fearon, james d, and david d laitin_2003_ethnicity, insurgency, and civil war

American Political Science Review Vol. 97, No. 1 February 2003

Ethnicity, Insurgency, and Civil WarJAMES D. FEARON and DAVID D. LAITIN Stanford University

An influential conventional wisdom holds that civil wars proliferated rapidly with the end ofthe Cold War and that the root cause of many or most of these has been ethnic and religiousantagonisms. We show that the current prevalence of internal war is mainly the result of a steadyaccumulation of protracted conflicts since the 1950s and 1960s rather than a sudden change associatedwith a new, post-Cold War international system. We also find that after controlling for per capita income,more ethnically or religiously diverse countries have been no more likely to experience significant civilviolence in this period. We argue for understanding civil war in this period in terms of insurgency orrural guerrilla warfare, a particular form of military practice that can be harnessed to diverse politicalagendas. The factors that explain which countries have been at risk for civil war are not their ethnic orreligious characteristics but rather the conditions that favor insurgency. These include povertywhichmarks financially and bureaucratically weak states and also favors rebel recruitmentpolitical instability,rough terrain, and large populations.

Between 1945 and 1999, about 3.33 million battledeaths occurred in the 25 interstate wars thatkilled at least 1,000 and had at least 100 deadon each side. These wars involved just 25 states thatsuffered casualties of at least 1,000 and had a medianduration of not quite 3 months. In contrast, in the sameperiod there were roughly 127 civil wars that killed atleast 1,000, 25 of which were ongoing in 1999. A con-servative estimate of the total dead as a direct result ofthese conflicts is 16.2 million, five times the interstatetoll. These civil wars occurred in 73 statesmore than athird of the United Nations systemand had a medianduration of roughly six years.1 The civil conflicts in thisperiod surely produced refugee flows far greater thantheir death toll and far greater than the refugee flowsassociated with interstate wars since 1945. Cases suchas Afghanistan, Somalia, and Lebanon testify to theeconomic devastation that civil wars can produce. Bythese crude measures, civil war has been a far greaterscourge than interstate war in this period, though it hasbeen studied far less.What explains the recent prevalence of violent civil

conflict around the world? Is it due to the end of theCold War and associated changes in the internationalsystem, or is it the result of longer-term trends? Whyhave some countries had civil wars while others havenot? and Why did the wars break out when they did?We address these questions using data for the period1945 to 1999 on the 161 countries that had a populationof at least half a million in 1990.

James D. Fearon and David D. Laitin are Professors, Departmentof Political Science, Stanford University, Stanford, CA 94305-6044([email protected] and [email protected]).

We wish to thank the many people who provided comments onearlier versions of this paper in a series of seminar presentations.The authors also gratefully acknowledge the support of the NationalScience Foundation (Grants SES-9876477 and SES-9876530); sup-port from the Center for Advanced Study in the Behavioral Scienceswith funds from the William and Flora Hewlett Foundation; valu-able research assistance from Ebru Erdem, Nikolay Marinov, QuinnMecham,David Patel, andTQShang; sharing of data by Paul Collier.1 The interstatewar data derive fromSinger and Small 1994, updatedto include the Kargil and Eritrean wars. The bases for the civil warestimates are discussed below.

The data cast doubt on three influential conventionalwisdoms concerning political conflict before and afterthe Cold War. First, contrary to common opinion, theprevalence of civil war in the 1990s was not due to theendof theColdWar andassociated changes in the inter-national system. The current level of about one insix countries had already been reached prior to thebreakupof theSovietUnionand resulted froma steady,gradual accumulation of civil conflicts that began im-mediately after World War II.Second, it appearsnot to be true that a greater degree

of ethnic or religious diversityor indeed any partic-ular cultural demographyby itself makes a countrymore prone to civil war. This finding runs contrary toa common view among journalists, policy makers, andacademics, which holds plural societies to be espe-cially conflict-prone due to ethnic or religious tensionsand antagonisms.Third, we find little evidence that one can predict

where a civil war will break out by looking for whereethnic or other broad political grievances are strongest.Were this so, one would expect political democraciesand states that observe civil liberties to be less civilwar-prone thandictatorships.Onewould further antici-pate that state discrimination againstminority religionsor languages would imply higher risks of civil war. Weshow thatwhen comparing states at similar levels of percapita income, these expectations are not borne out.The main factors determining both the secular trend

and the cross-sectional variation in civil violence in thisperiod are not ethnic or religious differences or broadlyheld grievances but, rather, conditions that favor insur-gency. Insurgency is a technology of military conflictcharacterized by small, lightly armed bands practicingguerrilla warfare from rural base areas. As a form ofwarfare insurgency can be harnessed to diverse politi-cal agendas,motivations, andgrievances. The concept ismost closely associatedwith communist insurgency, butthe methods have equally served Islamic fundamental-ists, ethnic nationalists, or rebels who focus mainlyon traffic in coca or diamonds.Wehypothesize thatfinancially, organizationally, and

politically weak central governments render insur-gency more feasible and attractive due to weak local

75

Ethnicity, Insurgency, and Civil War February 2003

policing or inept and corrupt counterinsurgency prac-tices. These often include a propensity for brutal andindiscriminate retaliation that helps drive noncombat-ant locals into rebel forces. Police and counterinsurgentweakness, we argue, is proxied by a low per capita in-come. Shocks to counterinsurgent capabilities can arisefrom political instability at the center or the suddenloss of a foreign patron. On the rebel side, insurgencyis favored by rough terrain, rebels with local knowl-edge of the population superior to the governments,and a large population. All three aid rebels in hidingfrom superior government forces. Foreign base camps,financial support, and training also favor insurgency.Ourdata show thatmeasuresof cultural diversity and

grievances fail to postdict civil war onset, while mea-sures of conditions that favor insurgency do fairly well.Surely ethnic antagonisms, nationalist sentiments, andgrievances often motivate rebels and their supporters.But such broad factors are too common to distinguishthe cases where civil war breaks out. Also, because in-surgency can be successfully practiced by small num-bers of rebels under the right conditions, civil war mayrequire only a small number with intense grievances toget going.Using data on about 45 civil wars since 1960, Collier

and Hoeffler (1999, 2001) find similarly that measuresof objective grievance fare worse as predictors thaneconomic variables, which they initially interpreted asmeasures of rebel greed (i.e., economicmotivation).2More recently, they argue that rebellion is betterexplained by opportunity than by grievance (cf.Eisinger 1973 and Tilly 1978) and that the main de-terminant of opportunity is the availability of financeand recruits for rebels. They proxy these withmeasuresof primary commodity exports and rates of secondary-school enrollment for males. We agree that financingis one determinant of the viability of insurgency. Weargue, however, that economic variables such as percapita income matter primarily because they proxyfor state administrative, military, and police capabil-ities. We find no impact for primary commodity ex-ports, and none for secondary schooling rates distinctfrom income. Our theoretical interpretation is moreHobbesian than economic. Where states are relativelyweak and capricious, both fears and opportunities en-courage the rise of would-be rulers who supply a roughlocal justice while arrogating the power to tax forthemselves and, often, for a larger cause.

CIVIL WAR SINCE 1945Building on similar efforts by other civil war resear-chers,3 we constructed a list of violent civil conflicts

2 There are 79 wars in their sample, but they lose about 34 due tomissing values on explanatory variables, which are mainly economic.Standard economic data tend to be missing for countries that arepoor and civil war-torn. This highly nonrandom listwise deletionmayaccount for some of the differences between our results.3 In particular, Doyle and Sambanis (2000), Esty et al. (1998),Gleditsch et al. (2002), the Institute for International and Strate-gic Studies (2000), Licklider (1995), Singer and Small (1994), Sivard(1996), and Valentino (2002).

that we presently believe tomeet the following primarycriteria. (1) They involved fighting between agents of(or claimants to) a state and organized, nonstate groupswho sought either to take control of a government, totakepower in a region, or to use violence to change gov-ernment policies. (2) The conflict killed at least 1,000over its course, with a yearly average of at least 100.(3) At least 100 were killed on both sides (includingcivilians attacked by rebels). The last condition is in-tended to rule out massacres where there is no orga-nized or effective opposition.4These criteria are broadly similar to those stated

by the Correlates of War (COW) project, Doyle andSambanis (2000), and several others.Wedeveloped ourown list (working from these and other sources)mainlybecause we wanted data for the whole 194599 periodand because of doubts about particular inclusions andexclusions in each list.5In one respect our data differ significantly frommost

others: We see no reason in principle to exclude anti-colonial wars, such as the French versus the NationalLiberation Front (FLN) in Algeria. We count these asoccuring within the colonial empire. Thus, the Frenchstate/empire looks highly civil war-prone by our list,with six colonial wars occuring in the 1950s. Butto drop such cases would be like dropping the cur-rent conflict in Chechnya as a civil war in Russiaif the Chechens succeed in gaining independence. Al-ternatively, it would make even less sense to includethem as wars within states that did not exist (such asAlgeria in 1954).There are both practical and theoretical consider-

ations pointing the other way, however. In practicalterms, to include the anticolonial wars in the analysisrequires that we form estimates of possible explana-tory factors for whole empires, such as gross domes-tic product (GDP) per capita, ethnic fractionalization,and democracy scores. Further, these estimates mustchange almost by year, as the colonial empires gradu-ally diminished in size.We are able to use country-leveldata to produce such estimates for ethnic fractionaliza-tion, but our estimates for per capita income are more

4 We used the following secondary criteria to deal with several othercoding issues. (4) The start year is the first year in which 100 werekilled or in which a violent event occurred that was followed by asequence of actions that came to satisfy the primary criteria. (5) If amain party to the conflict drops out, we code a new war start if thefighting continues (e.g., Somalia gets anewcivilwar afterSiadBarre isdefeated in 1991). (6)War ends are coded by observation of a victory,wholesale demobilization, truce, or peace agreement followed by atleast two years of peace. (7) Involvement by foreign troops does notdisqualify a case as a civil war for us, provided the other criteria aresatisfied. (8) We code multiple wars in a country when distinct rebelgroups with distinct objectives are fighting a coherent central stateon distinct fronts with little or no explicit coordination. (9) If a stateseeks to incorporate and govern territory that is not a recognizedstate, we consider it a civil war only if the fighting continues afterthe state begins to govern the territory (thus, Indonesia/East Timor1975, yes, and India/Hyderabad 1947, no).5 Sambanis (2002) discusses a number of conceptual and operationalambiguities and problems with the COW civil war data. Collier andHoeffler (2001) base their list on COW data, breaking a number ofCOW civil wars into multiple wars according to unspecified criteriaand including some colonial wars (coded in nonexistent states, suchas Angola 1961) but not others.

76

American Political Science Review Vol. 97, No. 1

FIGURE 1. Number and Percentage of Countries with Ongoing Civil Wars by Year from 1945 to 1999

problematic and the question of how to code the em-pires on a democracy index is vexed. Regarding theory,the colonial empires differed so radically from otherindependent states, and faced such an inhospitable in-ternational environment after the war (with pressurefrom the United States and the new United Nationssystem), that we need to be cautious about any empir-ical results that depend wholly on these cases. Thus weanalyze the data both with andwithout the anticolonialwars.

Descriptive StatisticsWe identified 127 conflicts that meet the above criteria,of which 13 were anticolonial wars.6 This makes for 127civil war starts in a sample of 6,610 country years, a rateof 1.92 per 100. The periods following major interna-tional systemic change had the highest onset rates. Civilwars broke out in the late 1940s and the 1950s at 4.6 and2.2 per 100 country-years, respectively, followed by the1990s, at 2.0. In absolute terms, the largest number ofcivil wars began in the 1990s (31), followed by the 1960sand 1970s (19 and 25, respectively).Omitting the anticolonial conflicts, most civil wars

occurred in sub-Saharan Africa (34) and Asia (33),followed by North Africa/the Middle East (17), LatinAmerica (15), Eastern Europe and the former SovietUnion (13), and the West (2). The rate of outbreakwas highest in Asia, at three per 100 country-years;Africa, North Africa/the Middle East, and LatinAmerica all had rates of about two per 100 country-years. France, Indonesia, and the Soviet Union/

6 See Fearon and Laitin 2003 for a list of the conflicts.

Russia are the most civil war-prone countries in thesample, with six onsets each.7

Trends over TimeFigure 1 shows the number of countries with ongoingcivil wars by year from 1945 to 1999. Since the numberof independent states grew sharply in this period, italso shows the proportion of countries with at least oneongoing war in each year.The graph indicates that, contrary to popular belief,

the prevalence of civil wars in the 1990s is not dueto effects of the end of the Cold War. The 1999 levelof 25 ongoing wars had already been reached by themid 1980s. Conflicts associated with the Soviet collapsewere partly responsible for the sharp increase in theearly 1990s, but a marked decline has followed.8One might conjecture that more and more civil wars

are breaking out over time, thus producing the secu-lar increase. This is incorrect. The rate of outbreak is2.31 per year since 1945, highly variable but showingno significant trend up or down. The secular increasestems from the fact that civil wars have ended at arate of only about 1.85 per year. The result has been asteady, almost-linear accumulation of unresolved con-flicts since 1945.Put differently, states in the international system

have been subject to a more or less constant risk

7 Four outbreaks are coded in the SovietUnion in 1946, in theBalticsand Ukraine, plus Russias two Chechen wars in the 1990s.8 Gurr (2000) notes the late-1990s decline in ethnic war and arguesthat the trend reflects improvedmanagement strategies by states andinternational organizations. The basic pattern in Figure 1 is not anartifact of thewaywe have coded civil war; it is observed in a broadrange of other data sets on violent domestic conflict for this period(e.g., Gleditsch et al. 2002).

77


of violent civil conflict over the period, but the con-flicts they suffer have been difficult to end. The aver-age duration of the civil wars in progress has increasedsteadily from two years in 1947 to about 15 yearsin 1999. From a policy perspective this suggests cau-tion about seeing as a temporary blip the sorts ofmilitary and political problems Western foreign policymakers have faced recently in Kosovo, Macedonia,Bosnia, Somalia, Haiti, East Timor, Colombia, andelsewhere.

ETHNICITY, DISCRIMINATION,AND GRIEVANCESDuring the Cold War, political scientists and sociolo-gists often sought to trace rebellion to economic in-equality (Muller 1985; Paige 1975; Russett 1964), torapid economic growth said to destabilize traditionalrural social systems (Huntington 1968; Scott 1976), orto frustrations arising from the failure to gain expectedbenefits of economic modernization (Gurr 1971). Afew scholars argued that the real source of rebellionwas often ethnic nationalism (Connor 1994), and a richliterature on the sources nationalist mobilization de-veloped in comparative politics (e.g., Anderson 1983,Deutsch 1953, and Gellner 1983). With the collapseof the Soviet Union and Yugoslavia, such culturalistperspectives became a dominant frame for interpretinginter- and intranational conflict (e.g.,Huntington1996).Using a broad brush, we can distinguish between

perennialist andmodernist (or constructivist) positionson the nature and sources of ethnic nationalism. Givenin purest form by nationalist politicians and journalistsreporting on nationalist conflicts in progress, perenni-alist arguments stress long-standing cultural practicessaid to define and distinguish ethnic groups. Differ-ences between these practices are argued to havemadeconflict more likely. Academics rarely make such ar-guments as baldly as do nationalist leaders. But au-thors who stress the long-standing, deep nature ofethnic differences and suggest that these make domes-tic peace difficult include Huntington (1996), Ignatieff(1993), Moynihan (1993), Rabushka and Shepsle(1972), and Smith (1986). Arguably, the main messageof Horowitzs (1985) influential book on ethnic conflictis that plural societies face a host of pathologies thatrender them especially prone to conflict and, at theextreme, violence.In contrast, modernist theories see the thorough-

going politicization of cultural difference that ethnicnationalism represents as a development of the last 200to 500 years. The core argument is that economic mod-ernization and the development of the modern statemake upward social mobility possible, but contingenton sharing the culture of the group that dominatesstate or society. When the state or society poses ascrip-tive barriers to upward mobility for minority groups,theymaydevelop separatist nationalistmovements:Allthe more so, according to Anderson (1983), Deutsch(1953), and Gellner (1983), the greater the preexistingcultural differences between the minority group and

the dominant group. (When preexisting differences areslight, assimilation is more likely.)The two approaches thus imply the same hypothesis

about the relationship between cultural diversity andcivil conflict.

H1: Measures of a countrys ethnic or religious di-versity should be associated with a higher risk ofcivil war.

Seeking to explain anticolonial nationalist move-ments in countries barely affected by economic mod-ernization, Gellner (1983, 108, n1) argued that eventhe advance shadow of modernization was sufficientto start nationalist dynamics in motion. For Anderson,nationalism quickly became a modular form thatcould be easily pirated, even under conditions quitedifferent from those during its origination. Nonethe-less, if one took the modernist approach literally onemight infer that more modernization should implymore discrimination and thus more nationalist con-tention in culturally divided countries.

H2: The effect of ethnic diversity on the probabilityof civil war should increase at higher levels of percapita income (a proxy for economic moderniza-tion).

Horowitz (1985) and several others argue that therelationship between ethnic diversity and severe ethnicconflict is nonmonotonic, with less violence for highlyhomogeneous andhighly heterogeneous countries. Thepolitics of a highly diverse country might be strate-gically similar to those of a homogeneous one, sincesmall ethnic groupsmust join coalitions tohavepoliticalinfluence, and there are many possibilities for cross-cutting and shifts in coalition membership. Horowitzsuggests that the most severe ethnic conflicts will arisein countries where a substantial ethnic minority facesan ethnic majority that can, given ethnic voting, win forsure in any national election. Such arguments yield thefollowing hypothesis.

H3: Countries with an ethnic majority and a sig-nificant ethnic minority are at greater risk for civilwar.

We consider severalmeasures for these concepts. ForH1 we use (1) the commonly employed ethnolinguis-tic fractionalization (ELF) index based on data fromAtlas Narodov Mira 1964, which gives the probabilitythat two randomly drawn individuals in a country arefrom different ethnolinguistic groups;9 (2) a measureof the share of population belonging to the largest eth-nic group that we constructed from the CIA Factbookand other sources (Fearon 2002); (3) the number ofdistinct languages spoken by groups exceeding 1% of

9 For this and several other variables we filled in values for missingcountry years whenever possible based on our own research; oftenthe sources were the CIA Factbook, Encyclopedia Brittanica, andthe Library of Congress Country Studies, though we used country-specific sources when necessary.

78

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven


the countrys population, based onGrimes andGrimes1996; and (4) a measure of religious fractionalization(analogous to the ELF) that we constructed using datafrom the CIA Factbook and other sources. For H2 weinteract these measures with per capita income. ForH3, we use a dummy variable marking the 74 countrieswhose largest and second-largest ethnic groups exceed49% and 7% of the population, respectively.10The mechanism that gives rise to nationalist con-

tention in modernist arguments is state or societal dis-crimination along the lines of cultural difference, whichis thought to create the grievances that motivate re-bellions. Grievances are difficult to measure indepen-dently of our knowledge of the actions we are tryingto explain (rebellions, civil war), but measures of aver-age levels of discrimination are feasible. Other thingsbeing equal, political democracy should be associatedwith less discrimination and repression along culturalor other lines, since democracy endows citizens with apolitical power (the vote) they do not have in dictator-ships. Evenmore directly,measures of state observanceof civil rights such as freedom of association, expres-sion, and due process should be associated with lessrepressionand thus lower grievances. Statepolicies thatdiscriminate in favor of a particular groups languageor religion should be associated with greater minoritygrievances. Finally, it is often argued that greater eco-nomic inequality creates broad grievances that favorcivil conflict (e.g., Muller 1985). Thus, we have the fol-lowing hypotheses.

H4: Measures of political democracy and civil lib-erties should be associated with lower risks of civilwar onset.

H5: Policies that discriminate in favor of a partic-ular language or religion should raise the risk ofcivil war onset in states with religious or linguisticminorities.

H6: Greater income inequality should be associ-ated with higher risks of civil war onset.

We consider both the Polity IV and the Przeworskiet al. 2000 democracy measures, along with the Free-dom House indicator of the observance of civilliberties.11 For income inequality we interpolated andextended as necessary the Gini coefficients assembled

10 Alternative thresholds for the second-largest group, such as 10%,make no difference in the results.11 Freedom House codes countries annually from 1972 on a seven-point scale based on a checklist that awards points for governmentobservance of a long list of civil liberties and rule of law, includingfreedomof the press, religion, and association, independent judiciary,equal treatment under the law, civilian control of police, protec-tion from political terror, secure property rights, and equality ofopportunity. Freedom from war and insurgency is one element oftheir checklist, so we are careful to lag this independent variable.For Polity IV, we use the difference between the 11-point democracyand autocracy scales. Following the Polity coders suggestions, weinterpolate values for transition period years (88), treat foreignoccupation years (66) as missing, and treat interruptions (77)as zeros. See http://www.bsos.umd.edu/cidcm/polity/.

by Deininger and Squire (1996); using the country-average of Gini values produced the same results. Forpolicies on religion, we coded, by decade, indicatorsfor whether the state had an official religion, gave re-sources to one religion not available to others, regu-lated missionary activities, required religious groupsto get official approval for religious activities, or sin-gled out for or permitted harassment of a particularreligious group. For policies on language, we coded byyear whether a language spoken by at least 5% of thepopulation received no official recognition at any levelof government.12One might argue, contrary to H1 above, that we

should expect an association between measures of eth-nic diversity and the occurence of ethnic strife, but notbetween ethnic diversity and civil war more broadly.In the limit this argument borders on tautology, sinceit would not be surprising if ethnic wars are rare incountries whose citizens think of themselves as ethni-cally homogeneous. However, one might ask if ethnicwars become more likely as ethnic diversity increasesamong those countries that have at least one nontrivialethnic minority.

H7:Among countries with an ethnicminority com-prising at least 5% of the population, greater eth-nic diversity should associate with a higher risk ofethnic civil war.

The measurement problem here is to say what anethnic civil war is, and we suspect that under anyplausible statement of the concept there will be numer-ous cases that are mixed or ambiguous. We coded asethnic conflicts in which the fighters were mobilizedprimarily along ethnic lines, marking off 58 (51%) asethnic, 20 (18%) as mixed or ambiguous, and 36 (32%)as not ethnic (excluding the anticolonial conflicts).

INSURGENCYIf many post-1945 civil wars have been ethnic ornationalist as these terms are usually understood,then even more have been fought as insurgencies. In-surgency is a technology of military conflict character-ized by small, lightly armed bands practicing guerrillawarfare from rural base areas. To explain why somecountries have experienced civil wars in this period oneneeds to understand the conditions that favor insur-gency, which are largely independent of cultural dif-ferences between groups and even group grievances.These conditions are best summarized byway of a briefstatement of the logic of insurgency.13The fundamental fact about insurgency is that insur-

gents are weak relative to the governments they are

12 Official recognition implies use by some public, officially sanc-tioned entity. For language we relied mainly on Asher 1994 andGunnemark 1992; and for religion, the U.S. State Departments 1999Annual Report on Religious Freedom.13 Though our formulations differ, we have been influenced hereby Stathis Kalyvass work on the Greek civil war. The literature onguerrilla warfare is extensive; see, for examples, Desai and Eckstein1990, Griffith 1961, and Laqueur 1976.

79

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven


fighting, at least at the start of operations. If govern-ment forces knew who the rebels were and how to findthem, theywould be fairly easily destroyedor captured.This is true even in states whose military and policecapacities are low. The total number of active rebels inmany wars in which thousands of civilians have beenkilled (through the actions of both governments andrebels) is often in the hundreds or low thousands.The numerical weakness of the insurgents implies

that, to survive, the rebels must be able to hide fromgovernment forces. Several hypotheses follow.

H8: The presence of (a) rough terrain, poorlyserved by roads, at a distance from the centersof state power, should favor insurgency and civilwar. So should the availability of (b) foreign, cross-border sanctuaries and (c) a local population thatcan be induced not to denounce the insurgents togovernment agents.

Much scholarly writing holds that ethnic or class sol-idarity and grievances are necessary for H8c, the lo-cal populations support of active rebels. In line withKriger (1992) and some analysts of communist insur-gencies (e.g., Clutterbuck 1967, Leites and Wolf 1970,and Thompson 1966), we argue that while grievancesand ethnic solidarity can be helpful in this regard, theyare not necessary. Instead, the key to inducing the lo-cal population not to denounce the active rebels islocal knowledge, or information about who is doingwhat at the village level. Local knowledge allows theactive rebels to threaten retribution for denunciationcredibly.14 Ethnic insurgents use this informational ad-vantage to great effect, often threatening and inflictingunimaginably harsh sanctions on their own people(Kalyvas 1999; Kriger 1992). The presence of an eth-nic insurgency does not imply that the members of theethnic group are of one mind in their determination tofight the state till they realize a nationalist dream. Theimmediate concern is how to survive in between gov-ernment forces using violence to gain information orpunish alleged rebel supporters and rebel forces usingviolence to punish alleged informants, moderates, orgovernment sympathizers.An empirical implication of the importance of lo-

cal knowledge is hypothesis H8d: Having a rural baseshould greatly favor insurgency. In the city, anony-mous denunciation is easier to get away with, givingthe government an advantage in its counterinsurgentefforts.Given the basic constraints posed by numerical

weaknessthe need to hide and not be denouncedvarious factors determine insurgents ability to wagewar. To survive, rebels need arms and materiel, moneyto buy them, or smugglable goods to trade for them.They need a supply of recruits to the insurgent way of

14 A second-ordermechanism by which ethnicity may favor insur-gency is that ethnic minorities are sometimes marked by dense socialnetworks that are isolated fromdominant groupnetworks, thusgivingan informational advantage to local rebels (Fearon and Laitin 1996).But such an advantage does not require ethnic distinctiveness.

life, and theymayalsoneed informationand instructionin the practical details of running an insurgency.15Most important for the prospects of a nascent insur-

gency, however, are the governments police andmilitarycapabilities and the reach of government institutions intorural areas. Insurgents are better able to survive andprosper if the government and military they opposeare relatively weakbadly financed, organizationallyinept, corrupt, politically divided, and poorly informedabout goings-on at the local level.Effective counterinsurgency requires government

forces to distinguish active rebels from noncombatantswithout destroying the lives and living conditions of thelatter. This is an extremely difficult political, military,and organizational problem even for well-equippedand well-paid modern militaries; witness the U.S. mil-itarys failures in Vietnam (Avant 1994; Krepinevich1986), early British efforts in Northern Ireland(Kennedy-Pipe 1997), or Soviet efforts in Afghanistan.For less well-financed and bureaucratically competentstates, the problem appears to be nearly insoluble. Suchstates either cannot prevent the abuse of local powersby field commanders or may even permit these abusesas a sort of tax farming to the military. That is, theypay the soldiers with the opportunity to loot and pil-lage, a practice that tends to sustain rather than endinsurgencies (see Keen 1998 for examples). Thus, wehave the following hypothesis.

H9: Proxies for the relativeweakness or strength ofthe insurgentstheir odds of being killed or cap-tured for a given level of counterinsurgent effortby the governmentshould be associatedwith thelikelihood that a country develops a civil war. Inparticular, a higher per capita income should be as-sociatedwith a lower riskof civilwaronset because(a) it is a proxy for a states overall financial, ad-ministrative, police, and military capabilities, and(b) it will markmore developed countries with ter-rain more disciplined by roads and rural societymore penetrated by central administration.

There is an additional reason why a lower per capitaincome should favor the technology of insurgency:(c)Recruiting youngmen to the life of a guerrilla is eas-ier when the economic alternatives are worse. Thoughwe trybelow, it is difficult tofindmeasures todistinguishamong these three mechanisms associating a low percapita income with civil war onset. We believe that thestrong results for per capita income reported below aredue largely to its acting as a proxy for state military andpolice strength relative to potential insurgents (a and bin H9). The fact thatmeasures such as the percentage ofyoung males and male secondary schooling rates pre-dict less well than per capita income is consistent withthis conjecture, though not definitive.

15 In the case literature one frequently finds either that rebels leadershave spent time at guerrilla training camps in, for example, Libya,Afghanistan, Lebanon, or Mozambique (in the 1970s) or that theygained guerrilla experience in one insurgency that they apply in pur-suing another.

80

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven


Additional factors that would be expected to affect(or proxy) the strength of an insurgent band relative toa state follow.

H10: The political and military technology of in-surgency will be favored, and thus civil war mademore likely, when potential rebels face or haveavailable the following.

(a) A newly independent state, which suddenlyloses the coercive backing of the former im-perial power and whose military capabilitiesare new and untested (Fearon 1998).

(b) Political instability at the center, which mayindicate disorganization and weakness andthus an opportunity for a separatist or center-seeking rebellion.

(c) A regime that mixes democratic with auto-cratic features, as this is likely to indicate polit-ical contestation among competing forces and,in consequence, state incapacity. (In contrast,pure autocracy tends to reflect the successfulmonopolization of state coercive and adminis-trative power by an individual or group.)

(d) A large country population, which makes itnecessary for the center to multiply layers ofagents to keep tabs onwho is doing what at thelocal level and, also, increases the number ofpotential recruits to an insurgency for a givenlevel of income.

(e) A territorial base separated from the statescenter by water or distancefor example,East Pakistan (now Bangladesh) from WestPakistan or Angola from Portugal.

(f) Foreign governments or diasporas willing tosupply weapons, money, or training.

(g) Land that supports the production of high-value, low-weight goods such as coca, opium,diamonds, and other contraband, which can beused to finance an insurgency.

(h) A state whose revenues derive primarily fromoil exports. Oil producers tend to have weakerstate apparatuses than one would expect giventheir level of income because the rulers haveless need for a socially intrusive and elaboratebureaucratic system to raise revenuesa po-litical Dutch disease (Chaudhry 1989; Karl1997; Wantchekon 2000). At the same time,oil revenues raise the value of the prize ofcontrolling state power.

Partially excepting f, none of these conditions cru-cially involves cultural differences, ethnic minoritystatus, or group grievances. We do not claim that thesefactors provide no help to would-be insurgents in spe-cific cases. But, to reiterate, grievances and ethnicdifferences are too common to help distinguish thecountries and years that see civil wars, and in any eventthe technology of insurgency does not require strong

popular grievances or ethnic solidarities to operate ef-fectively. The latter point suggests a contrast to H4H6.

H11: After controlling for per capita income (orother measures of state strength), neither politicaldemocracy, thepresenceof civil liberties, higher in-come inequality, nor nondiscriminatory linguisticor religious policies should associate strongly withlower odds of civil war. Given the right environ-mental conditions, insurgencies can thrive on thebasis of small numbers of rebels without strong,widespread, popular support rooted in grievancesand, hence, even in democracies.

As for measures, for rough terrain we use the pro-portion of the country that is mountainous accordingto the codings of geographer A. J. Gerard.16 This doesnot pick up other sorts of rough terrain that can befavorable to guerrillas such as swamps and jungle, andit takes no account of population distributions or foodavailability in relation tomountains; but it is the bestwehave been able to do forH8a. ForH9 we use PennWorldTables and World Bank data on per capita income, es-timating missing values using data on per capita en-ergy consumption.17 For H10a (new states) we markcountries in their first and second years of indepen-dence; for H10b (political instability) we use a dummyvariable indicating whether the country had a three-or-greater change on the Polity IV regime index in any ofthe three years prior to the country-year in question.18For countries that mix democratic and autocratic fea-tures (called anocracies or semidemocracies inthe international relations literature and praetorianregimes by Huntington 1968) we mark regimes thatscore between 5 and 5 on the difference betweenPolity IVs democracy and autocracy measures (thedifference ranges from 10 to 10). Country popula-tion (H10d) is based largely on World Bank figures. Foroil exporters we marked country-years in which fuelexports exceeded one-third of export revenues, usingWorld Bank data.19 We coded a dummy variable forstates with noncontiguous territory ourselves (H10e).20

16 Gerard produced this measure for the DECRG project on civilwars at the World Bank. Our sample of countries differs slightly,so we estimated values for 21 missing countries using the differencebetween the highest and the lowest point of elevation in each country,which is well correlated with the mountains measure (0.78 in logs).17 We used income growth rates from the World Development Indi-cators 2001 to extend the estimates in the PennWorld Tables 5.6 andthen used the per capita energy consumption estimates provided bythe COW project to estimate additional missing values. For detailssee Fearon and Laitin 2003.18 For this variable, transition periods and interruptions (whichindicate a complete collapse of central authority) are coded asinstability; foreign occupations are treated as missing.19 The data are for five-year intervals beginning in 1960; we inter-polated for years after 1960, set the value to that in 1960 for yearsprior to 1960, and used country-specific sources for a few countrieswithout World Bank coverage.20 Countries with territory holding at least 10,000 people and sep-arated from the land area containing the capital city either by landor by 100 km of water were coded as noncontiguous. Ignoring thecolonial empires, 25 of our 161 countries meet this criterion at sometime since 1945.

81

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven

Joshua Scriven


The remaining hypotheses (H8bH8d, H10f andH10g) present more difficult measurement challenges.Whether availability of a rural base favors civil war(H8d) is better tested in a research design where ethnicgroups are the unit of analysis, so that groups with dif-ferent geographic concentrations can be compared.21Although it is possible to code rebellions in progress

for whether the rebels receive shelter and support fromforeign countries (H8b, H10f), the potential availabilityof these aids to rebel strength is difficult to observeprior to the onset of fighting. In two special cases, thepotential availability of support from a foreign powerto governments is observablein Soviet policy (theBrezhnev doctrine) in Eastern Europe and Frenchpolicy with regard to its former colonies in subSaharanAfrica.22 We would expect such support to increase therelative advantage of government forces against po-tential insurgents and thus associate with lower ratesof civil war onset. We also consider a more tenuousmeasure of potential support to rebelsthe numberof civil wars ongoing in neighboring countrieswhichmight yield more easily available weapons, training, orthe presence of experienced guerrillas.23

EMPIRICAL ANALYSISOur central hypotheses concern the relationship be-tween ethnic and religious diversity or structure, onthe one hand, and the susceptibility of a country to civilwar, on the other. Several multivariate analyses of thecountry-year data are presented below, but the mainstory emerging from them is made clear by the contourplot in Figure 2.

Are More Diverse Countries Proneto Civil War?Figure 2 shows how probabilities of civil war onsetvary at different percentiles for country income (on thex axis, measured in lagged 1985 dollars) and ethnichomogeneity (on the y axis, measured by the popu-lation share of the largest ethnic group). The lines inthe plot show the probability of war onset in the nextfive years for a country at the given level of incomeand ethnic homogeneity. For example, countries at thetwentieth percentile in terms of the size of their largestethnic groupthus quite ethnically diversebut at theeightieth percentile on income have had about a 5%

21 Using the Phase III Minorities at Risk (MAR) data, Fearon andLaitin (1999) found that groupswithout a rural base areawere far lesslikely to be engaged in violent conflict with the state, even after con-trolling for various country- and group-specific factors. Toft (1996)was the first to note and examine the strong bivariate relationship inthe MAR data.22 U.S. support to rightist regimes in Latin America during the ColdWar might also qualify, although this was perhaps more offset bysupport for armed insurgency in this area from the Soviet Union andCuba.23 The presence of valuable minerals or the suitability of land forthe cultivation of narcotics is also codable in principle, but at presentwe lack such measures (H10e). Nor do we have measures for thecomparative disadvantage of governments in access to village-levelinformation (H8c).

chance of civil war outbreak in the next five years. Incontrast, countries at the eightieth percentile on ethnichomogeneity and at the twentieth percentile on incomehad a 15% chance of war in the next five years.24Note that for any level of ethnic diversity, as one

moves up the income scale (to the right in Figure 2),the odds of civil war decrease, by substantial factorsin all cases and dramatically among the most homoge-neous countries. The richest fifth is practically immuneregardless of ethnic composition. In contrast, for givenlevels of country income, no consistent effect is associ-ated with variation in ethnic homogeneity (i.e., movingup or down the figure). Among the poorest countrieswherewe observe the highest rates of civil war, the dataindicate a tendency formore homogeneous countries tobe more civil war-prone. Among the richest countriesthere may be a weak tendency for the most homoge-neous countries to have fewer civil wars, but the size ofthe effect, if any, is small.The empirical pattern is thus inconsistent with H1,

the common expectation that ethnic diversity is amajorand direct cause of civil violence. Nor is there strongevidence in favor of H2, which expects ethnic strife tobe activated as modernization advances. Ethnic diver-sity could still cause civil war indirectly, if it causes alow per capita income (Easterly and Levine 1997) ora weak state. But then the mechanisms that actuallyproduce the violence would more likely be those of theinsurgency perspective than the culturalist argumentsin either perennialist or modernist forms.The data for Figure 2 omit the anticolonial wars, and

the analysis does not control for other possible influ-ences on civil strife enumerated earlier. We considerthese issues next.

Multivariate ResultsWe coded a variable onset as 1 for all country-yearsin which a civil war started and 0 for all others.25Model 1 in Table 1 shows the results of a logit analysisusing onset as the dependent variable and a fairly fullspecification of independent variables discussed above.Prior war is a control variable indexing whether the

24 Thefigurewas produced usingRs locfit package,with a smoothingparameter of 0.9, and transforming annual probabilities of outbreaktofive-year equivalents. Thefigure looks highly similar ifweuseothermeasures of ethnic diversity, such as fractionalization.25 We do not code as ones years in which a civil war continues, whichwould be relevant if our focus were causes of war duration ratherthan onset. Nor do we drop country-years that have an ongoingwar (as in Collier and Hoeffler 2001), since this would omit the14 wars in our data that start while another war is in progress. Otherapproaches to the dependent variable and estimationsuch as mak-ing the dependent variable 1 for all war years and using dynamicprobit or Poisson models that interact independent variables with alagged dependent variable (Jackman 2001; Przeworski et al. 2001)produce virtually identical results. Only one country-year in the datahas more than one onset (Soviet Union 1946, with one Ukrainianand three Baltic rebellions). We treat this as a one in the analysesreported here, though there are no significant changes if we usePoisson regression instead (treating the dependent variable as acount) or weight this observation four times in the logit likelihood.Using rare events logit (King and Zeng 2001) had no appreciableimpact on any estimates reported.

82


FIGURE 2. Probability of Civil War Onset per Five-Year Period

country had a distinct civil war ongoing in the previousyear.26

Per Capita Income. Per capita income (measured asthousands of 1985 U.S. dollars and lagged one year) isstrongly significant inbotha statistical anda substantivesense: $1,000 less in per capita income is associatedwith41% greater annual odds of civil war onset, on average.Holding other variables at their median values, a coun-try in the tenthpercentile on incomehas an 18%chanceof a civil war outbreak over a decade, compared to an11% chance for a country at the median income anda 1% chance for a country at the ninetieth percentile($573, $1,995, and $9,505, respectively). The incomevariable is not just a proxy for theWest,whose states

26 With onset as the dependent variable, the data are grouped dura-tion data and we need to consider the possibility of temporal depen-dence between observations. One approach is dynamic probit; seefootnote 25. We also tried Beck et al.s (1998) method of includingdummies for each successive peace year prior to an onset, or fittingnatural cubic splines. This had no appreciable effect on the estimates,and the splines were jointly insignificant.

might have low rates of civil war for reasons of cultureor history that have little to do with income. The esti-mated coefficient drops only to 0.28 when a dummyfor theWest is included and remains strongly significantdespite the high correlation between the two variables.The relationship holds even within the poorest regions.Among the (mainly) former colonies of Africa, theMiddle East, and Asia, we estimate that $1,000 lessin income corresponds to 34% greater annual odds ofoutbreak.27

Ethnic and Religious Composition. The estimatesfor theeffectof ethnicand religious fractionalizationaresubstantively and statistically insignificant. Alternativemeasures of ethnic and religious diversitysuch as the

27 Contrary toHibbs 1973 andHegre et al. 2001, adding the square ofper capita income does not significantly improve the fit of the model.Regarding H2, the interaction of income and ELF has the expectedsign but is not significant. For lack of space, we report the details ofthese and other nonresultsmentioned below in the tables includedin Fearon and Laitin 2003.

83


TABLE 1. Logit Analyses of Determinants of Civil War Onset, 194599Model

(4) (5)(1) (2) (3) Civil War Civil War

Civil War Ethnic War Civil War (Plus Empires) (COW)Prior war 0.954 0.849 0.916 0.688 0.551

(0.314) (0.388) (0.312) (0.264) (0.374)Per capita incomea,b 0.344 0.379 0.318 0.305 0.309

(0.072) (0.100) (0.071) (0.063) (0.079)log(population)a,b 0.263 0.389 0.272 0.267 0.223

(0.073) (0.110) (0.074) (0.069) (0.079)log(% mountainous) 0.219 0.120 0.199 0.192 0.418

(0.085) (0.106) (0.085) (0.082) (0.103)Noncontiguous state 0.443 0.481 0.426 0.798 0.171

(0.274) (0.398) (0.272) (0.241) (0.328)Oil exporter 0.858 0.809 0.751 0.548 1.269

(0.279) (0.352) (0.278) (0.262) (0.297)New state 1.709 1.777 1.658 1.523 1.147

(0.339) (0.415) (0.342) (0.332) (0.413)Instabilitya 0.618 0.385 0.513 0.548 0.584

(0.235) (0.316) (0.242) (0.225) (0.268)Democracya,c 0.021 0.013

(0.017) (0.022)Ethnic fractionalization 0.166 0.146 0.164 0.490 0.119

(0.373) (0.584) (0.368) (0.345) (0.396)Religious fractionalization 0.285 1.533 0.326 1.176

(0.509) (0.724) (0.506) (0.563)Anocracya 0.521 0.597

(0.237) (0.261)Democracya,d 0.127 0.219

(0.304) (0.354)Constant 6.731 8.450 7.019 6.801 7.503

(0.736) (1.092) (0.751) (0.681) (0.854)N 6327 5186 6327 6360 5378

Note: The dependent variable is coded 1 for country years in which a civil war began and 0 in all others. Standard errors are inparentheses. Estimations performed using Stata 7.0. p < .05; p < .01; p < .001.a Lagged one year.b In 1000s.c Polity IV; varies from 10 to 10.d Dichotomous.

proportion of the largest group and the log of the num-ber of languages spoken by at least 1%prove to bejust as unrelated. The ethnic diversity measures showa strong bivariate relationship with civil war onset (notso for the religion measures), but this evaporates whenwe control for income.28Nor are countries that are ethnically or religiously

polarized in the sense of H3 more likely to experiencemajor civil violence.Whenwe add dummy variables forcountries that have an ethnic or religious majority anda minority of at least 8% of the countrys population,both are incorrectly signed and neither comes close tostatistical significance. This finding does not depend onwhich other variables are included in the model.

EthnicWar. The strong effect of per capita income re-mains even when we restrict attention to ethnic wars,

28 Collier and Hoeffler (2001) find the interaction of ethnic and reli-gious fractionalization to be negatively related to onset. Others findthat fractionalization has a nonmonotonic relation to conflict. Nei-ther relationship appears in these data; we suspect that one reasonmay be listwise deletion on missing economic data in other data sets.

while evidence of an independent effect of cultural di-versity is surprisingly weak. In Model 2 (Table 1), thedependent variable marks the onset of wars that wecoded as ethnicor partially ethnic, andwe consideronly countries with at least a 5% ethnic minority. Thecoefficient for ethnic fractionalization barely changes.The same is true for our othermeasures of ethnic diver-sity. Our twomeasures for religious diversity give weakand inconsistent results, and support for H7 diminishesfurther if we code the partiallyor ambiguously eth-nic wars as nonethnic.29

Democracy and Civil Liberties. Broad social and po-litical grievances should be lower, on average, in politi-cal democracies. But contrary to H4 and consistent withH11, civil war onsets are no less frequent in democraciesafter controlling for income, as shown by the positiveand statistically insignificant coefficient for democracy,

29 Coefficients for political instability and mountains diminish inthis subsample, partly because the associations appear to have beenweaker for ethnic wars and partly due to the omission of highly ho-mogeneous countries.

84


the (lagged)Polity IVmeasure.The results are the sameif we use Przeworski et al.s dichotomous measure in-stead (available for 195091) or the lag of FreedomHouses measure of observance of civil liberties (avail-able for 197299). In the former case the sign iswrong, while in the latter the sign is consistent withH4 but the coefficient on civil liberties is thoroughlyinsignificant.Some past studies of civil strife found an inverted-

U relationship with democracy and sought to explainthis by the observation that themost autocratic regimescan repress dissent and thus avoid civil violence despitefacing the highest levels of popular grievance (Hegreet al. 2001; Muller and Weede 1990). This observationdoes not explain why the leaders of a partially demo-cratic regime are not able to implement full autocracyor democracy to avoid conflict and opposition. As sug-gested in H10c, we suspect that the answer is often thatanocracies are weak regimes, lacking the resourcesto be successful autocrats or containing an unstablemix of political forces that makes them unable to moveto crush nascent rebel groups. Consistent with this hy-pothesis and with the prior studies, Model 3 (Table 1)shows that a dummy variable marking anocracies takesa positive coefficient. In substantive terms the estimatesuggests that these regimes have about 68% greaterodds of civil war outbreak in any given year than woulda full autocracy. This is so despite the fact that we arecontrolling for recent political instability, which ismuchmore common in anocracies.30

Linguistic and Religious Discrimination. Added toModel 3, our measures of state discrimination againstregional languages or against minority religions are notassociated with systematically higher risks of civil waronset.31 This nonresult persists when we restrict thesample to those countries with at least a 5% religiousor ethnic minority.

Inequality. Whether in a bivariate model or addedto Model 3, the Gini coefficient estimates of incomeinequality do not come close to either statistical or sub-stantive significance. The poor quality of the inequalitydata, available for only 108 countries, does not allowus to go beyond the claim that there appears to be nopowerful cross-national relationship between inequal-ity and onset, consistent with H11.

30 InModel 3Democracy is a dichotomous variablemarking regimesthat scored higher than 5 on the Polity scale; the excluded category isautocracies. ForModel 3we followedPolity IVs suggestion in codingregime interruptions as anocracies; if these values are instead in-terpolated, the resulting anocracy indicator gets a smaller estimatedcoefficient that is not quite significant at the 5% level, while that forpolitical instability increases. Another issue, raised by Hegre et al.(2001), is that the observation of some violence may lead Polity IVto code democracies as less democratic and autocracies as lessautocratic (since there is mobilized opposition). So if there is low-level violence preceding the reaching of our thresholds, some of theeffect of anocracy may be that it is picking up nascent civil war.31 We tried the several religion indicators both separately and as asinglemeasurebasedona factor analysis of the components.Discrim-ination against regional languages is in fact associated with slightlylower odds of civil war onset, consistent with Laitin 2000.

New States and Political Instability. Consistent withH10a, the odds of civil war onset are estimated as5.25 times greater in the first two years of a states in-dependent existence than in other years, a huge effect(Model 3). For the median country, this translates toan 11% chance in its first two years. Political instabilityat the center matters in other years as well. The oddsof onset in a given year are estimated to increase by67% if there was instability in governing arrangementsin any of the previous three years.32

Mountains andNoncontiguous Territory. Mountain-ous terrain is significantly related to higher rates ofcivil war. A country that is about half mountainous(ninetieth percentile) and otherwise at the median hasan estimated 13.2% chance of civil war over the courseof a decade. A similar country that is not mountainousat all (tenth percentile) has a 6.5% risk. The estimatedsign for stateswithnoncontiguous territory is consistentwith H10e, but its 95% confidence interval includes thepossibility of no effect.33

Population. Holding other variables at medians, theestimated risk of civil war over the course of a decadefor a country at the tenth percentile in population is6.4%, versus 16.4% for a country at the ninetieth per-centile (1.4 versus 56.3 million, e.g., Botswana versusIran). This effect is not due to large states being moreethnically diverse. We are controlling for ethnic diver-sity, and in any event it turns out that there is essen-tially no correlation between diversity measures andlog of size, even if we omit the relatively homogeneousChina.34

Oil, Islam, Young Males. Consistent with H10h, de-riving at least one-third of export revenues from fossilfuels is estimated tomore than double a countrys odds.Based on Model 3, the median country had a 10%chance of civil war over a decade, whereas the samecountry as an oil exporter would have an estimated21% chance.Oil producers are disproportionately Middle

Eastern, which raises the question of whether thisvariable might proxy some other relevant attribute ofthese countries. But a dummy for the North Africa/Middle East region is insignificant when added toModel 3, while the estimate for oil is barely affected.Huntington (1996) argues that Islam has bloodyborders and bloody innards because Islamicsocieties have cultural and demographic features thatmake them violence-prone. When we add a variable

32 Snyder (2000) argues that democratization puts states at greaterrisk for civil war, ethnic in particular. Coding variables for three-or-greater moves toward and away from democracy on the Polity index,we find in these data that moves away from democracy are muchmore strongly associated with civil war onset in the next year thanmoves toward it, though both forms of instability elevate the risk.Hegre et al. (2001) found the same thing using COW civil war data.33 Dropping this or other insignificant variables from Model 3barely affects the remaining estimated coefficients.34 Huntington (1996) argues that rapid population growth puts statesat greater risk for civil war. Using the average growth rate for thethree prior years, we find no significant impact.

85


measuring the percentage of Muslims in each countryto our main specification,35 it takes a positive sign butis not statistically significant. The effect of oil remainsstrong.Huntington also argues that societieswith a surfeit of

young males will be prone to civil violence, suggestingthis as a part of the explanation for Islams bloodyinnards. Given that young males have physical andperhaps psychological characteristics that make themapt guerrillas, our arguments on insurgency point in thesame general direction. If we include a lagged measureof the proportion of males aged 15 to 24 in the popula-tion in Model 3, it has the expected positive sign,36 butthe estimate is highly uncertain (p = .21); one problemis that percentage of young males has a strong negativecorrelation with income, and income gets the better ofitwhenboth are included.The same is true formale sec-ondary schooling rates (cf. Collier and Hoeffler 2001).

Foreign Support. One determinant of the prospectsfor insurgency is the availability of third-party supportto either the rebels or the government of the state inquestion. Availability is difficult to observe ex ante,however. A possible proxy is the number of civil warsongoing in neighboring countries in the previous year,which might increase the availability of arms, support,and seasoned guerrillas. Adding the lagged version ofthis variable to Model 3, its sign is as predicted, but itis not significant (p = .33). Civil war next door is toocommon, occuring in 44% of the country-years in thesample.Another possible proxy is foreign support to govern-

ments against domestic challengers. A dummy variablefor Soviet blocprior to 1991 cannot be included in thelogit analysis because it perfectly predicts peace. But byitself this observation is consistentwithour argument.37In the late 1980s, Mitterands government departed

from long-standing French foreign policy by support-ing, to a limited extent, democratization in some of itsformer sub-Saharan colonies. This involved encourag-ing national conferences, elections, and some politi-cal opposition, which all suggested that the prior policyof immediatemilitary support for French-speaking dic-tators might have changed. Civil war rates for for-mer French African colonies had been much lowerthan those for other sub-Saharan states. The 25 otherssaw 17 civil war onsets before 1990, a rate of 2.4 per100 country-years. The 16 former French colonies hadonly three onsets in the same period, for a rate of0.65 per 100 country-years (Chad in 1965, Senegal[Casamance] in 1989, and Mali in 1989). If a variablemarking former French colonies in sub-SaharanAfricaprior to 1990 is included in Model 3, its estimated co-efficient is 1.0, which implies an almost-threefold re-duction in the annual odds of civil war onset (p = .11;

35 Coded from the CIA Factbook and a variety of country-specificsources.36 Data from theWorld Bank, linearly interpolated for missing yearswithin countries.37 Foreign support may be at least as important as communist partystructure inexplainingHuntingtons (1968)observation that the com-munist countries rarely faced insurrections.

the other estimates remain stable). A dummy variablefor all former French colonies is unrelated to civil waronset, which suggests that the observed effect in sub-Saharan Africa is due to the mechanism behind H10frather than the effects of French colonial administra-tion or law.

Other Variables and Robustness ChecksAnticolonial Wars. Figure 2 and the multivariateanalyses above omitted the 13 anticolonial wars in fivecolonial empires (Britain, France, Belgium, Portugal,and The Netherlands). Including them is problematicfor both theoretical and empirical reasons, but so isignoring them. The main practical obstacle is the lackof comparable data. However, by using the ethnicfractionalization, population,mountainousness, and in-come data for each colony in its first year of inde-pendence, we were able to produce annual estimatesof these variables for each empire. The income esti-mates are systematically toohighweuse, for example,Nigerian GDP per capita in 1962 as an estimate of itsincome as a British colony in 1945. This creates somebias against the incomevariable and in favorof theELF,which has a median value of 0.83 for the 114 empireyears, compared to 0.34 for nonempire years.Model 4 reports a logit analysis parallel to our speci-

fication inModel 1 (Table 1), omitting only the democ-racy variable and religious fractionalization (which ishard to estimate for the empires due to shared religionsbetween colonies and metropoles). Given that the em-pires were quite civil war-prone and highly diverse, itis not surprising that the estimated coefficient for ELFincreases. Surprisingly, though, it remains statisticallyindistinguishable from zero, while, excepting nonconti-guity, the others change little. Noncontiguous states arenow estimated to have had 2.2 times the odds of civilwar in this period, other thingsbeing equal.Because theempires areall noncontiguousandbecausenoncontigu-ity otherwise does a better job than the ELF at pickingout civil wars onsets, it is given greater substantive andstatistical significance by the logit. If we drop noncon-tiguity, the estimate for ELF increases to 0.62whichcorresponds to a 59% increase in risk moving from thetenth to the ninetieth percentilebut still fails statis-tical significance at the 5% level (p= .07). Thus evenif we include the anticolonial wars, the evidence thatethnic diversity directly causes states to be more civilwar prone remains slim at best.38Stepping back from the statistical analysis, we ob-

serve that these five imperial states managed theirempires for many years before World War II withfairly little active rebellion despite the same levels ofethnic fractionalization, noncontiguity, and mountain-ousness. Nor did discrimination and oppression sud-denly increase in the empires after the war, causing

38 We created a democracy estimate for the empires by taking aweighted average of the metropoles lagged Polity democracy scoreand 10, weighting by the metropoles proportion of total imperialpopulation. Adding this or dummies for anocracy and democracyto Model 5 has no substantive impact.

86


anticolonial movements as a response. Rather, the wargreatly weakened the main imperial states materially,and the new international legal order (theU.N. system)gave support to anticolonial movements. These shocksto the relative power of the metropoles help explaincivil war onset in these cases (Fearon 2001).

Regional Effects. Different regions of the worldshare a variety of historical, cultural, and economictraits. It is reasonable to wonder if any of the variablesconsidered in the multivariate analysis just proxy forsuch factors. Further, if regional dummy variables donot add to the models explanatory power, then wehave managed to account for the influence of regionalcharacteristicswithourmoregeneral independent vari-ables.We find that if we add all (but one) regional dum-mies toModel 3, the coefficients and significance levelsfor our other variables are little affected. A likelihood-ratio test fails to reject the null hypothesis that re-gions do not matter beyond the included variables(p = .22). Including region dummies individually re-veals that none has a rate of civil war onset significantlydifferent from what one would expect on the basisof the country characteristics already included in themodel.

Period and Fixed Effects. If added to Model 3,dummy variables marking each decade (but one) arejointly significant in a likelihood-ratio test (p= .04),showing evidence of a general upward trend in civil warrisk after the 1940s. Adding a dummy for the 1940s anda variable marking the year indicates that from 1950the odds of civil war outbreak rose about 2% per year,controlling for the other variables. As noted earlier,there is no secular trend when we do not control forother factors. Part of the explanation is that themedianof country incomes roughly doubled from 1950 to 1999,which should havebased onModel 3reduced theonset odds for the median country by about 35%.The fact that the rate of outbreak remains fairly con-stant while incomes generally increased in effect pe-nalizes income in Model 3. When year is added, theeffect estimate for income increases to0.36 (the othereffect estimates stay steady, except for noncontiguity,which becomes statistically significant at p= .035). Wedo not knowwhat accounts for steady rates of outbreakdespite increasing incomes. Increased international in-equality (so that there is a set of persistently poor coun-tries at risk for civil war), greater availability of smallarms, and diffusion of insurgency know-how come tomind, but are just conjectures.This observation also raises the question of how

much the estimated impact of income is due to cross-country comparisons and how much to increasingincome within given countries associating with civilpeace. Applying conditional fixed effects logit toModel 3without the variables that have little or no tem-poral variation (mountains, noncontiguity, ethnic andreligious fractionalization) and with a control for year,we find that the estimated coefficient for income dropsonly to 0.24 but is no longer significant (p= .16).Using log of income, however, returns the same co-

efficient that log of income takes in the standard logiton the full sample (0.78), significant at p= .02.39 Inaddition, the effect estimates for the other variablesare, with one exception, virtually identical to the es-timates in the full, pooled model, the significanceremaining high for all but anocracy and oil (which haslittle temporal variation). The exception is population,which varies vastly more across countries than withinthem over time. While bigger countries are more civilwar-prone, there is no evidence that population growthwithin a country raises the risk.

Primary Commodity Exports. Using an expandedversion of the COW data, Collier and Hoeffler (2001)find that the risk of civil war onset is maximized whenprimary commodity exports comprise about 32%of thecountrys GDP, with risk declining on either side of thisfigure. They propose that, up to a point, higher levelsof primary commodity exports provide more motiva-tion and opportunity for rebels to support themselvesthrough looting.Beyond this point, the revenue gainsavailable to the state are said to be large enough tomake civil war less likely.We find little evidence of such a relationship in our

data. Neither the share of primary commodity exportsin GDP nor its square is remotely significant whenadded to Model 3.40We agree with the general argument behind Collier

and Hoefflers hypothesis. In our terms, insurgencyshould bemore feasible if sources of financing are read-ily available. But we doubt that primary commodityexports are a good measure of financing potential forrebels. This measure combines oil, agricultural prod-ucts, and metals and minerals, which, except for someminerals, are hard to exploit without control of a na-tional distribution system and ports.Oil exportsmay berelevant not somuch because they finance rebel groupsbut, as we argued, because they mark relative stateweakness at a given level of income.41 Abettermeasurefor financing potential would focus on the presence ofminerals or contraband that can reward control of asmall enclave with huge profits (Leonard and Strauss2001).

Trade Openness. Using a measure of state failurethat includesboth civilwar anddisruptive regime tran-sitions, Esty et al. (1998) found that trade share ofGDP strongly predicted civil peace. Using the PennWorld Tables measure of trade as a share of the GDP,wefindno such relationship in our data. Trade is relatedto civil peace in a bivariate logit, but this is becausesmaller countries have more trade and less civil war.

39 Whether we use income or logged income in the standard logitmakes for no significant differences.40 Thedata seriesonprimary commodityexportsbegins in1960and isavailable only at five-year intervals. We interpolated missing valuesand extended the 1995 values through 1999. We also tried Collierand Hoefflers procedure of grouping the data by five-year periodsbeginning in 1960, again finding no relationship.41 Our oil variable is moderately correlated with the primary com-modities measure, at r = .46, but the latter and its square remaininsignificant when oil is dropped from the model.

87


Alternative Definitions and Coding for Civil War.Our coding rules admit civil conflicts in which at least1,000 have been killed. Some might argue that thisthreshold is too low to distinguish properly betweencivil war and, say, terrorism. When we recode the 30conflicts in our sample estimated to have killed fewerthan 5,000 as zeros, the results for Model 3 are nearlyidentical.42We also checked to see how our results differed if

we based the coding of civil war onset on the listsgiven by Collier andHoeffler (2001), the COWproject,and Doyle and Sambanis (2000). As shown in Model 5(Table 1), which bases onset on the COW data (194592), the coefficients and significance levels for the in-surgency variables are remarkably stable.43 We obtainquite similar results for the other two coding schemes.Although these four civil war lists differ on many spe-cific start years and, to some extent, on what cases con-stitute civil wars at all, there proves to be substantialshared variation. For example, 53 of the 93 COW civilwars in the sample start in the same year as an onsetin our data, and 68 start within two years of one of ouronsets. At the country level, the bivariate correlationbetween our estimate of the total number of wars bycountry and the equivalent for the others ranges from0.71 with Collier and Hoeffler to 0.82 with Doyle andSambanis.

CONCLUSIONThe prevalence of internal war in the 1990s is mainlythe result of an accumulation of protracted conflictssince the 1950s rather than a sudden change associatedwith a new, post-ColdWar international system.Decol-onization from the 1940s through the 1970s gave birthto a large number of financially, bureaucratically, andmilitarily weak states. These states have been at riskfor civil violence for the whole period, almost entirelyin the form of insurgency, or rural guerrilla warfare.Insurgency is a mode of military practice that can beharnessed to various political agendas, be it commu-nism in Southeast Asia and Latin America, Islamicfundamentalism in Afghanistan, Algeria, or Kashmir,right-wing reaction in Nicaragua, or ethnic national-ism in a great many states. The conditions that favorinsurgencyin particular, state weakness marked bypoverty, a large population, and instabilityare betterpredictors of which countries are at risk for civil warthan are indicators of ethnic and religious diversity ormeasures of grievances such as economic inequality,lack of democracy or civil liberties, or state discrimina-tion against minority religions or languages.

42 The coefficient on income increases 21%, noncontiguity becomessignificant, and anocracy weakens slightly. We coded total deathsourselves, based on Brogan 1998, Institute for International andStrategic Studies 2000, Sivard 1996, Valentino 2002, and country-specific sources.43 The one exception is religious fractionalization, which just man-ages significance at the 5% level here. This does not occur in theother data sets, and even with the COW data the estimate for reli-gious fractionalization is highly sensitive to which other variables areincluded in the model.

How could democracy and cultural or religious ho-mogeneity fail to associate with civil peace across coun-tries? Viewing ethnic wars as a species of insurgencymay help explain this paradoxical result. If, under theright environmental conditions, just 500 to 2,000 ac-tive guerrillas can make for a long-running, destruc-tive internal war, then the average level of grievancein a group may not matter that much. What matters iswhether active rebels can hide from government forcesand whether economic opportunities are so poor thatthe life of a rebel is attractive to 500 or 2,000 youngmen. Grievance may favor rebellion by leading nonac-tive rebels to help in hiding the active rebels. But allthe guerrillas really need is superior local knowledge,which enables them to threaten reprisal for denun-ciation.If our analysis is correct, then policy makers should

not assume that civil wars and the failed states theysometimes produce are temporary phenomena of theimmediate post-Cold War world. Nor should policymakers or academics infer that ethnic diversity is theroot cause of civil conflict when they observe insur-gents in a poor country who mobilize fighters alongethnic lines. Instead, the civil wars of the period havestructural roots, in the combination of a simple, robustmilitary technology and decolonization, which createdan international systemnumerically dominatedby frag-ile states with limited administrative control of theirperipheries.Regarding policy implications, the spread of democ-

racy and tolerance for ethnic and religious minoritiesshould be major foreign policy goals because they aredesirable for their own sake, but not with the expec-tation that they are magic bullets for the preventionor resolution of civil war. Sometimes recommended asa general international policy for resolving ethnic civilwars (e.g., Kaufmann 1996), ethnic partitions should beviewed as having large international implications andhigh costs. International support for partition wouldincrease the expected benefits for rebels, who, we haveargued, may be able to get a nasty civil war going onthe basis of small numbers when the conditions for in-surgency are right.Policies to redress grievances, or, in the limit, parti-

tion, could be important to resolve ongoing conflicts.We cannot say on the basis of this research, which fo-cused on civil war onset rather than termination. Wefind little evidence that civil war is predicted by largecultural divisions or broadly held grievances. But itseems quite clear that intense grievances are producedby civil warindeed, this is often a central objectiveof rebel strategy. These could well pose obstacles tosettlement.Regarding prevention, our analysis suggests that

while economic growth may correlate with fewer civilwars, the causal mechanism is more likely a well-financed and administratively competent government.In specific terms, international and nongovernmen-tal organizations should develop programs that im-prove legal accountability within developing worldmilitaries and police, and make aid to governmentsfighting civil wars conditional on the state observing

88


counterinsurgency practices that do not help rebels re-cruit militias. Governments that follow horrible, war-perpetuating counterinsurgency practices or are so cor-rupt as to be hopeless should be left on their own or,when there aremajor implications for regional stabilityor international terrorism, be viewed as candidates forneotrusteeship under theUnitedNations or regionalmilitary and political organizations such as NATO andthe European Union. The latter system, which we al-ready see operating, in effect, in Bosnia, Kosovo, andEast Timor, should be rationalized so as to improveinternal coordinationamong themanyplayers involvedin such operations.

REFERENCESAnderson, Benedict. 1983. Imagined Communities. London: Verso.Asher, R. E., ed. 1994. The Encyclopedia of Language and Linguis-tics. Oxford: Pergamon Press.

Atlas Narodov Mira. 1964. Moscow: Glavnoe upravlenie geodezii ikartografii.

Avant, Deborah D. 1994. Political Institutions and Military Change.Ithaca, NY: Cornell University Press.

Beck, Neal, Jonathan N. Katz, and Richard Tucker. 1998. TakingTime Seriously: Time-Series-Cross-SectionAnalysis with a BinaryDependent Variable.American Journal of Political Science 42 (4):126088.

Brogan, Patrick. 1998. World Conflicts. 3rd ed. Lanham, MD:Scarecrow Press.

Chaudhry,KirenAziz. 1989. ThePriceofWealth:Business andStatein Labor Remittance and Oil Economies. International Organi-zation 43 (1): 101145.

Clutterbuck, Richard L. 1967. The Long, Long War: The Emergencyin Malaya. London: Cassell.

Collier, Paul, and Anke Hoeffler. 1999. Justice Seeking andLoot-seeking in Civil War. World Bank. Typescript. http://econ.worldbank.org/programs/conflict/library (November 18, 2002).

Collier, Paul, and Anke Hoeffler. 2001. Greed and Grievancein Civil War.World Bank. Typescript. http://econ.worldbank.org/programs/library (November 18, 2002).

Connor, Walker. 1994. Ethnonationalism. Princeton, NJ: PrincetonUniversity Press.

Deininger, Claus, and Lyn Squire. 1996. ANewData SetMeasuringIncome Inequality.World Bank Economic Review 10 (3): 56591.

Desai, Raj, and Harry Eckstein. 1990. Insurgency: The Transforma-tion of Peasant Rebellion.World Politics 42 (4): 44166.

Deutsch, Karl W. 1953. Nationalism and Social Communication.Cambridge, MA: MIT Press.

Doyle, Michael W., and Nicholas Sambanis. 2000. Interna-tional Peacebuilding: A Theoretical and Quantitative Analysis.American Political Science Review 94 (4): 778801.

Easterly, William, and Ross Levine. 1997. Africas Growth Tragedy:Policies and Ethnic Divisions. Quarterly Journal of Economics112 (4): 120350.

Eisinger, Peter. 1973. The Conditions of Protest Behavior inAmerican Cities. American Political Science Review 67: 1128.

Esty,Daniel C., JackA.Goldstone, TedR.Gurr, BarbaraHarff,MarcLevy, Geoffrey D. Dabelko, Pamela T. Surko, and Alan N. Unger.1998. State Failure Task Force Report: Phase II Findings. McLean,VA: Science Applications International Corporation.

Fearon, James D. 1998. Commitment Problems and the Spread ofEthnicConflict. InThe International SpreadofEthnicConflict, ed.David A. Lake and Donald Rothchild. Princeton, NJ: PrincetonUniversity Press.

Fearon, James D. 2001. Why Do Some Civil Wars Last So MuchLonger thanOthers?Presented at theWorldBankDECRGCon-ference on Civil War held, University of California, Irvine.

Fearon, James D. 2002. Ethnic Structure and Cultural Diversityaround the World. Presented at the Annual Meetings of theAmerican Political Science Association, Boston.

Fearon, JamesD., andDavidD. Laitin. 1996. Explaining InterethnicCooperation. American Political Science Review 4: 71535.

Fearon, James D., and David D. Laitin. 1999. Weak States, RoughTerrain, and Large-Scale Ethnic Violence since 1945. Presentedat the Annual Meetings of the American Political Science Associ-ation, Atlanta, GA.

Fearon, James D., and David D. Laitin. 2003. Additional Tables forEthnicity, Insurgency, and Civil War. http://www.stanford.edu/group/ethnic/.

Gellner, Ernest. 1983. Nations and Nationalism. Ithaca, NY: CornellUniversity Press.

Gleditsch, Nils, Havard Strand, Mikael Eriksson, MargaretaSollenberg, and Peter Wallensteen. 2002. Armed Conflict 19462001: A New Dataset. Journal of Peace Research 39 (5): 61537.

Griffith, Samuel B., ed. 1961. Mao Tse-Tung on Guerrilla Warfare.New York: Praeger.

Grimes, Joseph E., and Barbara F. Grimes. 1996. Ethnologue: Lan-guages of the World. 13th ed. Dallas, TX: Summer Institute ofLinguistics.

Gunnemark, Erik V. 1992. Countries, Peoples and Their Languages.Gothenburg, Sweden: Geolingua.

Gurr, Ted R. 1971. Why Men Rebel. Princeton, NJ: Princeton Uni-versity Press.

Gurr, Ted R. 2000. Peoples versus States. Washington, DC: UnitedStates Institute of Peace Press.

Hegre, Havard, Tanja Ellingsen, Scott Gates, and Nils PetterGleditsch. 2001. Toward a Democratic Civil Peace? Democracy,Political Change and Civil War, 19161992. American PoliticalScience Review 95 (March): 3348.

Hibbs, Douglas A. 1973.Mass Political Violence. New York: Wiley.Horowitz, Donald L. 1985.EthnicGroups in Conflict. Berkeley: Uni-versity of California Press.

Huntington, Samuel P. 1968. Political Order in Changing Societies.New Haven, CT: Yale University Press.

Huntington, Samuel. 1996.TheClash ofCivilizations and theRemak-ing of World Order. New York: Simon and Shuster.

Ignatieff, Michael. 1993. Blood and Belonging. London: NoondayPress.

Institute for International and Strategic Studies. 2000. The MilitaryBalance, 2000. London: IISS.

Jackman, Simon. 2001. In andOut ofWar and Peace: The StatisticalAnalysis of Discrete Serial Data on International Conflict. Ms.,Stanford University.

Kalyvas, Stathis. 1999. Wanton and Senseless? The Logic ofMassacres in Algeria. Rationality and Society 11 (3): 24385.

Karl, Terry L. 1997. The Paradox of Plenty. Berkeley: University ofCalifornia Press.

Kaufmann, Chaim. 1996. Possible and Impossible Solutions to Eth-nic Civil Wars. International Security 20 (4): 13675.

Keen, David. 1998. The Economic Functions of Violence in CivilWars. London: Oxford University Press.

Kennedy-Pipe, Caroline. 1997. The Origins of the Present Troublesin Northern Ireland. London: Longman.

King, Gary, and Langche Zeng. 2001. Logistic Regression in RareEvents Data. Political Analysis 9 (2): 13763.

Krepinevich, Andrew F. 1986. The Army and Vietnam. Baltimore,MD: Johns Hopkins University Press.

Kriger, Norma J. 1992. Zimbabwes Guerrilla War. New York:Cambridge University Press.

Laitin, David D. 2000. Language Conflict and Violence. In Inter-national Conflict Resolution After the Cold War, ed. Paul C. SternandDanielDruckman.Washington,DC:NationalAcademyPress.

Laqueur,Walter. 1976.GuerrillaWarfare.NewBrunswick,NJ:Trans-action Books.

Leites, Nathan, and Charles Wolf. 1970. Rebellion and Authority.Chicago, IL: Markham.

Leonard, David K., and Scott Strauss. 2001. Power without Respon-sibility: The International Community and Africas Failed Devel-opment. Book ms., University of California, Berkeley.

Licklider, Roy A. 1995. The Consequences of Negotiated Settle-ment in Civil Wars, 194593. American Political Science Review89 (3): 68190.

Moynihan, Daniel P. 1993.Pandaemonium: Ethnicity in InternationalPolitics. New York: Oxford University Press.

Muller, Edward N. 1985. Income Inequality, Regime Repres-siveness, and Political Violence. American Sociological Review50: 4761.

89


Muller, Edward N., and Erich Weede. 1990. Cross-National Varia-tions in Political Violence: A Rational Action Approach. Journalof Conflict Resolution 34: 62451.

Paige, Jeffery M. 1975. Agrarian Revolution. New York: Free Press.Przeworski, Adam, Michael E. Alvarez, Jose Antonio Cheibub,and Fernando Limongi. 2000. Democracy and Development.Cambridge: Cambridge University Press.

Rabushka,Alvin, andKennethA. Shepsle. 1972.Poli

fearon, james d, and david d laitin_2003_ethnicity, insurgency, and civil war

Documents