COMMENTARY: DEBATING ENVIRONMENT ... Report6-5.pdfThe editors of the ECSP Report wish to thank Tad Homer-Dixon, Dan Schwartz, Tom Deligiannis, Paul Diehl, Nils Petter Gleditsch, Leo

77ENVIRONMENTAL CHANGE & SECURITY PROJECT REPORT, ISSUE 6 (SUMMER 2000)

Commentary

COMMENTARY: DEBATING ENVIRONMENT,POPULATION, AND CONFLICT

The environment, population, and conflict thesis remains central to current environment and security debates. During the1990s, an explosion of scholarship and policy attention was devoted to unraveling the linkages among the three variables.While it can easily be argued that both the research and policy communities have made significant advances, the scholarlyfindings and policy lessons remain the subject of intense debate. The recent publication of a host of significant contributions tothis debate dictated a special commentary section to supplement the lengthy book reviews provided in this 2000 issue of theEnvironmental Change and Security Project Report.

In the first article, leading figure Thomas Homer-Dixon and his colleagues from the University of Toronto respond to theprominent critique enunciated by fellow peace researcher Nils Petter Gleditsch from the International Peace Research Institute,Oslo (see box on Gleditsch’s critique). Richard Matthew of the University of California, Irvine, comments on the five-yearNATO Committee on the Challenges of Modern Society pilot study entitled Environmental Security in an InternationalContext. Geoffrey D. Dabelko joins Richard Matthew to draw conclusions from a March 2000 environment, population, andconflict workshop with leading scholars. In the last commentary, University of California, Irvine researcher Ted Gaulin brieflycritiques Indra de Soysa and Nils Petter Gleditsch’s To Cultivate Peace: Agriculture in a World of Conflict, portions of whichwere reprinted in issue 4 of the ECSP Report.

The Environment and Violent Conflict:A Response to Gleditsch’s Critique and Some Suggestions

for Future Research

by Daniel M. Schwartz, Tom Deligiannis, and Thomas F. Homer-Dixon

INTRODUCTION

Nils Petter Gleditsch, senior researcher at the International Peace Research Institute, Oslo, has written awidely noted critique of recent research in the new field of environmental security (Gleditsch, 1998).Gleditsch’s critique echoes and builds upon criticisms leveled by skeptics of environment-conflict research(e.g., Deudney, 1991; Levy, 1995; and Rønnfeldt, 1997). He identifies a number of specific “problems” of theory,conceptualization, and methodology, sometimes singling out the work of the team led by Thomas Homer-Dixon ofthe University of Toronto (henceforth referred to as the Toronto Group). In this article, we respond to these con-cerns and propose avenues for future research.

ENVIRONMENTAL CHANGE & SECURITY PROJECT REPORT, ISSUE 6 (SUMMER 2000)78

Commentary

Daniel M. Schwartz and Tom Deligiannis are Ph.D. candidates at the University of Toronto. Thomas F. Homer-Dixonis the Director of the Peace and Conflicts Studies Program at the University of Toronto and author of Environment,Scarcity, and Violence (Princeton University Press, 1999). This article is drawn from the forthcoming November 2000edited volume by Paul F. Diehl and Nils Petter Gleditsch entitled Environmental Conflict (Westview Press). Publishedwith permission of the publisher (Copyright WestviewPress). The editors of the ECSP Report wish to thank Tad Homer-Dixon, Dan Schwartz, Tom Deligiannis, Paul Diehl, Nils Petter Gleditsch, Leo A. W. Wiegman, and Ian Gross for theirassistance.

Norwegian peace researcher Nils PetterGleditsch makes a nine-point critique of envi-ronment, population, and conflict literature inhis seminal 1998 article, “Armed Conflict andthe Environment: A Critique of the Literature.”These nine points, summarized below, spurredThomas Homer-Dixon and his colleagues at theUniversity of Toronto to pen the response pub-lished here. Gleditsch, senior researcher at theInternational Peace Research Institute, Oslo,(PRIO) maintains the literature has the follow-ing characteristics.

1. There is a lack of clarity over what is meantby “environmental conflict”;

2. Researchers engage in definitional and po-lemical exercises rather than analysis;

3. Important variables are neglected, notablypolitical and economic factors, which havea strong influence on conflict and mediatethe influence of resources and environmen-tal factors;

4. Some models become so large and com-plex that they are virtually untestable;

5. Cases are selected on values of the depen-dent variable;

6. The causality of the relationship is reversed;7. Postulated events in the future are cited as

empirical evidence;8. Studies fail to distinguish between foreign

and domestic conflict; and,9. Confusion reigns about the appropriate level

of analysis.

Nils Petter Gleditsch. “Armed Conflict and the

Environment: A Critique of the Literature.” Journal of

Peace Research Vol. 35, no. 3, 1998: 381-400.

Methodological issues underpin Gleditsch’s critique,and we therefore deal with them in detail. Gleditschasserts that much environment-conflict research is meth-odologically unsound and fails to qualify as “systematicresearch.” He contends it violates the rules of quasi-experimental methodology—used by conventional socialscientists in lieu of true experimental methods that arenot viable for many social scientific inquiries. This per-spective is his starting point for identifying many of thespecific problems in environment-conflict research. Asa result, he disregards the detailed findings of the TorontoGroup, the Swiss-based Environmental Conflicts project(ENCOP), and other research projects that do not meethis standards of evidence. We argue that Gleditsch’s pro-posed approach is a methodological straightjacket thatwould, if widely adopted, severely constrain research inthe field. We do not take issue with the quasi-experi-mental methodology per se. Rather, we show that thecase-study method used by the Toronto Group has quali-ties that complement quasi-experimental methods.

In Section One, we address some of the conceptualand theoretical “problems” identified by Gleditsch anddiscuss his selective critique of the literature on the rela-tionship between environmental scarcity and conflict.Gleditsch’s critique does not address the validity of thespecific findings that emerged from ENCOP and theToronto Group. Instead, he treats these projects with abroad brush, at times associating them with other, lessrigorous, research. Section two examines underlyingmethodological issues and addresses Gleditsch’s concernsarising from his methodological perspective. The finalsection of the article looks forward and suggests avenuesfor future research on the environment-conflict nexus.

I. CONCEPTUAL AND THEORETICAL ISSUES

Gleditsch identifies a number of common “prob-lems” with the literature on environmental stress andconflict. This section responds to conceptual and theo-retical criticisms aimed explicitly at the Toronto Group’sresearch.

B A C K G R O U N D


Commentary

Employing a Comprehensive Definition of ScarcityDisputes among scholars about how to conceptual-

ize environmental stress have long hindered research onthe links between this stress and violent conflict. Essen-tially, these are disputes about the delineation of theindependent, or causal, variable. Gleditsch faults muchof the literature for being “unclear as to whether thecausal factor is absolute resource scarcity or environ-mental degradation.” He criticizes Homer-Dixon’sconcept of environmental scarcity—which integrates sup-ply, demand, and distributional sources of the scarcityof renewable resources—suggesting it “muddies the wa-ters,” although he fails to explain why (Gleditsch, 1998:387).

Following Stephan Libiszewski, Gleditsch adopts adistinction between conflicts that result from “simpleresource scarcity” and those that result from “environ-mental degradation” (Libiszewski, 1992). Unfortunately,however, Libiszewski’s distinction is a wholly inadequatestarting point for research on the environmental causesof violence. First, as Gleditsch himself acknowledges,the two categories are not causally separate: degradationof an environmental resource, like cropland or freshwater supplies, can cause a straightforward—or “simple”scarcity of that resource. Second, degradation of an en-vironmental resource is only one of two possible sourcesof a decrease in a resource’s supply. “Degradation” refersto a drop in the quality of the resource; but cropland,fresh water, and the like can also be “depleted,” whichmeans the resource’s quantity is reduced. If we restrictour analysis to conflicts caused by degradation of envi-ronmental resources, we will omit a main source of thereduced supply of these resources in many poor coun-tries around the world.

Third, environmental degradation, the phenomenonGleditsch wants us to emphasize, is exclusively a sup-ply-side problem: if we degrade a resource, then there isless of it available. Any hypothesis linking environmen-tal degradation to violence is linking, essentially, thereduction in the resource’s supply to violence. However,if we want to explore the causes of violence, a resource’sabsolute supply is not interesting. What we should in-vestigate, rather, is the resource’s supply relative to, first,demand on the resource, and, second, the social distri-bution of the resource. The relationships between supplyand demand and between supply and distribution de-termine people’s actual experience of scarcity, and underany practical hypothesis, it is these relationships thatinfluence the probability of violence. This is the reasonthat we include demand and distributional aspects inour definition of environmental scarcity.1

Fourth and finally, focusing on environmental deg-

This article— “The Environment and Violent Con-flict: A Response to Gleditsch’s Critique andSome Suggestions for Future Research” byDaniel M. Schwartz, Tom Deligiannis, and Tho-mas Homer-Dixon—is drawn from the forthcom-ing November 2000 edited volume by Paul F.Diehl and Nils Petter Gleditsch entitled Environ-mental Conflict (Westview Press). This volumepromises to make a significant contribution tothe environment, population, and conflict litera-ture. Following, is the table of contents for Envi-ronmental Conflict:

Chapter 1. Controversies and QuestionsPaul F. Diehl & Nils Petter GleditschChapter 2. The Case of South AfricaVal Percival & Thomas Homer-Dixon

Chapter 3. Causal Pathways to ConflictWenche Hauge & Tanja Ellingsen

Chapter 4. Demographic Pressure andInterstate Conflict

Jaroslav Tir & Paul F. DiehlChapter 5. Demography, Environment, and

SecurityJack Goldstone

Chapter 6. Water and ConflictSteve Lonergan

Chapter 7. Resource Constraints orAbundance?

Bjørn LomborgChapter 8. Democracy and the Environment

Manus MidlarskyChapter 9. The Limits and Promise of

Environmental Conflict PreventionRodger Payne

Chapter 10. The Spratly Islands ConflictDavid Denoon & Steven Brams

Chapter 11. Environmental Cooperation andRegional Peace

Ken ConcaChapter 12. Armed Conflict and the

EnvironmentNils Petter Gleditsch

Chapter 13. The Environment and ViolentConflict

Daniel M. Schwartz, Tom Deligiannis &Thomas Homer-Dixon

To order a copy of this book, please contactWestview Press at:

Westview PressPerseus Books Group Customer Service

5500 Central AvenueBoulder, CO 80301

(800) 386-5656; Fax: (303) 449-3356E-mail: [email protected]


Commentary

radation alone tends to lead researchers to overlook orneglect key interactions—such as the processes we callresource capture and ecological marginalization—amongsupply, demand, and distributional pressures (Homer-Dixon, 1999: 73-80). Resource capture occurs when thedegradation and depletion of a renewable resource (adecrease in supply) interacts with population growth (anincrease in demand) to encourage powerful groupswithin a society to shift resource access (that is, to changethe resource’s distribution) in their favor. These groups

tighten their grip on the increasingly scarce resource anduse this control to boost their wealth and power. Re-source capture intensifies scarcity for poorer and weakergroups in society. Ecological marginalization occurs whenunequal resource access (skewed distribution) combineswith population growth (an increase in demand) to causelong-term migrations of people to ecologically fragileregions such as steep upland slopes, areas at risk of de-sertification, tropical rain forests, and low-quality pub-lic lands within urban areas. High population densitiesin these regions, combined with a lack of knowledgeand capital to protect the local ecosystem, cause severeresource degradation (a decrease in supply) (Homer-Dixon, 1999: 177). Resource capture and ecologicalmarginalization are often intimately inter-linked, withone leading to the other.

Some might argue that by including distributionalissues in our definition of environmental scarcity, theToronto Group makes the concept so broad as to beuseless, because the group classes conflicts solely overresource distribution as environmental conflicts.2 Theargument is misguided. Uneven distribution never actson its own: its impact is always a function of its interac-tion with resource supply and demand. In practicalterms, the reason resource distribution is important isthat the resources people want (that are, in other words,in demand) are in finite supply. Indeed, the TorontoGroup found in its research that problems of decliningresource supply and rising resource demand were alwaysintimately entangled with uneven resource distribution.

For these four reasons, an exclusive focus on envi-ronmental degradation in environment-conflict researchunreasonably restricts, distorts the scope of the research,

and misses crucial aspects of the environmental chal-lenges facing the developing world. It is better, we believe,to acknowledge explicitly that the fundamental issue isone of scarcity of renewable resources and that any treat-ment of this scarcity should encompass the exhaustiveset of scarcity’s sources: decreases in supply, increases indemand, and changes in distribution. The TorontoGroup incorporates these three facets of scarcity in itstripartite definition of scarcity.

Challenging Simonesque OptimismAt a more fundamental level, Gleditsch questions

the very idea that humanity is facing increasing envi-ronmental scarcities. His critique seems to be guided bythe assumption that the links between environmentalscarcity and violence are overstated, because humanityshows astonishing capacity to adapt to scarcities(Gleditsch, 1998: 383-384 and 395). Markets stimu-late human inventiveness and commerce that open upnew sources of scarce resources, encourage conservation,and create technologies that allow substitution of rela-tively abundant resources for scarce ones. These adaptiveprocesses certainly operate in many cases, as we havepreviously noted (Homer-Dixon, 1995; Homer-Dixon,1999: 31-5 and 107-32). But Gleditsch does not ac-knowledge that societies often fail to adequately adjustto scarcity, with poverty, migrations, and institutionalfailure the result. Environmental scarcities unquestion-ably have profoundly debilitating effects on someeconomies, societies, and social groups.3 Just becausehumans are remarkably adaptive in some cases does notmean that they are always adaptive.

Gleditsch seems particularly influenced by JulianSimon’s cornucopian thesis that, based on the historicalrecord, human societies can bring to bear on their re-source scarcities sufficient ingenuity to prevent anydecline in well-being over the long run (Simon, 1996).But Simon’s techno-optimist arguments are too simplis-tic, for three reasons. First, he tends to project the trulyextraordinary improvements in human well-being overthe past two centuries linearly into the future, withoutmuch questioning or reflection. Yet, if we look back fur-ther than two hundred years, it is clear that human affairs

“We argue that Gleditsch’s proposed approach is a methodologi-cal straightjacket that would, if widely adopted, severely con-strain research in the field.”


Commentary

have my been marked by many ‘non-linear’ events—sudden, sharp changes in economic and socialbehavior—some of which have had decidedly negativeeffects on human well-being. The progress of the lasttwo centuries is not the only evidence we should use toestimate our trajectory into the future.

Second, when Simon analyzes trends in human well-being, he usually uses highly aggregated data, such asstatistical averages for all of humankind. Yet, when thesedata are disaggregated—that is,broken into sub-categories—Simon’s optimism is lesspersuasive. For example, althoughboth the percentage and absolutenumber of hungry people havefallen globally in the last twentyyears, Latin America, South Asia,and especially sub-Saharan Africahave not seen reductions in theabsolute number of hungrypeople.4 Third and finally, closestudy of specific cases shows thatsocieties do not always generate theingenuity they need when andwhere it is needed (Lele and Stone, 1989). Althoughenvironmental and demographic stress often drives upthe requirement for ingenuity in poor countries, a num-ber of factors—including market failure, inadequatehuman capital, and political competition over scarcenatural resources among powerful groups—can impedethe flow of ingenuity (Homer-Dixon, 1999: 107-32).

Homer-Dixon and others have responded with con-siderable theoretical and empirical detail to thearguments of Simon and other techno-economic opti-mists (Homer-Dixon, 1999: 28-44; Cohen, 1995;Ahlburg, 1998). Nowhere does Gleditsch acknowledgethese responses. The determinants of adaptation to scar-city are a major outstanding issue for researchers in thisfield. Gleditsch would better advance our understand-ing if he engaged with the various positions on the issue,rather than appearing to accept the arguments of theoptimists at face value.

Bringing Nature into Social TheoryGleditsch’s skepticism about the seriousness of en-

vironmental scarcities is the starting point for a keyelement of his critique of the environment-conflict lit-erature. He argues that the literature overstates the impactof environmental scarcities on violent conflict and, inthe process, ignores other, perhaps more powerful causalvariables. “Far too many analyses of conflict and theenvironment are based on… overly simplistic reasoning,”

he writes. “The greatest weakness in this respect is thatmuch of this literature ignores political, economic, andcultural variables” (Gleditsch, 1998: 389). Bydeemphasizing environmental scarcities, Gleditsch cor-respondingly emphasizes other variables.

This approach implies that environmental stress maybe no more than an intermediate or intervening vari-able between dysfunctional political and economic in-stitutions and conflict. Thus, Gleditsch asks if

environmental conflict “may beprimarily an underdevelopmentproblem,” because environ-mental degradation or “load” isstrongly correlated with poverty(Gleditsch, 1998: 396). Heseems to argue that conflict indeveloping countries is best ex-plained by social causes, not bythe physical influences of thenatural environment. In theprocess, like many scholars ofcomparative development,Gleditsch marginalizes thephysical circumstances of hu-

man society as explanatory variables; he appears to con-sider them to be, at most, secondary causes of socialbehavior. When it comes to violent conflict, they aremerely aggravators of already existing social stresses. Ifthis is his position, Gleditsch is making a classicendogeneity mistake: he is claiming that environmentalproblems are a consequence of, and endogenous to, thebroader social system and that, therefore, any conflictcaused by environmental problems is ultimately causedby social factors.

It is unquestionably true that social variables mustbe central to any adequate explanation of human con-flict, whether in rich or poor countries. The TorontoGroup discusses at length the political, economic, andcultural factors that interact with environmental scar-city to cause violence. The societies most vulnerable toenvironmentally-induced violence are those simulta-neously experiencing severe environmental scarcity andvarious forms of institutional failure (especially failuresof states and markets) that hinder social adaptation tothe scarcity. The key role of social variables must there-fore be acknowledged. However, this requirement doesnot mean that physical variables should be made fullyendogenous to the social system and, consequently,turned into relatively uninteresting secondary causes ofsocial conflict and stress.5

As Homer-Dixon has noted, there are three reasonswhy environmental scarcity should be considered at least

“Environmental scarci-ties unquestionablyhave profoundly debili-tating effects on someeconomies, societies, andsocial groups.”


Commentary

partly an exogenous factor in social behavior and con-flict and why, therefore, environmental scarcity deservesresearch attention in its own right (Homer-Dixon, 1999:16-18 and 104-6). First, environmental scarcity is notonly influenced by social variables like institutions andpolicies; it can itself affect these institutions and poli-cies in harmful ways. This is the case when shortages ofa renewable resource, such as cropland or forests, moti-vate elites to seize control, through either legal or coer-cive means, of the resource’s remaining stocks (resourcecapture). In other words, we should not assume that so-cial variables are completely independent and externalstarting points in the causal chain; it turns out that theycan be affected by environmental scarcity, sometimesnegatively. Second, the degree of environmental scar-city a society experiences is not, as it turns out, wholly aresult of economic, political, and social variables, likefailed institutions and policies; it is also partly a func-tion of the particular physical characteristics of thesociety’s surrounding environment. These characteris-tics are, in some respects, independent of human activi-ties. For example, the vulnerability of coastal aquifers tosalt intrusion from the sea and the depth of upland soilsin tropical regions are physical “givens” of these envi-ronmental resources. Third, once environmental scar-city becomes irreversible—as when most of a country’svital topsoil washes into the sea—then the scarcity is,almost by definition, an external influence on society.Even if enlightened reform of institutions and policiesremoves the underlying social causes of the scarcity, be-cause the scarcity itself is irreversible, it will remain acontinuing burden on society.

The claim that environmental scarcity can be, inpart, an exogenous variable, should not be confused withthe claim (which we do not make) that environmentalscarcity can have a direct impact on conflict. We arguethat the link between environmental scarcity and con-flict is most often indirect. Nevertheless, environmentalscarcity can still have an exogenous impact on the socialconditions that eventually lead to conflict.

Identifying Key VariablesGleditsch claims that the Toronto Group and other

researchers overlook important variables like regime typeand democracy. However, the Group’s full model doesintegrate regime-type variables into its analysis of thesocial and economic effects of environmental scarcities.In the scholarly literature on the origins of revolutionsand civil violence, the variables of opportunity structureand state capacity, which are central to the TorontoGroup’s model, are recognized as integral aspects of re-gime type (Goodwin, 1997; Tarrow, 1994; Skocpol,

1979). Furthermore, in his recent work, Colin Kahl ex-plicitly builds on the Toronto Group’s model to furtherour understanding of how regime type affects the linksbetween environmental scarcity and violence (Kahl,1998).

More specifically, however, Gleditsch’s suggestionthat the Toronto Group is blind to the importance ofregime type is, on close reading, actually a call for theinclusion of a democracy variable in environment-con-flict models (Gleditsch, 1998: 389).6 We agree withGleditsch that a more explicit focus on democracy couldbe beneficial—as long as analysts are careful in their useof “democracy.” As Homer-Dixon has argued, “the termdemocracy is used too loosely by lay commentators andexperts alike. It commonly encompasses an extraordi-narily variegated set of social phenomena and institutionsthat have complicated and multiple effects on the inci-dence of social turmoil and violence” (Homer-Dixon,1999: 182).7 Gleditsch deserves credit for advancingenvironmental conflict literature along this importanttheoretical path. If future research can address the diffi-cult issues surrounding the precise definition andoperationalization of democracy, important findings mayyet emerge.8

Using Historical EvidenceFinally, Gleditsch claims that the Toronto Group’s

theory about the links between environmental scarcityand conflict is flawed, in part because it is founded oninferences about future scarcities. Gleditsch asserts that“Homer-Dixon, and many other authors. . .have stressedthe potential for violent conflict in the future” withoutproviding adequate empirical evidence of past or presentlinkages between environmental scarcities and violentconflict (Gleditsch, 1998: 393).9

Gleditsch is mistaken that the Toronto Group uses“the future as evidence” to substantiate its claims thatthere are links between environmental scarcities andconflict. In the process of developing its model, theGroup has undertaken more than a dozen detailed his-torical case studies. These include studies of the Chiapasrebellion, the Rwandan genocide, violence betweenSenegal and Mauritania, civil conflict in the Philippines,and ethnic violence in Assam, India.10 The historicalanalyses in these case studies were informed by the richliteratures on the causes of revolution, insurgency, andethnic strife. Taken together, they are a foundation forthe Toronto Group’s larger theoretical model about link-ages between environmental scarcity and violent conflict.None of the hypotheses in this model depends on eventsyet to come; rather, the model is informed by eventsthat have already taken place.11


Commentary

To support his claim that the Toronto Group usesthe future as evidence for its model, Gleditsch takes is-sue with commentators who argue that water scarcity inthe Middle East could lead to armed conflict in the fu-ture. Without referring to any research in particular, buthaving identified the Toronto Group by name at thebeginning of the paragraph, Gleditsch concludes thatthis is a hypothesis “based on controversial theory anddebatable extrapolations, rather than ‘data’ which mayconfirm the prediction” (Gleditsch, 1998: 394).Gleditsch thus conflates the findings of the TorontoGroup with largely unsubstantiated claims by other writ-ers regarding the potential for conflict over waterresources.

The specific findings of both the Toronto Groupand ENCOP are certainly worthy of detailed consider-ation in any discussion of links between environmentalscarcity and conflict. In this case, Gleditsch did not torefer to the Toronto Group’s thinking on the conse-quences of water scarcity. Had he done so, he wouldhave noted a number of interesting hypotheses worthyof testing. Homer-Dixon argues that the world is notabout to witness a surge of water wars. “Wars over riverwater between upstream and downstream neighbors arelikely only in a narrow set of circumstances,” Homer-Dixon writes. “The downstream country must bethreatening to restrict substantially the river’s flow; theremust be a history of antagonism between the two coun-tries; and, most importantly, the downstream countrymust believe it is militarily stronger than the upstreamcountry. . . .There are, in fact, very few basins aroundthe world where all these conditions hold now or mighthold in the future” (Homer-Dixon, 1999: 139). TheToronto Group’s research on water scarcity is, in fact, atodds with sensationalist claims about water wars.

II. FINDING OUR WAY IN THE WILDERNESS

Underpinning many of Gleditsch’s criticisms aredeeper methodological issues pertaining to the conductof social science inquiry. Gleditsch claims, for instance,that the Toronto Group fails to select cases appropri-ately, neglects to investigate the possibility of reversecausation, devises untestable models, overemphasizes thecomplexity of ecological-political systems, and lacks thetools to weight causal variables. These criticisms can onlybe understood in the context of Gleditsch’s unduly nar-row perspective on what constitutes “systematicresearch.”

In this section, we first show that process-tracingwithin single cases should be an integral part of system-

atic research in the social sciences; this method comple-ments more conventional quasi-experimentalapproaches. Drawing a distinction between causal ef-fects and causal mechanisms, we then show why Gleditsch’scriticisms of the Toronto Group’s research—as identi-fied in the previous paragraph—are unfounded. We alsorecap some of the key findings of the Toronto Groupthat Gleditsch overlooked as a result of his methodologi-cal bias. In short, we show that there are more than a“few lights in the wilderness” to guide future researchinto the relationship between environment and conflict.

Conducting Systematic ResearchGleditsch asserts that scholars have conducted little

systematic research to date on the link between envi-ronmental scarcity and violent conflict (Gleditsch, 1998:384-7). By “systematic” research, he seems to mean ei-ther experimental or quasi-experimental analyses. (Hediscusses statistical analyses and controlled-case com-parisons in particular, but quasi-experimental methodscan include counterfactual analyses and congruence pro-cedures.) Gleditsch additionally contends that pastresearch into the links between environment and con-flict consisted merely of “exploratory case-studies” thatfailed to demonstrate causal connections (Gleditsch,1998: 392).

In our opinion, Gleditsch has an overly circum-scribed view of what counts as systematic research inthe social sciences. Many social science methodologistshave long recognized that systematic research includesnot only experimental and quasi-experimental methods,but single-case methods as well.12 Highly influentialstudies in the social sciences—such as Graham Allison’sEssence of Decision (1971) and Arend Lijphart’s The Poli-tics of Accommodation (1975)—have used single casestudies to build and test theories.13

At issue in this debate over the merits of the case-study method are fundamental ontological andepistemological questions pertaining to the nature ofcausation. Among competing views on how causationcan be demonstrated, philosopher David Hume’s argu-ments remain influential.14 Hume asserted that causationcould be demonstrated only by showing a high degreeof covariance between types of events, which he termedconstant conjunction. Hume’s notion of constant con-junction underpins experimental and quasi-experimentalmethodologies in the social sciences; many researchers,including Gleditsch, appear to believe that it also viti-ates the single-case method.

However, Andrew Bennett (1997) shows convinc-ingly that Hume’s notion of causality underpins not onlyexperimental and quasi-experimental methods but the


Commentary

single case-study method as well. Bennett notes thatHume recognized three “sources of causality,” only oneof which was constant conjunction. The other two weretemporal succession and contiguity. Bennett argues thatwhile constant conjunction is related to what method-ologists term causal effect, temporal succession andcontiguity are related to causal mechanism. The causaleffect of an explanatory variable is defined by Bennettas “the change in probability and/or value of the depen-

dent variable that would have occurred if the explana-tory variable had assumed a different value.” Causalmechanism, on the other hand, is defined as “the causalprocess and intervening variables through which causalor explanatory variables produce causal effects” (Bennett,1997: 18-19). Both causal effect and causal mechanismare therefore essential and complementary facets of cau-sality. While the experimental and quasi-experimentalmethods aim to gauge causal effect, they say little aboutcausal mechanism. The single-case method, conversely,helps reveal causal mechanism but gives little indicationof causal effect. In short, neither the experimental andquasi-experimental nor the single-case method is suffi-cient to demonstrate causation with any finality. It isequally evident, however, that the single-case method isa necessary tool to demonstrate causation.

An example from the natural sciences illustrates thedistinction between causal effect and causal mechanism.Although the correlation between smoking and cancerhas been known for many years, only within the last fiveyears have researchers pinpointed exactly how smokingengenders cancer. That is, the causal effects were alreadyknown, but until recently the causal mechanisms re-mained unknown. The recent identification of thesemechanisms has put the tobacco industry on the defen-sive, because they now find it harder to retreat to theclaim that scientific proof is lacking.15

The distinction between causal mechanism andcausal effect is also cogent for the social sciences. Timo-thy McKeown notes that only by distinguishing betweencausal effect and causal mechanism can one begin to

understand why Allison’s Essence of a Decision andLijphart’s The Politics of Accommodation had such mo-mentous impact on the field of political science. Bothseriously challenged long-standing theories: Allison’sanalysis of decision-making during the Cuban missilecrisis undermined the notion of the state as a unitary,rational actor; and Lijphart’s analysis of politics in theNetherlands challenged prevailing ideas about the im-pact of political cleavages. The important processes these

authors identified in their case studies would have beenoverlooked in a statistical analysis. McKeown (1999:172-174) asserts these case analyses had a large impactprecisely because they highlighted how events unfoldedby identifying their causal mechanisms.16

Several leading philosophers of science have madesimilar points. Wesley Salmon (1984: 121), for example,argues in favor of explicating causal mechanisms: “Themere fitting of regularities into patterns has little, if any,explanatory force.” Andrew Sayer (1992: 106-7) statesthat “what we would like… is a knowledge of how theprocess works. Merely knowing that C has generally beenfollowed by E is not enough: we want to understand thecontinuous process by which C produced E…” AndAbraham Kaplan (1964: 329) asserts that “we see betterwhy something happens when we see better—in moredetail, or in broader perspective—just what does hap-pen.”17

Bennett notes that the distinction between causaleffect and causal mechanism has prompted a debateamong methodologists about which of these two sourcesof causality is more important. Although some analystssuggest that causal effects are “logically prior to the iden-tification of causal mechanisms” (King, Keohane, andVerba, 1994: 86), others insist that causal mechanismsare “ontologically prior” to causal effect (Yee, 1996: 84).Bennett dismisses this controversy, arguing that “cau-sality involves both causal effects and causal mechanismsand its study requires a diversity of methods, some ofwhich are better adapted to the former and some to thelatter” (Bennett, 1997: 25).

“[A]n exclusive focus on environmental degradation in envi-ronment-conflict research unreasonably restricts, distortsthe scope of the research, and misses crucial aspects of the environ-mental challenges facing the developing world.”


Commentary

Bennett’s reluctance to confer priority on eithercausal effect or causal mechanism, however, does notsuggest that the identification of one should not pre-cede the identification of the other in terms of thepractical task of puzzle solving in the social sciences.Indeed, when a research program is in its early stagesand the underlying theory is still largely undeveloped,

focusing first on causal mechanisms is probably the beststrategy. Once researchers have discovered these causalmechanisms and elaborated the theory, they can thenbegin to estimate causal effects.18 Thus George andMcKeown emphasize the role that single-case methods(involving process-tracing) can play in the developmentof theory (George and McKeown, 1985: 34-41).19

With this methodological underpinning, theToronto Group set out to perform a series of case stud-ies of the causal links between environment scarcity andconflict. Although the possibility of such links had beenrecognized by previous scholarship, theory was rudimen-tary. Using a process-tracing approach, the TorontoGroup conducted over a dozen case studies to betterunderstand the causal mechanisms that might connectenvironmental scarcity to conflict. The results producedby the Toronto Group reflect the methodology used:the Group does make general claims about causal mecha-nisms (for example, at the end of his second InternationalSecurity article on the subject, Homer-Dixon says ex-plicitly that “environmental scarcity causes violentconflict”), but it has been careful to avoid making suchclaims about causal effects (nowhere in the Group’s re-search reports are there any claims about the power ofenvironmental scarcity relative to other potential causesof conflict).20

Without undertaking research into causal mecha-nisms, estimates of causal effect are far less illuminating,for two reasons. First, researchers will not know whichpotentially confounding variables they should controlin their statistical tests; and, second, researchers mayoverlook key processes and causal relationships that arehidden in the data. In quasi-experimental methods of

social science, it is impossible to control all variablesthat may affect the dependent variable under study;therefore, researchers must pick and choose their con-trol variables carefully. Process-tracing helps identifythose particularly worthy of control.21 Also, process-tracing reveals variables and causal patterns that maynot emerge from statistical analysis. For instance, the

patterns of ecological marginalization and resource cap-ture, which were discovered by the Toronto Group usingprocess-tracing, are not obvious and would undoubt-edly have remained hidden from statistical analysis. TheGroup’s research suggests, however, that quasi-experi-mental and statistical methods should now be used toinvestigate these patterns.22 This more inclusive under-standing of systematic research helps us address fivefurther concerns raised by Gleditsch about contempo-rary environment-conflict research: selection of cases onthe independent and dependent variables; failure to con-sider that the dependent variable may in fact be animportant cause of the independent variable; a propen-sity to develop untestable models; overemphasis on thecomplexity of ecological-political systems; and an in-ability to gauge the relative power of environmentalscarcity as a cause of conflict.

Selecting Case StudiesFollowing Marc Levy (1995) and Carsten Rønnfeldt

(1997), Gleditsch contends that choosing cases in whichboth environmental scarcity and violent conflict wereknown a priori to exist, violates a fundamental principleof research design that applies to both qualitative andquantitative analyses. Consequently, Gleditsch(1998:391-92) asserts this practice produces nothingmore than “anecdotal evidence” to support its hypoth-eses.

Gleditsch’s approach to research design appears tohinge on the assumption that causality is little more thancausal effect. Causal mechanism is regarded as less im-portant or is simply not considered at all. Although weagree that researchers must allow for variation on both

“The societies most vulnerable to environmentally-induced vio-lence are those simultaneously experiencing severe environmen-tal scarcity and various forms of institutional failure… that hindersocial adaptation to the scarcity.”


Commentary

the independent and dependent variable if they want toestimate casual effect, we contend researchers will findsuch an approach less helpful in identifying causalmechanisms. If causal mechanism is believed to be anintegral aspect of causality, then selecting case studieson the independent and dependent variable is hardly anegregious error in research design. Indeed, in order tounderstand whether there are causal links between envi-ronmental scarcity and violent conflict—and, if thereare, how these variegated links work—it will be some-times necessary to select cases in this manner. TheToronto Group therefore intentionally selected cases inwhich environmental scarcities and violent conflict wereknown a priori to exist (Homer-Dixon, 1999: 169-76).The Group then used process-tracing to determine if theindependent and dependent variables were actually caus-ally linked, and, if they were, to induce from a closestudy of many such cases the common mechanisms ofcausality and the key intermediate variables that charac-terized these links.

A related objection to selecting cases on both theindependent and dependent variables is that the re-searcher might as a result, overlook possible confoundingvariables and spurious relationships. The researchermight, for instance, believe that data show a causal linkbetween variables A and B, a link that fits the researcher’shypothesis nicely. But the researcher might fail to lookfor variable C, a variable that is linked to both A and Band is a cause of both. For example, environmental scar-city might appear to be a cause of conflict, but, in reality,not be a cause, if poverty is actually a cause of changesin both these variables. This concern, however, is mis-placed, because vigilant case-study researchers shoulddetect such situations. Eckstein (1975: 125-26) contendsthat such researchers can test “countertheories”—thatis, theories about other likely causes of changes in thevalue of the dependent variable.23 Just as the quasi-ex-perimental researcher must anticipate variables tocontrol, the case-study researcher must anticipate po-tentially spurious causal mechanisms.

Investigating Reverse CausationA distinction between causal effect and causal

mechanism helps us address Gleditsch’s concern thatviolent conflict (the dependent variable in most research)may in fact be an important cause of environmental scar-city (the independent variable). Gleditsch (1998: 393-3)claims that environment-conflict researchers have ne-glected this possibility of reverse causation and havelikewise failed to consider the possibility that environ-mental scarcity and violent conflict are related to eachother in a positive feedback loop—that is, a vicious circle.

We do not deny that conflict may exacerbate envi-ronmental scarcity, but this possibility was not the focusof the Toronto Group’s research. Nevertheless, we wouldargue that process-tracing offers an excellent way to dis-cover reverse causality, because it unearths causalmechanisms. It allows researchers to trace causal mecha-nisms that unfold over long periods of time and therebyto investigate the impacts of past conflicts on subsequentenvironmental conditions. An approach that focuses oncausal effects, however, cannot reveal reverse causationas easily. Although simultaneous equations can be usedto model reverse causation, and although quasi-experi-mental methods, using lagged variables or congruenceprocedures, can be used to span time, a far more intui-tive approach is to focus on causal mechanisms, becausethey will tell the researcher exactly how past conflictsexacerbated environmental scarcity.

Moreover, the quasi-experimental method can pro-duce ambiguous results when attempting to differentiatebetween cause and effect. Consider the following ex-ample: When a barometer falls, deteriorating weather islikely to follow. Although it precedes the change inweather, the falling barometer clearly does not cause thischange. Thus, we can not distinguish between cause andeffect. If we understand the mechanism that causes thebarometer to fall, however, we understand that causeand effect can only be differentiated once weather con-ditions prior to the barometer’s fall are controlled (Miller,1987: 34).

Constructing Testable ModelsWhile Gleditsch contends that much of the envi-

ronment-conflict literature to date is overly simplistic,he asserts that the Toronto Group is guilty of just theopposite mistake—that is, of developing overly com-plex models that are not testable (1998: 391-92). Webelieve that Gleditsch contradicts himself here by de-manding a strict adherence to conventional researchdesign while simultaneously agitating for an incremen-tal and modular approach to theory building.Conventional research design forbids the omission ofvariables that are correlated with the key independentvariable. Such an omission creates what Gary King,Robert Keohane, and Sidney Verba (1994: 168-176)term an omitted variable bias. Many of the variables con-sidered by the Toronto Group are correlated with thekey independent variable of environmental scarcity. IfGleditsch is suggesting that we drop these variables outof the equation in the name of testable models, he isalso suggesting that we contravene a fundamental canonof conventional research design.

Since the Toronto Group did not adopt such a re-


Commentary

search design, however, this internal contradiction doesnot directly concern us. Nevertheless, Gleditsch’s agita-tion for less complexity is disturbing. If environmentalscarcity were either a necessary or a sufficient cause ofconflict, it would be possible to reduce our model’s com-plexity. Of course, environmental scarcity is neither anecessary nor sufficient cause (there are few, if any, such

causes of conflict). If, therefore, researchers are to makea nomothetic claim about the relationship between en-vironmental scarcity and conflict, environmental scarcitymust be part of what philosopher J. L. Mackie (1965)terms an INUS condition: it must be an insufficient butnecessary component of a condition that is itself an un-necessary but sufficient cause of conflict.

Discovering INUS conditions is the goal of the case-study researcher. For environmental-conflict researchers,this entails unearthing the myriad and variegated waysin which environmental scarcity interacts with othersocial, economic, and political factors to engender con-flict. We do not mean to suggest that a process-tracingapproach eclipses the important goal of parsimony.Rather, by focusing on relevant causal mechanisms, pro-cess-tracing helps the environment-conflict researcherdetermine the boundaries of the INUS condition. With-out a clear picture of these boundaries, simply droppingvariables in the name of parsimony becomes a haphaz-ard affair. Once these boundaries have been defined,however, estimating causal effects becomes a more pre-cise procedure.

Dealing with Complex SystemsAccording to Gleditsch, the Toronto Group claims

that ecological-political systems are more complex thanstrictly social or physical systems. He goes on to arguethat this claim is unwarranted because “any social sys-tem is as complex as the theory developed to study it”(1998: 392). In other words, the complexity is in themind of the beholder, rather than in the phenomenonitself. Actually, the Toronto Group does not argue thatecological-political systems are more complex. They ar-gue simply that these systems are, intrinsically,exceedingly complex. No doubt many social, biologi-

cal, and physical systems are just as complex or evenmore complex (although some unquestionably are not).

The problem of complexity exists in the real world.It cannot be wished away by assuming that it residesonly in the mind of the researcher. Gleditsch’s extraor-dinarily strong constructivist position on this issue isquestionable both empirically and philosophically

(Rescher, 1998). Researchers in a variety of fields in-creasingly acknowledge the reality of complexity and aredeveloping powerful theories to understand complexsystems. These theories raise serious questions aboutconventional (often mechanistic) explanations of socialphenomena and about the conventional methodologiesused to study these phenomena (Cowan, Pines, andMeltzer, 1994). Rather than denying complexity’s exist-ence, Gleditsch and other social scientists shouldexplicitly acknowledge the problems it creates for theirresearch and try to develop methods—such as those fo-cusing on causal mechanisms—for dealing with it.

Weighting Causal VariablesGleditsch implies that process-tracing within single

case-studies does not allow researchers to gauge the rela-tive weights of causal variables (1998: 384-386). He alsosuggests that the quantitative analysis by Wenche Haugeand Tanja Ellingsen (1998) is one of the few attempts totest systematically the relationship between environmen-tal scarcity and conflict. These researchers, he notes, didfind a statistically significant relationship between envi-ronmental degradation and violent conflict, but theyconcluded that economic and political variables weremore important than environmental variables. Thus,Gleditsch implicitly accepts the notion that indepen-dent variables can be assigned weights that indicate theirrelative causal power. Gleditsch, of course, is hardly alonehere. Causal weighting is widely considered to be theultimate goal of statistical analyses, and the lack of abil-ity to weight variables using single case studies isconsidered this method’s foremost drawback.

The practice of causal weighting, however, has itsproblems. Elliott Sober (1988) contends that the stan-dard statistical technique of analysis of variance

“[T]he Toronto Group and ENCOP, among others, is not onlytheoretically and conceptually intact, but also rests on soundmethodological pillars.”


Commentary

(ANOVA) does not actually yield causal weights. Rather,it identifies the difference that various causes can makein an observed effect. Ascertaining the difference, Sobermaintains, is distinct from ascertaining a causal weight.Similarly, Richard Lewontin (1976) argues that, althoughcausal weighting may be appropriate when the relation-ships among variables are additive, it is misguided whenthe relationships are interactive. Lewontin contends thatanalysis of variance produces uninterpretable resultswhen dealing with interactive variables.

If environmental scarcity is one component of anINUS condition, as argued above, then environment-conflict researchers are not dealing with additive relationsamong causal variables. Rather, these relationships areinteractive. Environmental scarcity, for example, inter-acts with a society’s ability to supply social and technicalingenuity. If the society can supply abundant ingenuityin response to its environmental problems, then severesocial disruptions will probably be avoided; if it cannot,then negative outcomes, including conflict, are muchmore likely.

Interactivity is hardly limited to the relationshipsamong variables in ecological-political systems. Mostsocial systems exhibit interactivity among variables. Thatso many researchers treat the relationships among vari-ables in social systems as additive does not reflect thereality of these systems. Rather, it reflects misguided at-tempts by researchers to avoid dealing with the realityof the complexity of these systems.

III. SUGGESTIONS FOR FUTURE RESEARCH

In Gleditsch’s final section, entitled “The WayAhead,” he asserts that “critique will serve to advancethe field only if it stimulates more satisfactory research”(Gleditsch, 1998: 395). Although we do not agree thatall work on environment and conflict has been unsatis-factory, we do agree that debates in the field, such as theone we are engaged in here, can provide the spark fornew research agendas. In this spirit, we draw on the aboveremarks to make some suggestions for future work. Thesesuggestions fall in five categories: filling data gaps,operationalizing key variables, specifying contextual fac-tors, dealing with complexity, and encouragingmethodological pluralism.

Filling Data GapsWe agree with Gleditsch that serious data gaps im-

pede research on the links between environment andconflict. There is a particular lack of good data on the

extent and degree of soil, water, and forest degradationin developing countries; data on resource distributionand resource-use practices are also poor. The field there-fore needs a more systematic and rigorous approach todata collection. Because this research crosses so manydisciplinary boundaries, systematic data collection mustinvolve intimate collaboration with experts in a widerange of disciplines, including soil science, hydrology,forest ecology, and the political economy of commu-nity resource use.

In our efforts to improve the foundation of data onwhich we build our environment-conflict research, how-ever, we must recognize that not all good data arequantitative: process-tracing of single cases, in fact, gen-erates thick descriptions of environment-conflictlinkages—descriptions rich with qualitative data. Morelocal case studies are needed, which build upon researchdone to date, and test and refine existing hypotheses atthe local level.

Operationalizing Key VariablesIf environment-conflict researchers want to estimate

causal effect, it is essential that they include in their analy-ses key variables identified by environment-conflictresearch. In order to include these variables, efforts mustturn towards their operationalization.

The Toronto Group has identified a number of vari-ables that play a pivotal role in the link betweenenvironment and conflict. For instance, as noted above,the quantity of ingenuity a society supplies in responseto environmental scarcity can play a key role in deter-mining its ability to adapt to that scarcity. The supplyof ingenuity, then, is an independent variable that shouldbe included in any statistical analysis attempting tomeasure the causal effect of environmental scarcity onconflict (Homer-Dixon, 1999: 107-114).

But, operationalizing this variable is not a straight-forward task. Researchers need an adequate measure ofingenuity. The Toronto Group has identified other mea-sures that should be included in any complete statisticalanalysis—and that therefore require operational-ization—including state capacity and socialsegmentation, as well as the aforementioned processesof resource capture and ecological marginalization(which can potentially be represented as single vari-ables).24

Specifying Contextual FactorsEmpirical research has now identified some causal

mechanisms linking environmental scarcity and violence.However, much more work remains to be done to de-termine precisely the intervening and interacting


Commentary

variables—the contextual factors—affecting the strengthof these processes. Under what circumstances, exactly,do these processes unfold? In the following, we refer tosome of the specific findings of the Toronto Group, andsuggest some contextual factors worthy of further in-vestigation.

• By setting in motion processes of resource captureand ecological marginalization, environmental scar-city often increases the wealth gap between thoseelites that take advantage of the opportunities scar-city offers and those marginal groups that suffer thebrunt of scarcity. How does the degree of state au-tonomy affect these two processes? To what extentwould better-defined and enforced property rightsreduce the predatory behavior of elites?

• The multiple effects of environmental scarcity in-crease demands on the state, stimulate intra-elitebehavior, and depress state tax revenues. Such pres-sures can weaken the administrative capacity andlegitimacy of the state. How does institutional de-sign affect state capacity in the presence of thesepressures? How do international economic forcesboth aggravate and mitigate these pressures?

• Narrow distributional coalitions (e.g., coalitions ofrent-seekers that work to redistribute the economy’swealth in their favor) often block institutional re-form—including reform of markets, property rights,judicial systems, and the state’s resource-manage-ment regimes—essential to reducing environmentalscarcity or alleviating its harsh effects. To what ex-tent does scarcity provoke such behavior? Can arobust civil society counteract the obstructionistbehavior of these narrow distributional coalitions?

Dealing with ComplexityAt the methodological level, we need to explore how

causation works at the interface between the physical/ecological and social worlds. Environment-conflict re-search brings us face to face with some of the mostintractable issues in philosophy of science, specificallywhether causal generalizations describing the social worldhave the same status as those describing the naturalworld. Because systems in both these domains are fun-damentally complex—characterized by huge numbersof components, causal interactions, feedback loops, andnonlinearity—environment-conflict researchers can gaininsights from complexity theory. We urge greater recep-tivity to the concepts and findings of this rapidlydeveloping field.

Encouraging Methodological PluralismIn order to deal with the research challenges de-

scribed above, we encourage our colleagues to accept adegree of methodological pluralism. The various meth-ods available to us make up a diverse set of arrows in thequiver of the social scientist, and we should choose thearrow most likely to hit our target. Statistical and quasi-experimental methodologies are needed to identifycorrelations and causal effects; process-tracing of singlecases is needed to specify causal mechanisms. These twogeneral approaches should not be used in isolation fromeach other; rather, we should try to exploit the synergiesthat are possible when they are used in parallel by col-laborating researchers. For instance, statistical analysiscan identify outliers and anomalous cases that deservefocused attention using process-tracing; in turn, pro-cess-tracing can identify key interacting variables andscope conditions that should be incorporated into sta-tistical tests of the environment-conflict hypothesis.

Methodological pluralism, however, is not a licensefor shoddiness. Researchers should be held to high stan-dards of evidence. This paper has demonstrated that theenvironment-conflict research of the Toronto Group andENCOP, among others, is not only theoretically andconceptually intact, but also rests on sound method-ological pillars. We hope that future researchers will usethis body of evidence to deepen our understandings ofthe linkages between environmental scarcity and vio-lent conflict.

REFERENCES

Ahlburg, Dennis. “Julian Simon and the PopulationGrowth Debate.” Population and Development Review24:2 (June 1998): 317-27.

Allison, Graham. Essence of Decision: Explaining theCuban Missile Crisis. Boston, MA: Little, Brown, 1971.

Baechler, Günther, Volker Böge, Stefan Klötzli, StephanLibiszewski, and Kurt R. Spillmann. KriegsursacheUmweltzerstörung, Band I, Ökologische Konflikte in derDritten Welt und Wege ihrer friedlichen Bearbeitung. Chur/Zürich: Verlag Rüegger, 1996.

Baechler, Günther and Kurt R. Spillman, eds.Kriegsursache Umweltzerstörung [Environmental Causesof War, in German]. Vol. 2, Regiona und Länderstudienvon Projektmitarbeitern [Regional and Country Studies of


Commentary

Research Fellows, in German]. Chur/Zürich: VerlagRüegger, 1996.

Baechler, Günther and Kurt R. Spillman, eds.Kriegsursache Umweltzerstörung [Environmental Causesof War, in German]. Vol. 3, Länderstudien von externenExperten [Country Studies of External Experts, in Ger-man]. Chur/Zürich: Verlag Rüegger, 1996.

Baumrind, Diana. “Specious Causal Attributions in theSocial Sciences.” Journal of Personality and Social Psy-chology 45 (1983): 1289-1298.

Bennett, Andrew. “Lost in the Translation: Big (N) Mis-interpretations of Case Study Research.” Paper presentedat the 38th Annual Convention of the International Stud-ies Association, Toronto, 18-22 March 1997.

Bennett, Andrew and Alexander George, eds. Case Studiesand Theory Development. Cambridge, MA: MIT Press,2000 (forthcoming).

Campbell, Donald T. “Degrees of Freedom and the CaseStudy.” Comparative Political Studies 8:2 (July 1975):178-193.

Cartwright, Nancy. “Causal Laws and Effective Strate-gies.” Nous 13:4 (November 1979): 419-438.

Cohen, Joel. How Many People Can the Earth Support?New York: W.W. Norton, 1995.

Collier, David. “The Comparative Method,” In Politi-cal Science: The State of the Discipline II, ed. Ada W.Finifter. Washington, DC: The American Political Sci-ence Association, 1993.

Collier, David and Steven Levitsky. “Democracy withAdjectives: Conceptual Innovation in Comparative Re-search.” World Politics 49:3 (April 1997): 430-51.

Cowan, George, David Pines, and David Meltzer, eds.Complexity: Metaphors, Models, and Reality. Santa FeInstitute Studies in the Sciences of Complexity, Proceed-ings Vol.19, Reading, MA: Addison-Wesley, 1994.

Dasgupta, Partha. An Inquiry into Well-Being and Desti-tution. Oxford: Clarendon Press, 1993.

Dessler, David. “Beyond Correlations: Toward a CausalTheory of War.” International Studies Quarterly 35:3(September 1991): 337-355.

Dessler, David. “Making Research on EnvironmentalScarcity and Violent Conflict More Useful toPolicymakers.” Paper prepared for the ‘Roundtable onthe University of Toronto’s Project on Environment,Population and Security’ and presented at the 38th An-nual Convention of the International StudiesAssociation, Toronto, 18-22 March 1997.

Deudney, Daniel H. “Environment and Security:Muddled Thinking.” The Bulletin of the Atomic Scien-tists (April 1991): 22-28._________. “Bringing Nature Back In: GeopoliticalTheory from the Greeks to the Global Era,” In Con-tested Grounds: Security and Conflict in the NewEnvironmental Politics, eds. Daniel H. Deudney andRichard A. Matthew. Albany, NY: State University ofNew York Press, 1999.

Durham, William H. Scarcity and Survival in CentralAmerica: Ecological Origins of the Soccer War. Stanford:Stanford University Press, 1979.

Eckstein, Harry. “Case Study and Theory in PoliticalScience,” In Strategies of Inquiry: Handbook of PoliticalScience Volume 7, eds. Fred I. Greenstein and Nelson W.Polsy, Reading, MA: Addison-Wesley, 1975.

Esty, Daniel C., Jack A. Goldstone, Ted Robert Gurr,Barbara Harff, Marc Levy, Geoffrey D. Dabelko, PamelaT. Surko, and Alan N. Unger. “State Failure Task ForceReport: Phase II Findings” Environmental Change andSecurity Project Report 5 (Summer 1999): 49-72.

Food and Agricultural Organization of the United Na-tions. The State of Food and Agriculture. Rome: FAO,1999a._________. The State of Food Insecurity in the World.Rome: FAO, 1999b.

Garavito, R. Michael. “Working Knowledge: Aspirin.”Scientific American 280:5 (May 1999): 108.

George, Alexander L. “Case Studies and Theory Devel-opment: The Method of Structured, FocusedComparison,” In Diplomacy: New Approaches in History,Theory, and Policy, ed. P.G. Lauren. New York: The FreePress, 1979.

George, Alexander and Timothy McKeown. “Case Stud-ies and Theories of Organizational Decision-Making,”In Advances in Information Processing in Organizations,eds. Robert Coulam and Richard Smith. Vol. 2, Lon-


Commentary

don: JAI Press, 1985

Gleditsch, Nils Petter. “Armed Conflict and the Envi-ronment: A Critique of the Literature.” Journal of PeaceResearch 35:3 (May 1998): 381-400.

Glymour, Clark, et al. Discovering Causal Structure: Ar-tificial Intelligence, Philosophy of Science, and StatisticalModeling, Orlando, FL: Academic Press, 1987.

Goodwin, Jeff. “State-Centered Approaches to SocialRevolutions,” In Theorizing Revolutions, ed. John Foran.New York: Routledge, 1997.

Grady, Denise. “So, Smoking Causes Cancer: This isNews?” New York Times, (October 27, 1996).

Harré, Rom. The Philosophies of Science. Oxford: Ox-ford University Press, 1985.

Hauge, Wenche and Tanja Ellingsen. “Beyond Environ-mental Security: Causal Pathways to Conflict.” Journalof Peace Research 35:3 (1998): 299-317.

Homer-Dixon, Thomas. “The Ingenuity Gap: Can PoorCountries Adapt to Resource Scarcity?” Population andDevelopment Review 21:5 (September 1995): 587-612.

Homer-Dixon, Thomas F. and Marc A. Levy. “Corre-spondence: Environment and Security.” InternationalSecurity 20:3 (Winter 1995/1996): 189-198.

Homer-Dixon, Thomas F. and Jessica Blitt, eds.Ecoviolence: Links among Environment, Population, andSecurity. New York: Rowman and Littlefield PublishersInc, 1998.

Homer-Dixon, Thomas F. Environment, Scarcity, andViolence. Princeton, NJ: Princeton University Press,1999.

Kahl, Colin H. “Population Growth, EnvironmentalDegradation, and State-Sponsored Violence: The Caseof Kenya, 1991-93.” International Security 23:2 (Fall1998): 80-119.

Kaplan, Abraham. The Conduct of Inquiry: Methodologyfor Behavioral Science. San Francisco, CA: Chandler,1964.

Kincaid, Harold. “Defending Laws in the Social Sci-ences,” In Readings in the Philosophy of Social Science,

eds. Michael Martin and Lee C. McIntyre. Cambridge:MIT Press, 1994.

King, Gary, Robert Keohane and Sidney Verba. Design-ing Social Inquiry: Scientific Inference in QualitativeResearch. Princeton, NJ: Princeton University Press,1994.

Lele, Uma and Steven Stone. Population Pressure, theEnvironment and Agricultural Intensification: Variationson the Boserup Hypothesis, MADIA Discussion Paper 4,Washington, DC: World Bank, 1989.

Levy, Mark. “Is the Environment a National SecurityIssue?” International Security 20:2 (Fall 1995): 35-62.

Lewontin, Richard C. “The Analysis of Variance andthe Analysis of Causes,” In The IQ Controversy: CriticalReadings, eds. J.J Block and Gerald Dwarkin. New York:Pantheon, 1976.

Libiszewski, Stephan. “What Is an Environmental Con-flict?” Occasional Paper 1 Bern: Swiss Peace Foundationand Zurich: Center for Security Studies and ConflictResearch, Swiss Federal Institute of Technology (July1992).

Lijphart, Arend. The Politics of Accommodation: Plural-ism and Democracy, Berkeley, CA: University ofCalifornia Press, 1975.

Mackie, J.L. “Causes and Conditions.” American Philo-sophical Quarterly 2:4 (October 1965): 245-264.

McKeown, Timothy J. “Case Studies and the StatisticalWorldview.” International Organization 53:1 (Winter1999): 161-190.

Midlarsky, Manus I. The Evolution of Inequality. Stanford,CA: Stanford University Press, 1999.

Miller, Richard. Fact and Method. Princeton, NJ:Princeton University Press, 1987.

Repetto, Robert. The ‘Second India’ Revisited: Population,Poverty, and Environmental Stress Over Two Decades.Washington, DC: World Resources Institute, 1994.

Rescher, Nicholas. Complexity: A Philosophical Overview.New Brunswick, NJ: Transaction, 1998.

Rrnnfeldt, Carsten. “Three Generations of Environment


Commentary

and Security.” Journal of Peace Research 34:4 (November1997): 473-482.

Salmon, Wesley C. Scientific Explanation and the CausalStructure of the World. Princeton, NJ: Princeton Univer-sity Press, 1984.

Sayer, Andrew. Method in Social Science: A Realist Ap-proach. New York: Routledge, 1992.

Simon, Julian. The Ultimate Resource 2. Princeton, NJ:Princeton University Press, 1996.

Skocpol, Theda. States and Social Revolutions: A Com-parative Analysis of France, Russia, and China. Cambridge:Cambridge University Press, 1979.

Smil, Vaclav. China’s Environmental Crisis: An Inquiryinto the Limits of National Development. Armonk, New

York: M.E. Sharpe, 1993.

Sober, Elliot. “Apportioning Causal Responsibility.” TheJournal of Philosophy 85:6 (June 1988): 303-318.

Suliman, Mohamed, ed. Ecology, Politics, and ViolentConflict. New York: Zed Books, 1999.

Tarrow, Sidney. Power in Movement: Social Movements,Collective Action, and Politics. Cambridge: CambridgeUniversity Press, 1994.

World Bank. World Development Report 1992: Develop-ment and the Environment. Washington, DC: WorldBank, 1992.

Yee, Albert. “The Effects of Ideas on Policies.” Interna-tional Organization 50:2 (Winter 1996): 69-108.

1 The three main sources of environmental scarcity includereduced supply, increased demand, and skewed distribution.See Homer-Dixon (1999: 47-72).

2 An anonymous reviewer of this article raised this objec-tion, as well as David Dessler (1999: 100-101).

3 The literature supporting this claim is so vast it cannot besummarized. An excellent survey can be found in, World Bank(1992). See also Midlarsky (1999) who provides compellingempirical evidence on the intimate connections between scar-city (including resource scarcity), inequality, and social conflict.Dasgupta (1993) provides an economic analysis of the effectof resource scarcity on communities in the developing world.Good and relatively current surveys of the state of the envi-ronment in China and India, which together constitute aboutforty percent of the world’s population, are Smil (1993) andRepetto (1994).

4 Recent data from the Food and Agriculture Organiza-tion (FAO, 1999a) shows that the percentage ofundernourished people in all three regions has either remainedsteady (sub-Saharan Africa) or fallen (South Asia and LatinAmerica). However, the absolute number of undernourishedpeople over the past thirty years has either grown (sub-Sa-haran Africa and South Asia), or remained relatively stableoverall (Latin America). See also FAO, 1999b.

5 Dan Deudney (1999), has recently coined the phrase so-cial-social theory for theories that presume social events haveonly social causes; he uses nature-social theory for theories in

which natural variables play a significant causal role.6 Gleditsch writes that ‘words such as “democracy” or “au-

tocracy” do not occur in the model. In view of the extensivetheoretical literature “relating the degree of democracy to civilviolence . . . a democracy variable should have been includedexplicitly.”

7 For an excellent treatment of the variegated nature of de-mocracy, see Collier and Levitsky (1997).

8 The differentiation of the “democracy variable,” in theState Failure Task Force’s Phase II report, represents an at-tempt to move in this direction. See Esty, et al (1999: 52-53).

9 “Much of the literature,” Gleditsch writes, “deals withconflicts of interest involving potential violence rather thanwith actual violence. . . . The argument is entirely in terms offuture wars, which may happen.” (Italics in original.)

10 Several of the Toronto Group’s historical case studies arereproduced in Homer-Dixon and Blitt (1998).

11 ENCOP similarly relied upon a large number of histori-cal case studies during the course of the project research. Thesecase studies are published, along with their theoretical find-ings in a three volume work, Bächler, et al (1996). This volumecontains: M. Abdul Hafiz and Nahid Islam, “EnvironmentalDegradation and Intra/Interstate Conflicts in Bangladesh;”Mohamed Suliman, “Civil War in Sudan: the Impact of Eco-logical Degradation;” Mohamed Suliman, “War in Darfur orthe Desert versus the Oasis Syndrome;” Peter B. Okoh, “En-

NOTES


Commentary

vironmental Degradation, Conflicts, and Peaceful Resolutionin Nigeria and between Nigeria and Neighboring States;”Stephan Klotzli, “The Water and Soil Crisis in Central Asia:A Source for Future Conflicts?”; Stephen Libiszewski, “WaterDisputes in the Jordan Basin Region and Their Role in theResolution of the Arab-Israeli Conflict;” Gunther Bächler,“Rwanda: The Roots of Tragedy, Battle for Elimination on anEthno-Political and Ecological Basis.”

12 See, for example, Eckstein (1975); Campbell (1975);George (1979); George and McKeown (1985); Dessler(1991); Yee (1996); Bennett (1997); and McKeown (1999).

13 This point is made in McKeown (1999: 172-174). Non-experimental methods have also been widely used in thenatural sciences. See McKeown (1999: 171), and Eckstein(1975: 114-115).

14 Hume was, in fact, highly skeptical of our ability to showcausation. His analysis of causation was meant to ascertainthe bare epistemic facts that undergird our intuition of cau-sality.

15 See Grady (1996: 3). A similar distinction between causaleffect and causal mechanism has implications for other areasin the natural sciences as well. For instance, although scien-tists have known for nearly a century that aspirin relieves pain,it is only within the last few years that they have discoveredthe causal mechanisms behind this pain relief. See Garavito(1999: 108).

16 Although qualitative quasi-experimental methods, suchas the comparative case study, can also detect causal mecha-nisms, the single-case method is often a more efficient meansof discovering these processes. Moreover, because control isextremely difficult to achieve in the comparative method, it isquestionable if causal mechanisms can be more accuratelydetected than with the single-case method.

17 Kaplan also argues that descriptions, which are often dif-ferentiated from explanation, may themselves be explanatory,‘because the “how” may provide a “why” and not just a “what”.’Noted philosopher of science, Rom Harré (1985: 40), makesa similar point when he asserts, “In practice we never restcontent with laws for which there is no explanation.” Onenotable criticism of this approach is made by Kincaid (1994:117), who argues that causal mechanisms can always be dis-covered at deeper levels (e.g. at psychological or evenneurophysiological levels). King, Keohane, and Verba make asimilar point (1994: 86). We believe this criticism ultimatelyfails, however, because a researcher must always conduct theirresearch at a chosen level of analysis, and the causal mecha-nisms they seek should correspond to the level of analysis oftheir research. Moreover, if deeper causal mechanisms are dis-covered, and if they support the theory, then the theory willonly be more robust.

18 Research is sometimes sparked by a preliminary correla-tion analysis that offers a promising avenue for further research(e.g. the Democratic Peace). Nevertheless, we maintain that afull-blown statistical analysis of these preliminary findingswould benefit greatly from case-study research into causalmechanism. The research process then, should be viewed asan iterative one, with quasi-experimental and case-study meth-ods complementing one another.

19 Although the logic for separating the testing and build-ing of theories in quantitative methodologies is sound,Campbell (1975:178-193) shows that this partition is notnecessary in case-study research. Campbell convincingly dem-onstrates that the problem of ex post facto hypothesizing isovercome in the ‘pattern matching’ methodology—from whichprocess-tracing was conceived—because this procedure opensthe possibility that an hypothesis initially generated by a par-ticular case could subsequently fail to be supported by thesame case. Also, see Collier (1993: 115).

20 Although the above quote from Homer-Dixon’s Interna-tional Security article does not refer explicitly to causalmechanism, the underlying approach taken throughout thearticle consists of an explicit attempt to discover these pro-cesses. It is therefore reasonable to assume that the nometheticclaim made is this article refers to causal mechanisms.

21 The common mistake among researchers is to omit a vari-able that should be controlled in a statistical analysis. Thiscan result in what statisticians refer to as a Type I error, wherethe null hypothesis is true but researchers decide it is false.However, it can be equally dangerous to include a variablethat should not be controlled. This can result in a Type IIerror, where the null hypothesis is false but researchers decideit is true. Cartwright (1979: 429-32) points out that an “ir-relevant” control variable can always be found that annuls orreduces a true relationship. A Type II error can also be com-mitted by failing to include a suppressor variable; that is, avariable that, once controlled for, unmasks a true relation-ship. To avoid both Type I and Type II errors, we suggest thatresearchers use process-tracing to determine the appropriatecontrol variables.

22 A parable recounted by Diana Baurmind (1983:1297)illustrates why research into causal mechanisms can be in-valuable in discovering control variables. “The number ofnever-married persons in certain British villages is highly in-versely correlated with the number of field mice in thesurrounding meadows. Marital status of humans was consid-ered an established cause of field mice by the village eldersuntil the mechanism of transmission were finally surmised:Never-married persons bring with them a disproportionatenumber of cats relative to the rest of the village populace andcats consume field mice. With the generative mechanisms un-derstood, village elders could concentrate their attention on


Commentary

increasing the population of cats rather than the proportionof never-married persons.” Glymour et al. (1987:19-21) op-pose Baumrind’s “generative” account of causality. They arguethat in fact “never-married persons” do cause variation in fieldmice, even if the causation is indirect, and nothing in thestory prevents the use of covariance analysis on uncontrolledsamples to discover that the intervening variables is the den-sity of cats. But this belies the process that social scientists useto discover control variables. Without an investigation of thecausal mechanisms, it is doubtful that the density of cats wouldhave been included in a statistical analysis.

23 Although it is not possible for case-study researchers toconsider all possible spurious relationships, neither is it pos-sible to include all possible confounding variables in a statisticalmodel. David Dessler (1999: 101) adopts this approach whenhe suggests that environment-conflict researchers “Test causalclaims not against the null hypothesis but against rival sub-stantive accounts of political violence in the cases analyzed.”

24 The State Failure Task Force’s Phase II report makes sig-nificant strides in this direction. Unfortunately, datalimitations seriously impeded their ability to adequatelyoperationalize some key variables. See Esty, et al (1999).

PECS NEWS

A TRI-ANNUAL NEWSLETTER ON ISSUES OF POPULATION, ENVIRONMENTAL CHANGE, AND SECURITY

PUBLISHED BY THE ENVIRONMENTAL CHANGE AND SECURITY PROJECT.

FEATURES OF EACH ISSUE:

Meeting reports from recent ECSP meetingsPast synopses have covered meetings that explored thedemographic, environmental, and food security conditions inWest Africa; the connection between population,consumption, and wood resources; and migration andenvironmental issues.

Book reviewA solicited review of a recently released publication looking atpopulation, environment, and/or security issues.

Lessons from the FieldThis section is dedicated to sharing field reports from theUniversity of Michigan Population Fellows Programs.

Project NewsThis final section includes information on new ECSPpublications, website features, and staff activities.

Early feedback on the first issue of the PECS News:

“You have really put together a first-rate piece. The layout is crisp, clean, highly readable… You have a fineassortment of articles, all at just the right length for all-in-one-sitting readability. I loved the book review yousolicited… it was refreshing to read a true critique, with both positive and constructive comments… I alsoespecially liked the report highlights and project news sections—they really give the reader a sense of thebreadth of your work at a glance.”

The next issue of PECS News will be published in October 2000. If you are interested in receiving a copy, pleasecontact the Project at [email protected] or by telephone at 202-691-4130. All issues of the newsletter areavailable on-line at http://ecsp.si.edu/PECS-News.


Commentary

Environment and Security in an International Context:Critiquing a Pilot Study from NATO’s Committee on the

Challenges of Modern Society

by Richard A. Matthew

In Pakistan’s North-West Frontier Province, the combination of severe environmental stress and diffuse, violentconflict along the Afghanistan border might seem to provide an excellent example of the relationships examinedin the NATO Committee on the Challenges of Modern Society’s (CCMS) new report on Environment andSecurity in an International Context.1 At first blush, one might conclude that environmental stress in this region isescalating the conflict. However, on closer examination it seems that the effect of environmental stress is mixed.Sometimes it acts to reinforce or trigger conflict, while other times it motivates people to reduce tensions andcooperate to solve problems. It has proven very difficult to model the varying effects of environmental stress in anaccurate way. The co-chairs of this CCMS pilot study have undertaken the arduous task of developing a generalmodel of environmental stress and conflict applicable to the entire world. In my experience, modeling a smallregion of a country abounding with many factors pulling in different directions is so complex that it is impossibleto be satisfied with the results.

Although the problem is an exceedingly complex one, the new CCMS report on environment and security isvery lucid and well organized. It asks several explicit questions: What is environmental security? How can we modelit? What sort of information would be useful to policymakers? And what sort of responses are available?

Much of the value of the NATO/CCMS and Science Programme work on environmental issues is in theprocess of bringing people together to focus on certain challenges, view them from new perspectives, and, perhaps,come to a shared understanding. These benefits may add value to the task force experience beyond what is writtenin the report, but I shall consider only the text of the report, recognizing that it is a small part of a valuable andimportant initiative.

In many ways, it is a bold report. It is bold not only in its recommendations but also in that it addresses thequestion of complexity very directly, unlike many other policy documents. This is a source of strength, but it alsointroduces some theoretical problems, which are the focus of my comments.

The report begins by noting that NATO has a long-standing interest in non-traditional security issues, and thatenvironmental stresses are emerging as one of these issues. As anybody who studies environmental history knows, anumber of scholars believe that environmental stress has been the driving force behind many events in humanhistory.2 The report is not introducing new issues. What is new is that a set of institutions that traditionally have notlooked at these issues are now starting to examine them.

Among the questions that have divided scholars and policymakers recently are whether security institutions

Richard A. Matthew is Assistant Professor of International Relations and Environmental Policy at the University ofCalifornia, Irvine.


Commentary

ought to be studying this set of issues, whether theseissues are best addressed by security institutions, andwhether these institutions bring a certain value to theprocess that is not found elsewhere. It is important tomention these concerns since many of the conclusionsin this report are similar to those of work undertaken bydevelopment organizations and other organizations.

One might be encouraged by the apparent conver-gence of thinking among institutions that ostensibly havelittle in common. Indeed, one possible value of rethink-ing security is that this can provide a framework in whichdialogue can take place among different groups and in-stitutions, largely free of preconceptions and prejudices.Of course, one might also raise the concern that mili-tary institutions are moving into areas already occupiedby the development and other communities, and won-der if this will lead to better outcomes in the long run.

One of the fears of those critical of security institu-tions examining environmental stress is that they willstudy it from a perspective or with a mindset that sim-ply reinforces or extends their traditional mandates.Critics worry that particular environmental issues willbe identified as important, while others will be neglectedbecause they do not fit well into a traditional securityframework. Environmental problems identified as se-curity issues, and hence likely to receive support fromgovernments and publics, could begin to monopolizeresources. Critics are concerned that the neglected is-sues might be those most important to the developingworld, and that the ultimate result of military involve-ment will be to reinforce inequalities in the internationalsystem

A different type of concern is that once the state ofthe environment has been characterized as a securityproblem in ways relevant to security institutions, a path-way has been created towards the eventual use of forceand coercion.

These familiar concerns do not seem justified bythe CCMS report. Instead of narrowing the concept ofenvironmental security to cases that could benefit mili-tary interests, this report opens the concept up. Theauthors do this by taking a very liberal position on thesocial impacts of environmental stress by defining con-flict very broadly. By being so broad, the authors areable to be very inclusive. There are many roles for manydifferent institutions. There are clear suggestions of theimportance of inter-agency dialogue and cooperation.There is no sense that now is the time for the military tobecome involved and start solving problems others havenot been able to address effectively.

Being broad, however, does introduce some prob-lems in terms of modeling the phenomenon of

environmental stress and conflict or insecurity. The au-thors of the report are clearly aware of this. Early on,they write, “It is not environmental stress in i

Documents

COMMENTARY: DEBATING ENVIRONMENT ... Report6-5.pdfThe editors of the ECSP Report wish to thank Tad Homer-Dixon, Dan Schwartz, Tom Deligiannis, Paul Diehl, Nils Petter Gleditsch, Leo