline → Greater propensity
2. Realist	Pursue military power, which requires resources	Resource decline → Greater propensity
3. Malthusian Discontent	Rally ’round the flag effect	Resource decline → Greater propensity
4. Godwinian	Markets, progress, and democracy solutions	Resource decline → Smaller propensity
5. Resource Curse Discontent	Rally ’round the flag syndrome	Resource increase → Greater propensity
6. Abundance to War	Militarize, leader autonomy, risk taking	Resource increase → Greater propensity
7. Abundance to Peace	War is bad for resource business	Resource increase → Smaller propensity
8. Malthusian Trap	Less mental energy and capabilities	Resource decline → Smaller propensity

The Malthusian, Realist, and Malthusian Discontent theories in Table 1 expect that a decline in a country’s resource is associated with a higher overall propensity to engage in war. The Godwinian and Malthusian Trap theories predict a decline in resources is linked to a decline in war propensity. The Resource Curse Discontent and Abundance to War theories suggest that more resources are related to a higher war propensity, and the Abundance to Peace position expects that more resources are associated with a smaller propensity.

The term propensity for war is treated here as probabilistic, which is to say these theories are not deterministic, but statements of increased or decreased likelihood. We thus develop a model for the likelihood of war in light of resources. We begin by assuming a government faces an international dispute that it must attempt to resolve through war or otherwise. The nature of the dispute is not a concern here, as in most conflict models. Our modeling effort centers on how an increase (or decrease) in a resource in a country increases or decreases the likelihood that a government will go to war over the disputed issue, not why they fight. When we refer to a country’s resources, we mean all the resources found in the country, not only state-owned or controlled natural assets.

We model the government as a unitary actor, a simplification widely used in modeling literature. This actor is assumed to be rational, implying the government chooses the option (war or no war) that offers a larger utility. This assumption is also implicitly or explicitly included in effectively all models of conflict, and much of social science as a whole. Formally, the government of country i (i = 1, 2…) expects to gain utility $U_{i w t}$ from going to war and utility $U_{i p t}$ should it choose no war. Regression analysis typically employs a linear form, so we assume the utilities depend linearly on the country’s resources (R), other observed attributes (X), and unobserved or unmeasured forces ( $ε$ ):

U_{i p t} = α_{p} + X_{i t}^{'} β_{p} + R_{i t}^{'} γ_{p} + ε_{i p t}

U_{i w t} = α_{w} + X_{i t}^{'} β_{w} + R_{i t}^{'} γ_{w} + ε_{i w t}

The column vector $R_{i t}^{'}$ (prime denotes transposed) holds resource values at time t. The column vector $X_{i t}$ holds values of other forces that influence the war decision. The elements of the column vectors $β_{p}$ and $β_{w}$ give the marginal utilities of the X forces, the utility change due to a unit increase in each of the forces when all else is constant. The elements of the column vectors $γ_{p}$ and $γ_{w}$ hold the marginal utilities of the resources. The random variables $ε_{i p t}$ and $ε_{i w t}$ give the sums of the unobserved or unmeasured influences. We assume these forces are independent and identically distributed (IID) with a Gumbel probability density function (PDF) for a reason discussed below. Finally, the terms $α_{p}$ and $α_{w}$ give the base utilities when all the other stimuli are zero.

The utility levels are unobserved, but we observe the war/no war choice. A binary variable $w_{i t}$ equals 1 if the government chooses war at time t and equals 0 if it chooses peace. Since the government is a rational actor, a choice of war indicates its perception that $U_{i w t} > U_{i p t},$ and vice versa if it chooses no war. Using the notation $P r (e v e n t)$ to denote probability of an event, we can thus write expression (2) for the probability that a government chooses war:

\begin{aligned} P r_{i t} (w_{i t} = 1) = P r (U_{i w t} > U_{i p t}) \\ = P r (U_{i w t} - U_{i p t} > 0) \end{aligned}

The term $U_{i w t} - U_{i p t}$ represents the propensity of the government of country i to engage in war in time t. Substituting the utilities from (1a) and (1b) into (2) and rearranging gives:

\begin{aligned} P r_{i t} (w_{i t} = 1) = P r (ε_{i w t} - ε_{i p t}) > \\ - (α_{w} - α_{p} + X_{i t}^{^{'}} (β_{w} - β_{p}) \\ + R_{i t}^{^{'}} (γ_{w} - γ_{p})) \end{aligned}

Equation (3) tells us that the probability of war is given by the area under the PDF of ( $ε_{i w t} - ε_{i p t}$ ).

A known result indicates the difference between two IID random variables with the Gumbel PDF is a random variable with a logistic PDF. Since the logistic PDF is symmetric, we get:

\begin{aligned} P r_{i t} (w_{i t} = 1) = P r (ε_{i w t} - ε_{i p t} > \\ - (α_{w} - α_{p} + X_{i t}^{^{'}} (β_{w} - β_{p}) + R_{i t}^{^{'}} (γ_{w} - γ_{p})) \end{aligned}

\begin{aligned} = P (ε_{i w t} - ε_{i p t}) < (α_{w} - α_{p} + X_{i t}^{^{'}} (β_{w} - β_{p}) \\ + R_{i t}^{^{'}} (γ_{w} - γ_{p}))) \end{aligned} .

The probability of war is thus given by the familiar logistic cumulative distribution function:

P r_{i t} (w_{i t} = 1) = \frac{e^{α_{w} - α_{p} + X_{i t}^{^{'}} (β_{w} - β_{p}) + R_{i t}^{^{'}} (γ_{w} - γ_{p})}}{1 + e^{α_{w} - α_{p} + X_{i t}^{^{'}} (β_{w} - β_{p}) + R_{i t}^{^{'}} (γ_{w} - γ_{p})}} .

Equation (5) says the effect of a change in some resource j (where j = 1, 2…), in country i, on the likelihood of war depends on the relative sizes of the elements of the vectors $γ_{p}$ and $γ_{w}$ for that resource. Stated in terms of the theories in Table 1, we get three cases.

C a s e (A), γ_{j w} > γ_{j p}

A decline in country i’s resource j reduces the probability of choosing war, and an increase in this resource increases the probability.⁴ This case represents the Godwinian, Malthusian Trap, Resource Curse Discontent, and Abundance to War theories in Table 1.

C a s e (B), γ_{j w} < γ_{j p}

A decline in country i’s resource j increases the probability of choosing war, and an increase in this resource reduces the probability.⁵ This case represents the Malthusian, Realist, Malthusian Discontent, and Abundance to Peace theories in Table 1.

C a s e (C), γ_{j w} = γ_{j p} :

A change in country i’s resource j does not affect the probability of choosing war. This case represents a more or less similar effect on the war and peace utilities, resulting in a small or no effect on war.

This formal model suggests it is not possible to predict based on theoretical consideration alone the effect of a change in some resource in a country on the probability that it will be more war-prone overall. Such theoretical indeterminacy often occurs in the social sciences, particularly when issues are complex and there are potentially competing forces at play.⁶

While we do not state an initial hypothesis for the effects of change in a given resource on the likelihood of choosing war, we can statistically examine changes. The competing effects may vary across resources, but this is not an insoluble problem, for the model can include many resources. Indeed, since national economies use many resources, it is reasonable to assume a government considers all of them as it decides whether to go to war. Our model thus needs to include measures of many resources, though there are obviously practical limits to what can be included.

IV Statistical model

We first manipulate equation (5) to get a linear form on the right hand side. This gives:

ln (\frac{P r_{i t}}{1 - P r_{i t}}) = α + X_{i t}^{^{'}} β + R_{i t}^{'} γ

The term $l n (\frac{P r_{i t}}{1 - P r_{i t}})$ is the logit function of war. This section specifies the elements of X and R vectors, and discusses econometric issues.

All the variables are measured for a country, per year. The variable names are capitalized below and appear as such in the tables reporting the estimation results. We estimate models for War Onset and War Involvement, both of which are binary variables, in line with equation (6). We set War Onset to 1 for the first year of a war, and War Involvement to 1 for each year of a war. These variables are coded based on the MIDs data in COW (2010), which defines war as interstate violence killing at least 1000 combatants in a year.

The vector R includes terrestrial and water-related renewable resources, and non-renewable resources. The terrestrial renewable resources include Arable Land, Cropland, Timber, and Agriculture. Arable Land is the size of the productive land in hectares per capita. Timber is the total profit from wood extraction, expressed as a share of national income. The data for these variables come from the World Bank (2005). Cropland is the share of the total area permanently used for growing crops. Agriculture is an index of agricultural output produced per capita using 1999–2001 as a base year. The data for these variables come from GEO (2010).

The water-related renewable resources include Precipitation and Freshwater. Precipitation is the average yearly precipitation expressed in millimeters, as recorded by GEO (2010). Freshwater is the total flow of internal (in a country) and external (into a country) freshwater, expressed in cubic meters per capita, as recorded by FAO (2010).

For non-renewable resources, we include Minerals and Fuel. The Minerals variable is the share of the total profit from selling materials such as metals, gemstones, and phosphates in the national income. Fuel is the share of the export of all types of fuels (mostly fossil, by far; but also other types), out of a country’s total export. The data for these variables come from the World Bank (2005).

Though our resource list is not exhaustive, economic theory suggests that many of them stand as proxies for several other resource variables. For example, larger shares of timber or mineral profits in the national income likely indicate a smaller need to import timber or minerals, or lower levels of dependence on import of timber and minerals, respectively. Larger share of exporting fossil fuels in the total export indicates a smaller need to import fuels, or a lower level of fuel import dependence; countries that export fuels, in other words, should not depend much, if at all, on importing them. The per capita measures for arable land, agricultural production, and freshwater likely indicate greater food availability. A larger share of the total area permanently used for growing crops likely indicates greater availability of crops in a country. Further investigation of these points is better taken by future research, which we revisit in the last section.

For the X vector of control variables, we include typical forces and indicate our rationale for their inclusion, though we naturally cannot spend much time discussing their expected effects. Population is the logged population size. Population Growth is the population’s yearly growth rate. The data for both variables come from the World Bank (2005). A larger population may mean a larger army, increasing the ability to fight. But a larger army may raise a sense of security, reducing the willingness to fight. A larger population may also mean greater poverty and/or pressure on resources, raising the willingness to fight, but these forces may reduce the will and ability to fight.

The variable Borders is the number of borders a state has with other nations from COW (2010).⁷ A larger number of borders may come with more contentious issues and enable fighting due to proximity. But it may also promote peaceful behavior to prevent attacks by adjacent states.

The variable Disasters is the number of persons killed by extreme weather events (e.g. storms, floods) per million people. The data come from GEO (2010).⁸ A rise in Disasters may reduce ability to fight by damaging properties, but may also increase willingness to attack others in order to divert public discontent and/or replace those properties by seizing those of other countries.

GDPpc is the logged gross domestic product (GDP) per capita, expressed in constant 1985 dollars. The data come from the University of Pennsylvania and the World Bank, as compiled by Fearon and Laitin (2003). A richer state may have more means to fight, but it may be more averse to war since it has more to lose. We add a quadratic logged GDPpc term to allow for a non-linear effect.

Trade Openness is the sum of the export and import values of a country, divided by the country’s GDP. The data come from the World Bank (2005). Larger Trade Openness can promote economic growth, increasing means to fight, but could also promote peace since war can harm business.

National Capability is an index of material capabilities from COW (2010). More capable states are more able to fight, but they may also feel more secure, reducing their willingness to fight. Democracy ranges from –10 (full autocracy) to 10 (full democracy). The data come from Polity (2010). A higher Democracy may reduce willingness to fight due to norms of peaceful dispute resolution, but may also enable pro-war interest to affect the government. Political Instability is coded 1 if Democracy changes at least three points over its level in the last three years, and 0 otherwise.⁹ A rise in Political Instability may promote war via the rally ’round the flag effect, but may also discourage war by reducing the ability to fight.

Since we aim to make a general statement, we estimate the model using a panel of many countries and years. We modify the specification slightly to account for the possible effect of war on resources and/or control variables. While this effect may be weak, we take a conservative approach and replace variables by their first lag. Models with resources lagged two or three periods are also estimated under the assumption that resource activity might change in the expectation of war in the next period, necessitating longer than one period lag.

The coefficient estimates for the resource variables give marginal effects on war due to a one-time unit rise in their levels. We also estimate models with distributed resource lags (include several lags), as one-time unit rise in a resource may continue to affect war in the following periods due to habit formation or adjustment time.

We center the linear and quadratic terms for GDPpc at their means to mitigate collinearity. Finally, in panel data, the error term may be heteroskedastic and/or serially correlated. We deal with these possibilities by estimating robust standard errors clustered by country. These standard errors should also alleviate the risk of serial correlation due to the prevalence of peace years in the sample, but we include the method of Beck et al. (1998) to further mitigate this issue.¹⁰

V Results

This section presents the estimation results. The estimation sample includes 123 countries from 1975 to 2001.¹¹ The sample size varies from 1711 to 1944, depending on the model estimated. Appendixes 1 –3 list the countries and summary statistics. For the estimates, recalling section III’s competing effects, a positive and statistically significant coefficient estimate for some resource would indicate that for an increase in that resource, the war-promoting forces outweigh the war-inhibiting forces, on average, and vice versa. A statistically insignificant estimate would tell us that the competing effects are about equal in size, on average. We could use one-tailed significance tests under the reading that competing effects come from different theories, or the two-tailed test under the reading that effects come from one theory. We use the more conservative two-tailed test.

Table 2 presents results for resources lagged one period. The pseudo R ² is 0.237 for war onset and 0.392 for war involvement, suggesting a relatively good fit. The $χ^{2}$ statistics are jointly significant, further indicating the model fits the data. The mean VIF score is 1.66, the max VIF score is 2.88, and the condition index is 3.57, so multi-collinearity is not a concern, a finding supported by the correlation matrix in Appendix 3. In the tables presenting the coefficients, bolded and underlined estimates are significant at the 1% level for the two-tailed test, bolded estimates are significant at the 5% level, and underlined estimates at the 10% level. We report signs of estimates significant at the 5% level and report p-values for estimates significant at the 10% level.

Table 2.

Resources lagged one period.

	Onset		Involvement
Variable	Coefficients	Size of effects %	Coefficients	Size of effects %
Arable Land	0.826	99.6	0.573	88.3
	0.039		0.045
Agriculture	0.00986	0.0	0.00989	0.0
	0.210		0.149
Cropland	–0.035	0.0	–0.0029	0.0
	0.543		0.942
Timber	0.516	175.2	0.397	142.8
	0.000		0.004
Minerals	–0.686	83.0	–0.441	20.9
	0.103		0.080
Fuel	–0.0203	42.5	–0.0184	40.0
	0.011		0.002
Freshwater	–0.0177	28.6	–0.0161	27.8
	0.017		0.000
Precipitation	–0.00207	27.7	–0.00170	26.8
	0.012		0.002
Population	0.787	183.0	0.801	176.0
	0.001		0.000
Population Growth	0.519	203.6	0.380	149.6
	0.000		0.003
Disasters	–0.0115	0.0	0.00165	107.6
	0.197		0.000
Trade Openness	0.00875	0.0	0.00515	0.0
	0.373		0.477
GDPpc	1.214	158.9	0.668	110.9
	0.005		0.055
GDPpc²	–0.562	46.5	–0.178	0.0
	0.056		0.339
Borders	0.0171	0.0	–0.0584	0.0
	0.846		0.306
Political Instability	–1.692	23.3	–2.138	17.3
	0.037		0.014
National Capability	–13.027	58.5	–14.219	63.7
	0.076		0.015
Democracy	0.0132	0.0	0.0136	0.0
	0.754		0.655
N	1944		1944
Pseudo R²	0.237		0.392
χ2	136.61		304.61
P > χ²	0.000		0.000
Log Likelihood	–133.694		–183.412

Note: p-level below, two-tailed test: bold and underline p ≤ 0.01; bold ≤ 0.05; underline p ≤ 0.1; spline terms, peace years and a constant terms are included but are not reported. The coefficient estimates for water are multiplied by 1000.

The signs of the significant coefficient estimates are largely similar for War Onset and War Involvement. The estimates for Arable Land and Timber in Table 2 are positive, indicating that rise in arable land per capita or the share of timber profits in the national income make war more likely. The estimates for Agriculture and Cropland are insignificant, suggesting their first lag does not play a role in war or that the positive and negative forces are about equal. The estimates for Freshwater and Precipitation in Table 2 are negative, meaning a rise in terrestrial freshwater per capita or precipitation makes war less likely. The estimate for Minerals is negative and almost significant at the 10% level (p = 0.103) for War Onset, and negative and significant at the 10% level for War Involvement. A rise in the share of mineral profits out of the national income weakly reduces the likelihood of war. Finally, the estimate for Fuel is negative, telling us that a rise in the share of fuel export makes war less likely.

Among the controls, an increase in weather Disasters does not affect War Onset in the same year, but raises the likelihood of a War Involvement. The War Involvement result (and others like it in Table 4 below) specifically seems to call for additional research into the effect of climate change and war, as it is expected to increase the intensity and frequency of weather disasters.

The War Onset estimate is positive for GDPpc and negative at the 5.6% p-level for GDPpc², suggesting the probability of War Onset rises with GDPpc until about 14,250 in current dollars and then falls. For War Involvement, the probability rises with GDPpc at the 5.5% p-level. A rise in Political Instability inhibits the likelihoods of both War Onset and War Involvement. The estimate for National Capability is negative at the 7.6% p-level for War Onset and negative for War Involvement.

Changes in Borders and Democracy do not affect War Onset and War Involvement, but the coefficient estimates for Population and Population Growth are positive for both War Onset and War Involvement, suggesting the overall war propensity of a country rises with these variables. This initially seems to uphold the Malthusian thesis, but it also supports ideas suggesting states with larger populations are more able to wage war, and are more willing to fight to change the status quo due to their poverty, as discussed in section IV. These results may also initially seem to run against the decline in interstate war since 1945 despite population growth, but this may not necessarily be the case. These estimates are marginal effects, assuming nothing else changes, but many forces have changed for many states since 1945. For example, material capabilities grew, political instability rose (e.g. civil strife), and GDPpc rose above the inverted U threshold, all of which reduce war propensity. Additional mitigators reducing war propensity may include United Nations mediation, foreign aid growth, and increase in efficient use of resources.

For the magnitude of effects, we compute relative risks, as in most studies of war. The relative risk measure is defined as the ratio of the predicted probability of war when one variable is increased, holding all the other variables at some base values, to the predicted probability of war when all the variables are at their base values. For our base values, we choose the means for the continuous variables and 0 for the binary political instability. The continuous variables are raised by one standard deviation above their mean, and the binary from 0 to 1, one change at a time. We report results for the coefficient estimates significant at the 10% p-level and better. The relative risks for the insignificant coefficient estimates are set to zero, indicating no effect.

Columns 2 and 4 of Table 2 report the relative risks for the significant coefficient estimates. A rise of one standard deviation in Arable Land above its mean nearly doubles the relative risks of War Onset and War Involvement, compared with the base level risk. Increases in Freshwater or Precipitation reduce the War Onset and War Involvement risks about 28%. For Timber, the relative risks rise 175% for War Onset and 140% for War Involvement, and for Fuel the risks decline around 40%. For Minerals, the relative risk falls almost 21% for War Involvement at the 10% significance level and 83% at the 10.3% p-level for War Onset. These relative risks are substantial. The largest relative risk in the model (203.6% for Population Growth) divided by the smallest resource relative risk (20.9% rise for Minerals) is less than 10, indicating the resource relative risks have the same order of magnitude of the relative risks obtained for the non-resource variables.

The thrust of these results holds in Tables 3 and 4, so hereafter we focus on the resource effects. The results for resources lagged two or three periods in Table 3 resemble those in Table 2 with a few differences. For Arable Land, the War Onset estimate is still positive, though insignificant for two lags and more significant (7.5% p-level) for three lags. The Arable Land War Involvement estimates are more significant (8.7% p-level) for two lags and less significant (2.3% p-level) for three lags. For Agriculture, the two and three lags coefficient estimates are significant, unlike in Table 2. For Minerals, the involvement estimates are still negative, though insignificant.

Table 3.

Resources lagged two or three periods.

	Onset coefficients		Involvement coefficients
Variables	Two periods	Three periods	Two periods	Three periods
Arable Land	0.372	1.044	0.370	0.481
	0.184	0.075	0.087	0.023
Agriculture	0.016	0.015	0.015	0.013
	0.082	0.067	0.013	0.046
Cropland	0.017	−0.047	0.028	0.001
	0.822	0.386	0.558	0.971
Timber	0.334	0.635	0.266	0.442
	0.024	0.000	0.048	0.001
Minerals	−0.199	−0.904	−0.103	−0.157
	0.261	0.216	0.350	0.298
Fuels	−0.019	−0.014	−0.022	−0.011
	0.027	0.120	0.000	0.094
Freshwater	−0.0128	−0.0176	−0.0122	−0.0120
	0.070	0.021	0.011	0.006
Precipitation	−0.002	−0.002	−0.001	−0.001
	0.015	0.011	0.009	0.011
Population	0.819	0.782	0.849	0.770
	0.001	0.001	0.000	0.000
Population Growth	0.509	0.525	0.418	0.164
	0.002	0.003	0.001	0.331
Disasters	−0.014	−0.012	0.000	0.000
	0.282	0.181	0.516	0.432
Trade Openness	0.005	0.007	0.010	0.011
	0.679	0.459	0.153	0.098
GDPpc	0.975	1.288	0.484	0.387
	0.021	0.010	0.103	0.183
GDPpc²	−0.183	−0.774	0.015	−0.111
	0.548	0.030	0.933	0.524
Borders	0.044	−0.015	−0.018	−0.032
	0.580	0.862	0.739	0.517
Political Instability	−1.229	−0.997	−1.182	−0.916
	0.171	0.114	0.127	0.145
National Capability	−11.546	−10.948	−10.911	−9.005
	0.077	0.155	0.040	0.074
Democracy	−0.013	0.043	−0.011	0.007
	0.788	0.292	0.721	0.820
N	1913	1880	1913	1880
Pesudo R²	0.210	0.214	0.402	0.392
χ2	103.48	104.07	304.02	298.90
P > χ²	0.000	0.000	0.000	0.000
Log Likelihood	−134.837	−127.422	−185.310	181.871

Table 4.

Distributed lags for resources.

	Onset coefficients		Involvement coefficients
Variables	Two lags	Three lags	Two lags	Three lags
Arable Land	0.918	1.292	0.771	1.061
	0.054	0.075	0.009	0.008
Agriculture	0.019	0.029	0.017	0.024
	0.100	0.039	0.040	0.007
Cropland	−0.048	−0.050	−0.019	−0.015
	0.410	0.386	0.622	0.692
Timber	1.103	1.229	0.821	1.010
	0.001	0.003	0.000	0.000
Minerals	−0.832	−1.506	−0.599	−0.740
	0.120	0.189	0.041	0.078
Fuels	−0.024	−0.021	−0.023	−0.024
	0.014	0.057	0.000	0.003
Freshwater	−0.0210	−0.0207	−0.0179	−0.0177
	0.004	0.005	0.000	0.000
Precipitation	−0.003	−0.002	−0.002	−0.002
	0.028	0.031	0.004	0.005
Population	1.137	1.175	1.027	1.146
	0.000	0.001	0.000	0.000
Population Growth	0.656	0.663	0.532	0.499
	0.001	0.006	0.002	0.006
Disasters	−0.012	−0.017	0.002	0.002
	0.276	0.308	0.000	0.000
Trade Openness	0.015	0.015	0.011	0.015
	0.134	0.140	0.132	0.063
GDPpc	1.801	1.805	1.128	1.245
	0.003	0.017	0.008	0.008
GDPpc²	−0.514	−0.713	−0.206	−0.325
	0.181	0.116	0.359	0.197
Borders	0.014	−0.021	−0.045	−0.084
	0.894	0.842	0.488	0.195
Political Instability	−1.146	−1.144	−1.300	−1.095
	0.059	0.057	0.039	0.070
National Capability	−21.255	−21.098	−17.288	−18.375
	0.006	0.026	0.006	0.004
Democracy	0.007	0.035	0.013	0.020
	0.877	0.435	0.664	0.523
N	1813	1711	1813	1711
Pseudo R²	0.270	0.246	0.435	0.432
χ2	144.35	303.21	351.04	589.53
P > χ²	0.000	0.000	0.000	0.000
Log Likelihood	−117.486	−107.729	−162.137	−149.418

Table 4 presents results from models representing resources using distributed lag structures. For each resource, we report the sums of lagged resource coefficient estimates and their significance levels. These sums give the longer run marginal resource effects on war. For the non-resource variables, we report the coefficient estimates. We show results for two and three distributed resource lags. Longer resource lag structures were also tried but the results were increasingly less significant.

The estimated long run marginal resource effects on war in Table 4 have the same signs as the short run marginal resource effects in Table 2, but are larger and more statistically significant across the board, with one small exception. For Arable Land the long run marginal effects in Table 4 are still larger than in Table 3, but they are significant at the 5.4% p-level for two distributed lags and at the 7.5% p-level for three distributed lags case. Regardless, we see that the effects of one time changes in resources continue over time for several periods, underlining their importance.

Table 5 presents the long-run marginal effects for War Involvement reported in Table 4 and indicates the theories they support. Column 3 shows results for more of a given resource, and column 4 for less. For example, the coefficient estimate for Minerals is negative, so a rise in Minerals in a country reduces its overall propensity toward war, and a decline increases it.

Table 5.

Expected and estimated effects of resource changes on interstate war propensity.

Theory	Expected effect	Supported by estimates stated for more resources	Supported by estimates stated for less resources
1. Malthusian	Resource decline → Greater propensity		Minerals, Fuel, Water, Precipitation
2. Realist	Resource decline → Greater propensity		Minerals, Fuel, Water, Precipitation
3. Malthusian Discontent	Resource decline → Greater propensity		Minerals, Fuel, Water, Precipitation
4. Godwinian	Resource decline → Smaller propensity		Arable land, Timber, Agriculture
5. Resource Curse Discontent	Resource increase → Greater propensity	Arable land, Timber, Agriculture
6. Abundance to War	Resource increase → Greater propensity	Arable land, Timber, Agriculture
7. Abundance to Peace	Resource increase → Smaller propensity	Minerals, Fuel, Water, Precipitation
8. Malthusian Trap	Resource decline → Smaller propensity		Arable land, Timber, Agriculture

The estimates for Minerals, Fuel, Freshwater, and Precipitation are negative, suggesting the effects of changes in these resources support the Malthusian, Realist, Malthusian Discontent, and Abundance to Peace positions. The results for Arable land, Timber, and Agriculture are positive, suggesting support for the Godwinian, Resource Curse Discontent, Abundance to War, and Malthusian Trap positions. Our models do not indicate relative contributions of these theories. We revisit this point below.

VI Summary and implications

The effect of resources on interstate war has attracted growing scholarly and governmental attention. The quantitative literature, by contrast, has been slow to develop. This paper examines the causal role of a country’s resources on its overall propensity for war in world politics. Our formal model offered competing effects, and so we turned to empirical analysis. Our statistical models included measures of eight types of resources per country and other variables. The estimation used a large N sample of countries and years. We find that a country’s resources play statistically significant roles in its overall propensity to engage in war.

Next, we consider implications of these findings for the coming decades, assuming that past trends in our model will continue into the period on which we plan to comment. In so doing, we return to the NIC (2012) projection of growing international conflict over resources increasing the likelihood of interstate war by 2030. Regardless of whether this projection will hold true, it is useful to use it as a reference point for our discussion if only because of the stature of the NIC in the USA and the influence it can exert on its government (and possibly others).

Our first step is to forecast the average levels of the resources included in our models by 2030. This is a major undertaking, for any numbers pertaining to one resource are linked to the supply and demand patterns of other resources. Such a forecast would need to venture into geology and environmental science (for resource availability), and into economics and sociology (for resource extraction and consumption). In place of such extensive additional work, we rely on the NIC’s resource change expectations for 2030 and use our model to examine implications for war.

To summarize, we assume: (1) the effects identified by our models will hold to 2030; and (2) the availability of resources will evolve according to the NIC’s projection. The NIC expectation for declining availability of energy, freshwater, and minerals by 2030 implies Fuel, Freshwater, and Minerals will decline. By 2030, the NIC expects rainfall will generally decline due to climate change, indicating less Precipitation. The NIC expectations of rising demand for food and declining food supply imply that Agriculture will rise to meet the demand, but still fall short. The NIC also highlights the precedence for countries’ reliance upon increased deforestation as means to support growing populations, suggesting Timber, Arable Land, and Cropland could increase.

We recognize these are strong assumptions and revisit them shortly. Taking the NIC projections at face value, our finding that increases in Freshwater, Precipitation, Minerals, and Fuel in a country reduce its overall propensity for interstate war, coupled with the NIC projected worsening outlook for these resources, suggests higher chances for interstate war by 2030. Our finding that rises in Arable Land, Timber, and Agriculture raise war propensity, taken alongside the NIC’s projection of increases in these resources, also suggests increased chances of warfare. We find these effects grow over time and have the same order of magnitude of those obtained for the non-resource variables. Summarizing, our finding for the NIC resource scenario to 2030 supports an expectation of growing chances of interstate war for the average country. This prospect is worrying, but it assumes, as all regressions do, that the other variables in the model remain constant. This assumption rarely holds, so let us consider next which of our non-resource variables could change by 2030.

The NIC, the UN, and many others, expect the populations of the large majority of countries to increase by 2030, though by a declining rate. Our model suggests a larger Population will increase propensity for war, but a smaller Population Growth rate will reduce this propensity. By 2030, GDPpc will likely rise for all countries, though we do not know if it will rise above the US$14,250 (in current dollars) threshold for the inverted U we see for the one-year lagged results. If it does, further rises in GDPpc would make warfare less likely; otherwise, or if the effect proves to be linear as in Table 4’s distributed lag results, the chances of war will rise. The NIC assumes climate change will increase the frequency and intensity of weather Disasters. This too would raise the chances of war in the model. National Capability may climb, reflecting the likely growth of its components. This effect would reduce the chances of war. Considering these modifications, increased chance for warfare still seems possible, provided, again, the future trends will resemble their estimated precedents and that resource stocks will follow the NIC’s projected trajectory, neither of which is assured.

A third possibility exists for differentiation of the future from our projection. States could take proactive steps to avert this future. As Godwin argued, innovations may still be sufficient to alleviate resource problems, provided they are launched skilfully and with adequate support. These innovations take the form of adaptations or mitigations. Adaptation could increase resource-use efficiency and offer ways to better cope with weather disasters. Mitigation could cap greenhouse emissions to slow down climate change, reduce resource consumption, intensify resource discovery, find ways to extract resource stocks that have been too costly to get, and create resource substitutes.

Scholars have only begun to model effects of resources on interstate war. In future research, our modeling approach could be expanded to examine, for example, war escalation, duration, location, or termination. Studies may calibrate relative contributions to the net effect (this would require innovative data collection) or include more resources.¹² Our approach could also be applied to questions within other levels of analysis (e.g. country-pair, system). Regardless of the direction future research takes, continued growth of this nascent field demands high priority.

Footnotes

Appendix 1: Countries in the sample

Algeria	Angola	Argentina
Armenia	Australia	Austria
Azerbaijan	Bahrain	Bangladesh
Belarus	Benin	Bhutan
Bolivia	Brazil	Bulgaria
Burkina Faso	Burundi	Cameroon
Canada	Central African Republic	Chad
Chile	China	Colombia
Congo	Costa Rica	Cyprus
Denmark	Dominican Republic	Ecuador
Egypt	El Salvador	Estonia
Estonia	Ethiopia	Fiji
Finland	France	Gabon
Gambia	Georgia	Ghana
Greece	Guatemala	Guinea
Guinea-Bissau	Guyana	Haiti
Honduras	Hungary	India
Indonesia	Iran	Ireland
Israel	Italy	Ivory Coast
Jamaica	Japan	Jordan
Kazakhstan	Kenya	Kuwait
Kyrgyzstan	Latvia	Libya
Lithuania	Madagascar	Malawi
Malaysia	Mali	Mauritania
Mauritius	Mexico	Moldova
Mongolia	Morocco	Mozambique
Nepal	Netherlands	New Zealand
Nicaragua	Niger	Nigeria
Oman	Pakistan	Panama
Papua New Guinea	Paraguay	Peru
Philippines	Poland	Portugal
Romania	Russia	Rwanda
Saudi Arabia	Senegal	Sierra Leone
South Africa	South Korea	Spain
Sri Lanka	Sudan	Sweden
Switzerland	Syria	Tanzania
Thailand	Togo	Trinidad Tobago
Tunisia	Turkey	Turkmenistan
Uganda	Ukraine	United Arab Emirates
United Kingdom	United States of America	Uruguay
Venezuela	Zambia	Zimbabwe

Appendix 2: Summary statistics

Variable	Mean value	Standard deviation	Minimum value	Maximum value
Arable Land	0.34	0.36	0.00	3.50
Agriculture	106.12	79.79	9.50	1457.70
Cropland	2.66	3.66	0.00	17.61
Timber	0.50	1.49	0.00	18.52
Minerals	1.49	5.25	0.00	82.59
Fuels	16.13	35.24	0.00	723.00
Freshwater	28467.47	59550.23	8.00	748217.00
Precipitation	1113.25	747.33	51.20	3141.70
Population	9.04	1.47	5.40	14.03
Population Growth	2.07	1.71	–44.41	21.76
Disasters	13.22	233.13	0.00	7630.01
Trade Openness	60.45	34.14	1.53	282.40
GDPpc	0.00	1.05	–2.84	3.38
Borders	3.49	2.34	0.00	20.00
Political Instability	0.14	0.35	0.00	1.00
National Capability	0.01	0.02	0.00	0.22
Democracy	–0.61	7.53	–10.00	10.00

Note: GDPpc is demeaned log(GDP per capita).

Appendix 3: Correlation matrix

Variable	No.	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18
Arable Land	1	1.00
Agriculture	2	–0.10	1.00
Cropland	3	–0.25	0.06	1.00
Freshwater	4	0.11	0.04	–0.20	1.00
Precipitation	5	–0.27	0.08	0.30	0.33	1.00
Population	6	0.05	–0.24	–0.03	–0.24	–0.07	1.00
Population Growth	7	–0.10	0.01	–0.05	0.05	0.02	–0.10	1.00
Timber	8	–0.08	–0.01	0.14	–0.09	0.13	0.06	0.12	1.00
Minerals	9	0.06	–0.03	–0.07	0.35	0.14	–0.13	0.03	–0.04	1.00
Fuel	10	–0.05	–0.04	–0.10	0.13	–0.14	–0.06	0.23	–0.12	–0.12	1.00
Disasters	11	–0.02	0.01	0.01	0.00	0.08	0.01	0.02	0.02	–0.01	0.00	1.00
Trade Openness	12	–0.19	0.23	0.15	0.12	0.00	–0.63	0.02	–0.07	0.14	0.15	–0.02	1.00
GDPpc	13	0.17	0.02	–0.05	–0.02	–0.18	–0.03	–0.31	–0.39	–0.16	0.08	–0.03	0.13	1.00
GDPpc²	14	0.27	–0.03	–0.23	–0.13	–0.29	0.01	–0.17	–0.03	–0.20	–0.07	–0.05	0.00	0.59	1.00
Borders	15	–0.03	–0.11	–0.29	–0.04	–0.30	0.34	0.11	0.02	–0.05	0.09	–0.01	–0.24	–0.20	–0.14	1.00
Political Instability	16	-0.03	-0.06	0.01	-0.02	0.03	0.05	0.06	0.15	0.02	-0.07	0.04	-0.16	-0.26	-0.19	0.07	1.00
National Capability	17	0.06	-0.10	-0.13	-0.10	-0.10	0.61	-0.12	-0.03	-0.08	-0.08	-0.02	-0.33	0.20	0.26	0.17	-0.09	1.00
Democracy	18	0.08	0.02	0.13	0.00	0.19	0.07	-0.41	-0.12	-0.03	-0.33	-0.01	-0.06	0.47	0.32	-0.24	-0.05	0.14	1.00

Notes

References

Andrews-Speed

Bleischwitz

Boersma

. (2012) The Global Resource Nexus: The Struggles for Land, Energy, Food, Water, and Minerals. Washington, DC: Transatlantic Academy.

Australian Government, Department of Defence (2009) Defending Australia in the Asia Pacific Century: Force 2030. Canberra: Department of Defence.

Barro

Sala-i-Martin

(2004) Economic Growth. Cambridge, MA: The MIT Press.

Beck

Katz

Tucker

(1998) Taking time seriously: Time-series–cross-section analysis with a binary dependent variable. American Journal of Political Science 42(4): 1260–1288.

Bennett

Stam

(2004) The Behavioral Origins of War. Ann Arbor, MI: University of Michigan Press.

Boserup

(1981) Population and Technological Change. Chicago, IL: University of Chicago Press.

Braithwaite

(2006) The geographic spread of militarized disputes. Journal of Peace Research 43 (5): 507–522.

Cashman

(2014) What Causes War? Lanham, MD: Rowman and Littlefield.

Center for Naval Analyses (CNA) (2014) National Security and the Accelerating Risks of Climate Change. Alexandria, VA: CNA Military Advisory Board.

10.

Choucri

North

(1975) Nations in Conflict. San Francisco, CA: Freeman.

11.

Choucri

North

(1989) Lateral pressure in international relations: Concept and theory. In: Midlarsky

(ed.) Handbook of War Studies. Boston, MA: Unwin Hyman, 289–326.

12.

Choucri

North

Yamakage

(1992) The Challenge of Japan Before World War II and After. London: Routledge.

13.

Colgan

(2013a) Fueling the fire: Pathways from oil to war. International Security 38(2): 147–180.

14.

Colgan

(2013b) Petro-Aggression: When Oil Causes War. Cambridge: Cambridge University Press.

15.

Conca

(2001) Environmental cooperation and international peace. In: Gleditsch

Diehl

(eds) Environmental Conflict. Boulder, CO: Westview, 225–250.

16.

Correlates of War (COW) (2010) Correlates of war. Urbana, IL: University of Illinois. Available at: http://www.correlatesofwar.org.

17.

Couttenier

Soubeyran

(2013) An overview of the roots of civil wars: Natural factors and economic conditions. G-MonD Policy Paper 2. Paris: Paris School of Economics.

18.

De Soysa

Gartzke

Lie

(2011) Oil, blood and strategy: How petroleum influences interstate conflict. Trondheim: Norwegian University of Science and Technology.

19.

Downes

Sechser

(2012) The illusion of democratic credibility. International Organization 66(3): 457–489.

20.

Durham

(1979) Scarcity and Survival in Central America: Ecological Origins of the Soccer War. Stanford, CA: Stanford University Press.

21.

Fearon

Laitin

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

22.

Food and Agriculture Organization (FAO) (2010) Aquastat. Rome: FAO.

23.

Frankel

(2012) The natural resource curse: A survey, of diagnoses and some prescriptions. Working Paper RWP12-014. Cambridge, MA: Harvard University.

24.

Furlong

Gleditsch

Hegre

(2006) Geographic opportunity and neomalthusian willingness: Boundaries, shared rivers, and conflict. International Interactions 32(1): 79–108.

25.

Garrett

Piccinni

(2012) Natural resources and conflict: A new security challenge for the European Union. A report for SIPRI by Resource Consulting Services. Available at: http://www.sipri.org/research/conflict/pko/PKO_archive/resources/RCSfull1206.pdf.

26.

Gleditsch

Furlong

Hegre

. (2006) Conflicts over shared rivers: Resource scarcity or fuzzy boundaries? Political Geography 25(2): 361–382.

27.

Gleick

(1993) Water and conflict: Fresh water resources and international security. International Security 18(1): 79–122.

28.

Global Environmental Outlook (GEO) (2010) Global environmental outlook. Nairobi: United Nations Environment Programme (UNEP). Available at: http://www.unep.org/geo.

29.

Godwin

(1793) An Enquiry Concerning Political Justice and its Influence on General Virtue and Happiness. Volume 2. London: Robinson.

30.

Heske

(1987) Karl Haushofer: His role in German geopolitics and in Nazi politics. Political Geography Quarterly 6(2): 135–144.

31.

Klare

(2002) Resource Wars: The New Landscape of Global Conflict. New York: Owl.

32.

Klare

(2005) Blood and Oil: The Dangers and Consequences of America’s Growing Dependency on Imported Petroleum. New York: Holt.

33.

Klare

(2012) The Race for What’s Left. New York: Metropolitan.

34.

Koubi

Spilker

Böhmelt

. (2014) Do natural resources matter for interstate and intrastate armed conflict? Journal of Peace Research 51(2): 227–243.

35.

Le Billon

(2012) Wars of Plunder: Conflicts, Profits and the Politics of Resources. London: Hurst.

36.

Lebow

(2010) Why Nations Fight: Past and Future Motives for War. Cambridge: Cambridge University Press.

37.

Lenin

(1917) Imperialism: The Highest Stage of Capitalism. London: Pluto.

38.

Levy

Thompson

(2011) Causes of War. Malden, MA: Wiley-Blackwell.

39.

Lomborg

(2001) The Skeptical Environmentalist. New York: Cambridge University Press.

40.

Mabey

Gulledge

Finel

. (2011) Degrees of Risk: Defining a Risk Management Framework for Climate Security. Brussels: Third Generation Environmentalism Ltd (E3G).

41.

Maddow

(2014) Why We Did It? MSNBC video. Available at: http://www.msnbc.com/maddow-why-we-did-it.

42.

Malthus

(1798) An Essay on the Principle of Population. Mineola, NY: Dover.

43.

Mearsheimer

(2001) The Tragedy of Great Power Politics. New York: Norton.

44.

Morgenthau

Thompson

Clinton

(2005) Politics Among Nations. New York: McGraw-Hill.

45.

Myers

(1993) Ultimate Security: The Environmental Basis of Political Stability. New York: Norton.

46.

National Intelligence Council (NIC) (2012) Global Trends 2030: Alternative Worlds. Washington, DC: Office of the Director of National Intelligence.

47.

North

(1995) Life on a Modern Planet: A Manifesto for Progress. Manchester: Manchester University Press.

48.

Payne

(1995) Freedom and the environment. Journal of Democracy 6(3): 41–55.

49.

Polity (2010) Polity IV Project. Fairfax, VA: George Mason University. Available at: http://www.systemicpeace.org/polity/polity4.htm.

50.

Richardson

(1960) Statistics of Deadly Quarrels. Pittsburgh, PA: Boxwood.

51.

Russett

(1981) Security and the resource scramble: Will 1984 be like 1914? International Affairs 58(1): 42–58.

52.

Schneider

(2010) Economics and conflict. In: Denmark

(ed.) International Studies Encyclopedia. Malden, MA: Wiley-Blackwell, 1284–1300.

53.

Simon

(1989) Lebensraum: Paradoxically, population growth may eventually end wars. Journal of Conflict Resolution 33(1): 164–180.

54.

Simon

(1998) The Ultimate Resource 2. Princeton, NJ: Princeton University Press.

55.

Singer

(1969) The Level-of-Analysis Problem in International Relations. In: Rosenau

(ed.) International Politics and Foreign Policy. New York: Free Press, 20–29.

56.

Soltanov

(2011) Beyond the Resource Curse. In: Shaffer

Ziyadov

(eds) Natural Resources, Domestic Instability, and International Conflicts. Philadelphia, PA: University of Pennsylvania Press, 313–329.

57.

Stalley

(2003) Environmental scarcity and international conflict. Conflict Management and Peace Science 20(2): 33–58.

58.

Strüver

(2010) Too many resources or too few? What drives international conflicts? GIGA Working Paper 147. Hamburg: German Institute of Global and Area Studies.

59.

Swatuk

(2007) Seeing the forest for the trees: Tropical forests, the state, and violent conflict in Africa. In: De Jong

Donovan

Abe

(eds) Extreme Conflict and Tropical Forest. Dordrecht: Springer, 93–116.

60.

Tir

Stinnett

(2012) Weathering climate change: Can institutions mitigate international water conflict? Journal of Peace Research 49(1): 211–225.

61.

Van der Poleg

(2011) Natural resource: Curse or blessing? Journal of Economic Literature 49(2): 366–420.

62.

Van Evera

(2001) Causes of War: Power and the Root of Conflict. Ithaca, NY: Cornel University Press.

63.

UK Ministry of Defence (2010) Global Strategic Trends – Out to 2040. London: Ministry of Defence.

64.

Waltz

(1979) Theory of International Politics. Long Grove, IL: Waveland.

65.

Wasson

(2007) Natural resources and interstate conflict: Beg, borrow, or steal? Paper presented at the International Studies Association Conference, 28 February–3 March, Chicago.

66.

Westing

(1986) Environmental factors in strategic policy and action: An overview. In: Westing

(ed.) Global Resources and International Conflict. New York: Oxford University Press, 3–20.

67.

World Bank (2005) World Development Indicators. Washington, DC: World Bank.

68.

Wright

(1965) A Study of War. Chicago, IL: University of Chicago Press.