te,1

3SQ

178

n wwesidegs pncarc

sanittion, wine, public order, irrigation, roadsand public health, what have the RomanAttendee: Brought peace?

(Monty

tal scarcarcewa, 200Lacin

Mitche

alleviate grievances. Thus, our article focuses on how political andinstitutional responses can mediate between resource scarcity andthe risk of intrastate conict.

analysis we discuss our empirical ndings from the pooled dataanalysis of 98 countries from 1980 to 2001. The nal sectionconcludes.

Water scarcity and conict potential

In the late 1980s, neo-Malthusian scholars began exploringenvironmental scarcity and conict relationships, most arguingthat natural resource scarcities, together with social inequalities,

* Corresponding author. Department of Government, University of Essex,Wivenhoe Park, Colchester, Essex CO4 3SQ, UK. Tel.: 44 1206 872648; fax: 441206 873234.

E-mail addresses: tig@essex.ac.uk (T.-I. Gizelis), amanda.wooden@bucknell.edu(A.E. Wooden).

1

Contents lists availab

Political Ge

journal homepage: www.els

Political Geography 29 (2010) 444e453Tel.: 1 570 577 3464; fax: 1 570 577 3536.Homer-Dixon 1991, 1994, 1999; Lowi, 1995; Toset, Gleditsch, &Hegre, 2000; Wolf, 1995). Most studies have focused on proving ordisproving that a direct, deterministic relationship exists betweenscarcity and conict. Likewise, non-conict studies scholars studyingthese issues often do not recognize important patterns in conictonset anddonot differentiate between conicts of interest thatmightarise in conditions of environmental scarcity and the actual envi-ronmentally-driven onset of a violent conict.We argue that politicalinstitutions and democratic governance can mitigate disputes and

promote better conservation and more efcient use. These factors allsuggest that governance can play a crucial role in whether waterscarcity is likely to give rise to violent conict. Responsive govern-mentsmaybebetterable todealwithproblems likewater scarcity, andthus avert discontent and civil strife caused by environmental factors.

The next section discusses the link between water scarcity andthe probability of intrastate conict. We present our argument andcontribution in section The mediating role of conict potential. Insection A model of environmental scarcity, governance, and conictwe discuss our methodology and data, and in section Empirical2003; Gleditsch, Owen, Furlong, &Hauge & Ellingsen, 1998; Hensel,Existing literature on environmentomixed conclusions aboutwhether sto generate violent conict (De Soys0962-6298/$ e see front matter 2010 Elsevier Ltd.doi:10.1016/j.polgeo.2010.10.005, the freshwater systems ever done for us?

Python's Life of Brian)

city and conict has ledater resources are likely2; Furlong & Gleditsch,a, 2004; Gleick, 1993;ll, & Sowers II, 2004;

resource scarcity, although there are important distinctions betweentypes of scarcity, or as Gleditsch (1998), notes, between simple scarcityand degradation. Likewise, sub-national level variation in scarcity anddegradation and the specic decision-making inuences at thoselevels are important, but beyond the scope of this analysis.

Effective governance can address problems of water supply, forexample by improving storage, preservation, and water quality.Governance can also help ensure a fair and equitable distribution ofwater resource, as well as limiting total demand through efforts toReg: All right, but apart from the ation, medicine, educa-andspecicallywater scarcity, is an importantdeterminantofwhetherviolent intrastate conictwill arise. For this analysis,we focus onwaterIntroduction The ability of governments to manage environmental grievances,Water resources, institutions, & intrasta

Theodora-Ismene Gizelis a,b,*, Amanda E. Wooden c

aDepartment of Government, University of Essex, Wivenhoe Park, Colchester, Essex CO4bCentre for the Study of Civil War, PRIO, NorwaycBucknell University, Environmental Studies Program, 102 Coleman Hall, Lewisburg, PA

Keywords:Environmental scarcityIntrastate conictDomestic institutionsDemocracy

a b s t r a c t

Although linkages betweequalitative case studies asature has generally not conand indirect effects linkinintervening factors, such aprobability of conict. Wetionships between water sAll rights reserved.conict

, UK

37, USA

ater scarcity and conict have received a great deal of attention, both inll as quantitative studies, the relationship remains unclear since the liter-red the effectiveness of governance. We distinguish between direct effectswater resource scarcity and conict by systematically examining howolitical institutions, might inuence the impact of water scarcity on thed support for our hypotheses postulating both direct and indirect rela-ity, governance, and conict.

2010 Elsevier Ltd. All rights reserved.

le at ScienceDirect

ography

evier .com/locate/polgeo

ical Gindependently and directly increase the probability of conict.Although Homer-Dixon (1999) and others (Schwartz, Deligiannis, &Homer-Dixon, 2000) discuss social variables such as ingenuity andadaptability as possible conict-mitigating factors, they alsoemphasize the physical limits of such social variables. That is, mostneo-Malthusians scholars circle back to deterministic linksbetween environmental scarcity and conict.

Gleditsch (1998) and others make the case that neo-Malthusianstudies are overly complex and deterministic and highlight theabsence of important political and economic variables in theseanalyses. Neo-Malthusian theorists have responded to critiques ofdeterminism and complexity by pointing out the exogenous factorof environmental scarcity and thus the automatic response ofconict to such scarcity. While the neo-Malthusians make animportant point in highlighting the complexity of ecosystems andhuman interactions (Matthew, 2000), a common neo-Malthusianassumption that conict arises automatically out of scarcity isproblematic. This automatism argument is problematic because itassumes that institutions are either irrelevant or minimallyimportant, that adaptation possibilities are limited and determinedat base by environmental context, and that politics does not matteras much as other factors. Some of these deterministic and pessi-mistic neo-Malthusian claims are not supported by recent micro-level studies (Horowitz, 2009; Turner, 2004).

Resource optimists or cornucopianistsdexemplied by Simon(1996) and Lomborg (2001a, 2001b)dclaim that scarcity is notthe condition that actually applies tomost natural resources (Urdal,2005). Humans automatically adapt to resource scarcities, either byusing market mechanisms which respond to scarcities or throughpurposeful technological innovation which drives adequateresponses (Boserup & Schultz, 1990, Urdal, 2005).

The argument that innovation drives environmental improve-ment is also important in the literature on environmental Kuznetscurves. Scholars claim that environmental degradation will displayan inverted u-shaped relationship with economic development.Although economic growth in very poor countries at the outset willlead to greater environmental problems, such as water pollution,higher economic development beyond some threshold will yielddecreasing degradation either through industrial and technologicalshifts, increasing concerns over environmental degradation oncecore material needs are satised, or more income to expend onanti-pollution efforts (Grossman & Krueger, 1994; Shak &Bandopadhyay, 1992). Work focusing on water resource avail-ability and Kuznets curves has considered outcomes such as waterconsumption (Goklany, 2002; Rock, 1998), improved irrigationefciency (Bhattarai, 2004), and water withdrawals and virtualwater (Katz, 1998). However, contrary to the cornucopianistexpectations, this line of research does not consistently ndsupport for the hypothesized inverted u-shaped relationshipbetween income and water resource outcomes. For example, whileKatz (1998) nds that water withdrawals exhibit a Kuznets curve edecreasingwith increasing income after some levele he also showsthat total water use does not conrm this relationship. Katzdemonstrates that water footprints, which incorporate virtualwater through imported goods and therefore all water consump-tion, do not exhibit Kuznets curves. The fact that there is not a oneto one relationship between income and effective governance mayaccount for some of these ndings.

In addition to cornucopianist arguments about income andresource scarcity and the tests of these hypotheses, the resourceabundance hypothesis posits that a plethora rather than the scar-city of natural resources increases the risk of violent conict,although the resource abundance literature does not address waterresources (Collier, 2000; Collier & Hoefer, 1998; De Soysa, 2002; Le

T.-I. Gizelis, A.E. Wooden / PolitBillon, 2001; Urdal, 2005).Conict studies scholars have moved to a more systematicexamination of the competing arguments about the impact ofdecreasing water resource availability primarily on the probabilityof interstate conict (Collier & Hoefer, 1998; Gleditsch, 1998;Gleditsch & Urdal, 2002; Hensel et al., 2004; Tir & Diehl, 2001;Vasquez, 1993). These studies have sought to identify whether ornot patterns of relationships between nations sharing waterresources exhibit the hypothesized neo-Malthusian relationshipbetween conict and increasing scarcity (Ackerman & Tir, 2003,Furlong & Gleditsch, 2003, Gleditsch et al., 2004, Hensel et al.,2004).

Hauge and Ellingsen (2001) nd a positive and signicantimpact of freshwater availability on both intra- and interstateconict, with more signicant impacts on intrastate conicts.However, their ndings have not yet been replicated. Toset et al.(2000) nd that countries with shared rivers have a slightlyhigher frequency of armed conicts than those that do not shareriver basins. Hensel et al. (2004) argue that regional conictmanagement institutions develop more often and are more effec-tive in resource abundant regions. Basing their work on thehypotheses of the neo-Malthusian approach by Critchley and Terriff(1993) and Lateral Pressure Theory proposed by Choucri andNorth (1975), Hensel et al. (2004) offer a hypothesized indirectrelationship between water scarcity and peace through regionalinstitutions and democracy, rather than just the direct relationshippostulated by most other studies. Their results indicate that inter-national institutions dealing with water scarcity can account forvariation in interstate conict and can reduce the probability ofsuch conict.

The systematic studies of water resource conict have beenimportant for shedding light on the role of institutions at theinternational level. Thewide consensus in the literature is thatwaterscarcity does not induce violent international conicts; however,the literature pays little attention to the impact of water scarcity onthe most common type of violent conict, namely intrastate wars.Whilemuchof the systematic literature in search of general patternsuses the Homer-Dixon groups work as a starting point for provingthat this relationship is not automatic, to our knowledge no quan-titative work approaches the level of complexity of most neo-Malthusian arguments and models. Thus, much of the quantitativeliterature does not adequately reect the complex relationshipsidentied by the case study literature. Asmost conicts tend to be atthe intrastate level, the analysis of water resource conict potentialshould be at the sub-state level, where environmental factorsinteract with domestic institutional structures.

The mediating role of domestic institutions

In order to better understand when and how conicts overwater resources occur, we look at the institutional capacity of statesto adapt to environmental constraints and public dissatisfactionwith increasing environmental constraints. Our argument thatintrastate water resource conicts are institutionally driven isbased on a set of assumptions: regime type inuences conictpotential, resource distribution, and perception of resource avail-ability; intrastate conict has an impact on institutional effective-ness; international spillover effects are possible in regionallyresource-dependent countries; and conict, political institutionsand natural resources interrelate both directly and indirectly in anendogenous system.

Structural scarcity e the unequal distribution of resources ematters for the emergence of environmental conict, as neo-Malthusians suggest, because it puts pressure both on naturalecosystems and social structures (e.g. Burton, 1990a, 1990b; Galtung,

eography 29 (2010) 444e453 4451996; Lederer & Galtung, 1980). Nevertheless, since we assume that

cal Ginstitutions affect distribution and access, conict will not alwaysoccur inwater scarce nations, and in fact, themodal tendencymay befor it not to occur. What matters is how governments respond to thepressure of natural and structural scarcity. Case studies and in-depthanalysis of herder-farmer conicts inAfrica also support the argumentthat these conicts are often driven by social tensions that primarilyreect institutional failures and mismanagement at the local andcommunity level. Environmental degradation is the outcome ratherthan the cause of institutional break down which leads to socialconict (Turner, 2004: 883). In the context of these social conicts,environmental factors interact with political decisions.

Government policies and their effectiveness are not onlya function of state strength, but also inuenced by the formal andinformal relationships between political leaders and their constit-uents. In democracies, political elites need to accommodate andsatisfy a larger part of the electorate to ensure political survival(Bueno de Mesquita & Root, 2000; Bueno de Mesquita, Morrow,Siverson, & Smith, 1999). The degree of a political regimesdemocracy is not always synonymous with good governance; veryautocratic regimes often have high levels of institutional capacity,and may be able to distribute public goods more efciently thanweak democratic political systems (e.g. Singapore and China).However, even though there are cases of highly effective autocraticregimes, it is generally the case that democracies tend to havebetter and more responsive governance systems, especially atmeeting the basic needs of the electorate.

Regime type matters for environmental conict not becausedemocratic rulers are more enlightened to environmental issues,but because democracies are better at alleviating public concernsand developing trust in government responsiveness. Democraciespay greater attention to local actors needs and concerns, or haveformal mechanisms that are likely to hold governments account-able when such needs are not met or rights protected. Conicts ofinterest are likely to be present in scarce resource distributionsituations. However, disputes can be addressed in the early stagesbefore escalating into violence or intrastate war. Domestic institu-tional structures can mitigate the social polarization that environ-mental scarcities may give rise to. Research has demonstrated thatin democracies institutions tend to have a mitigating impact on theprobability of intrastate conict more so than in autocratic regimesall things equal (Benson & Kugler, 1998; Ellingsen, 2000; Hegre,Ellingsen, Gates, & Gleditsch, 2001).

Overall, democracies are less likely to experience violent intra-state conict, since they can provide a political outlet to theexpression of grievances (Hegre & Austvoll Nome, 2009; Muller &Weede, 1990), including demands over distribution of resourcesand environmental scarcity. These distribution and scarcity issueswill not necessarily be addressed or resolved to anyones satisfac-tion, but the existence of an outlet to express dissatisfaction willreduce violent conict potential regardless of actual governmentresponsiveness. Thus, democracies may or may not be moreeffective or efcient at dealing with water resources, but they arebetter at dealing with internal conict. It is the ability to effectivelyalleviate grievances and adapt to change that matters.

Democracies tend to develop these pressure-valve mechanismsand are better at distributing public goods, such as health,compared to autocratic political systems (Lake & Baum, 2001).Although freshwater is not a pure public good, it is a common good,which is necessary for the survival of communities, and we antic-ipate that democracies will be more responsive to demands for theprovision of adequate water resources to the majority of the pop-ulation (Gleditsch, 1998; Gleditsch & Sverdrup, 2002; Hauge &Ellingsen, 2001; Urdal, 2005).

Colorado River ow disputes in theWestern US demonstrate the

T.-I. Gizelis, A.E. Wooden / Politi446importance of political institutions in water disputes. Urbanizationstrained natural resource per capita availability and reprioritizeddistribution from one sector e agricultural to municipal e andregion in the Colorado basin to another. This environmental conicthas been contained in intensity and scope partly because of thepresence of well-established institutional structures that allow forlitigation and mediation of competing demands between urbanand rural communities and between states, even as water rightssystems have not changed to reect new realities. This type ofinterdependent, tense relationship between US states over theColorado River is magnied for regionally resource-dependentcountries without established mediating institutions and a highergoverning body to manage the disputes.

International water politics in Central Asia exemplify the waterresource dependency and institutional issue for internal andinternationalized disputes. Two of the downstream nations in theAmu Darya and Syr Darya basinsdUzbekistan and Turkmenis-tandare not categorized as water scarce or water-stressed nations.But these semi-arid and arid countries are highly dependent ontheir upstream neighborsdKyrgyzstan and Tajikistandfor waterresources due to the downstream countries important agriculturalsectors. Internally and in relations with their neighbours, Uzbeki-stan and Turkmenistan frame themselves as water scarce nationsbecause of their dependency.

The nexus between water scarcity, domestic institutions, and conict

Domestic institutions are the mitigating factors between envi-ronmental scarcity and conict. However, violent intrastate conicthas also a direct and measurable impact on the formation andcapacity of domestic institutions. Countries with weak andunstable political institutions and low economic development(Fearon & Laitin, 2003; Collier et al., 2003) are of higher risk toexperience intrastate conict, all things equal. If countries fail torebuild their domestic structures and improve economic develop-ment, more often than not, they enter the vicious circle of conictand underdevelopment that is identied as the conict trap (Collieret al., 2003). Thus, domestic institutions and intrastate conict aretwo potentially endogenous outcomes, and the role of conict andinstitutions must be considered in a simultaneous framework.

One question that is raised by the Malthusian literature andunderstudied in quantitative analyses is whether there exists anecological trap e analogous to the conict trap e linking environ-mental scarcitieswith conicts and institutions directly in a feedbackcycle. Bio-geographic endowments have been found to have a deepinuence on economic development (Hibbs & Olsson, 2004; Sachs,2005), and ecological factors at initial stages of development plusstate formation inuence the type of political institutions thatemerge, particularly a countrys prospects for democracy (Crenshaw,1995, Midlarsky, 1998). As we have argued, effective institutions canaddress environmental and structural scarcity through policies suchas conservationor redistribution. Yet,will countrieswithunfavorableinitial endowments be stuck in a trap where bad institutions exac-erbate rather than alleviate scarcity and scarcity in turn underminesprospects for political reform? Stated differently, can undemocraticinstitutions create policies that lead to greater scarcity e such asthrough unchecked population growth placing pressure on waterresources e or alternatively increasing perceptions of scarcity?

Conict can clearly lead to the deterioration of existing waterresources or inability to distribute them adequately through thedestruction of infrastructure or the lack of control over key regions.Countries where water resource infrastructure has been under-mined by internal conict include Sudan, Afghanistan, Iraq, theDemocratic Republic of Congo, the former Yugoslavia, Columbia,Tajikistan and Nepal. The decline of institutional capacity could

eography 29 (2010) 444e453make it more difcult for states to address environmental stress,

In sum, political institutions and external dependence matter.Political institutions, together with conict and water resource

scarcity as an exogenous variable. More specically, we estimate

ical Gthe following system of equations:

C gC IRFWR xCG bCXC 3CG gGIRFWR xGC bGXG 3G

(1)

where C intrastate and internationalized intrastate conict;G State Capacity/Governance; XC water scarcity (IRFWR),previous conict, total population, log of GDP per capita, agricul-tural productivity, lag of urbanization, lag of rural populationdensity, interactive term between democracy and IRFWR; XG logof GDP per capita, lag of urbanization, urbanization square, andnumber of peace years; e error terms.

Operationalization

We now detail the specic measures used for our key variables.Our variable selection is based on generally accepted practice andprecedents in the literature reviewed and extended by ourassumptions and theoretical framework. Included also are factorsthat are likely to inuence both the dependent and independentavailability, are interrelatedandmustbeconsidered inanendogenoussystem. Based on these and our previously outlined assumptions, wederive and test the following direct and indirect relationships amongpolitical institutions, conict and water scarcity:

H1: Water resource scarcity increases the probability of intra-state conict.H2: Democratic institutions reduce the probability of intrastateconict.H3: Water resource scarcity contributes to the emergence ofautocratic regimes.

A model of environmental scarcity, governance, and conict

We assess our three hypotheses using data for 98 countries forthe time period 1981e2000 (for a list of the countries included inthe analysis see Appendix II). The timeframe of our data wasdetermined by the availability of most indicators on the countriesincluded in the sample. Similarly, missing data on many countrieslimit the sample to 98 countries. Despite this limitation, our sampleincludes a very diverse set of countries which guarantees signi-cant variation in terms of both the dependent and the independentvariables. Since we assume that reciprocal causation exists, weneed to consider a system of simultaneous equations to take intoaccount the potential endogeneity of scarcity, governance, andconict. For example, treating institutions as pre-determined orxed when estimating their effects on conict would be inappro-priate here given our hypothesized relationships between waterscarcity, institutional capacity, and the probability of conict.Simultaneous equations methods are therefore appropriate in ouranalysis to take into account the potential endogeneity suggestedby our theoretical argument. We use a two-equation model toevaluate our central argument that regime type impacts conictpotential in water scarce conditions. In this model, we treat waterwhich then in return reduces the ability of institutions to mitigatethe link between resources and conict. Therefore, these questionsabout the direct links e and potential endogeneity e betweenpolitical institutions and water resource scarcity are important toconsider, in addition to our arguments about the indirect rela-tionship between these factors.

T.-I. Gizelis, A.E. Wooden / Politvariables of interest.Water availability variable

One difculty faced by scholars is reliablymeasuring socially andpolitically constrained natural resource availability at the nationallevel. As we are interested in scarcity (rather than sanitation,infrastructure, or distribution) to properly test the arguments madein the literature, we utilize a measure that is widely used by envi-ronmental and conict scholars (Furlong&Gleditsch, 2003;Hauge&Ellingsen, 2001; Hensel et al., 2004; Midlarsky, 2001; Toset et al.,2000). This measurement is to a limited extent a proxy for thehuman-impacted side of natural availability, specically in terms ofpopulation growth (and thus per capita distribution). Thus, wemeasure water availability as internally renewable freshwaterresources per capita (IRFWR). This includes the average annual owof rivers and the recharge of groundwater (aquifers) generated fromprecipitation occurring within a countrys borders. While this is themost widely used measure of water resource scarcity, there areother possible measures to capture scarcity, such as access tofreshwateror available drinkingwater (Molle&Mollinga, 2003). Themeasure used in this study has themost comprehensive coverage interms of countries and time periods. We expand the available databy extrapolation using linear regression. Extrapolating data e inparticular water availability which can be stochastic e is somewhatproblematic, but the benets of increasing the number of observa-tions outweigh the potential problems of extrapolation in thissetting. The data are from FAO Aquastat (2001), provided for inter-vals of ve years (FAO 2001). We extrapolated the in-between yearsto obtain annual data and take the natural log of the variable for thepresented models. While it may be useful to measure trends inscarcity, thus considering moving averages of water scarcity levels,we are limited by gaps in the data in conducting such an analysis.

Differentiation between totally available freshwater resourcesand internally derived resources is made by FAO and in the WorldDevelopment Indicators to incorporate political and economicrealities along with ecological ones. Internal water resources reectpopulation changes that affect natural availability. Likewise, thismeasure does not reect purely natural availability given theinexact match between national boundaries and watersheds.Internally available water resources constrains our measurementto non-shared water sources or those not originating internation-ally, as well as reecting population changes impacted by govern-ment policies, economic growth, and social welfare. This waterresource measure therefore is a partly politically and sociallycontrived one, rather than purely natural. Moreover, we use aninteractive term of democracy and freshwater resources per capitato capture the combined effect of political institutions and waterscarcity on the probability of civil conict.

Endogenous variables

ConictAlthough conict in terms of incompatibilities may not always

give rise to violence, we here focus on violent conict where armedforce is used. For our measurement we use the Uppsala ConictData Programme (UCDP) (Uppsala Conict Data Programme, 2008)which uses as threshold 25 battle-related deaths per year. This isthe lowest possible threshold available cross-nationally for armedconicts. Thus, UCDPs threshold allows us to capture a largernumber of violent intrastate and internationalized intrastateconicts than alternative measures of intrastate conict such as theCOW dataset. We do not include interstate conicts, since our focusis on intrastate conict and the sample is for the period 1981e2000,where very few interstate wars occurred (on the data see Eriksson& Wallensteen, 2004; Gleditsch, Wallensteen, Eriksson, Sollenberg,

eography 29 (2010) 444e453 447& Strand, 2002). Based on the Uppsala dataset we create

(CPI) over the percentage of irrigated land (CPI/%irrigated land). Thedata are from the FAO reported by the World Bank Development

cal Ga dichotomous variable for conict, which takes the value of 1 if wehave either a type 3 (intrastate) or type 4 (internationalizedintrastate) conict breaking out in a particular year. International-ized intrastate conicts are dened as civil wars where externalactors are involved. We also exclude from our analysis cases wherethe location of the incompatibility or conict is different from thegeographical location of the ghting (for more on location seeUCDP, 2008).

GovernanceMeasuring governance effectiveness remains challenging, and

most available suggestedmeasures such as GDP per capita (Fearon &Laitin, 2003) and tax extraction rate of governments (Arbetman &Kugler, 1997) are considered controversial as they may reecta host of factors other than governance or the ability of the state toenact responsive policies (Jackman,1993).Moreover, theWorld Bankindicators on governance have limited coverage starting from 1996(Kaufmann, Kraay, &Mastruzzi, 2007). Therefore, we use democracyor typeof political regimeas a proxy for better governance, as argued.To measure the type of political system we use the Polity institu-tionalized democracy measure from Polity IV (on the data seeJaggers &Gurr,1995),which takes onvalues ranging from10 for themost autocratic regime to 10 for the most democratic politicalsystems. We converted irregular polity scores (e.g. values of 77,88, 60) to 10 for the empirical analysis. Although high polityscores indicate democratic regimes and very low scores indicateautocracies, it is possible there is not always a linear relationship inthe index, making the interpretation difcult (Przeworski, Alvarez,Cheibub, & Limongi, 2000). Thus, we created a dummy variable fordemocracy (if Polity score is above 7 then the dummy variable getsa value of 1, otherwise it gets a value of 0). Further distinction inregime types than degree of democracy can also matter, but the lackof systematic data means that these are better addressed throughcase studies (see Chatrchyan & Wooden, 2005).

We also used Vanhanens (2000) index of democracy as analternative measurement for political institutions. Vanhanensindex of democracy and the institutionalized democracy measurefrom Polity are positively correlated (r 0.66). We replicated all themodels presented in this article by using Vanhanens index insteadof Polity. Including the Vanhanens index indicates a decrease in theprobability of conict and an increase in freshwater resources percapita. All the other coefcients remained the same. The replicationdata are available upon request.

Exogenous variables

In a system of simultaneous equations the search for the properinstruments as well as parsimony is always challenging. We chooseexogenous variables based on plausible specications for each of theendogenous variables and the need to consider relevant controlvariables. In this context, we use as instruments variables that explainthe variance of all three endogenous variables (conict, waterresource scarcity, and regime type) and their hypothesized relation-ships. We run separate OLS regressions with each one of the endog-enous variables to examine if the applied instruments are adequate toexplain variation. Each one of the OLS regressions has adjusted R-squares between 0.49 (for the total internal renewable resources) and0.28 (for the conict equation). The exogenous variables can bedivided into two categories: environmental and structural.

Environmental variables

The most commonly used control variable of renewable fresh-water resources is the amount of annual precipitation, a proxy for

T.-I. Gizelis, A.E. Wooden / Politi448the natural factors in the complicated hydrological cycle whichIndicators. This measure captures demand side water use andindicates how productive a countrys agriculture (worldwide themost water intensive sector) is relative to the amount of land beingirrigated e thus how efciently a country uses its water resourceswhen precipitation is not enough at the right time.

For urbanization, which is dened as the percentage of the totalpopulation that lives in urban centers, we also use data from theWorld Bank Development Indicators. Urbanization is lagged by oneyear to avoid problems of endogeneity. In the institution equationwe also include the square of urbanization, as per the literature ondemocratization that nds a curvilinear relationship with urbani-zation continuing without increases in economic growth (Barro,1999; Fay & Opal, 1999). Urbanization is also squared to capturethe rapid urbanization effects in developing countries and the dragit can play on social and political institutions (Barro, 1999).

As a measure of population distribution, urbanization is distinctfrom and has a different impact on water resources than ruralpopulation density, which is also used in our analysis. Also, ruraland urban groups tend to engage in distinct manifestations ofpolitical actions. The data for rural population density are also fromtheWorld Bank Social Indicators and we use the one year lag of thisvariable.

The natural log of GDP per capita is included to capture overalllevels of development, which is widely held to be associated withprovides an alternative test of an exogenous impact of freshwaterresources on political outcomes. FAO Aquastat provides averagevalues of precipitation for every 5 years, measured as Km3. Annualprecipitation has little variance across years for the time period andcountries evaluated here; thus, we cannot make any inferences.Due to limited observations, we extrapolated the data using linearregression.

The second environmental control variable we are using is thedependency ratio capturing water resources that originate fromneighboring countries. Countries that are dependent derive a largeportion of their total available water resources externally, i.e. theyare located downstream in a river basin, share a river border, waterused is delivered from another nation. The dependency ratio vari-able reects a governments controllability of resources andexternal pressures, and is important for evaluating the perceptionof scarcity as well as the internationalized component of intrastateconicts (in other words, when external factors inuence domesticpolitics or when domestic conicts spillover national borders).Theoretically the ratio varies from 0 to 100. The ratio is estimated asfollows: Dependency ratio IRWR/(IRWR IWR) 100 percent,where IRWR stands for internal renewable water resources andIWR stands for total volume of incoming water resources fromneighbouring countries (FAO 2001).

Structural variables

Urbanization, economic development, agricultural productivityand density of the rural population put consumption pressure onwater resources and lead to additional asymmetry in distribution(beyond geographically-based maldistribution). Moreover, due topolitical choices these processes and sectors (e.g. urbanerural,agricultural and municipal uses) are often competing, creating theunderlying conditions for intrastate conict. Thus, changes in thesestructural variables reect the interaction between domesticinstitutions and environmental pressures, and how these processesare linked to intrastate conict (Matthew & Gaulin, 2001).

Agricultural productivity is the ratio of crop production index

eography 29 (2010) 444e453better prospects for democracy and lower levels of intrastate

are also interested in exploring how water availability conditionalon political institutions affects the probability of conict. We thusintroduce an interactive term of these two variables to ourbaseline model.

Interactive term model

The effect of the interactive term between democracy and wateravailability on the probability of conict in Table 3 is signicant atthe 1 percent level and negative. When a multiplicative interactionterm between the two variables is included in a regression model,the individual terms for these variables cannot be interpreted

ical Gconict (Fearon & Laitin, 2003). The data for GDP per capita comefrom the World Bank Development Indicators.

Although urbanization and GDP per capita are mildly correlated(0.55), the Variance Ination Factor (VIF) does not indicate thatmulticollinearity is a problem in our analysis. The conventionalwisdom is that countries with higher income are also more urban-ized; yet, some of the large and developed countries included in thissample (e.g. US andCanada) havemuch lower levels of urbanizationsthan the average. All of these variables are estimated at the nationallevel; thus, they cannot explain regional variation within countries.

Regarding intrastate conict we use two key control variables asthese are identied by the literature on intrastate conict to controlfor alternative explanations to the environmental scarcity argu-ment. We include total population and the occurrence of conict inprevious years (Fearon & Laitin, 2003; Hegre & Sambanis, 2006).Since our sample is from 1981 to 2000, we estimate the presence ofconict in previous years by averaging the years from 1970 to 1980and create a dummy variable that takes the value of 1 if a countryhas experienced conict during this time period and takes the valueof 0 if a country has not experienced any intrastate conict for thesame time period.

This is not an exhaustive list of all the factors that mightcontribute to intrastate conict independently from environmentalpressures and concerns. Nevertheless, we are condent thatconict in previous years, together with GDP per capita as anindicator of social and economic development, captures many ofthe structural factors that are usually associated with the onset andrecurrence of intrastate conict (Collier et al., 2003; Ellingsen,2000; Lacina, 2006).

Finally, in the institutions equations we include as an additionalcontrol variable the number of consecutive peace years since thebeginning of our sample. Existing studies suggest that democracydoes not break down in the presence of conict, but there isevidence suggesting that democracy is less likely to emerge inconict prone regions (Thompson, 1996). Therefore we also includea count of the number of consecutive years of peace to assesspotential long run consequences of conict ondomestic institutions.For a list of all the variables included in the analysis, see Appendix III.

Empirical analysis

The rst step in our analysis is a preliminary view of the data toidentify if water scarcity indeed is more prevalent in countries thathave experienced internal conict and have democratic regimes.Since internal renewable water resources (water availability) isa continuous variable we look in Table 1 at how the median valuesvary conditional to the other main explanatory variables of interest.

We observe in Table 1 that the median value of the wateravailability variable is higher among the countries that are demo-cratic than non-democracies and higher for countries that do notexperience intrastate conict than among countries that do. Thesedescriptive statistics are clearly consistent with our hypotheses.However, democracy, conict, and freshwater could also be inu-enced by other features that may be correlated with the otherindependent variables of interest, and these descriptive statisticsdo not take into account the potential endogeneity between thevariables. Hence, we move to the next step, which is the estimationof the simultaneous equations model and the hypothesized causallinks among water scarcity, democracy, and intrastate conictwhile taking into account the relevant control variables.

Simultaneous equation models

Institutional structures are the intervening mechanisms inu-

T.-I. Gizelis, A.E. Wooden / Politencing the signicance and direction of the water scarcity andintrastate conict relationship. It is important to distinguishbetween short-term and net effects in order to evaluate our results.We make this distinction throughout our empirical analysisdiscussion. We present rst the results for our baseline model,which does not include the interactive term between democracyand water availability. Then we turn to presenting the results forour interactive term model. Finally, we introduce preliminaryresults of a three-equation model where water availability isincluded as an endogenous variable.

Baseline model

As previously discussed, the literature on water scarcity andconict almost entirely focuses on direct effects between thesevariables. Earlier in this paper, we hypothesize that water scarcityadversely affects the prospects of democracy to emerge. Thesehypothesized short-term effects of water availability on regimetype are supported by our empirical analysis. In the politicalregime equation, the coefcient for water resources is positive andstatistically signicant at the 1% level. These results indicate thatcountries that are water scarce are less likely to have democraticinstitutions, all else being equal. Similarly, countries that experi-ence civil conict are less likely to be democratic; the coefcientfor civil conict is statistically signicant at the 1% level. Thus,intrastate conict increases the likelihood of autocracy. Regardingthe control variables, economic development increases the like-lihood that a country will be democratic (see Table 2), as isanticipated by the literature on democracy and development(Lipset, 1960).

When it comes to the results for the conict equation,hypothesis one e i.e. that water resource scarcity increasesthe probability of intrastate conict e is not supported. Thecoefcient is positive but insignicant. Hypothesis 2 e democraticinstitutions reduce the probability of intrastate conict e is alsonot supported in our baseline model. Regime type does not havea signicant effect on the probability of conict when controllingfor water availability. This nding at rst does not seem to t wellwith most of the literature on intrastate conicts (Buhaug, 2006;Hegre et al., 2001). Thus, further exploration is required. Sincewater scarcity has an impact on the type of political regime, we

Table 1Median water availability by independent variables with 95% CI.

Median water availability, democracies0.0041 (0.0038, 0.0055)Median water availability, non-democracies0.0031 (0.0029, 0.0034)Median water availability, conict cases0.0032 (0.0029, 0.0036)Median water availability, non-conict cases0.0038 (0.0036, 0.0042)

eography 29 (2010) 444e453 449independently and the net effect of a marginal change in one

way reciprocal causation between political regime, civil conict,and water availability.

Discussion

Despite the complexity of our arguments, there is a clear themein our empirical results: institutions inuence the ability of states toadapt to their freshwater needs by mitigating possible conicts ofinterest that could potentially escalate to intrastate wars. Similarly,the availability of water resources can affect the nature and effec-tiveness of domestic institutions. Environmental conditions tend tobe linked to and interact with the development of political andeconomic institutions (Diamon, 1997; Wittfogel, 1957). In fact, theecological trap hypothesis is granted support in this analysis, asanticipated (Crenshaw, 1995; Midlarsky, 1998).

The neo-Malthusian arguments about direct and largelyunavoidable linkages between resource scarcity and conict arenot supported by the current analysis. In fact, in the two-equationmodels presented in this article, the level of water resources doesnot have a signicant impact on the probability of conict, anempirical nding that suggests that the actual distribution, ratherthan the available stock, of resources might be more relevant tothe onset of conict. The convex relationship found betweenurbanization and democratic institutions renders further supportto our argument about the complex interaction between pop-ulation dynamics, environmental structures, and institutions.Micro and meso-level studies of the links between urbanization,population distribution and environmental issues, such as White

cal Geography 29 (2010) 444e453variable will depend on the value of the other (Braumoeller, 2004).In other words, the estimated coefcients from the model with

Table 2Two-equation base model, 1981e2000.

Equations Variable Coef. S.E. z P > |z|

Probabilityof civil conicta

Political regime(Democr/Polity)

0.056 0.156 0.360 0.720

Log of (TIRFWR)b 0.300 0.268 1.120 0.262Conict inprevious years

0.398*** 0.026 15.420 0.000

Population >0.001 0.000 0.970 0.333Natural log ofGDP per capita

0.047* 0.025 1.910 0.056

Productivity

d a g lc

l ipths s ta t n lts l rs s s ais o e e -tu ection, to evaluate indirect as well asd t saa t r o n .F t o -ti f d rim

A

q him rG ,In l a F ,R n i lc

A

3

intrastate conict

Conict in previous years 0.381**** 0.030 12.720 0.000

ical Ga more nuanced understanding of the mechanism through whichwater scarcity might be linked to intrastate conict.

Water scarcity undermines governance and the establish-ment of democratic institutions within a country, all thingsequal. Thus, countries that experience water scarcity mightenter a vicious cycle of scarcityeautocracyeconict. Theimplication of this nding can also be linked to climate change,which can threaten not only the availability of water resources,but also undermine the overall capacity of states to respond toincreasing demands and the deterioration of freshwaterresources (Barnett & Adger, 2007). However, the impact ofclimate change on political institutions will vary in differentareas (Hendrix & Glaser, 2007), and there is a great deal ofuncertainty over what future changes may look like and thespecic response required. Regardless, it seems that domesticinstitutions become the key to the solution of this puzzle.

Our results are at the level of nation-states. This is generally anappropriate level of analysis for studying governances inuenceon water resource distribution and internal conicts. However,given the widely recognized fact of sub-national and regionalvariations in water resource availability (see Feitelson &Fischhendler, 2009), future studies should incorporate thedegrees of variation across a nation in the range of water resourceavailability as well as links to other states. In fact, case studies atthe community level support our main argument and suggest thatthe environmental conict literature needs to further explore andunderstand the role of institutions in shaping the pathways toviolent conict in the context of environmental disputes (Bogale &Korf, 2009; Turner, 2004). Thus, greater attention to variationswithin countries would also highlight the role of the quality ofgovernance, as a well-governed state would be more capable todeal with these differences and the conict-generating possibili-ties. The presence of democratic institutions by itself might not beenough to mitigate environmentally-related conict; the quality ofgovernance can be important, in particular at the local level.Evidence from the resource-related conict in New Caledoniasuggests that conicts that are often considered to be related toenvironmental factors might also be related to political factors,such as legitimacy and responsiveness to the needs of localcommunities (Horowitz, 2009).

We expect that our results here are conservative estimates giventhe national aggregate nature of the data used for water resourcesand other factors. Likewise, we utilize the commonly acceptedavailable measure of water resources availability per capita, but onethat reects mostly physical scarcity rather than human-inducedscarcity. Hence, the fact that we are able to identify both direct andindirect links among our main dependent variables at the nationallevel leads us to believe that these relationships may well be evenstronger at the sub-national level and with improved measures ofwater resource availability.

Conclusions

Our main argument focuses on the role of political and institu-tional factors e primarily, governance e and the ability ofgovernments to mitigate or exacerbate water scarcity and itsimpact on conict. Overall, we nd evidence that supportsa nuanced resource abundance argument. But this is only part ofthe story, as the net effect of water availability on the probability ofconict through the interaction between democracy and waterresources suggests that institutions mediate the impact of envi-ronmental factors on the onset of conict. Democratic state insti-tutions reduce the likelihood of intrastate conict; democraticinstitutions are also more likely present in countries with greater

T.-I. Gizelis, A.E. Wooden / Politavailability of freshwater resources per capita. Thus, regime typeRural population density (lag)

ACountries included in the an

Algeria ilippinesAngola land

un

uu

BotswanaBrazil i

Cameroon

rn

goi

El Salvador ited States of America

German Federal RepublicGhana

A

S les.

Meas

Interfresh sper cDum l),based xVanh> 25death

Km3

IRWR 100% of tthat lcenteRatioprodu erperce edland land)GDP

Total

1 if ahas econ

cal G1970e1980/0otherwise

Peace Years # of consecurtive yearssince the 1980

Calculatedbased on UCDP, 2008ict in the period

xperiencedPrevious Conict history country UCDP, 2008

DevelopmentPopulation Population World Bank

DevelopmentRural population density World Bank

DevelopmentDevelopment per capita World Bank

(CPI/%irrigated

ntage of irrigat Indicators/FAO

ction Index ov DevelopmentAgricultural Productivity of crop World Bank

rs Indicators

ives in urban DevelopmentUrbanization otal population World Bank

Dependency ratio /(IRWRIWR) % FAO, Aquastat

Annual Precipitation FAO, AquastatIntrastate Conict

Internationalized s per yearIntrastate Conict & battle-related UCDP, 2008

anen Index Vanhanen, 2000

on Polity indePolitical regime my Variable (0/ Jaggers & Gurr, 1995

apita [TIRFWR]

water resourceWater availability nally renewable FAO, AquastatVariable urement Sourceummary of included variabppendix IIIGuatemala

GreeceFinland Pakistan ZambiaFrance Panama ZimbabweGabon ParaguayGambia PeruEthiopiaFijiNetherlands UnNew Zealand UnNicaragua UrNigeria VeuguaynezuelaEcuadorEgyptMyanmar TuNepal Uganda

ited KingdomCyprusDenmarkMorocco TunisiarkeyColombiaCongoCosta RicaMauritius ToMexico Tr nidad and TobagoChile Mali TaMalta ThzaniaailandChad Malaysia Sy ia

CanadaKuwait SuMadagascar SwMalawi SwrinameazilanditzerlandBurkina FasoBurundiJamaica SoJapan SpJordan SrKenya SuainLankadanBeninBoliviaIvory Coast So

rra Leoneth Africath KoreaBangladeshBelgiumIsrael SeItaly Siedi ArabiaegalAustriaBahrainIndonesia RwIran SamaniaandaArgentinaAustraliaalysis.

Haiti PhHonduras PoHungary PoIndia Rortugalppendix IIT.-I. Gizelis, A.E. Wooden / Politi452References

Ackerman, J., & Tir, J. (2003). To share or not to share: Conditions affecting signing ofriver-managing treaties between Riparian states. Presented at the annual meetingof international studies Association, Portland, Oregon.

Arbetman, M., & Kugler, J. (1997). Political capacity and economic behaviour. Boulder:Westview Press.

Barnett, J., & Adger, W. N. (2007). Climate change, human security and violentconict. Political Geography, 26(6), 639e655.

Barro, R. J. (1999). Determinants of democracy. Journal of Political Economy, 107(S6),S158eS183.

Benson, M., & Kugler, J. (1998). Power parity, democracy, and the severity of internalviolence. Journal of Conict Resolution, 42(2), 196e209.

Bhattarai, M. (2004). Irrigation Kuznets curve, governance and dynamics of irrigationdevelopment: A global cross-country panel analysis from 1972 to 1991. ResearchReport 78. Colombo, Sri Lanka: International Water Management Institute.

Bogale, A., & Korf, B. (2009). To share or not to share? (non-)violence, scarcity andresource access in Somali region, Ethiopia. Development Studies, 43(4), 743e765.

Boserup, E., & Schultz, T. P. (Eds.). (1990). Economic and demographic relationships indevelopment. Baltimore: Johns Hopkins University Press.

Braumoeller, B. F. (2004). Hypothesis testing and multiplicative interaction terms.International Organization, 58(4), 807e820.

Bueno de Mesquita, B., Morrow, J. D., Siverson, R. M., & Smith, A. (1999). Policyfailure and political survival: the contribution of political institutions. Journal ofConict Resolution, 43(2), 147e161.

Bueno de Mesquita, B., & Root, H. L. (2000). When bad economics is good politics. InB. Bueno de Mesquita, & H. L. Root (Eds.), Governing for prosperity (pp. 1e16).New Haven: Yale University Press.

Buhaug, H. (2006). Relative capability and rebel objective in civil war. Journal ofPeace Research, 43(6), 691e708.

Burton, J. (1990a). Conict: Resolution and prevention. New York: St. Martins Press.Burton, J. (1990b). Conict: Basic human needs. New York: St. Martins Press.Chatrchyan, A. M., & Wooden, A. (2005). Linking rule of law and environmental

policy reform in Armenia and Georgia. In CP. M. Waters (Ed.), The state of law inthe South Caucasus (pp. 144e178). London: Palgrave Macmillan.

Choucri, N., & North, R. C. (1975). Nations in conict. San Francisco: W.H. Freemanand Company.

Collier, P. (2000). Doing well out of war: an economic perspective. In M. Berdal, &D. M. Malone (Eds.), Greed & grievance: Economic agendas in civil wars (pp.91e111). Boulder: Lynne Rienner.

Collier, P., & Hoefer, A. (1998). On economic causes of civil war. Oxford EconomicPapers, 50(4), 563e573.

Collier, P., Elliott, V. L., Hegre, H., Hoefer, A., Reynal-Querol, M., & Sambanis, N. (2003).Breaking the conict trap: Civil war and development policy. New York:World Bank.

Crenshaw, E. M. (1995). Democracy and demographic inheritance: the inuence ofmodernity and proto-modernity on political and civil rights, 1965 to 1980.American Sociological Review, 60(3), 702e718.

Critchley, W. H., & Terriff, T. (1993). Environment and security. In R. Shultz,R.Godson., & T. Greenwood (Eds.), Security studies for the 1990s (pp. 327e352).Washington, D.C.: Brasseys.

Diamon, J. (1997). Guns, germs, and steel: The fates of human societies. New York:W.W. Norton & Co.

De Soysa, I. (2002). Ecoviolence: shrinking pie or honey pot? Global EnvironmentalPolitics, 2(4), 1e36.

Ellingsen, T. (2000). Colorful community or Ethnic Witches Brew? Multiethnicityand domestic conict during and after the Cold war. Journal of Conict Reso-lution, 44(2), 228e249.

Eriksson, M., & Wallensteen, P. (2004). Patterns of major armed conicts 1990e2003.In SIPRI Yearbook 2004 armaments, disarmament and international security.Oxford: Oxford University Press. 132e147.

Fay, M., & Opal, C. (1999). Urbanization without growth: A not-so-uncommonphenomenon. World Bank policy research working paper No. 2412. SSRN. http://ssrn.com/abstract632483 Accessed 03.11.08.

Fearon, J. D., & Laitin, D. D. (2003). Ethnicity, Insurgency, and civil War. AmericanPolitical Science Review, 97(1), 75e90.

Feitelson, E., & Fischhendler, I. (2009). Spaces of water governance: the case of Israeland her neighbors. Annals of the Association of American Geographers, 99(4),728e745.

Food and Agriculture Organization of the United Nations Water Resources, Devel-opment and Management Service. (2001). Statistics on water resources bycountry in FAOs AQUASTAT programme. Rome: FAO. Available on-line www.fao.org/ag/aglw/aquastat/water_res/index.htm Accessed 02.11.08.

Furlong, K., & Gleditsch, N. P. (2003). The boundary dataset: description anddiscussion. Conict Management and Peace Science, 20(1), 92e117.

Galtung, J. (1996). Peace by peaceful means: Peace, conict, development and civili-zation. Thousand Oaks: Sage Publications.

Gleditsch, N. P. (1998). Armed conict and the environment: a critique of theliterature. Journal of Peace Research, 35(3), 381e400.

Gleditsch, N. P., Owen, T., Furlong, K., & Lacina, B. A. (2004). Conicts over sharedrivers: Resource wars or Fuzzy boundaries presented at the 45th annual Conven-tion of the international studies association. Montreal.

Gleditsch, N. P., & Sverdrup, B. O. (2002). Democracy and the environment. In

eography 29 (2010) 444e453M. Redclift, & E. A. Page (Eds.), Human security and the environment (pp. 45e70).Cheltenham: Edward Elgar.

Gleditsch, N. P., & Urdal, H. (2002). Ecoviolence? Links between population-growth,environmental scarcity and violent conict in Thomas Homer-Dixons work.Journal of International Affairs, 56(1), 283e302.

Gleditsch, N. P., Wallensteen, P., Eriksson, M., Sollenberg, M., & Strand, H. (2002).Armed conict 1946e2001: a new dataset. Journal of Peace Research, 39(5),615e637.

Gleick, P. H. (1993). Water and conict: freshwater resources and internationalsecurity. International Security, 18(1), 79e112.

Goklany, I. (2002). Comparing twentieth century trends in US and global agricul-tural water and land use. Water International, 27(3), 321e329.

Grossman, G. M., & Krueger, A. B. (1994). Economic growth and the environment.Working Paper No. 4634. Cambridge: National Bureau of Economic Research.

Hauge, W., & Ellingsen, T. (1998). Beyond environmental scarcity: casual pathwaysto conict. Journal of Peace Research, 35(3), 299e317.

Hauge, W., & Ellingsen, T. (2001). Causal pathways to peace. In P. F. Diehl, &N. P. Gleditsch (Eds.), Environmental conict (pp. 26e57). Westview Press.

Hegre, H., & Austvoll Nome, M. (2009). Disentangling democracy and development asdeterminants of armed conict. Typescript, University of Oslo.

Hegre, H., Ellingsen, T., Gates, S., & Gleditsch, N. P. (2001). Toward a democratic civilpeace? Democracy, political change, and civil war 1816e1992. American PoliticalScience Review, 95(1), 16e33.

Lomborg, B. (2001b). Resource constraints orabundance? InP. F. Diehl, &N. P. Gleditsch(Eds.), Environmental conict (pp. 125e154) Boulder: Westview Press.

Lowi, M. R. (1995). Water and power: The politics of a scarce resource in the Jordanriver basin. Cambridge: Cambridge University Press.

Matthew, R. A. (2000). Environment and security in an international context:critiquing a Pilot study from NATOs committee on the challenges of modernsociety. Environmental Change and Security Project Report (6), 95e98.

Matthew, R. A., & Gaulin, T. (2001). Conict or cooperation? The social and politicalimpacts of resource scarcity on small Island states. Global Environmental Politics,12, 48e70.

Midlarsky, M. I. (1998). Democracy and the environment: an empirical assessment.Journal of Peace Research, 35(3), 341e361.

Midlarsky,M. I. (2001). Democracyand the environment. In P. F. Diehl, &N. P. Gleditsch(Eds.), Environmental conict (pp. 155e178). Boulder: Westview Press.

Miguel, E., Satyanath, S., & Sergenti, E. (2004). Economic shocks and civil conict: aninstrumental variables approach. Journal of Political Economy, 112(4), 725e753.

Molle, F., & Mollinga, P. (2003). Water poverty indicators: conceptual problems andpolicy issues. Water Policy, 5(5), 529e544.

Muller, E. N., & Weede, E. (1990). Cross-national variation in political violence:a rational action approach. Journal of Conict Resolution, 34(4), 624e651.

Przeworski, A., Alvarez, M. E., Cheibub, J. A., & Limongi, F. (2000). Democracy and

T.-I. Gizelis, A.E. Wooden / Political Geography 29 (2010) 444e453 453Hegre, H., & Sambanis, N. (2006). Sensitivity analysis of the empirical literature oncivil war onset. Journal of Conict Resolution, 50(4), 508e535.

Hendrix, C. S., & Glaser, S. M. (2007). Trends and triggers: climate, climate changeand civil conict in sub-Saharan Africa. Political Geography, 26(6), 695e715.

Hensel, P. R., McLaughlin Mitchell, S., & Sowers, T. E., II (2004). Conict managementof Riparian disputes: A regional comparison of dispute resolution. Paper presentedat the 2004 Annual Meeting of the International Studies Association, Montreal,Canada.

Hibbs, D. A., & Olsson, O. (2004). Geography, biogeography, and why some countriesare rich and others are poor. Proceedings of the National Academy of Sciences, 101,3715e3720.

Homer-Dixon, T. F. (1991). On the threshold: environmental changes as causes ofacute conict. International Security, 16(2), 76e116.

Homer-Dixon, T. F. (1994). Environmental scarcities and violent conict: evidencefrom cases. International Security, 19(1), 5e40.

Homer-Dixon, T. F. (1999). Environment, scarcity, and violence. Princeton, N.J.:Princeton University Press.

Horowitz, L. S. (2009). Environmental violence and crises of legitimacy in NewCaledonia. Political Geography, 28, 248e258.

Jackman, R. W. (1993). Power without force: The political capacity of nation-states.Ann Arbor: The University of Michigan Press.

Jaggers, K., & Gurr, T. R. (1995). Tracking democracys third wave with the polity IIIdata. Journal of Peace Research, 32(4), 469e482.

Kaufmann, D., Kraay, A., & Mastruzzi, M. (2007). Governance matters VI: Governanceindicators for 1996e2006. World Bank Policy Research Working Paper No 4280.Washington, D.C.: World Bank.

Katz, D. L. (1998). Water, economic growth, and conict: Three studies. Ph.D.Dissertation. University Of Michigan.

Lacina, B. (2006). Explaining the severity of civil wars. Journal of Conict Resolution,50(2), 276e289.

Lake, D. A., & Baum, M. A. (2001). The invisible hand of democracy: political controland the provision of public services. Comparative Political Studies, 34(6),587e621.

Le Billon, P. (2001). The political ecology of war: natural resources and armedconicts. Political Geography, 20(5), 561e584.

Lederer, K., & Galtung, J. (Eds.). (1980). Human needs: A contribution to the currentdebate. Cambridge: Oelenschlager Press.

Lipset, S. M. (1960). Political man: The social bases of politics. Garden City, NY:Anchor.

Lomborg, B. (2001a). The skeptical environmentalist: Measuring the real state of theworld. Cambridge: Cambridge University Press.development: Political institutions and well-being in the world, 1950e1990.Cambridge: Cambridge University Press.

Rock, M. T. (1998). Freshwater use, freshwater scarcity, and socioeconomic devel-opment. The Journal of Environment and Development, 7(3), 278e301.

Sachs, J. (2005). The end of poverty: Economic possibilities for our time. New York:Penguin Press.

Schwartz, D. M., Deligiannis, T., & Homer-Dixon, T. F. (2000). The Environment andviolent conict: a response to Gleditschs critique and some suggestions forfuture research. Environmental Change and Security Project Report, Issue, 6,77e94.

Shak, N., & Bandopadhyay, S. (1992). Economic growth and environmental quality:Time series and cross-country evidence. Background Paper for the WorldDevelopment Report. Washington, DC: World Bank.

Simon, J. (1996). The ultimate resource (3rd ed.). Princeton: PrincetonUniversity Press.Thompson, W. (1996). Democracy and peace: putting the cart before the horse?

International Organization, 1(50), 141e174.Tir, J., & Diehl, P. F. (2001). Demographic pressure and interstate conict: linking

population growth and density to Militarized disputes and wars. In P. F. Diehl, &N. P. Gleditsch (Eds.), Environmental conict (pp. 58e83). Boulder:Westview Press.

Toset, H. P. W., Gleditsch, N. P., & Hegre, H. (2000). Shared rivers and interstateconict. Political Geography, 19(8), 971e996.

Turner, M. D. (2004). Political ecology and the moral dimensions of Resourceconicts: the case of farmer-herder conicts in the Sahel. Political Geography,23, 863e889.

Uppsala Conict Data Program (UCDP). (2008). UCDP database. Department of Peaceand Conict Research, Uppsala University www.ucdp.uu.se/database Accessed12.08.10.

Urdal, H. (2005). People vs. Malthus: population pressure, environmental degra-dation, and armed conict revisited. Journal of Peace Research, 42(4), 417e434.

Vanhanen, T. (2000). A new dataset for measuring democracy, 1810e1998. Journalof Peace Research, 37(2), 251e265.

Vasquez, J. A. (1993). The war puzzle. Cambridge: Cambridge University Press.Wittfogel, K. A. (1957). Oriental despotism, a comparative study of total power. New

Haven: Yale University Press.White, M., Andrzejewski, Awusabo-Asare, K., Kumi-Kyereme, A., Nixon, S. W.,

Buckley, B., et al. (2009). Urbanization and environmental quality: insights fromGhana on sustainable policies. In A. de Sherbinin, A. Rahman, A. Barbieri,J. C. Fotso, & Y. Zhu (Eds.), Urban population-environment dynamics in thedeveloping world: Case studies and lessons learned (pp. 153e180). Paris: CICRED.

Wolf, A. T. (1995). Hydropolitics along the Jordan river: Scarce water and its impact onthe Arab-Israeli conict. New York: United Nations University Press.

Water resources, institutions, & intrastate conflictIntroductionWater scarcity and conflict potentialThe mediating role of domestic institutionsThe nexus between water scarcity, domestic institutions, and conflict

A model of environmental scarcity, governance, and conflictOperationalizationWater availability variableEndogenous variablesConflictGovernance

Exogenous variablesEnvironmental variablesStructural variables

Empirical analysisSimultaneous equation modelsBaseline modelInteractive term model

DiscussionConclusionsAcknowledgmentsAppendix IAppendix IIAppendix IIIReferences