Trends and triggers redux: Climate change, rainfall, and interstate conflict

Political Geography xxx (2014) 1e13 Contents lists available at ScienceDirect Political Geography journal homepage: www.elsevier.com/locate/polgeo ...

Download PDF

1005KB Sizes 0 Downloads 42 Views

Report

PDF Reader
Full Text

Political Geography xxx (2014) 1e13

Contents lists available at ScienceDirect

Political Geography journal homepage: www.elsevier.com/locate/polgeo

Trends and triggers redux: Climate change, rainfall, and interstate conﬂict Colleen Devlin a, Cullen S. Hendrix b, *, 1 a b

Desert Development Center, American University in Cairo, New Cairo AUC Avenue, P.O. Box 74, New Cairo 11835, Egypt Korbel School of International Studies, University of Denver, 2201 S. Gaylord St., Denver, CO 80210, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Available online xxx

Given freshwater is crucial to sustaining life and forecasted to decline in relative abundance under most climate change scenarios, there is concern changing precipitation patterns will be a cause of future interstate conﬂict. In theorizing the impact of climate change for interstate conﬂict, we distinguish between trends (long-term means) that may affect the baseline probability of conﬂict, and triggers (shortterm deviations) that may affect the probability of conﬂict in the short run. We jointly model the effects of mean precipitation scarcity and variability (trends) and year-to-year changes in precipitation (triggers) on militarized interstate disputes between states. We ﬁnd higher long-run variability in precipitation and lower mean levels of precipitation in dyads are associated with the outbreak of militarized interstate disputes (MIDs). Contra neo-Malthusian expectations, however, we ﬁnd joint precipitation scarcity e deﬁned as both countries experiencing below mean rainfall in the same year e has a conﬂict-dampening effect. These ﬁndings push the literature in a direction that more closely aligns our modeling of human impacts with our understanding of the physical impacts of climate change. © 2014 Elsevier Ltd. All rights reserved.

Keywords: Climate change Conﬂict River basins Precipitation Bargaining Water War

Introduction Water is necessary for sustaining human, animal and plant life, providing a variety of ecosystem services, and it is an increasingly important source of electrical power. Despite its centrality to human existence, nearly one billion people lack reliable access to clean drinking water. While economic development promises to reduce the burden of disease associated with contaminated drinking water and lack of access to basic sanitation, the same economic development and population growth will increase global demand. A 2009 report by the Water Resources Group projects that by 2030, annual global freshwater needs will reach 6.9 trillion cubic

* Corresponding author. E-mail addresses: [email protected] (C. Devlin), cullen.hendrix@du. edu (C.S. Hendrix). 1 This paper was begun while Devlin was an undergraduate student at the College of William & Mary. We thank Jaroslav Tir and Marit Brochmann for providing replication data and Stu Hamilton at the Center for Geospatial Analysis for help with data processing. Idean Salehyan, Erika Weinthal, Scott Wolford, three anonymous reviewers and the editors provided helpful comments and suggestions. Brittany Franck provided able research assistance. This material is based upon work supported by, or in part by, the US Army Research Laboratory and the US Army Research Ofﬁce under contract/grant number W911NF-09-1-0077.

meters, which is 64 percent more than the existing accessible, reliable, and sustainable supply (Water Resources Group, 2009). This forecast, while alarming, likely understates the magnitude of the challenge, as it does not account for the impacts of global climate change on hydrological systems. While the Intergovernmental Panel on Climate Change (IPCC) forecasts an increase in total precipitation at the global level, regional patterns will vary signiﬁcantly. Rainfall will likely decline (>20 percent) across North Africa and the Middle East, central Mexico and Central America and the Caribbean, Southern Africa, the eastern Amazon basin, and Western Australia, leading to declines in water availability of 10e30 percent (IPCC, 2007, 2013). The IPCC also forecasts a 90 percent likelihood that variability in rainfall will increase, leading to more numerous dry spells, but also more extreme precipitation events and ﬂooding. Thus, climate change may increase aggregate surface water availability, but it will do so at the cost of increasing stress in some already vulnerable regions and making access to surface water less reliable due to increasingly variable precipitation patterns. Given water is crucial to sustain human life and declining in relative abundance, there is concern that water will be both a cause of future conﬂict and a source of bargaining power for states that control access to surface and groundwater supplies (Gleick, 1993; Klare, 2001). Climate change has crept on to the security agenda

http://dx.doi.org/10.1016/j.polgeo.2014.07.001 0962-6298/© 2014 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

2

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13

for a variety of reasons, but perhaps the most obvious is the potential relationship between climate change, water resources and political conﬂict, at the communal, intrastate, interstate, and civilizational levels (see special issues in Political Geography (2007) and Journal of Peace Research (2012); Fjelde & von Uexkull, 2012; Hsiang, Burke, & Miguel, 2013; Hsiang, Meng, & Cane, 2011). Policy discussions of climate change impacts on water security have tended to focus on declining stocks of freshwater resources e absolute scarcity e as the primary driver of conﬂict. The United States National Intelligence Council (2012) concludes that while water-related interstate conﬂict is unlikely over the next decade, serious water shortages will, over the medium term, destabilize already tense bilateral relationships. Speciﬁcally, they assert, “as water shortages become more acute beyond the next 10 years, water in shared basins will increasingly be used as leverage; the use of water as a weapon or to further terrorist objectives also will become more likely beyond 10 years” (NIC 2012, 3). This discourse is rooted in a neo-Malthusian characterization of the relationship between carrying capacity and violence: declining or degraded stocks of natural resources e for which no substitutes are available e spark distributional conﬂicts (Ehrlich, 1969; Homer-Dixon, 1999). The prospect of interstate conﬂict over water resources is most clear in shared river basins, in which surface freshwater is shared between two or more states. In these cases river water constitutes a common pool resource whose consumption is rival: one country's increasing consumption necessarily leaves the other country(ies) with less (Hardin, 1968). Demand for water is relatively supply-inelastic and substitutes are, for most applications, nonexistent. In preceding work (Tir & Stinnett, 2012) and in policy discussions, the link between climate change and conﬂict operates through changes in declining mean levels of freshwater abundance from other sources: “In response to reduced precipitation, states will likely increase their reliance on other water sources, including transboundary rivers” (Tir & Stinnett, 2012, 213). Competition between “water ‘haves’ and ‘have nots’” exists at multiple geographical scales (Agnew, 2011, 466). Whether due to technological differences, the physical positions of actors relative to renewable freshwater resource, or economic resources, differential access to water resources has been a source of conﬂict in contexts as varied as the Colorado River basin, the US Great Lakes region, the Middle East, and East and Southeast Asia. Climate change has the potential to exacerbate existing water resource competition. Climate change will likely affect levels of precipitation, with some countries growing more arid and others wetter. However, climate change will also result in increasing climatic variability: more frequent dry spells and ﬂooding, more erratic rainfall patterns, and larger year-to-year variability in precipitation levels (IPCC, 2013; Tompkins, 2005). That is, climate change is likely to affect both the means and variability of precipitation in a given area. Conjectures about whether these changes will have security implications, however, should be rooted in an understanding of the role (if any) that precipitation levels, variability, and short-term scarcity have played in past conﬂicts. We extend previous work on water and conﬂict in two ways. First, following Hendrix and Glaser (2007), we distinguish between trends, longer-term mean states that may affect the baseline probability of conﬂict, and triggers, or acute scarcity or abundance, that may affect the probability of conﬂict in the short run, and apply this distinction to the study of interstate conﬂict. Second, we move past neo-Malthusian explanations for resource-based conﬂict and instead look at how climatic factors may affect bargaining between states more generally. This leads us to hypothesize differential effects for precipitation scarcity at differing time scales: while over

the long term, more scarce rainfall may be associated with a greater probability of conﬂict, over the short term, acute scarcity should have a pacifying effect. The same model leads us to hypothesize conﬂict will be more likely in dyads characterized by higher variation in rainfall. We jointly model the effects of trends (mean precipitation level and precipitation variability), and triggers (year-to-year changes in precipitation), on militarized interstate disputes, basing our models on reanalyses of Tir and Stinnett (2012) and Brochmann and Gleditsch (2012). Our main ﬁndings are two. First, we ﬁnd precipitation variability is more strongly associated with the outbreak of conﬂict, operationalized as militarized interstate disputes (MIDs), than mean levels of rainfall. Our ﬁndings indicate higher variability of rainfall will have more substantial effects on peace and stability between states than absolute scarcity per se, assuming past trends (1950e2002) continue into the future. Second, we ﬁnd states which are both experiencing lower than normal precipitation are less, not more, likely to enter into disputes. These ﬁndings suggest the policy emphasis on increasing water scarcity should be accompanied by a focus on variability. In particular, our modeling indicates variability plays a role apart from an increase in extreme, acute climatic events such as ﬂoods and droughts e short-term triggers that may affect the probability of conﬂict in a given year. The remainder of the paper proceeds as follows. The following section addresses recent scholarship on conﬂict and cooperation among riparian states and the ﬁndings regarding the effect of absolute water scarcity on conﬂict, as well as the growing literature on precipitation shocks and conﬂict. Then, we present a theory of bargaining and conﬂict over shared water resources under uncertainty about water availability before presenting our hypotheses. The next section describes our data, methodology, and presents our results. The ﬁnal section concludes by arguing for the importance of aligning our modeling of the human impacts of climate change with our understanding of its physical impacts. Water and interstate conﬂict Despite seemingly dire warnings that shared water resources will lead to conﬂict in the future (Gleick, 1993; Klare, 2001), the general ﬁnding is cooperation over shared freshwater resources e cultural and scientiﬁc agreements, treaty signature, comanagement e is more common than conﬂict (Yoffe, Wolf, & Giordano, 2003). Still, several studies ﬁnd countries sharing river basins are at an increased risk of conﬂict relative to other dyads, or pairs of states (Gleditsch, Furlong, Hegre, Lacina, & Owen, 2006; Toset, Gleditsch, & Hegre, 2000). The most recent study on the subject (Brochmann & Gleditsch, 2012) ﬁnds river sharing is almost omnipresent, and does not increase the propensity for conﬂict between countries beyond the increase in conﬂict propensity due to proximity, i.e., having a shared border. However, Brochmann and Gleditsch ﬁnd upstream/downstream relationships e relationships where countries have asymmetric ability to affect the ﬂow of water to other states due to their upstream position e to be more conﬂictual. Among countries sharing rivers, however, there is general consensus that water scarcity increases the likelihood of conﬂict (Hensel, Mitchell, & Sowers, 2006; Tir & Stinnett, 2012; Toset et al., 2000), though Hensel, Mitchell and Sowers ﬁnd water scarcity increases both conﬂict and cooperation, perhaps suggesting a more general relationship between scarcity and interaction over water resources, consistent with Yoffe et al. (2003) and Stinnett and Tir (2009). Tir and Stinnett (2012) ﬁnd water scarcity e operationalized as log-transformed renewable freshwater resources per capita in the water-poorer member of the dyad e to be a robust covariate of conﬂict: as water becomes more scarce, the probability of conﬂict increases. They defend this

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13

choice of operationalization on grounds that their intention is to inform policy debates over climate change adaptation, and climate projections are most convergent (i.e., in agreement with one another) in forecasting overall levels of water availability at regional scales. The question of whether climate models are convergent with respect to precipitation variability at regional scales is separate from whether precipitation variability itself has contributed to conﬂict in the past. There may be other relevant attributes of water resources e other than relatively stable long-term trends in freshwater abundance (see Appendix 2) e that may affect conﬂict propensity. By focusing on scarcity, we are disregarding potentially valuable information: for any mean level of freshwater availability, there may be important differences e for water policy at the national and international level, and for water withdrawal from common pool resources e between dyads with low year-to-year rainfall variability and those with high variability. A separate strand of literature investigates the impacts of both short-term climatic triggers and longer-term climatic trends on intrastate, interpersonal, and social conﬂict. In recent years, more research has focused on short-term triggers, such as drought, ﬂooding, and changes in precipitation patterns that may more precisely explain the timing of conﬂict. Water abundance/scarcity is a relatively persistent, i.e., time invariant, feature of an ecosystem, while actual conﬂict is signiﬁcantly more episodic: even the most conﬂictual dyad (IndiaePakistan) only witnessed MID onsets in roughly half of the yearly observations. Thus, focusing only on other, more time-variant attributes of water resources e or the variability of the water resources as a more durable feature of dyadic relationships e may give us more purchase on identifying which dyads will be most conﬂict-prone and may better explain the timing of conﬂict outbreak. Regarding the effects of acute water stress, ﬁndings in the literature are not convergent. Building on the work of Miguel, Satyanath, and Sergenti (2004) and Hendrix and Glaser (2007), which found rainfall scarcity to increase the likelihood of civil conﬂict, contributions to a recent Journal of Peace Research special issue and several recent studies ﬁnd rainfall and natural disasters have little or no effect on conﬂict (Bergholt & Lujala, 2012; Koubi, Bernauer, Kalbhenn, & Spilker, 2012; Sletteback, 2012; Theisen, Holtermann, & Buhaug, 2011/2012); that rainfall abundance increases conﬂict (Hendrix & Salehyan, 2012; Witsenberg & Adano, 2009 (with respect to intrastate conﬂict); Theisen, 2012); that rainfall scarcity increases conﬂict (Fjelde & von Uexkull, 2012); or that there is a curvilinear effect (Hendrix & Salehyan, 2012 (with respect to social conﬂict); Raleigh & Kniveton, 2012). Burke, Miguel, Satyanath, Dykema, and Lobell (2009) ﬁnd evidence of a positive relationship between temperature and conﬂict in Sub-Saharan Africa, with hotter years experiencing increased incidence of armed conﬂict, but this ﬁnding has been disputed by Buhaug (2010) on the grounds that the original ﬁndings were an artifact of their use of ﬁxed-effects models, which sap well known explanatory variables in conﬂict studies (poverty, weak state institutions) of their explanatory power. These disparate ﬁndings are in part attributable to differences in operationalizing conﬂict: armed vs. nonviolent, small-scale vs. large-scale. A recently published meta-analysis (Hsiang et al. 2013) ﬁnds deviations from normal (i.e., mean) precipitation and temperatures are robustly linked to violent conﬂict at a variety of scales, both spatial (ranging from cities to countries to continent-spanning empires) and temporal (hourly to millennia). The study has been criticized on the grounds of sample selection bias in selecting studies to be included (Buhaug et al., in press) and the conﬂation of weather events with climate (Solow, 2013). Setting these criticisms aside, the inclusion of studies at such disparate spatial and

3

temporal levels of aggregation e and the fact that the directions of the effects of climate variables in their study are both positive, negative, and curvilinear, depending on the study at hand e strains the ability of researchers to develop a coherent theoretical argument for how and why climatic factors affect conﬂict. Hsiang et al. close by noting, “It seems likely that climatic changes inﬂuence conﬂict through multiple pathways that may differ between contexts” (12). This statement is similar to the challenge laid out by Salehyan's contribution to this issue, which seeks to focus our attention on the potentially differential impacts of climatic factors at different spatial and temporal scales. Theory: trends and triggers Climate refers to the longer-run, stable patterns of variation in precipitation, temperature, and other meteorological variables that persist in a given region. With respect to precipitation, a particular climate will be characterized by a relatively stable a) mean level of annual precipitation, and b) variability in annual precipitation levels around that mean. These are climatic trend variables. In addition to these longer-term climatic means, we are interested in the effects of precipitation anomalies e lower than normal rainfall in a given year e that might trigger conﬂict. Fig. 1 demonstrates the relationship between mean water availability, current water availability, and water variability for Sri Lanka, Nigeria, and Somalia. The solid lines represent annual rainfall levels (from Dell, Jones, & Olken, 2012) at time t for each of the three countries, while the dashed lines represent country mean values. Finally, the bar chart to the right shows the coefﬁcient of variation e the ratio of the standard deviation to the mean e for the three countries. Sri Lanka has the highest mean value for precipitation and also the largest absolute deviations. Somalia has comparatively small year-to-year deviations in rainfall from the panel mean, but because the panel mean is so much lower, these smaller absolute deviations represent larger proportional changes from the mean. The following section develops hypotheses relating precipitation means, variability, and acute abundance/scarcity to the propensity for conﬂict between states. Trends While competing claims to territory, maritime boundaries, and water rights are virtually omnipresent e Brochmann and Gleditsch (2012) ﬁnd virtually all contiguous pairs of states share at least one river e actual militarized conﬂict over them is not. This may be due in part to the variable salience states attach to shared water resources. The issue salience school of world politics (Diehl, 1992; Hensel, Mitchell, Sowers, & Thyne, 2008; Randle, 1987) argues the value participants attach to a particular issue determines whether they will be willing to commit scarce resources and bear costs associated with conﬂict in order to achieve their goals. States marshal their resources in pursuit and defense of their interests, but not all interests are equal. States will be more willing to bear the costs of conﬂict over highly salient issues. While water is a highly salient issue for all states, the salience of shared water resources varies according to the degree of dependence states have on those water resources. For instance, the Nile River is salient for both Uganda and Egypt. However, the salience of the Nile as a source of freshwater is much higher for Egypt, an arid country with extremely low precipitation (~0.4 dm/year) and for which 96.9% of its renewable freshwater emanates outside its borders, than for Uganda, a tropical country with much higher average levels of precipitation (12.5 dm/year) and which is mostly reliant on internally sourced renewable freshwater (FAO, 2014).

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

4

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13

Fig. 1. Precipitation Means, Variability, and Year-to-Year Fluctuations. Base data are from Dell et al. (2012).

Pairs of states with higher mean levels of precipitation should attach less salience to shared water resources, as a larger proportion of their water needs will be met by rainfall. In contrast, pairs of countries with lower mean levels of precipitation are more dependent on groundwater and/or freshwater from external sources. Shared water resources should thus be more salient to comparatively rainfall-scarce countries than those whose freshwater needs are met adequately by domestic, renewable surface sources. This suggests dyads with lower overall levels of mean rainfall will be more likely to come into conﬂict over water resources. H1. Conﬂict will be more likely in dyads characterized by lower overall levels of rainfall. If the mechanism linking lower precipitation to interstate conﬂict were to operate through incentives to appropriate scarce resources, there must be a shared resource to appropriate. The corollary to this expectation, then, is that the effect of mean rainfall scarcity should be conditional on dyads sharing a river basin. H1a. Conﬂict will be more likely in dyads characterized by lower overall levels of rainfall in which the dyad members share a river basin. This issue-salience model is premised in the same logic as the neo-Malthusian claims about resource competition. However, conﬂict cannot create new resources (Maxwell & Reuveny, 2000), and conﬂict itself entails signiﬁcant costs for belligerent parties. Violent conﬂict generally only occurs when negotiations have failed to produce an outcome that both parties prefer to armed conﬂict; conﬂict can thus be understood as a bargaining failure (Fearon, 1995; Powell, 2002). While neo-Malthusian concerns may establish motive, they do not by themselves explain increased propensity for conﬂict. Rather, we focus on the potential for rainfall to affect the bargaining context. Variability in precipitation may contribute to tensions between states by making their withdrawal needs from shared water resources less predictable, as year-to-year changes in rainfall are proportionately larger. Under conditions of uncertainty, both

explicit and implicit agreements governing the use of shared resources are more difﬁcult to achieve and maintain (Dietz, Ostrom, & Stern, 2003; Ostrom, 1999). Uncertainty can be due to limited scientiﬁc understanding of the biophysical processes governing a common-pool resource, but can also be due to natural variability of weather. Because weather can affect both access to and withdrawal needs from common-pool resources, highly variable weather conditions introduce an additional form of uncertainty into the process of creating contracts governing shared use (Grafton, Squires, & Fox, 2000). Bargaining may be more likely to break down between riparian states characterized by high year-to-year variability in, and therefore uncertainty about, rainfall. One source of bargaining failure is an inability to credibly commit to not using force in the future (Fearon, 1995). If actors are perfectly informed about their maximum withdrawal needs, then any bargain which guarantees them this share of the water resource will be preferable to costly ﬁghting. The bargaining will be self-enforcing in the sense that it is in the interest of each actor to abide by it. Variability in rainfall makes each actor's reservation payoff e the portion of beneﬁt from the common-pool resource below which the actor prefers to mobilize and ﬁght e variable from year to year. Thus, each actor will be uncertain about whether their negotiated portion of the shared resource will satisfy their withdrawal needs. As a result, they cannot credibly commit to abiding by the terms of their agreements in the future, rendering present bargains noncredible and increasing the likelihood of conﬂict. Climate change may exacerbate this problem to the extent that changes in “resource environments” e i.e., the ecology of the resources themselves e occur faster than formal and informal agreements governing their use can be modiﬁed or adapted (Giordano, Giordano, & Wolf, 2005). Thus, we expect conﬂict will be more likely in dyads characterized by higher year-to-year variability in rainfall. H2. Conﬂict will be more likely in dyads characterized by higher variability in rainfall.

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13

Triggers Acute precipitation scarcity e periods when rainfall is low relative to historic means of water availability e may also affect conﬂict. Demand for water is relatively supply-inelastic. During periods of lower-than-normal rainfall, states have no alternative but to increase their demands on shared water resources (Tir & Stinnett, 2012). The probability of conﬂict outbreak should be highest when co-riparian states simultaneously increase their demands on shared resources due to abnormally low rainfall in both states. If the Neo-Malthusian hypothesis were correct, we would expect conﬂict to be most likely in dyads when both countries are experiencing lower than normal rainfall. Moreover, we would expect the probability of conﬂict would be greatest between dyads where both countries are experiencing lower than normal rainfall, share a river, and border one another. Since territorial control of the water resource is crucial to guaranteeing access, geographic contiguity is important. Most countries lack the ability to project military force over large distances via naval and air power. Most of the prominent candidates for “water wars” identiﬁed in the literature eIsrael-Jordan over the Jordan, PakistaneIndia over the Indus, BangladesheIndia over the Ganges, Egypt-Ethiopia-Sudan over the Nile, the Central Asian former Soviet Republics over the Aral Sea, etc. e are comprised of geographically contiguous countries (Gleick, 1993; Starr, 1991). Thus, these conditions (scarce precipitation and water stress, sharing a water resource, geographically contiguous) most closely approximate the concept of a water war: violent, appropriative conﬂict over shared water resources. H3. : Conﬂict will be more likely in dyads when both members are experiencing lower than normal rainfall. H3a. : Conﬂict will be more likely in dyads when both members are experiencing lower than normal rainfall and those countries share a border and/or a river basin. Instead of focusing on scarce precipitation as a motive for conﬂict, lower than normal rainfall can be viewed alternately as a factor affecting the perceived costs of ﬁghting. In the standard bargaining model of conﬂict, ﬁghting is ex-post inefﬁcient: because mobilization and ﬁghting are costly, there are always some outcomes that both parties prefer to conﬂict. As the perceived costs of conﬂict increase, the range of outcomes both parties prefer to war increases. These costs can be real, in terms of “blood and treasure”, but also take the form of opportunity costs: the economic and social losses stemming from diversion of productive resources into ﬁghting. In bargaining models of conﬂict, opportunity costs are typically characterized as a “guns and butter” tradeoff, in which states decide between allocating resources to satisfying internal ends and allocating resources to military purposes (Powell, 1993). The more a state allocates to addressing internal demands, the fewer resources are available to invest in the military and vice versa. The opportunity costs associated with ﬁghting should be increasing with lower than normal rainfall for three reasons. First, lower-than-normal rainfall leads to slowed or negative economic growth, with particularly severe impacts on the agricultural sector. The most obvious mechanism is through the reduced availability of water, a key agricultural input. However, there is some evidence for knock-on effects that operate through higher costs for raw materials and diminished electricity generation in economies that are dependent on hydropower. Battisti and Naylor (2009) and Loayza, Olaberría, Rigolini, and Christiaensen (2012) ﬁnd droughts are associated with overall economic contraction in developing countries and contraction in the agricultural economy in developed

5

countries. This relationship is evident even in resource-rich countries, such as Angola and Zambia, which have revised downward their 2014 GDP growth estimates in the face of a prolonged drought (Reuters, 2013). Second, demand for state resources, in the form of drought response activities such as emergency feeding programs and the provision of crop insurance, increases, while state revenues, especially in poorer countries, contract (Chantarat, Barrett, Mude, & Turvey, 2007; Noy & Aekkanush, 2011). Under these conditions, states have strong incentives to shift a greater proportion of their resources toward addressing domestic issues at the same time that the resource base is dwindling. One of the explanations for the lack of an effect of drought on aggregate economic growth in developed countries is due to offsetting domestic expenditures on emergency relief (Loayza, Olaberría, Rigolini, & Christiaensen, 2006). The US government budgeted $18.8 billion dollars to agricultural relief programs alone in 2012, part of a total disaster relief appropriation of $34.2 billion in FY 2012 (Weiss & Weidman, 2013). Third, droughts and local food shortages also may make countries more dependent on international markets and conﬂicts with neighboring countries can interrupt trade, placing even greater strain on the domestic economy (Anderton & Carter, 2001; Glick & Taylor, 2010). Opportunity costs to ﬁghting thus increase with acute water scarcity. As opportunity costs are increasing for both countries, the range of bargained outcomes preferred to conﬂict increases. The neo-Malthusian hypothesis suggests conﬂict should be more likely to occur between countries jointly experiencing acute precipitation scarcity. The opportunity cost hypothesis suggests the opposite: conﬂict will be less likely when both countries are experiencing acute precipitation scarcity. H4. Conﬂict will be less likely in dyads where both members are experiencing acute water scarcity. Data, methods and results We test our hypotheses via a re-analysis of two recent studies of conﬂict between co-riparian states, Tir and Stinnett (2012) and Brochmann and Gleditsch (2012), with some modiﬁcations and extensions. The unit of analysis for both is the dyad-year. We restrict our re-analysis of both to the period 1950e2002. Tir and Stinnett's sample covers signatories of river cooperation agreements for the period 1950e2002, while Brochmann and Gleditsch analyze all dyads occupying the same continent. For this reason, the Brochmann and Gleditsch sample is much larger, though the availability of the precipitation measures restricts our sample sizes somewhat (n ¼ 72,583 vs. n ¼ 5774). Conducting a reanalysis of both should help mitigate concerns that our ﬁndings are overly model- or sample-dependent. Dependent variable: MID occurrence We draw on data from the Militarized Interstate Dispute (MID) project (Ghosn, Palmer, & Bremer, 2004) to operationalize militarized conﬂict between riparian states. MIDs are “united historical cases of conﬂict in which the threat, display or use of military force short of war by one member state is explicitly directed towards the government, ofﬁcial representatives, ofﬁcial forces, property, or territory of another state” (Jones, Bremer, & Singer, 1996, 163). Fatal MIDS are those MIDs that resulted in at least one fatality. MIDs occur in 4.6% of the dyad-years of the Tir and Stinnett sample, while fatal MIDS occur in only 0.5% of dyad-years in the Brochmann and Gleditsch sample when analysis is restricted to dyads on the same continent.

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

6

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13

Independent variables: precipitation trends and triggers The preceding section emphasizes both long-term precipitation levels and variability and short-term, acute water scarcity as factors that may affect the propensity for conﬂict between states. We operationalize these trends (means, variability) and trigger (acute precipitation scarcity) using data from the Terrestrial Air Temperature and Precipitation: 1900e2006 Gridded Monthly Time Series, Version 1.01. The raw data are at a resolution of 0.5 0.5 and cover the time period 1900e2006. Dell et al. (2012) then aggregate population-weighted average temperature and precipitation levels to the country-year. From these values, we calculate mean levels of precipitation at the dyad level. Mean levels are calculated over the period 1950e2002. The variable ranges from 0.6 dm (EgypteQatar) to 33.8 dm (Costa RicaePanama). We expect the coefﬁcient on mean precipitation to be negative: ceterus paribus, a greater volume of renewable freshwater should be associated with a reduced probability of conﬂict. We operationalize within dyad variability in precipitation using the coefﬁcient of variation for the dyad. Coefﬁcient of variation for dyad d ¼ sd/xd, where s is the standard deviation of water scarcity for dyad d and x is the dyad mean. The coefﬁcient of variation is a useful statistic for representing dispersion of values around the mean, with higher values corresponding to greater within-dyad variability. The variable ranges from 0.05 to 0.84. Dyads with lower mean levels of rainfall tend to have higher variability (r ¼ 0.36). We expect the coefﬁcient of variation to be positively correlated with conﬂict. Finally, we estimate the effect of joint precipitation scarcity by creating a dummy coding which takes on a value of 1 when both members of the dyad are experiencing drier than average conditions, i.e., those dyad-years in which both dyad member's precipitation levels are lower than their ten-year moving averages. Moving averages are used to account for signiﬁcant temporal trending in the precipitation data, which would complicate time-series analysis due to the presence of unit roots in the panel data. These conditions obtain 29.7% of dyad-years in the sample. The theoretical expectations regarding acute water scarcity are ambiguous: the neo-Malthusian argument suggests a positive relationship between acute water scarcity and conﬂict, while the bargaining model suggests a negative relationship. The inclusion/exclusion of control variables is a contested issue among those studying the climate-conﬂict nexus. Some studies exclude control variables that might plausibly be affected by climatic factors and thus produce biased coefﬁcient estimates, and instead use time and country ﬁxed effects and panel-speciﬁc or region-speciﬁc time trends to identify the causal impact of contemporaneous climatic ﬂuctuations (Burke et al. 2009; Hsiang et al. 2013). Others (Buhaug, 2010; O'Loughlin, Linke, & Witmer, 2014; Theisen et al., 2012) argue control variables are necessary to place the relative causal weight of climatic factors in context and to address the confounding effects of factors (such as economic development) that may also trend over time. We address this debate by modeling the impact of dyadic precipitation scarcity both ways, reporting results for “as random” models with time and dyad ﬁxed effects and region-speciﬁc time trends and models with controls. Our re-analyses are conducted using random-effects logistic regression and include controls to address temporal dependence. The random intercept represents the combined effect of all omitted dyad-speciﬁc covariates that cause some dyads to be more prone to conﬂict than others. Several controls are common to both studies. Upstream/downstream relationships, in which one country has an asymmetric bargaining power by virtue of its unilateral ability to interrupt the ﬂow of water to the downstream country, have been found to be

particularly contentious. Both studies include controls for distance between dyad members, level of economic development in the dyad, and population. Countries that are further from one another, more economically developed, and that have smaller populations are less likely to be involved in militarized disputes. Both include controls for joint democracy. Democratic peace theory suggests that politically similar dyads, i.e., dyads composed of either two democracies or two autocracies should be less conﬂict-prone than mixed dyads, composed of one autocracy and one democracy (Gleditsch & Hegre, 1997; Oneal & Russett, 1997; Werner, 2000). Tir and Stinnett (2012) include several additional controls in order to model other elements of the “Kantian peace” (trade interdependence and joint participation in international organizations and treaties), as well as controls for power disparities e countries with more uneven military capabilities are less likely to come into conﬂict e and joint military alliance membership. Finally, we include a time trend. Due to the cross-sectional coverage of the precipitation data, the resulting sample sizes are smaller than in the original studies, though the consistent statistical and substantive effects of control variables indicate the samples are not biased due to non-random exclusion of cases. Table 1 presents the “as random” results for joint precipitation scarcity for the Tir and Stinnett (model 1.1) and Brochmann and Gleditsch (model 1.2) samples. Table 2 presents the full re-analysis of Tir and Stinnett, while Table 3 presents the full re-analysis of Brochmann and Gleditsch. Model 2.1 includes all of the controls in the original Tir and Stinnett model, with time splines omitted to save space. Model 2.2 includes river treaty institutionalization, which they ﬁnd to be a signiﬁcant determinant of conﬂict behavior between co-riparian states. Models 3.2e3.6 extend the basic Brochmann and Gleditsch model (3.1) by introducing interaction terms to untangle whether any relationship between acute water scarcity, mean dyadic precipitation, and precipitation variability and conﬂict behavior is conditional on the two countries either sharing a water source and being geographically contiguous and/or having an upstream/downstream relationship with respect to transboundary rivers. Results and discussion Across all seven models, MIDs and fatal MIDS are signiﬁcantly (p < 0.05) less likely to occur when both members of the dyad are experiencing lower than average rainfall. Though the “as random” results are not directly comparable to the samples with controls, the direction of the effect is similar across operationalizations of the dependent variable (MIDs vs. fatal MIDS) and robust to separate sets of controls, samples, and modeling strategies. Holding all other

Table 1 “As-Random” estimates of joint precipitation scarcity on interstate conﬂict, 1950e2001. Coefﬁcients estimates via linear probability models (OLS) with dyad ﬁxed effects, time ﬁxed effects, and region-speciﬁc time trends. Variables

Joint precipitation scarcity Constant R-squared N Dyads

(1.1)

(1.2)

Mid onset Tir and Stinnett (2012)

Fatal mid onset Brochmann and Gleditsch (2012)

0.011* (0.005) 0.815 (2.373) 0.05 6552 232

0.002*** (0.001) 1.167*** (0.373) 0.004 79,266 2568

Robust standard errors, clustered on dyads, in parentheses. *p < 0.05, **p < 0.01, ***p < 0.001.

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13 Table 2 Precipitation trends and triggers and interstate conﬂict, Reanalysis of Tir and Stinnett (2012). Coefﬁcients estimated using random effects logistic regression. Variables

Joint precipitation scarcity Mean Precipitation, Dyad Precipitation Variability, Dyad Upstream/downstream relationship

(2.1)

(2.2)

Mid

Mid

0.338* (0.152) 0.003 (0.034) 10.183** (3.231) 0.210 (0.393)

0.045 (0.043) 0.736* (0.339) 0.259** (0.097) 0.125 (0.247) 0.265* (0.108) 0.464* (0.233) 0.641*** (0.192) 0.937*** (0.139) 0.170* (0.072) 7.180*** (1.727)

0.317* (0.153) 0.013 (0.036) 8.842** (3.343) 0.154 (0.397) 0.288* (0.138) 0.023 (0.044) 0.744* (0.345) 0.276** (0.098) 0.227 (0.254) 0.310** (0.111) 0.567* (0.241) 0.792*** (0.209) 0.980*** (0.143) 0.170* (0.073) 5.736** (1.877)

5774 215

5643 212

River treaty institutionalization Number of treaties Joint democracy Level of economic development (Larger dyad member) Trade interdependence Relative power Alliance Distance Population, dyad Peace years Constant

Observations Number of dyads Standard errors in parentheses. p < 0.05,**p < 0.01,***p < 0.001.

covariates constant at their means, conditions of joint acute scarcity associated with, on average, a 33.7% diminution in the probability of conﬂict relative to years in which at least one of the dyad members received better-than-average precipitation. To put the magnitude of this effect in context, the average effect of joint democracy is a 47% decrease in the probability of MID/fatal MID initiation. What would happen if the threshold for scarcity were set higher? We attempted to re-estimate the models with a higher threshold (two standard deviations below mean), but at that threshold, joint scarcity predicts peace (i.e., the non-onset of a fatal MID) perfectly in the larger (Brochmann and Gleditsch) sample, and thus cannot be estimated. That no fatal MIDS occurred in the 316 dyad years when both members were experiencing extreme water scarcity is consistent with the main ﬁndings reported here. This ﬁnding does not appear to be conditional on states sharing a renewable water resource. The ﬁnding is of similar magnitude in the Tir and Stinnett sample e where all included dyads are parties to river management treaties, and thus by deﬁnition share a renewable water source e and the expanded Brochmann and Gleditsch sample. The interaction terms (joint scarcity*shared basin*contiguity, joint scarcity*upstream/downstream) introduced in models 3.2 and 3.3 are not statistically signiﬁcant, and thus provide no evidence that the effect of joint acute scarcity is conditional on states actually sharing a river basin or on the relative position of countries which share a basin. Thus, this ﬁnding is consistent with the opportunity cost

7

hypothesis: dyads in which states are jointly experiencing acute water scarcity are less, not more, likely to engage in conﬂict, and this effect is not contingent on the states in question sharing a water source. Short-term precipitation triggers are found to affect interstate conﬂict propensity, but the effect is in the opposite direction of neo-Malthusian expectations. This ﬁnding may be counterintuitive from a neo-Malthusian perspective, but it is consistent with research on river treaties, which ﬁnds water scarcity to be associated with cooperative outcomes (Hensel et al., 2006; Stinnett & Tir, 2009). Evidence for the effects of mean levels of precipitation is mixed. The coefﬁcients on dyadic mean precipitation are insigniﬁcant and close to zero in the Tir and Stinnett model, but strongly signiﬁcant (p < 0.01) and negative in the Brochmann and Gleditsch sample: mean rainfall abundance has a pacifying effect on interstate relations. Using model 3.1 as the baseline speciﬁcation, a one standard deviation (4.4 dL) increase in dyadic mean precipitation from the mean value (9.54 dL) is associated with a 41.0% decrease in the probability of MID occurrence. Fig. 2 demonstrates the effect is statistically signiﬁcant over a broad range of values (>99%) for mean dyadic precipitation. The effect becomes indistinguishable from zero at values above 26 dL per year. Thus, we ﬁnd some e but not overwhelming e evidence that lower mean levels of rainfall are associated with greater conﬂict propensity. We ﬁnd consistent evidence of a correlation between precipitation variability and conﬂict: as precipitation variability in a dyad increases, conﬂict becomes more likely. The effect is statistically signiﬁcant (p < 0.01 in models 2.1, 3.1e3.3, p < 0.05 in model 2.2) and of sizable magnitude across all ﬁve speciﬁcations. With models 2.1 and 3.1 as the baseline speciﬁcations, a one standard deviation (0.06) increase in dyadic precipitation variability from the mean value (0.13) is associated with a 45.7% increase in the probability of MID occurrence and a 28.9% increase in the probability of fatal MID occurrence. Fig. 3 plots the marginal effects of dyadic precipitation variability on the probability of a MID, holding values for all other covariates constant at their means. These plots indicate the effect is once again statistically signiﬁcant for most (>95%) values of the independent variable. Greater rainfall variability is associated with greater conﬂict propensity. We ﬁnd no evidence that the relationships between precipitation means and variability are contingent on countries sharing a water resource. Models 3.4e3.6 introduce interaction terms for dyadic precipitation means and variability and whether the countries share a river basin or are territorially contiguous. The coefﬁcients on mean precipitation are in the opposite direction as the main, un-interacted effects, though the effects are not statistically signiﬁcant. The inclusion of the interaction terms diminishes the statistical signiﬁcance of the un-interacted terms for precipitation variability, though the main effects are still signiﬁcant at the 0.10 level. The coefﬁcients on precipitation variability were mostly (three of four) in the same direction as the uninteracted effects, though the effects were never statistically signiﬁcant. With respect to the conditional effects of dyadic precipitation means and variability, we investigated the interactions with Brambor, Clark, and Golder (2006) style plots in order to test for interactive effects across the range of observed values. We conﬁrmed the effects were not statistically signiﬁcant: the conﬁdence intervals overlapped for all substantively meaningful values of the independent variables. The control variables performed largely as expected. Across samples, distance, dyadic economic development, and joint democracy were associated with a decreased probability of conﬂict, while dyads with larger populations were more conﬂict prone. Consistent with Brochmann and Gleditsch's earlier results, we do

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

8

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13

Table 3 Precipitation trends and triggers and interstate conﬂict, Reanalysis of Brochmann and Gleditsch (2012). Coefﬁcients estimated using random effects logistic regression. Variables

Joint precipitation scarcity

(3.1)

(3.2)

(3.3)

(3.4)

(3.5)

(3.6)

Fatal mid

Fatal mid

Fatal mid

Fatal mid

Fatal mid

Fatal mid

0.430** (0.134)

0.603* (0.243) 0.253 (0.291)

0.430** (0.134)

0.428** (0.134)

0.429** (0.134)

0.120*** (0.032) 5.682** (2.059) 0.062 (0.404) 1.548*** (0.347) 0.523 (0.323)

0.120*** (0.032) 5.714** (2.058) 0.054 (0.405) 1.496*** (0.352) 0.524 (0.323)

0.636* (0.250) 0.202 (0.304) 0.169 (0.285) 0.120*** (0.032) 5.704** (2.058) 0.060 (0.405) 1.506*** (0.353) 0.477 (0.333)

0.196*** (0.054) 4.726 (2.462) 1.075 (0.966) 1.538*** (0.346) 0.532 (0.327) 0.118 (0.065) 1.492 (4.326)

0.179*** (0.046) 4.823* (2.443) 0.107 (0.415) 0.270 (0.993) 0.516 (0.328)

0.199*** (0.054) 4.737 (2.532) 0.375 (1.370) 0.635 (1.351) 0.531 (0.329) 0.063 (0.093) 0.231 (5.697) 0.074 (0.087) 2.121 (5.420) 1.072*** (0.190) 0.387* (0.165) 0.561** (0.184) 0.302** (0.104) 0.478*** (0.122) 0.718 (0.371) 0.099 (0.185) 0.006 (0.005) 1.298*** (0.178) 12.772 (10.239) 72,583 2476

Joint Scarcity*Contiguous*Shared Basin Joint Scarcity*Upstream/Downstream Mean precipitation, dyad Precipitation variability, dyad Shared river basin Contiguous dyad Upstream/downstream Mean Precipitation*Shared Basin Precipitation Variability*Shared Basin Mean Precipitation*Contiguous

1.094*** (0.185) 0.367* (0.164) 0.559** (0.183) 0.297** (0.103) 0.465*** (0.120) 0.741* (0.370) 0.093 (0.184) 0.006 (0.005) 1.297*** (0.178) 11.190 (10.156)

1.095*** (0.185) 0.368* (0.164) 0.557** (0.183) 0.298** (0.103) 0.464*** (0.120) 0.747* (0.370) 0.089 (0.184) 0.006 (0.005) 1.296*** (0.178) 11.160 (10.152)

1.095*** (0.185) 0.366* (0.163) 0.559** (0.183) 0.297** (0.103) 0.464*** (0.120) 0.746* (0.370) 0.089 (0.184) 0.006 (0.005) 1.297*** (0.178) 11.072 (10.153)

1.073*** (0.189) 0.385* (0.165) 0.557** (0.183) 0.300** (0.103) 0.478*** (0.122) 0.718 (0.370) 0.104 (0.184) 0.006 (0.005) 1.299*** (0.178) 12.499 (10.204)

0.114 (0.061) 2.442 (4.152) 1.069*** (0.189) 0.382* (0.165) 0.561** (0.184) 0.302** (0.104) 0.471*** (0.122) 0.731* (0.370) 0.090 (0.185) 0.006 (0.005) 1.297*** (0.178) 12.322 (10.217)

72,583 2476

72,583 2476

72,583 2476

72,583 2476

72,583 2476

Precipitation Variability*Contiguous ln distance ln GDP per capita, Larger Dyad Member ln GDP per capita, smaller dyad member ln population, larger dyad member ln population, smaller dyad member Joint democracy One democracy Time trend Peace history Constant

Observations Number of dyads Standard errors in parentheses. *p < 0.05, **p < 0.01, ***p < 0.001.

not ﬁnd a signiﬁcant effect of sharing a river basin beyond the effect of geographical contiguity, which signiﬁcantly increases the probability of conﬂict (p < 0.01). Power preponderance and joint alliance membership were negatively associated with conﬂict in the Tir and Stinnett sample, but neither trade integration nor the extent of shared treaty membership exerted an impact on conﬂict outcomes. This is likely a function of the restricted sample size due to the precipitation data and differences in addressing unit effects (robust errors, as reported by Tir and Stinnett, vs. random effects). Consistent with Tir and Stinnett's earlier ﬁnding, however, we ﬁnd river treaty institutionalization e the extent to which treaties delineate rules governing joint monitoring of water resources, joint enforcement, delegate water issues to an intergovernmental organization, and provide alternative dispute mechanisms e is associated with lower probability of conﬂict occurrence between coriparian states. Taken together, these ﬁndings suggest three broad conclusions. First, we ﬁnd evidence for signiﬁcant impacts of precipitation e

both longer-term means and variability and shorter-term ﬂuctuations e on interstate conﬂict. Much recent scholarship has been dedicated to investigating the impacts of climatic conditions on “new” foci in security studies: civil war, non-state and one-sided conﬂict, and social conﬂict. Comparatively few studies have addressed the potential for climatic impacts on interstate conﬂict (see Bernauer & Siegfried, 2012; Gartzke, 2011, 2012 for exceptions). Hsiang et al.’s meta-analysis of 30 studies of intergroup violence included only one (Tol & Wagner, 2010) that analyzed trans-boundary (i.e., interstate) conﬂict. In a recently published literature review on the links between climate change and armed conﬂict, Theisen, Gleditsch, and Buhaug (2013) explicitly set aside discussion of potential links between interstate conﬂict and climatic conditions. We ﬁnd relatively strong evidence for signiﬁcant climatic impacts on interstate conﬂict, over both shorter and longer time scales. Second, the effects of joint precipitation scarcity, mean precipitation and precipitation variability do not seem to be

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13

9

Fig. 2. Probability of fatal mid onset as function of mean dyadic precipitation.

contingent on the countries in question sharing water resources or being geographically contiguous. While the ﬁndings regarding joint acute scarcity and dyadic precipitation variation in the Tir and Stinnett model clearly apply to co-riparian states (the only states in the sample), similar relationships are found in the broader sample of dyads analyzed by Brochmann and Gleditsch. In the case of joint acute scarcity, this ﬁnding can be interpreted in light of the bargaining model: irrespective of whether the countries share a water source, below-normal rainfall places strains on each country's economies and implies a larger role for the state in responding to crisis. The opportunity cost to ﬁghting is thus higher during these periods, and conﬂict relatively less likely. The ﬁndings regarding precipitation mean levels and variability, however, do not have as straightforward an interpretation. These ﬁndings could be evidence of a general effect of dyadic water stress e countries dealing with water scarcity and high levels of water variability, ceterus paribus, face greater domestic ecological pressures than those which are not similarly stressed. These stresses cause states to behave more aggressively in the international sphere. However, this is essentially a statelevel (i.e., monadic) explanation for conﬂict behavior: states experiencing water stress are more bellicose than those that do not. The ﬁnding, however, is dyadic in nature. Shared water scarcity/abundance and variability affect conﬂict behavior between states even when those states do not share rivers and/or are not contiguous, i.e., there is no shared water over which to war. This ambiguity relates to a broader debate in the environmental security literature over whether we should think about environmental scarcity of a resource as an ultimate cause of conﬂict (Libiszewski, 1992), or in the broader concept of environmental impacts on conﬂict. Most of the environmental security literature presupposes that environmental conditions primarily affect motives for engaging in conﬂict. However, climatic conditions may affect the speciﬁc timing or intensity of conﬂict irrespective of

whether environmental conditions were ultimately the root “cause”. Third, scale matters. We ﬁnd at shorter time scales, acute rainfall scarcity is pacifying. Over the longer-term, however, rainfall scarcity and variability are linked to increased propensity for conﬂict. As Salehyan notes in the introductory article, one of the sources of ambiguity in all the quantitative literature on climate change and conﬂict e and a source of signiﬁcant criticism of meta-studies on the subject e has been the proliferation of dependent variables and scales of analysis. This problem is ampliﬁed by the understandable desire to reduce the impacts of environmental scarcity/abundance to operating through a single mechanism: Malthusian pressures (Homer-Dixon, 1999; Kahl, 2006), livelihood decline leading to decreased opportunity cost to engaging in conﬂict (Hendrix & Glaser, 2007; Miguel et al. 2004), periods of abundance coinciding with better opportunities to loot (Witsenberg & Adano, 2009), etc. Each of these causal mechanisms is well suited to addressing a particular temporal scale and explaining conﬂict between irregular forces. However, none offers a uniﬁed framework for understanding climate impacts on conﬂict and multiple temporal scales and involving state actors. The bargaining model provides a theoretical framework for understanding the effects of short-term precipitation scarcity and longer-term variability on interstate conﬂict behavior, but does not provide any additional leverage over the question of why dyadic rainfall scarcity is associated with increased conﬂict propensity. Here, the standard neo-Malthusian/issue-salience model seems adequate. Conclusions While the 2007 IPCC Fourth Assessment Report contained several references to potential links between climate change and conﬂict, the Fifth Assessment Report, is the ﬁrst since a large body of peer-reviewed research on the subject has emerged. The

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

10

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13

Fig. 3. Probability of mid onset as function of dyadic precipitation variability.

report concludes “Climate change can indirectly increase risks of violent conﬂicts in the form of civil war and inter-group violence by amplifying well-documented drivers of these conﬂicts such as poverty and economic shocks” (IPCC, 2014, 20). On the question of interstate conﬂict, however, the report is silent, reﬂecting a paucity of research on the relationship between climate change and interstate conﬂict. In the absence of rigorous analysis to inform planning, policy makers are likely to ﬁll the void with interesting, if not fanciful, conjectures. In 2003, Peter Schwartz and Doug Randall published An Abrupt Climate Change Scenario and Its Implications for United States National Security, an extreme climate change scenario commissioned by the US Department of Defense. In it, they challenged US strategic planners to “Envision Pakistan, India, and China e all armed with nuclear weapons e skirmishing at their borders over refugees, access to shared rivers, and arable land … With over 200 river basins touching multiple nations, we can expect conﬂict over access to water for drinking, irrigation, and transportation” (18). This thought exercise was premised on an extremely unlikely climate change scenario. While not apocalyptic, the Working Group I Contribution to the IPCC Fifth Assessment Report: Climate Change 2013: The Physical Science Basis tells us the future will be much warmer and wetter at high latitudes while warmer and drier at most lower latitudes. Often lost in popular discussions of climate change, however, are its forecast impacts on climatic variability: an increase in extreme precipitation events, higher year-to-year variability in precipitation, and changing mean levels of precipitation in different regions. Assuming the estimated relationships between precipitation means, variability, and short-term scarcity hold in the future, higher latitudes will likely experience a continued decline in interstate conﬂict (due to higher mean levels of precipitation) that is partially offset by increased variability. Lower latitudes, where precipitation is forecast generally to become scarcer and more variable, would see an increase in the frequency of interstate conﬂict.

This study makes two novel contributions to the study of climatic impacts on conﬂict. First, we attempt a comprehensive test of the effects of precipitation on interstate conﬂict. We focus not just on co-riparian states, but analyze a much larger set of cases. We also investigate the effects of longer-term mean levels of precipitation and precipitation variability on baseline conﬂict propensity between states, as well as short term “shocks” that may affect year-to-year variation in the probability of conﬂict. Second, we develop a theoretical framework, premised in the bargaining model, which moves beyond simple neo-Malthusian conjectures about the relationship between resource scarcity and conﬂict. Our ﬁndings indicate lower mean levels of precipitation and higher variability of precipitation in dyads make conﬂict more likely, while joint precipitation scarcity has a short-term pacifying effect. Moreover, we ﬁnd these effects are not conditional on the pairs of countries sharing a river or being geographically contiguous. We see two promising paths for future research on this topic. The ﬁrst will be to model more explicitly environmental impacts on conﬂict, not in terms of their effects on actor motives, but rather on the degree to which environmental conditions facilitate or inhibit conﬂict behavior. By presuming that environment affects conﬂict through its effects on actor motives, the literature has tended to focus on most-likely cases, i.e., cases of signiﬁcant demographicecological stress (Kahl, 2006), and thus may have missed the more general impacts of climatic conditions on state behavior. Models that address the impact of resource scarcity not just on motive but on resource mobilization potential e the ability to sustain armed operations e would constitute a novel contribution to the ﬁeld. The second will be to look at the effects of ground water resources on conﬂict. Tir and Stinnett note “stress on surface water may also result in increasing conﬂict over groundwater” (2012, 215, fn 6), but their study does not consider the roll of groundwater resources, nor do those of Toset et al. (2000), Gleditsch et al.

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13

(2006), or Brochmann and Gleditsch (2012). The Nile and Indus river basins are highly politicized in no small part because some basin countries (Egypt and Sudan, Pakistan) are almost entirely dependent on water sources that originate either outside their borders or in non-renewable subsoil aquifers. Given the prominence of these cases in the literature, research in this area is necessary. The security implications of climate change have become the subject of much conjecture, often without signiﬁcant grounding in empirical research. This is due both to the inherent difﬁculties of forecasting human innovation and adaptation to shifting climatic realities and to a relative paucity of research in key areas. While much work has focused on absolute resource scarcity as a source of conﬂict, comparatively little has looked at one of the potentially more powerful effects of climate change: increasingly frequent “shocks” and increased variability in ecosystems. This study contributes to our understanding of climate change as a potential source of friction between countries by focusing attention not just on water scarcity, but also on water resource variability. In doing so, we hope to push the literature in a direction that more closely aligns our modeling of human impacts with our understanding of the physical impacts of climate change. Appendix 1. Descriptive statistics

Variable Tir and Stinnett (2012) MID onset Joint precipitation scarcity Mean precipitation, dyad Precipitation variability, dyad Upstream/downstream relationship Number of treaties Joint democracy Level of economic development (larger dyad member) Trade interdependence Relative power Alliance Distance Population, dyad Peace years River treaty institutionalization Brochmann and Gleditsch (2012) Fatal MID onset Joint precipitation scarcity Shared river basin Contiguous dyad Upstream/downstream Mean precipitation, dyad Precipitation variability, dyad ln distance ln GDP per capita, larger dyad member ln GDP per capita, smaller dyad member ln population, larger dyad member ln population, smaller dyad member Joint democracy One democracy Year Peace history

Obs

11

(Aquastat variable #4190), from the FAO's Aquastat database. Per Aquastat, these data are calculated as:

½Water resources : total renewable per capita ðactualÞ ¼ ½Water resources : total renewable ðactualÞ = ½Ttotal population; where [Water resources: total renewable (actual)] is deﬁned as “The sum of internal renewable water resources (IRWR) and external actual renewable water resources (ERWR_actual). It corresponds to the maximum theoretical yearly amount of water actually available for a country at a given moment” (FAO, 2014), and is time-invariant by country. Thus, all of the variation in total renewable freshwater per capita is due to within-country changes in population (i.e., the denominator). This is problematic for a variety of econometric reasons (non-stationarity in the per capita data, collinearity with population), but also because it mischaracterizes the actual availability of water. Figure A1 shows the temporal trends in total renewable freshwater, per Aquastat, as well as the annual Palmer Drought Severity Index score (PDSI) (Dai et al. 2004) and population-weighted precipitation levels (Dell et al., 2012) for Somalia, 1961e2005. The PDSI is an extensively validated measure of meteorological drought, including information on precipitation, temperature, and soil conditions, which ranges from þ10 (saturation) to 10 (extreme drought), while the precipitation data focus only on precipitation levels and are

Mean

Std. Dev.

Min

5774 5774 5774 5774 5774 5774 5774 5774 5774 5774 5774 5774 5774 5774 5643

0.0465882 0.3557326 9.707514 0.120701 0.1271216 4.114652 0.1821961 7.857947 0.2846161 1.624558 0.5051957 6.3553 10.42908 13.18358 1.948254

0.2107733 0.4787762 4.166587 0.04297 0.333138 3.534887 0.3860395 1.123231 0.7612913 1.303743 0.5000163 0.8743637 1.211084 11.50512 1.109152

0 0 1.852428 0.0466941 0 1 0 5.056628 0 0 0 3.78419 7.015712 0 0

72583 72583 72583 72583 72583 72583 72583 72583 72583 72583 72583 72583 72583 72583 72583 72583

0.0049185 0.3112437 0.2391745 0.1216952 0.1001199 9.538086 0.1239345 7.331971 8.322703 7.446081 9.873403 8.231943 0.1175482 0.2668945 1982.113 0.0570743

0.0699599 0.4630054 0.4265824 0.3269357 0.3001618 4.425192 0.0555142 0.7923705 1.003134 0.8892621 1.332831 1.204654 0.3220746 0.4423398 13.11496 0.1874022

0 0 0 0 0 0.648078 0.0295722 1.609438 6.050394 5.639279 5.680172 5.326953 0 0 1951 1

Appendix 2. Assessing renewable freshwater per capita Several studies on the relationship between water availability and conﬂict (Hendrix & Glaser, 2007; Tir & Stinnett, 2012) use measures of total renewable freshwater availability per capita

Max 1 1 21.4044 0.2959095 1 25 1 10.79891 9.534624 6.212126 1 8.580168 14.17782 52 4 1 1 1 1 1 33.40139 0.5207072 8.59822 10.7378 10.6778 14.04569 13.83131 1 1 2001 4.81e35

expressed in deciliters. The Aquastat measure takes on the same value during periods of acute drought (1974e1975, during which famine may have killed up to one million people) as during periods of record precipitation (1997e1998, during which ﬂooding in the Middle and Lower Juba provinces affected over 100,000 residents).

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

12

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13

Fig. A-1. Comparing Total Renewable Freshwater, PDSI and Precipitation for Somalia, 1961e2005.

Thus, total renewable freshwater e either aggregate or per capita e is inappropriate for assessing the impact of transient climatic shocks or climatic variability on interstate conﬂict, and likely misleading as measures of the underlying theoretical concept: total aggregate water availability, which is presumably time-variant. References. Dai, A., Trenberth, K. E., & Qian, T. (2004). A global dataset of Palmer Drought Severity Index for 1870e2002: relationship with soil moisture and effects of surface warming. Journal of Hydrometeorology, 5, 1117e1130. Dell, M., Jones, B.F., and Olken, B.A. (2012). Temperature shocks and economic growth: evidence from the last half century. American Economic Journal: Macroeconomics, 4, 66e95. Food and Agriculture Organziation. (2014). Aquastat database. http://www.fao.org/nr/water/aquastat/data/query/index.html. Accessed 02.03.14. References Agnew, J. (2011). Waterpower: politics and the geography of water provision. Annals of the Association of American Geographers, 101(3), 463e476. Anderton, C. H., & Carter, J. R. (2001). The impact of war on trade: an interrupted times-series study. Journal of Peace Research, 38(4), 445e457. Battisti, D. S., & Naylor, R. L. (2009). Historical warnings of future food insecurity with unprecedented seasonal heat. Science, 323(5911), 240e244. Bergholt, D., & Lujala, P. (2012). Climate-related natural disasters, economic growth, and armed civil conﬂict. Journal of Peace Research, 49(1), 147e162. Bernauer, T., & Siegfried, T. (2012). Climate change and international water conﬂict in Central Asia. Journal of Peace Research, 49(1), 227e239. Brambor, T., Clark, W. R., & Golder, M. (2006). Understanding interaction models: Improving empirical analyses. Political Analysis, 14(1), 63e82. Brochmann, M., & Gleditsch, N. P. (2012). Shared rivers and conﬂict e a reconsideration. Political Geography, 31(8), 519e527. Buhaug, H. (2010). Climate not to blame for African civil wars. Proceedings of the National Academy of Sciences, 107(38), 16477e16482. € hmelt, T., Brzoska, M., Busby, J. W., et al. Buhaug, H., Nordkvelle, J., Bernauer, T., Bo (2014). One effect to rule them all? A comment on climate and conﬂict. Climatic Change. in press. Burke, M. B., Miguel, E., Satyanath, S., Dykema, J. A., & Lobell, D. B. (2009). Warming increases risk of civil war in Africa. Proceedings of the National Academy of Sciences, 106(49), 20670e20674. Chantarat, S., Barrett, C. B., Mude, A. G., & Turvey, C. G. (2007). Using weather index insurance to improve drought response for famine prevention. American Journal of Agricultural Economics, 89(5), 1262e1268. Dell, M., Jones, B. F., & Olken, B. A. (2012). Temperature shocks and economic growth: evidence from the last half century. American Economic Journal: Macroeconomics, 4(3), 66e95.

Diehl, P. F. (1992). What are they ﬁghting for? the importance of issues in international conﬂict research. Journal of Peace Research, 29(3), 333e344. Dietz, T., Ostrom, E., & Stern, P. C. (2003). The struggle to govern the commons. Science, 302(5652), 1907e1912. Ehrlich, P. R. (1969). The population bomb. New York, United States: Sierra Club. Fearon, J. D. (1995). Rationalist explanations for war. International Organization, 49(3), 379e414. Fjelde, H., & von Uexkull, N. (2012). Climate triggers: rainfall anomalies, vulnerability and communal conﬂict in Sub-Saharan Africa. Political Geography, 31(7), 444e453. Food and Agriculture Organziation. (2014). Aquastat database. Rome: FAO Accessed 02.03.14 http://www.fao.org/nr/water/aquastat/data/query/index.html. Gartzke, E. (2011). Blame it on the weather: seasonality in interstate conﬂict. Typescript. La Jolla, CA http://dss.ucsd.edu/~egartzke/papers/seasonality_of_ conﬂict_102011.pdf. Gartzke, E. (2012). Could climate change precipitate peace? Journal of Peace Research, 49(1), 177e192. Ghosn, F., Palmer, G., & Bremer, S. (2004). The MID3 data set, 1993-2001: procedures, coding rules, and description. Conﬂict Management and Peace Science, 21(2), 133e154. Giordano, M. F., Giordano, M. A., & Wolf, A. T. (2005). International resource conﬂict and mitigation. Journal of Peace Research, 42(1), 47e65. Gleditsch, N. P., Furlong, K., Hegre, H., Lacina, B., & Owen, T. (2006). Conﬂicts over shared rivers: resource scarcity or fuzzy boundaries? Political Geography, 25(4), 361e382. Gleditsch, N. P., & Hegre, H. (1997). Peace and democracy: three levels of analysis. Journal of Conﬂict Resolution, 41(2), 283e310. Gleick, P. H. (1993). Water and conﬂict: fresh water resources and international security. International Security, 18(1), 79e112. Glick, R., & Taylor, A. M. (2010). Collateral damage: trade disruption and the economic impact of war. Review of Economics and Statistics, 92(1), 102e127. Grafton, R. Q., Squires, D., & Fox, K. J. (2000). Private property and economic efﬁciency: a study of a common-pool resource. Journal of Law & Economics, 43(2), 679e713. Hardin, G. (1968). The tragedy of the commons. Science, 162(3859), 1243e1248. Hendrix, C. S., & Salehyan, I. (2012). Climate change, rainfall, and social conﬂict in Africa. Journal of Peace Research, 49(1), 35e50. Hendrix, C. S., & Glaser, S. M. (2007). Trends and triggers: climate, climate change, and civil conﬂict in Sub-Saharan Africa. Political Geography, 26(6), 695e715. Hensel, P. R., Mitchell, S. M., & Sowers, T. E. (2006). Conﬂict management of riparian disputes. Political Geography, 25(4), 383e411. Hensel, P. R., Mitchell, S. M., Sowers, T. E., & Thyne, C. L. (2008). Bones of contention: comparing territorial, maritime, and river issues. Journal of Conﬂict Resolution, 52(1), 117e143. Homer-Dixon, T. F. (1999). Environment, scarcity, and violence. Princeton, United States: Princeton University Press. Hsiang, S. M., Burke, M., & Miguel, E. (2013). Quantifying the inﬂuence of climate change on human conﬂict. Science, 341(6151), 1235367. Hsiang, S. M., Meng, K. C., & Cane, M. A. (2011). Civil conﬂicts are associated with the global climate. Nature, 476(7361), 438e441. IPCC. (2007). Climate change 2007: Impacts, adaptation and vulnerability. Cambridge, Switzerland: Cambridge University Press, Intergovernmental Panel on Climate Change.

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

C. Devlin, C.S. Hendrix / Political Geography xxx (2014) 1e13 IPCC. (2013). Working Group I Contribution to the IPCC Fifth Assessment Report Climate Change 2013: The Physical Science Basis. Final Draft underlying scientiﬁcTechnical assessment. Geneva, Switzerland: IPCC Secretariat. IPCC. (2014). Working Group II Contribution to the IPCC Fifth Assessment Report Climate Change 2014: Impacts, Adaptation, and Vulnerability. Summary for Policymakers. Geneva, Switzerland: IPCC Secretariat. Jones, D. M., Bremer, S. A., & Singer, J. D. (1996). Militarized interstate disputes, 1816e1992: rationale, coding rules, and empirical patterns. Conﬂict Management and Peace Science, 15(2), 163e212. Kahl, C. H. (2006). States, scarcity, and civil strife in the developing world. Princeton, United States: Princeton University Press. Klare, M. T. (2001). Resource wars: the new landscape of global conﬂict. New York, United States: Metropolitan. Koubi, V., Bernauer, T., Kalbhenn, A., & Spilker, G. (2012). Climate variability, economic growth, and civil conﬂict. Journal of Peace Research, 49(1), 113e127. Libiszewski, S. (1992). What is an environmental conﬂict? Occasional Paper No. 1. Bern, Switzerland: Swiss Peace Foundation and Zurich: Center for Security Studies and Conﬂict Research. Loayza, N. V., Olaberría, E., Rigolini, J., & Christiaensen, L. (2006). Natural disasters and growth: Going beyond the averages. World Bank Policy Research Working Paper 4980. Washington, DC, United States: World Bank. Loayza, N. V., Olaberría, E., Rigolini, J., & Christiaensen, L. (2012). Natural disasters and growth: going beyond the averages. World Development, 40(7), 1317e1336. Maxwell, J., & Reuveny, R. (2000). Resource scarcity and conﬂict in developing countries. Journal of Peace Research, 37(3), 301e322. Miguel, E., Satyanath, S., & Sergenti, E. (2004). Economic shocks and civil conﬂict: an instrumental variables approach. Journal of Political Economy, 112(4), 725e753. National Intelligence Council. (2012). Global water security, ICA 2012e08. Washington, DC, United States: Ofﬁce of the Director of National Intelligence. Noy, I., & Aekkanush, N. (2011). Fiscal storms: public spending and revenues in the aftermath of natural disasters. Environment and Development Economics, 16(1), 113e128. O'Loughlin, J., Linke, A. M., & Witmer, D. W. (2014). Modeling and data choices sway conclusions about climate-conﬂict links. Proceedings of the National Academy of Sciences, 111(6), 2054e2055. Oneal, J. R., & Russett, B. M. (1997). The classical liberals were right: democracy, interdependence, and conﬂict, 1950e1985. International Studies Quarterly, 41(2), 267e294. Ostrom, E. (1999). Coping with tragedies of the commons. Annual Review of Political Science, 2(1), 493e535. Powell, R. (1993). Guns, butter, and anarchy. American Political Science Review, 87(1), 115e132. Powell, R. (2002). Bargaining theory and international conﬂict. Annual Review of Political Science, 5(1), 1e30.

13

Raleigh, C., & Kniveton, D. (2012). Come rain or shine: an analysis of conﬂict and climate variability in East Africa. Journal of Peace Research, 49(1), 51e64. Randle, R. (1987). Issues in the history of international relations. New York, United States: Praeger. Reuters. (2013). Southern Africa drought woes hurt GDP growth in region. 24 October 2013. New York, United States. Schwartz, P., & Randall, D. (2003). An abrupt climate change scenario and its implications for United States national security. Pasadena, United States: California Institute of Technology Jet Propulsion Laboratory. Slettebak, R. (2012). Don't blame the weather! Climate-related natural disasters and civil conﬂict. Journal of Peace Research, 49(1), 163e176. Solow, A. R. (2013). Global warming: a call for peace on climate and conﬂict. Nature, 497(7448), 179e180. Starr, J. R. (1991). Water wars. Foreign Policy, 82, 17e36. Stinnett, D. M., & Tir, J. (2009). The institutionalization of river treaties. International Negotiation, 14(2), 229e251. Theisen, O. M., Holtermann, H., & Buhaug, H. (2011/2012). Climate wars? Assessing the claim that drought breeds conﬂict. International Security, 36(3), 79e106. Theisen, O. M., Gleditsch, N. P., & Buhaug, H. (2013). Is climate change a driver of armed conﬂict? Climatic Change, 117(3), 613e625. Theisen, O. M. (2012). Scarcity and organized violence in Kenya, 1989-2004: a ‘ﬁtting’ or a ‘mis-ﬁtting’ case of environmental security theory. Journal of Peace Research, 49(1), 81e96. Tir, J., & Stinnett, D. M. (2012). Weathering climate change: can institutions mitigate international water conﬂict? Journal of Peace Research, 49(1), 211e225. Tol, R. S. J., & Wagner, S. (2010). Climate change and violent conﬂict in Europe over the last millennium. Climatic Change, 99(1e2), 65e79. Tompkins, E. L. (2005). Planning for climate change in small islands: insights from national hurricane preparedness in the Cayman Islands. Global Environmental Change, 15(2), 139e149. Toset, H. P. W., Gleditsch, N. P., & Hegre, H. (2000). Shared rivers and interstate conﬂict. Political Geography, 19(8), 971e996. Water Resources Group. (2009). Charting our water future: Economic frameworks to inform decision-making. Washington, DC, United States: McKinsey & Co. Weiss, D. J., & Weidman, J. (2013). Disastrous spending: federal disaster-relief expenditures rise amid more extreme weather. Washington, DC, United States: Center for American Progress. Werner, S. (2000). The effects of political similarity on the onset of militarized disputes, 1816e1985. Political Research Quarterly, 53(2), 343e374. Witsenberg, K. M., & Adano, W. R. (2009). Of rain and raids: violent livestock raiding in northern Kenya. Civil Wars, 11(4), 514e538. Yoffe, S., Wolf, A. T., & Giordano, M. F. (2003). Conﬂict and cooperation over international freshwater resources: indicators of basins at risk. Journal of the American Water Resources Association, 39(5), 1109e1126.

Please cite this article in press as: Devlin, C., & Hendrix, C. S., Trends and triggers redux: Climate change, rainfall, and interstate conﬂict, Political Geography (2014), http://dx.doi.org/10.1016/j.polgeo.2014.07.001

Trends and triggers redux: Climate change, rainfall, and interstate conflict

Trends and triggers redux: Climate change, rainfall, and interstate conflict

Recommend Documents