Secessionism in Multicultural States: Does Sharing …othello.alma.edu/~cartrite/abm/lusmioeid.pdfSecessionism in Multicultural States: Does Sharing Power Prevent ... (Cederman 2001;MacyandWiller2002).7

American Political Science Review Vol. 98, No. 2 May 2004

Secessionism in Multicultural States: Does Sharing Power Preventor Encourage It?IAN S. LUSTICK, DANMIODOWNIK, and ROY J. EIDELSONUniversity of Pennsylvania

Institutional frameworks powerfully determine the goals, violence, and trajectories of identitarianmovements—including secessionist movements. However, both small-N and large-N researchersdisagree on the question of whether “power-sharing” arrangements, instead of repression, are more

or less likely to mitigate threats of secessionist mobilizations by disaffected, regionally concentratedminority groups. The PS-I modeling platform was used to create a virtual country “Beita,” containingwithin it a disaffected, partially controlled, regionally concentrated minority. Drawing on constructivistidentity theory to determine behaviors by individual agents inBeita, themost popular theoretical positionson this issue were tested. Data were drawn from batches of hundreds of Beita histories produced underrigorous experimental conditions. The results lend support to sophisticated interpretations of the effects ofrepression vs. responsive or representative types of power-sharing. Although in the short run repressionworks to suppress ethnopoliticalmobilization, it does not effectively reduce the threat of secession. Power-sharing can be more effective, but it also tends to encourage larger minority identitarian movements.

In a world of states and dominated by states it is un-surprising that themaintenance of state boundarieswould appear as a vital problemand that “state con-

traction,” “secession,” or “partition” would be figured,by most scholars and politicians, as evidence of publicpolicy failure or as desperately exercised options of lastresort. To be sure, in the 1990s some scholars revivedinterest in territorial self-determination via partitionof existing states as a sometimes useful policy optionfor individual states and for the international commu-nity.Against a backgroundof severepolitical instabilityin the Balkans, central Asia, Africa, the Middle East,and the former Soviet Union, it has been suggestedthat some political conflicts might be managed best by

Ian S. Lustick is Bess W. Heyman Professor, Political ScienceDepartment, University of Pennsylvania, and Associate Direc-tor, Solomon Asch Center for Study of Ethnopolitical Conflict,University of Pennsylvania. Address: Department of Political Sci-ence, University of Pennsylvania, Philadelphia, PA 19104-6215([email protected]).Dan Miodownik is a Ph.D. candidate in Political Science, Uni-

versity of Pennsylvania. Address: Department of Political Sci-ence, University of Pennsylvania, Philadelphia, PA 19104-6215([email protected]).Roy J. Eidelson, Ph.D., is a clinical psychologist and Execu-

tive Director, Solomon Asch Center for Study of EthnopoliticalConflict, University of Pennsylvania. Address: Solomon Asch Cen-ter, 3819-33 Chestnut Street, Philadelphia, PA 19104 ([email protected]).The authors gratefully acknowledge the crucial contributions pro-

vided by Vladimir Dergachev and Ben Eidelson in the software de-velopment and applications used in this research. In addition, weexpress our thanks for their comments to Amel Ahmed, KanchanChandra, Edward Gibson, Jason Kirk, David Laitin, BrendanO’Leary, David Rousseau, Maurits van der Veen, participants in thePenn Agent-Based Modeling Laboratory (PAMLA), members ofthe Laboratory in Comparative Ethnic Processes (LiCEP), and thisjournal’s anonymous reviewers. For their financial support we thankthe Carnegie Corporation of New York (Award B-7081), the Na-tional Science Foundation (Award SES-0218397), and the SolomonAsch Center. Color versions of the figures in this article are avail-able at http://www.polisci.upenn.edu/abir/SecessionFigures.pdf. Forreplication purposes, the software, manual, templates, and scriptsused to produce the data analyzed in this article are available athttp://www.polisci.upenn.edu/abir/SecessionReplication.zip.

“rightsizing” states, i.e., adjusting their boundaries orcreating new states (Hoppe 1998;O’Leary, Lustick, andCallaghy2001).Othershaveemphasized theoccasionalnecessity for forcible partitions and even populationtransfers to achieve a correspondence between eth-nopolitical or sectarian allegiances and the contours ofstates legitimized by them (Kaufmann 1998; Tullbergand Tullberg 1997). On balance, however, the bulk ofthe scholarly and public policy community continuesto oppose such approaches on moral, legal, practical,and other grounds. Instead they direct much more at-tention to how conflict, especially violent conflict, canbe managed while protecting existing state boundaries(Carley 1997;Horowitz 1985, 588–92; 1997, 435; Kumar1997; Sambanis 2000).Many factors, considered independently or in inter-

action, have been prominent in recent studies of the eti-ology of secessionism. They include the implications ofeconomic advantage/disadvantage, topography, worldregion, demographic patterns, globalization, culturaldistinctiveness, inter-group antipathy, type of identitiesin conflict, political entrepreneurship, and outside in-tervention by irredentist or culturally related powers.In this paper, however, we focus specifically on one keythread in this sprawling conversation—the relation-ship between institutionalized empowerment of poten-tially secessionist groups and the appearance of seces-sionism.Indeed it can be argued that the single most popular

line of argument offered by scholars to policy makershas been to suggest political and institutional arrange-ments to satisfy demands by whatever regional, reli-gious, ethnic, or other groups with secession potentialappear to threaten the integrity of the existing terri-torial states. Under this rubric various techniques andapproaches have been elaborated, including affirma-tive action, multicultural liberalism, federalism, auton-omy, cantonal arrangements, and power-sharing (e.g.,Danspeckgruber 1996; Lapidoth 1996; Lijphart 1977,1985; Sambanis 2000). The general view here is that byrespondingpositively and integratively, if only partially,

209

Secession in Multicultural States May 2004

to the demands of disgruntled minorities, secessionismcanbeabatedand secessionpreventedwhilepreservingthe predominance and stability of the central state. Bymaking governmentmore responsive to the concerns ofdisgruntled minorities, potentially secessionist groupswill be encouraged to feel confident of representationand protection for their most vital concerns. Such in-stitutional responses by the central state are deemedcapable, if designed and implemented properly, of re-ducing the intensity of separatist demandsby thosewhootherwise might make them. In Hirschman terms, theimpetus for exit is to be blunted by providing oppor-tunities for voice and reasons for loyalty (Hirschmann1970).However the opposite view is also strongly argued—

that creating autonomous, federal, or otherwisedevolved institutions of self-government or self-administration, especially if they allow regionally con-centrated groups to mobilize within them, is liableto contribute to secessionism by affording elites andgroups the political resources they need to undertakemass mobilization and wage separatist struggles. Thisliterature supports a widely acknowledged fear amongstate elites that granting regionally concentrated mi-norities special forms of autonomy, devolved powers,or privileges within a system of asymmetric federalismwould not so much assuage demands for representa-tion, control, or resources, as lead the country down aslippery slope to separatism (Cornell 2002; Hale 2000;Mozaffar and Scarritt 2000; Roeder 1991).Adistinct but related line of argumentwas spurred in

part by the collapse of the Soviet bloc, successful seces-sion (or partition) in Yugoslavia and Czechoslovakia,and eruption of secessionist conflicts in many parts ofthe former Soviet Union. Some researchers during thelast decade have argued that democratization mightnot be able to be fine-tuned enough to prevent devo-lution and empowerment from producing more ratherthan less secessionism and attendant violence. Fromthis position there is a tendency to minimize the im-portance of minority group dissatisfaction with cen-tral state policies as an explanation for secessionism.Instead, a “supply-side” interpretation is favored. Onthis account, illegal mobilization against the state forsecessionist purposes does not increase as a function ofdemands for more responsiveness, democracy, or au-tonomy. Rather it arises when the human and materialwherewithal to make such risky behavior profitable isavailable to political entrepreneurs and the small groupof activists actually engaged inmobilizational or violentactivities (Collier 2000; Fearon and Laitin 2003; Laitin2001).Another formof theposition that dangerous instabil-

ities arise from the abilities of peoples to rebel, ratherthan from the depredations of states, has been articu-lated by researcherswho point to apparent correlationsbetween democratization and the eruption of severeethnic violence and secessionism. Jack Snyder offers aneo-Huntingtonian argument that stresses the impor-tance of establishing strong political institutions priorto democratization to prevent demands for participa-tion from disrupting the political stability of the state.

Whereas Donald Horowitz used peaceful interethnicaccommodation in Malaysia and violent secessionisminSriLanka to support his thesis that properly designeddemocratic institutions prevent ethnic violence and se-cessionism, Snyder (2000), writing in the late 1990s,argues that Malaysia illustrates the “advantages of au-thoritarianism” while Sri Lanka shows the “dangers ofdemocratization” (275, 280).1This line of analysis reflects a subtheme in the litera-

ture on containing potentially secessionist and violentethnic mobilizations via repression, control, or domi-nation. The idea that “repression works” is implicit inthe argument that states can maintain their borders,prevent secessionism, and preserve the political posi-tion of ascendant groups by refraining from offers ofinstitutional compromises or added resources to ac-commodate out-group demands (Collier 2000; Laitin2001).2The current state of the scholarly debate regarding

the general relationship between institutionalized em-powerment of potentially secessionist regional groupsis confused. Hechter and Okamoto address this head-on in their 2001 metastudy. They observe that “thereis little consensus about the kinds of political insti-tutions that are most likely to contain nationalism”(203). As an example, they describe three positionsheld by important groups of scholars regarding the ef-fect of federalism on nationalist mobilization: (1) fed-eralism reduces nationalism, (2) federalism increasesnationalism, and (3) federalism itself does not deter-mine strength of nationalism (204). Similar to otherof the more sophisticated approaches to this question,Hechter and Okamoto conclude that causal relation-ships between institutionalized empowerment of po-tentially secessionist minorities and the rise of seces-sionism are complex, nonlinear, and highly sensitive tocontext.Thus inhis ownworkHechter (2000) suggests apossible reconciliation of the two opposing argumentsin the literature: “Whereas decentralization may pro-vide cultural minorities with greater resources to engagein collective action . . . at the same time it may erode thedemand for sovereignty.”3

A similarly nuanced view of the curvilinear patternsassociated with different combinations of pairs of im-portant variables is advanced by Atul Kohli. He ar-gues that if the central authority structure is strongbut willing to be accommodating and responsive todemands by potentially secessionist minorities, the re-sult will be short-term increases in ethnopolitical mobi-lization but long-term decreases in likelihood that thestate will face potent secessionist threats. Unaccom-modating strong states can expect continuing cycles ofmobilization and repression.Weakbut accommodating

1 Snyder makes a similar argument about patterns of political vi-olence and secessionism in India and the “perils of pluralism andpower-sharing” in Rwanda and Burundi. See also Cornell (2002),McGarry and O’Leary (1994, 94), and Saideman (1998).2 For broader consideration of strategies of control and dominationsee Adam (1971), Fearon (1998), Fearon and Laitin (2000), Lustick(1979, 1980), McGarry and O’Leary (1993, 23–26), and Yiftachel(2000).3 Emphasis in the original.

210

American Political Science Review Vol. 98, No. 2

states face increased possibilities of peaceful breakup.Weak and unaccommodating states can be expectedto experience turbulence or secession/collapse (Kohli1997).4 More common,however, among scholars study-ing the relationship between autonomy, devolution,power-sharing, federation, affirmative action, elec-toral reform, democratization, etc., and ethnicallycharged secessionism is the admission, a laHecther andOkamoto, that available findings are contradictory andinconclusive (Freeman 1999, 369–70; Hechter 2000, 9;Spencer 1998, 3).

AN AGENT-BASED MODELING APPROACHTO SECESSIONISM: ISSUES OF VALIDITYAND THE ROLE OF CONSTRUCTIVISTTHEORY

We turn now to consideration of the methods used toproduce these incomplete but intriguing results. Twomethods are prominent in scholarly work on this sub-ject. One is to conduct process tracings of a significantcase or cases, using synchronic or diachronic compar-isons. The other is to draw on one or more large datasets describing events, countries, or minority groups.Here the technique is to examine whether or not hy-potheses about general relationships between institu-tional practices and outcomes are consistent with thepatterns discernible from consideration of the data inthese arrays.There is much to be learned in these ways. Clearly

there isno substitute for studying thephenomena them-selves, in the real world, through careful scrutiny ofcrucial cases, through artfully structured comparisonsof small numbers of episodes, and through statisticallysophisticated treatments of highly processed, standard-ized, but rigorously conceived data sets. But it is ourview that with so many variables involved and so manyinteraction effects present, with the great difficulty ofgatheringdata relevant to thosevariablesdeemed theo-reticallymost interesting, andwith the extreme scarcityof episodes of secessionism and secessionist conflict,compared to the array of potential contexts for theiremergence, a third technique—agent-based computersimulation—has a crucial role to play in advancing thestudy of questions about the conditions that shape thelikelihood of serious internal threats to the integrityand stability of states.Among the most daunting challenges to scholars us-

ing the small-N approach, drawn from one case or fromstructured focused comparisons of a small number ofcases, is the stringent limit on the array of possiblenatural experiments that can be arranged. This limitarises from the rarity of full-blown secession and theinfrequency of severe ethnic conflict relative to thenumber of interethnic encounters that might have butdid not produce such conflict. Typically the investiga-tor must settle for some level of contamination of the

4 For similarly nuanced views on the difficulty of making linear pre-dictions about the effect on secessionism of more inclusive electoralarrangements, see Crawford (1998), Reilly andReynolds (1999), andScarritt, McMillan, and Mozaffar (2001).

comparison based on factors that were either incon-veniently different or similar about the cases or aboutthe episodes being compared within a single case study.Ideally the investigator would be able to select cleanlysorted comparisons, leaving all possibly relevant valuesidentical except for differences on either the depen-dent or the independent variable. In the real world,investigators have to settle for best available, partial,confounded, and therefore clouded comparisons.5Among the most daunting challenges facing schol-

ars using large-N approaches is the inevitable gap thatarises between the imputed meaning of the variablesunder investigation, as specified in theories under test,and the real measurements used to indicate the valuestaken on by those variables in specific cases. These gapsare in part produced by the need to use best availablequantifiable surrogates for the complex sentiments,opinions, behavior patterns, and circumstances that areof theoretical interest. Such gaps are then widened bythe need to treat data gathered in different ways indifferent countries as comparable across cases and byserious differences across cases and time periods in thereliability of the information gathered. Such problemsare then further compounded by the challenge of es-tablishing and enforcing coding routines that avoid is-sues of selection bias, pass tests of intercoder reliability,and are updated regularly based on new or improvedknowledge available about cases or historical episodes.Combined, these challengespose serious risks that find-ings may be driven as much or more by artifactual as-pects of data collection and processing as by the under-lying patterns putatively reflected in the data collected.It is probably safe to assume that a substantial portionof the disagreements that arise among researchers us-ing large-N approaches springs from these problems(Chandra 2001, 10; King and Zeng 2001).6The kind of “bottom-up” simulations offered by

agent-basedmodeling provides researchers a thirdway.Using computers to produce simulations of social andpolitical phenomena based on widely distributed butinteractive processes is not new. Important work hasbeen done on mobilization or repression in politicalcontexts (Bhavnani and Backer 2000; Epstein, Stein-bruner, and Parker 2002; Lustick andMiodownik 2002;Srbljinovic et al. 2003), identity diffusion (Axelrod1997; Hoffmann 2003; Lustick 2000; van der Veen2002), the emergence of ethnocentrism (Axelrod andHammond 2003), and the endogenization of borders(Cederman 1997, 2002). Excellent metaanalyses haveidentified emerging trends, challenges, and opportuni-ties in this cluster of research programs (Cederman2001; Macy and Willer 2002).7

5 This is the case even as scholars pick and choose from an arrayof historical accounts that is considerably more differentiated thanwhatever it is that those accounts are seeking to describe (Lustick1996).6 Important work is underway to improve the reliability of thesetechniques. See Doyle and Sambanis (2000), Laitin (2000), Laitinand Posner (2001), and Wilkinson (2000, 2001).7 An older tradition of computer simulation relies on differentialequations rather than on interactions among multitudes of adaptive

211


If theoretical expectations are relatively clear, butdata are hard to find that reliably match theoreticalcategories, if available natural experiments do not al-low crucial questions to be posed cleanly because ofinconvenient confounds, and if key aspects of the phe-nomenon of interest are relatively rare, computer sim-ulation should be considered a logical complement toother techniques of analysis. Such simulation involvescreating a virtual world in which the basic theoreticalrelationships among individuals or groups are imple-mented directly, obviating the need for surrogate mea-sures or indices of key variables. Large batches of “his-tories” or “futures” of these worlds can be producedby randomizing initial conditions or the pattern of per-turbations to which the world is subjected. Standardstatistical tests can then be run on the distributions ofoutcomes produced under specified conditions.By controlling initial conditions and/or the valences

and sequence of the streams of perturbations, the ef-fects of change in individual parameters of interest orthe interaction between two or more specific variablesof interest can be identified.8 Such theoretically signif-icant variables could include predominance of differ-ent identities across the population, geographical con-centration patterns, indices of difference or similarityacross groups of agents, amount of variation in agentinfluence, etc.By randomizingperturbations and/or ini-tial conditions and collecting data on the trajectoriesproduced by the “landscape” as it moves forward intime (with “agents” interacting and taking on or main-taining values depending on the algorithms with whichthey are endowed), researchers can systematically con-duct the thought experiments that they cannot conductor observe in the real world and cannot perform intheir heads because of the hundreds of thousands oreven millions of calculations involved in every step.It is common and natural for questions to be raised

about the “empirical validity” of computer simulationmodels. The important version of this question is notabout whether or not the theory being tested with themodel is true (a question that is obviously central in theevaluation of anymodel—whether computer deployedor not) but about whether the assumptions built intothe virtual world are so radically different from con-ditions in the “real world” that inferences about thelatter from the former cannot be justified. To under-stand why the problem of establishing the “empiricalvalidity” of agent-based computer models is no moredifficult, and possibly even less difficult, than in non-computer simulation models, it is useful to recall that

agents. We have not found useful applications of this “systems dy-namics” approach to ethnopolitical mobilization or secessionism.We do know of a substantial amount of work in this area done byRussians scholars. It is based largely on computer operationaliza-tions of Parsonian categories and variables. See, for example, Laptev(http://www.univer.omsk.su/MEP/). For a discussion of this earliertradition in relation to agent-based modeling approaches, see Macyand Willer (2002).8 By keeping the initial conditions and pattern of perturbations con-stant, and changing the algorithms or microrules controlling agentbehavior, problematic aspects of the theories of identity change, po-litical behavior, and elite recruitment these rules reflect can be testedas well.

such questions pertain to the “generalizability” of thefindings produced with the model. Can findings pro-duced in a virtual world, with a model whose terms areimplemented systematically and exactly via computer,be generalized to cognate phenomena the real world?Of course all applications of scientific models to the

real world must include, or should include, discussionof limits on their generalizability. Such limits are associ-ated with any kind of research project. Hypotheses arealways tested, not against the “real world,” but againsta proxy for it that arises from the assumptions of thetheoretical approach adopted and the rules used to ob-serve and code data. Accordingly, there is no intrinsicdifference between the use by agent-based modelersof an explicitly constructed virtual world as that proxyand the use by researchers employing traditional small-N or large-N techniques of the implicitly virtual worldsthat arise from their assumptions and coding rules asthe backgrounds against which their hypotheses aretested (or “validated”). For each of these approacheslimits to generalizability must be identified in terms ofthe versimilitude believed to exist between the surro-gate world created for the testing of hypotheses andthe “real world.” Indeed, for each of these approaches,including agent-based simulation modeling, an impor-tant source of validation is productive communicationacross theoretical assumptions and methodological di-vides. If researchers using diverse methods, assump-tions, and coding rules produce findings that corre-spond with or usefully articulate with one another,that cumulative discussion itself constitutes validation.It represents a kind of triangulation of a stable “realworld” to which each participating theoretical effortwould appear to have access.If the validation question is fundamentally the same

across approaches, each of which tests its hypothesesagainst a surrogate for the real world, what does dis-tinguish an agent-based modeling approach from oth-ers? It is the completeness and understandability of thevirtual world it is using as its surrogate. Unlike otherapproaches, its machinery can be both complex andtransparent, for everything in it is decidable, visible, andregistered. Some approaches (such as rational choice)are transparent but radically simple in terms of thenumber of interacting parts and variables. Other ap-proachesusingmore complexmodels obscure the limitsimposed by their substantive theories and conceal thevast number of arbitrary, uncontrolled, or unknown“parameter values” under a sweeping ceteris paribusassumption.In work with PS-I in general, and specifically in

our studies of secessionism, we use specific substantivetheories and “pretheories”—theories of constructivistidentity, social identity, and collective identity mobi-lization.We find, just as virtually all social scientists do,that even the best of these theories and the most care-ful of research designs fail to produce an experimentalworld in which every element in any way related to thephenomenon under study is assumed as a constant oris incorporated in the research design as a controlledvariable. An obvious example is the constructivist posi-tion that individuals and collectives have “repertoires”

212


of identities that are relatively easily presented or con-cealed. Exactly how much easier it is to present a con-cealed identity than to integrate anew identity formerlynot within the repertoire is seldom if ever explored, letalone set out as a theoretical proposition or researchfinding. The advantage of agent-based modeling is notthat it eliminates stipulated parameter values, but that,as in our work described below, every stipulation isnoted. In contrast to other approaches, especially non-formal approaches, there are no unrevealed or unsee-able assumptions. An added value of computer sim-ulations of this sort is the opportunity for systematicsensitivity tests in the face of good (i.e., theoreticallybased) arguments that specific parameters might beparticularly liable to produce artifactual results.The constructivist basis of political identity and

identity change is now well established.9 Studies ofidentity—ethnic, national, or otherwise—begin almostas commonly with a statement decrying the practice ofcomparing primordialist and constructivist approachesas they do with that comparison itself. As KanchanChandra (2001) put it in her introduction to a sym-posium on the constructivist consensus prevailing inpolitical science:

The constructivist approach, developed across the disci-plines of anthropology, sociology, political science, historyand literature, has discredited the primordialist approachby showing that ethnic groups are fluid and endogenousto a set of social, economic and political processes. Thosewho subscribe to the constructivist approach agree on twobasic propositions: First, individuals havemultiple, not sin-gle ethnic identities; and second, the identity with whichthey identify varies depending upon some specified causalvariable. Changes in the value of these causal variables arelikely to lead to changes in individual identifications. (7)

From the now standard constructivist position, iden-tities at both the individual and the collective lev-els are ultimately fluid, chosen, instrumentalizable, re-sponsive to change in relevant incentive structures,and susceptible to manipulation by cultural or polit-ical entrepreneurs. Examples of more or less stan-dard positionswithin this approach areAronoff (1998),Brass (1980), Laitin (1998), and Nagel (1994). Ex-tremes within this general perspective are anchoredon one side by arguments of the sort made by DavidBrown (1988), Walker Connor (1998), A. D. Smith(1981, 1986), and some sociobiologists such as Gary R.Johnson (1997). These scholars emphasize the psycho-logical affinity of identity appeals based on kinship pat-terns or tropes to explain why communities imaginedin common descent terms are so prevalent, stable, andpolitically potent (i.e., why the effects of opportunitiesfor choice are so limited). At the other extreme arethose scholars more impressed with the extraordinaryvariability of ethnic and other forms of cultural identifi-cation over time and in different circumstances. Indeed

9 For early examples of the constructivist approach to ethnic andnational identity see Barth (1959), Brass (1974), andHechter (1975).For representative applications and expositions of the approach seeEley and Suny (1996), Kowert and Legro (1996), Nagel (1994), andVerdery (1991).

some sociologists, anthropologists, and cultural studiesscholars operating in this mode shift their attentionentirely from the individuals and groups themselves(and their putative “identities”) to the prevailing cate-gories of identification and the relationship of changingcircumstances to “performances” of those categories.Rogers Brubaker (1996) is one of the most influentialscholars taking this perspective. Some political econ-omy approaches to identity also tend toward extremepositions of the sort that implicitly imagine any iden-tity as presentable by any actor or group dependingon incentives and the choices those incentives makeattractive (Hardin 1995).To be sure, constructivism performs its fundamental

theoretical role not by producing final answers but byframing questions. For example, how large are reper-toires of identities? How easily can they change?Whatis the effect on larger political processes of variation inthe size of identity repertoires, the fluidity with whichidentities are traded, or the volatility and turbulence ofthe incentive structure affecting competing identities?These questions are implicit in constructivist identitytheory, but until the development of agent-based mod-eling approaches they have been extremely difficult toinvestigate, or even pose properly (Lustick 2000). Butthe fact that there are good but unanswered questionsimplied by constructivism is evidence of its fruitfulnessas a framework for research, not of its inability to pro-vide guidelines for designing mechanisms to animate avirtual world of competing political identities carriedin the repertoires of multiple, but different, agents.

BEITA—A VIRTUAL MULTIETHNIC STATE

To exploit constructivist identity theory for the explo-ration of patterns of secession and secessionism a vir-tual statewas created, named, for convenience, “Beita.”Beita was designed using the PS-I simulation platformto capture in composite form certain common featuresof multicultural or multiethnic states that might en-counter threats of secession (Dergachev 2003; Lustick2000, 2002; Lustick and Miodownik 2000, 2002).As displayed in Figure 1, Beita is a square with 66

cells, or agents, per side (4356 cells). The external edgeof Beita is comprised of a fixed, unbroken, and im-permeable array of 260 black-colored “border” cells.Within these borders are located 4,096 agents that com-prise the Beita polity.10 Each square-shaped cell, oragent, in this array is endowed at time zero (t = 0) witha repertoire or portfolio of identities, one of which is“activated,” i.e., visible to the agents in its neighbor-hood (the eight agents bordering it on its four sides andits four corners). Different colors represent differentidentities.AsBeitamoves forward in time the rules gov-erning agent behavior permit the rotation and tradingof identities as functions of changing advantages anddisadvantages associated with individual identities andwith local conditions. These identity substitutions are

10 Agents can be imagined as modeling individuals, families, villages,or any unit of political aggregation that may seem appropriate.

213


FIGURE 1. Beita—A Virtual Multiethnic State

NW

SW

NE

SE

in turn reflected in patterns of color change and cluster-ing as some identities coalesce into control of particu-lar regions or lose their grip on those regions. Statis-tics describing changing characteristics of Beita areautomatically collected for diachronic and synchronicanalysis.The polity is divided into four quadrants, not by

boundaries but by changes in patterns of overlap-ping and shared political identities. The upper left, or“northwest” (NW), quadrant is the core of the state.A national bureaucracy, comprised of a web of agentsdrawn from one of three “agent classes” of “influen-tials” (see Table A1, Appendix) and having double,triple, or quadruple the “influence” of “basic” agents,radiates out from this quadrant into the other quad-rants. In the “standard” version ofBeita, used as a base-line for experimental purposes, these “bureaucrats” (or“officials”) all have the currently dominant, i.e., incum-bent, identity, identity “5,” in their repertoires and al-most all of them are activated on identity 5 at t = 0.Two other identities, identities 4 and 13, are “loyal op-position” national identities. All national bureaucratshave these identities in their repertoires.Top-echelon bureaucrats have an influence level of

4, compared to the influence of a basic agent, whoseinfluence level is 1. Top-echelon bureaucrats are few innumber, relatively centrally located within the radiat-ingbureaucraticweb, comprised (initially, at least, sinceall identities in Beita are tradable) of only the threenational identities, and marked visually with a circleinside the normal agent square. Mid-echelon bureau-crats are marked with a spiral, have an influence level

of 3, and a slightly larger identity repertoire than top-echelon bureaucrats—reflecting the regionally preva-lent identities as well as national identities. Lower-echelon bureaucrats, those with influence level 2,have within their repertoires both regionally preva-lent and parochial identities in their region along withthe three national identities. Thus, for example, lower-echelon national officials in the northeast (NE; upperright) quadrant have regionally prevalent identities 3and 15 in their repertoire, along with a less prominent“parochial” identity, identity 9, and all three nationalidentities, 4, 5, and 13. For detailed information aboutthe exact representation of different kinds of agentsin Beita and their identity complexions, see Table A2(Appendix); for color coding see Figure 3.Beita was produced as a composite rendering of a

multinational ormultiethnic country, corresponding di-rectly to no one particular country but containing com-mon aspects of many. In the NE and southwest (SW)quadrants regionally prevalent identities have sepa-rate small authority structures. These can be identifiedas local arrays of officials activated on colors associ-ated with regionally prevalent identities—bureaucratswhose loyalties to the central state are reflected in thepresence of national identities in their repertoires evenas they begin Beita histories activated on their par-ticular ethnic identity. The intent in these quadrantswas to model relationships of multinational democ-racy based on principles of federalism and/or multicul-tural liberalism. Thus national identities were includedwithin the repertoires of both bureaucrats and ordi-nary inhabitants, regardless of their activation onmore

214


particularist identities or the local prevalence of suchidentitieswithin the repertoires of agents in that region.The radiating web of the national bureaucracy in eachof these quadrants is substantial, though not as denseas in the state’s “national core” in theNW. The tolerantand accepting relationship between the national stateand the ethnic groups in theNEandSWis also reflectedin the presence of regionally prominent ethnic identi-ties in the repertoires of mid- and low-echelon nationalbureaucrats in those regions.But things are different in the southeast (SE). The

SE is modeled as a region controlled by the state butinhabited by a disgruntled regional minority (DRM)whose identity (10) is present in the repertoires of79% of the agents in this quadrant.11 The fearful andalienated relations between the state and this groupare reflected in the low activation rate of this identityand by the complete absence of identity 10 from therepertoires of the national bureaucrats stationed in SE.Beita also features a minority group in the SE region,associated with identity 16, that is historically distinctfrom and even antagonistic to the regionally dominantidentity (10). As is quite typically the case, it is thusattractive as an ally against identity 10 by the nationalcenter.12Such features in Beita were designed to make it a

kind of “ideal type”—a country that actually does notexist but that, in the deliberate clustering of elementsheld to be crucial by relevant theories, bears strongerresemblances to countries relatively predisposed to se-cessionism linked to processes of identity-based self-determination movements. Indeed Beita is not pre-sented as a universal template for simulating the manykinds of economic, political, legal, international, andcultural pressures that may be involved in the entirerange of ethnopolitical mobilization and secessionistactivity. It should be regarded as a specialized tool, use-ful for exploring the extent to which some patterns ofethnopolitical mobilization and secessionism could beaccounted for by focusing specifically on “identitarian”processes and pressures.In addition, particular aspects of Beita’s institutional

design render it more like some countries than oth-ers. For example, Beita features a dominant but notunitary regime authority structure, emanating from asecure core. The institutions of the regime within theborders of Beita include a variety of relatively de-centralized structures with overlapping loyalties in ar-eas of identitarian diversity. Beita also includes onearea of the country in which the regime appears as

11 In the SE quadrant, identity 10 is activated at t = 0 by 12% of theagents with that identity in their repertoire compared, for example,to an activation rate of 24% for identity 15 in theNEquadrant, whereit is a regionally prominent identity.12 Reflecting the favor it enjoys from the regime, identity 16 has avery high activation rate (54% of agents in SE with identity 16 intheir repertoire are activated on that identity at t = 0). It also has ahigh rate of overlapwith the incumbent national identity—identity 5.Indeedmore than 80%of SEagents activated on 16 have all three na-tional identities present within their repertoires. In contrast, identity16 is present within the repertoires of only 8%of SE agents activatedon 10 at t = 0.

a rigid, unresponsive, and alien set of institutions incontrast to a regionally predominant identity group de-prived of any substantial authority structure of its own.Countries and their disgruntled regional minorities towhich our findingsmay best apply include Iraq/Kurdishregion; Turkey/Kurdish region; Spain/Basque coun-try; Canada/Quebec; Yugoslavia under Milosevic;Sri Lanka/Tamil Northeast; United Kingdom/Ireland(19th century); France/Algeria (late 19th to mid-20thcenturies); Pakistan/East Pakistan; Pakistan/regionssuch as the Sindh or the Northwest FrontierProvince; India/Kashmir, Assam, or the Punjab;France and Corsica; and Indonesia/Papua or Aceh.Needless to say, none of these countries correspondsexactly toBeita, and in someof these cases, clearly, vari-ables other than the political/institutional variables in-vestigated in this article are decisively atwork.13 Never-theless, for different reasons these cases do contain theelements and dynamics of the kind of countries Beitais designed to simulate more than, for example, theUnited States/Native Americans, Iraq/southern Shiaareas, France/Brittany or Provence, Morocco/WesternSahara, and Israel/Palestinian areas.AsBeitamoves forward in time, the patterns of iden-

tity activation and hence the patterns of visible coloracross the landscape change. Individual cells changeidentity activation as a result of interactions with theirneighbors (agents directly touching their sides or cor-ners). Each agent registers the activated identities andinfluence levels of its neighbors but not the composi-tion of their repertoires. Simple calculations of relative“identity weight” lead each agent to either remain ac-tivated on its currently activated identity; rotate intoactivation an alternative identity from its repertoire;substitute an identity fromoutside its repertoire for oneinside its repertoire; or, in cases of a fairly overwhelm-ing discrepancy in favor of an identity not in its reper-toire, actually substitute andactivate on an identity pre-viously absent from its repertoire. (See the Appendixfor technical details.)Clustersof agents activatedon thesame identities form—a process usually accelerated inregions featuring webs of “influential” agents activatedon a common identity or having that identity in theirrepertoires. The clusters can then expand to includeagents who bring that identity into their repertoiresafter being surrounded or nearly surrounded by agentsactivated on that identity. For purposes of illustration,Figure 2 is provided, showing a typical run, or history,of Beita at time step 50 (t = 50).Just as in the real world the origin of signals that

shape the identity-basedbehavior of people and groupsis not wholly local, so too can agents in Beita (and inPS-I models in general) have access to some nonlocal

13 One important variable excluded from the sets of experimentswith Beita reported here is the direct influence of cross-border pres-sures fromneighboring countries possibly sympathetic to secessionisttendencies within Beita. The impermeable border surrounding Beitaprevents such pressures, though in other work, unreported here, wehave explored the implications of increasing the porosity of the bor-der separating the southeastern quadrant from variously constitutedneighboring states.

215


FIGURE 2. Beita Baseline—Typical Run (Time = 50)

NW

SW

NE

SE

information. At every point in time each identity is as-signed a “bias,” that is, a negative number, a positivenumber, or zero.14 According towhat can be thought ofas a “mass media” mechanism, each agent is aware ofthese signals and how they change. Each agent updatesits activation by adding these signals to its calculationsabout local identity weights.Figure 3 displays all the identities present in the spec-

trumofBeita (that is, present as a subscribed identity inat least one agent). Each identity is labeled by numberand color. The figure also shows, for each identity, thebias assigned to it at time step 50. The assignment of bi-ases is random, though for experimental purposes thatrandomness can be adjusted to introduce more or lessvolatility in the way the world is changing, a greater orsmaller range within which it can change, and more orless predictability in the way it will change.15 At everyeven-numbered time step each identity is eligible forthe assignment of a new bias. In the Beita history fromwhich these illustrations are taken, the volatility wasset so that the probability of being eligible for a biaschange at any one time step was .005% (with a biasrange of between −2 and +2).

14 A bias of “+1,” for example, would count in the identity weightcalculation of an agent as much as would one extra basic agent in itsneighborhood activated on that identity.15 Predictability is a measure, assuming that a change of bias assign-ment takes place, of how likely it is that that changewill entail amoveof more than one integer step away from its original bias assignment.As with most experiments reported here, this illustrative history wasset at a low predictability setting, such that large bias shifts were justas likely as small ones.

STUDYING SECESSIONISM WITH BEITA

Beita was designed as a state in which secession isnot ruled out in any region but, in which under typi-cal conditions of governance and political relations, in

FIGURE 3. Beita Identity Display

216


most regions, it does not occur.16 Secessionism amongidentity 10 agents in the southeast region is modeled inBeita as the emergence of boundaries within the statethat separate a sizable region dominated by identity 10from the rest of the state—a region featuring substan-tial cultural homogeneity in terms of both activated andsubscribed identities. Borders appear as some cells con-taining agents are transformed into small zones of sepa-ration thatmaybe considered either as “immutable andinactive agents” or as “border cells” (black squares).17Such transformations can be considered “secession-ist” activity in the sense that the appearance of suchzones of separation corresponds to the crystallizationof identitarian and political differences associated withregionally based emergent processes of alienation andout-group mobilization.The results of various factors (such as institutional

responsiveness, repression, and devolution) on the oc-currence of secessionism can then be investigated byadjusting those aspects of Beita and comparing the re-sults by analyzing large batches of runs under other-wise tightly controlled conditions. Apart from studyingsecessionism (how many border cells are produced inwhat regions), the emergence of a sufficient number ofborder cells combined with a sufficiently high level ofactivated homogeneity within the regions dominatedby the secessionist group can be used as a measure ofsecession itself.18The rules governing the circumstances in which indi-

vidual basic agents become border cells were designedto conform to basic and consensual understandings ofsecessionism as they appear in the scholarly literature.For although there is precious little agreement on thesufficient conditions for producing secessionism, i.e.,the relative contribution of variables such as regionaleconomic standing, degree of ethnic or identity differ-ence, irredentist influences, and institutional features,there is broad, if usually implicit, consensus on the nec-essary conditions for secessionism. These conditionsare (1) polarization or alienation of the potentially se-cessionist identity group from the dominant identity

16 Indeed in the thousands of histories we have generated, secessionby groups other than identity 10, or in regions other than the SE,where identity 10 is concentrated, have almost never been observed.17 Technically the cells in the Beita array are “positions” that mayor may not contain agents of different kinds at any given time. Sincealmost every cell in Beita is always inhabited by an agent with par-ticular characteristics, we refer to them as “agents.” But since thetransformation of a basic agent into a “border agent” is actually thetransformation of an inhabited cell into a cell that is uninhabitedbut rather serves as a small “zone of separation,” we use the phrase“border cells” rather than “border agents” to describe the emergenceof local manifestations of secessionism.18 By using the complexion of identity activation (via low tensionlevels within the newly bounded region) as a part of our indexof secessionism, we do not reject the role of a reorganization ortransformation of authority structures in the process of secession.Indeed, though we have not reported the data, secessions in Beitaare frequently associated with the emergence of separate authoritystructures. That is, portions of the webs of influential agents previ-ously activated on regime identities are captured by the secessionistmovement and transformed into webs of influentials activated on thesecessionist identity and separated from the rest of Beita by walls ofborder cells.

groups in the state; (2) regional concentration of theidentity on a scale that is not negligible; and (3) somesignificant level of tension or disharmony in the rela-tions between individuals expressing the “alienated”identity and others in their social environment. Indeed,few if any analysts have produced theories of seces-sion that do not assume the satisfaction of these threeconditions—alienation, nontrivial size, and individual-level tension with others.The conditions under which such behavior occurs

thus pertain to both macro circumstances of which theindividual agents are unaware and local realities ofwhich they are aware. When for any particular agentin any particular time step each of these conditions ismet, a low but nontrivial probability is created that thatagent will be transformed into a border cell.19 The ruleswe have implemented to operationalize these condi-tions governing the production of border cells can bestated, nontechnically, as follows:

1. Polarization/alienation: Secessionist activity can beexpected to be unlikely or impossible to the extentthatmembers of a potentially secessionist group alsoharbor the identity of the dominant group. Accord-ingly, no agent, at any particular time, can transforminto a border cell if 20% or more of the agents acti-vated on that identity at that time have the dominantidentity within their repertoires.

2. Size of a qualifying identity: Secession of the lead-ing group in a society is excluded from these ex-periments insofar as the leading group is consideredto be the identity activated at any particular timeby a plurality of agents in the polity. Similarly, se-cessionism cannot be produced by a group unless itconstitutes a substantial proportion of its region. InBeita no agent is allowed to transform into a bordercell unless its activated identity is activated by atleast 10% of Beita agents.20

3. Individual action: Some otherwise qualifying agentsare more likely than others to engage in secessionistactivity, and those lacking very much contact with

19 In the experiments reported in this paper the probability of anotherwise qualifying basic agent turning into a border cells was 20%per time step in which the basic agent remained qualified. This prob-ability can easily be adjusted for experimental purposes. Sensitivitytests indicated that adjusting this probability value between 15%and25% did not alter the frequency with which border cells appeared.The effects on these adjustments on the number of border cells pro-duced were linear, relatively small, and in the expected directions.20 Since the size of countries and the sizes of disaffected populationswithin those countries varywidely, the 10%rule for theminimumsizeof what we refer to as a “subordinate identity” cannot be consideredabsolute. Instead the rule used to calculate the minimum size of aregionally disgruntled minority capable of producing secessionism is40%of the ratioof thepopulationof the region to the total populationof the state. Thus, the SE quadrant of Beita represents 25% of theentire state. Forty percent of that is 10%, so the minimum size of anactivated group in Beita capable of producing secessionism is 10%ofthe size of Beita or 409 agents. Roeder (2003) presents an extensivediscussion of the statistical issues involved in coding secession forcomparisons across large and small states. Roeder’s findings, thougharrived at with different techniques and for somewhat different pur-poses, are consistent with our coding rules. We are grateful to PaulBrass for posing the question that clarified our understanding of thisgeneral point.

217


agents activated on identities other than their owncan reasonablybeexpected tobe less inclined to takethe risks of secessionist action than liminal agents,exposed to other identities but not harboring thoseidentities within their repertoires. Accordingly, noagent can transform into a border cell unless halfor more of the agents it is in direct contact with areactivated on an identity other than its own activatedidentity.

With these rules implemented for the operation ofBeita, we proceeded to investigate the relationshipamong variables of interest with respect to secession-ismand institutional schemesof repression, responsive-ness, and representation. This was done by comparinghistories produced under slightly different conditionsor with slight changes introduced into the polity to op-erationalize such institutional strategies.

SIMULATION EXPERIMENTS

Several groups of simulation experiments designed toexplore relationships among institutions, ethnopolit-ical mobilization, secessionist activity, and secessionwere conducted. Each series of experiments focusedon changing aspects of the authority structure in theSE region. The first series studied the effects of repres-sion, modeled as increasing the range and density ofthe distribution of regime officials in SE. The secondfocused on the effects of increasing the responsivenessof the regime officials to the demands of the potentiallysecessionist identity. The third group of experimentswas aimed at investigating the effect of increasing therepresentativeness of the authority structure, whetherby increasing the proportion of existing regime officialsexpressing the potentially secessionist identity or byexpanding the authority structure by adding networksof regime officials expressing that identity.21Each series of experiments entailed creation of dis-

tinctive versions of Beita—each captured in a different“snapshot.” Each snapshot featured a carefully con-trolled group of parameter settings and/or changesin the complexion of agents in SE. Each version ofBeita was used to produce 100 separate and uniquehistories—histories run through 508 time steps to time508 (t = 508).22 The uniqueness of each specific historywas ensured by the application of a distinctive stream

21 As a manipulation test for our baseline template we ran severalexperiments entailing the removal of key operationalizations (vari-able bias values, variable agent influence levels, the impermeableboundary surrounding Beita, and rules for the production of bordercells). We then added these elements and observed the results inbatches of 100 runs each. The results conformed to our expectationsof the role each of these operationalizations, both individually andin combination, plays in producing the overall effect of Beita as asimulated country disposed toward, but not necessarily destined for,separatism. The results of these manipulations are not reported herebut are available from the authors.22 The length of each history is considered to be 500 time steps. Ourexperience with Beita strongly suggests that only rarely would ex-tending the examinationof dynamicprocesses beyond this point yieldstatistically different results. The reason observations are conductedat t = 508, and not t = 500, is that each run includes eight initial timesteps during which the likelihood of change of a bias value associated

of randomly generated “biases” affecting the sequenceand extent of changes in the relative attractiveness ofactivating on different identities.23 Our analysis com-pared the distributions of histories produced by incre-mental changes in these variables to the distributionof histories produced by the standard Beita landscape.Table 1 presents a summary of all of the experimentalsettings.In the first series of experiments, the effect of re-

pression was studied by extending the size of the stateauthority structure in the SE region by an additional25%, 50%, 75%, or 100% of its original size. The ad-ditional regime bureaucrats were created by adjustingthe identity complexions of agents in the SE regionwith existing identity complexions that excluded iden-tity 10, activating them on the identity of the regimeofficials in SE and the dominant incumbent identity att = 0 in Beita (5), and endowing them with an influ-ence level of 2. The combination of a larger numberof influentials, arrayed in relatively close proximity toone another, activated on the originally activated dom-inant regime identity (5), and not including within theirrepertoires theDRM (DisgruntledRegionalMinority)identity (10), implements “repression” in that it makesit much more difficult for agents in the region of thisweb of influentials to activate on identities other thanthe identity activated by the regime bureaucracy. Inother words, in a region with a repressive bureaucracymore agents will be constrained to activate on an iden-tity that is not assigned a bias as high as that assignedto some other identity in those agents’ repertoires. Re-pression thus interferes with the ability of agents todo the best they can for themselves with the politicalresources under their control.These repression manipulations produced four dis-

tinct versions of Beita at t = 0, i.e., four different snap-shots. Running each of these snapshots, each with alarger regime bureaucracy in SE, produced 400 histo-ries (withone set of 100observations at t = 508 for eachof the four renderings of the regime officials in SE).In the second series of experiments we studied the

effects of increasing responsiveness by adding identity10 to the repertoires of 25%, 50%, 75%, and 100%

with any given identity is very high (50%, as opposed to 0.005% dur-ing the history itself). This has the effect of “scrambling” the biasesand allowing the history to begin “in media res,” rather than froman artificially “calm” position. Other settings that remained constantthrough all manipulations as well as in the standard or baseline Beitaincluded a bias range of −2, +2, a moderated jump factor of 10,000(meaning that the likelihood of a bias changing from one value toanother was equal, regardless of the size of the increment entailed inthat change), and a sight radius for all agents of 1, meaning that allagents could see the activated identities of the eight agents adjacentto them and no others.23 Given exactly the same parameter settings, exactly the same arrayof agents operating by the same rules, given the same complexionof identities in agent repertoires, and assuming that all agents areactivated on the same identities, applying any particular stream ofbias valueswill always produce exactly the same (unique) history. Ex-periments reported here are based on running “treated” landscapes100 times using the same set of 100 randomly generated streams ofbias values and comparing the distribution of results to the histo-ries generated from the standard Beita landscape when those samestreams of biases were applied.

218


TABLE 1. Experimental ManipulationsNumber of Simulations

Condition Manipulation Identity Complexion Snapshots per SnapshotBaseline None 1 100Repression Extending the presence of No change compared 4 100

low- and mid-echelon RO to baselinein the SE by a share of theiroriginal size (25%–100%)

Responsiveness Increasing affiliation with DRM Affected agents remain active 4 100among increasing proportions on the regime’s identity (ID 5)(25%–100%) of low-echelon and add the DRM (ID 10)RO in the SE to their repertoire

RepresentativenessPower-sharing Increasing affiliation with DRM Affected agents activate DRM 4 100

among increasing proportions (ID 10) and keep the regime’s(25%–100%) of low-echelon identity (ID 5) in theirRO in the SE repertoire

Semiautonomous Increasing the presence of Same as power-sharing 4 100institutions low-echelon DRM in the SE by

a share of the regime’s SEauthority structure (25%–100%)

Note: RO, regime official; DRM, disgruntled regional minority; SE, southeast region.

of the original set of regime officials in SE. These re-sponsiveness operationalizations entailed the use offour separate snapshots at t = 0 and the generationof 400 histories, yielding four sets of 100 observationseach. This was an attempt to operationalize a kind ofregime bureaucracy less repressive in its effects andmore oriented toward the provision of public goods tothe inhabitants of SE. Thus, even though activated onan identity alien to that of the majority of agents in theregion, the influentials in SE comprising the regime au-thority structure were nevertheless disposed to enablethat population to realize its potential for activation onits own identity under favorable conditions, i.e., whenthe bias for identity 10 might be relatively high.24The third series of experiments included simulations

of (a) power-sharing and (b) the granting of differ-ent degrees of autonomy to the potentially secessionistidentity. Power-sharingwas operationalized by increas-ing the proportion (25%–100%) of SE regime officialsactivated on the identity of the disgruntled regionalminority (10). The effect of semiautonomous institu-tions was studied by extending the size of the authoritystructure. In this set of experiments the existing au-thority structure in SEwas unchanged, but increasinglysubstantial webs of officials were added to it. Theseadded arrays were comprised of agents with identity10 not only in their repertoires, but activated. Each ofthese new officials was also endowed, in its repertoire,with identity 5, the identity of the regime officials in SEand the dominant incumbent identity at t = 0 in Beita.The four levels of manipulation (25%, 50%, 75%, and100%) and two types of representativeness produced800 histories (eight sets of 100 histories each).Three criterion measures were developed to study

the questions of interest. First, the number of agents ac-

24 We are grateful to one anonymous reviewer for this journal whosecomments on our operationalizations helped us clarify their intent.

tivated on identity 10 at t = 508 was used as a measureof ethnopolitical mobilization—the extent to which thepotential for aggregatedpublic expressionsof aparticu-lar collective identity was realized. Second, the amountof secessionist activity—the extent to which the neces-sary conditions for secessionism were met—measuredas the number of border cells at t = 508, was used.Third,we identified those specific histories thatmet twocriteria for outright secession.Histories at t = 508werecoded as secession if they included a substantial andclearly demarcated zone (or zones) within SE that wasboth (a) heavily dominatedby agents activatedon iden-tity 10and (b) separated fromthe central stateby coher-ent though not necessarily fully closed boundaries ofborder cells. Figure 4 displays an example of secession.Figure 5 displays an example from the same group ofBeita histories that features some secessionist activitybut is not codedas an instanceof secession.25Within thesampleof 100baselinehistories the three criterionmea-sures were significantly intercorrelated (all p < .01).The correlation between ethnopolitical mobilizationand secessionist activity was r(100) = .68, the correla-tion between ethnopolitical mobilization and secessionwas r(100) = .43, and the correlation between seces-sionist activity and secession was r(100) = .78.In the following sectionswe compare the results from

specific manipulations to the results obtained from 100histories produced with Beita “standard” (our baseline

25 Specifically, a history was treated as secession if two conditionswere met at t = 508: (1) there were at least 84 border cells in the SEquadrant, and (2) the average tension of agents activated on iden-tity 10 in the SE quadrant was less than .80 (which means that, onaverage, agents activated on identity 10 had fewer than one adjacentagent activated on a different identity). These two benchmarks wereadopted because they effectively statistically identified those base-line histories that a majority of four trained raters independentlyjudged to clearly represent secession based on visual inspection ofthe landscapes at t = 508.

219


FIGURE 4. Beita Baseline: Secessionist Activity with Secession (t = 508)

NW

SW

NE

SE

FIGURE 5. Beita Baseline: Secessionist Activity without Secession (t = 508)

NW

SW

NE

SE

220


TABLE 2. Experimental ResultsDependent Measure

Amount of EthnopoliticalMobilization: Amount of Secessionist Frequency of Secession:

Manipulation Agents Activated on ID 10 Activity: Border Cells Secessions per 100Condition Size at t = 508 (Mean) at t = 508 (Mean) Histories at t = 508

Baseline 398.71 34.95 18Repression 25% 359.67 35.29 15

50% 324.95 34.44 1475% 302.93 30.04 8100% 279.26 27.07 1

Responsiveness 25% 404.33 35.78 1750% 416.31 33.71 1475% 422.49 33.36 14100% 430.18 31.18 13

RepresentativenessPower-sharing 25% 464.84 37.76 15

50% 549.76 37.60 975% 693.81 7.24 0100% 736.35 0.97 0

Semiautonomous 25% 488.87 26.10 6institutions 50% 521.75 16.27 1

75% 567.10 6.04 0100% 583.75 1.42 0

Note: Numbers in boldface are statistically different from the baseline (p < .005). All values are in reference to SE only.

condition). Table 2 presents our findings for all manip-ulations and for the standard baseline condition. Twotypes of tests were used. For the two continuous crite-rion measures, ethnopolitical mobilization and seces-sionist activity, statistical comparisonsweremade usingpaired sample t-tests (e.g., comparing the baseline his-tories with each set of experimental histories). For thedichotomous criterion measure, secession frequency,comparisons were made using McNemar’s chi-squaretest for nonparametric distributions.

Repression

We first examined repression—the effect of increas-ing the size of the regime bureaucracy in SE—and itseffect on ethnopolitical mobilization, secessionist ac-tivity, and secession by identity 10. The findings forall of the repression manipulations are summarizedin the first set of four rows in Table 2. In the columnlabeled “Amount of Ethnopolitical Mobilization” arethe mean number of agents activated on identity 10 inSE at t = 508 for separate sets of 100 Beita historiesusing four different snapshots (with regime bureau-cracies in SE that are, respectively, 25%, 50%, 75%,and 100% larger than the regime bureaucracy in SEwithin standard Beita). We see that increasing the sizeof the regimebureaucracy progressively and effectivelyreduced public expressions of attachment to the poten-tially secessionist minority community as measured byactivation on identity 10.However, the “success” of repression was not so

clearly demonstrated with regard to our measures ofsecessionist activity and outrightsecession. Examining

the column in Table 2 labeled “Amount of Secession-ist Activity,” we see that increasing repression did notsignificantly reduce the amount of secessionist activ-ity within SE compared to the baseline condition, al-though the trend was in this direction. As indicated inthe column labeled “Frequencyof Secession,”when thethreat of secessionism to the central state is measuredin terms of the frequency of actual secession, sufficientrepression did reduce its occurrence. This reduction ap-proached statistical significance at the 75% level andwas highly significant when the size of the original bu-reaucracy was doubled.

Responsiveness

This set of experiments with Beita was designed to in-vestigate the impact of institutional changes very dif-ferent from the repressive changes described above.Specifically, how would changes in the bureaucracy tomake it more responsive to the potentially secession-ist identity affect patterns of mobilization, secession-ist activity, and secession in SE? As discussed in theintroduction, various specific schemes of institutionalreform are quite commonly recommended to reducethe threat of confrontation and secessionist strugglesin the kind of situations Beita simulates.The results of the responsiveness manipulations are

summarized in the second set of four rows in Table 2.With respect to changes in the amount of ethnopo-litical mobilization, our findings indicate that increas-ing responsiveness modestly but significantly increasedthe prevalence of agents activated on identity 10 inSE. The average number of Beita inhabitants publicly

221


expressing identification with identity 10 increased byroughly 8% from the baseline condition, when 100%of the entire authority structure in SE was responsive.While these results may disappoint those who wouldhope to reduce tendencies towardmobilization of com-peting ethnic identities by strategies of bureaucraticcooptation, they support those who contend that of-fering positions of power in the state’s authority struc-ture to regionally predominantminority groups is likelyto result in more robust mobilizations. On the otherhand, results from our experiments seem contrary tothe worst fears of those who believe that any conces-sions to a potentially secessionist group will lead thestate down a slippery slope toward secession. Our re-sponsiveness manipulations produced only a modesteffect on the average amount of secessionist activityeven when the entire network of regime officials wasmade more responsive (see Table 2), and the direc-tion of this change was toward diminished rather thanheightened secessionism.At the same time,ourfindingsdo not provide support for those who believe that sim-ply reforming the governing bureaucracy to be moreattentive to the concerns of a disaffected minority cansubstantially reduce the likely incidence of secession-ist activity. The results obtained for our third criterionmeasure, the frequencyofoutright secession,paralleledthose reported for secessionist activity: a statisticallynonsignificant decline with increasing responsivenesscompared to the baseline condition.

Representativeness: Power-Sharing andBuilding Semiautonomous Institutions

Most proposals for mitigating tendencies toward se-cessionism and confrontations with regional minori-ties seeking national self-determination entail morethan reforming or reorienting the existing authoritystructure, whether through democratization or affirma-tive action programs. Almost all schemes for devolvingpower to potentially secessionist groups require an in-crease in the number of power-holders publicly repre-senting the identity of the disaffected group. This canbe achieved in either of two ways: (a) through power-sharing by recruiting representatives of the minoritygroup for positions within the existing authority struc-ture or (b) by building semiautonomous institutionsdesigned to provide representation, expression, access,resources, and symbolic satisfaction to members of theout-group. The claim is that by increasing the represen-tativeness and/or scope of institutions exercising powerover the lives ofmembers of the potentially secessionistgroup, challenges to the regime can be avoided. Eth-nopolitical mobilization that otherwise might lead toconfrontation with the regime would instead be chan-neled within boundaries constrained by loyalty to thecentral state and acceptance of the integration of theregion within its jurisdiction.The bottom half of Table 2 presents summary data

from the experiments designed to explore represen-tativeness using the operationalizations described ear-lier for the simulation of power-sharing and semiau-tonomous institutions. The “sharing” of power and the

“semi” autonomous aspect of these institutions is ex-pressed in the fact that all officials, whether or not ac-tivated on identity 10, have within their repertoires atthe beginning of these histories identity 5—the dom-inant governing identity most prominently associatedwith the central state. The picture here is of a small au-thority structure, formerly domineering and rigid, thatbecomes not only sensitive to, but representative of andcontrolled by, agents identifying with the disaffectedgroup. Examining first the impact on the amount ofethnopolitical mobilization, we found that both power-sharing and the building of semiautonomous institu-tions produced significant increases in mobilization ofthe DRM (10)—even with only relatively small in-creases in the representativeness of the existing regimeofficials (i.e., the 25% and 50% conditions). In the 75%and100%conditions, the resulting levels of public iden-tification with identity 10 were especially high. Overall,as the degree of representativeness increased, the im-pact of power-sharing became more and more potentin comparison to the parallel effects of building semi-autonomous institutions.Despite these increases in ethnopolitical mobiliza-

tion, we find striking evidence of the constraining ef-fect of power-sharing or building semiautonomous in-stitutions on our second criterionmeasure, secessionistactivity.When agents expressingDRM (10) were givenreal opportunities to exercise public power, secession-ist activity was significantly reduced even in the midstof robust increases in size of the population activelyassociating themselves with that identity. Interestingly,under conditions of “tokenism,” when only 50% orfewer of existing officials display identity 10, power-sharing did not diminish secessionist activity. However,a significant decline in secessionist activity did appearunder power-sharing arrangementswhen a clearmajor-ity of the regime officials in SE began Beita histories bypublicly expressing their attachment to identity 10. Incontrast, the building of semi autonomous institutionseffectively reduced secessionist activity even with onlylimited devolution of poilitical power.The data describing the frequency of secession itself

mostly recapitulate the results reported regarding thereductionof secessionist activity associatedwithpower-sharing and building semiautonomous institutions. It isworth emphasizing the support that these findings seemto offer to the argument that opening up very substan-tial opportunities for participation in local governance,while tending to increase rates of ethnopolitical mo-bilization, can greatly reduce secessionist activity andmake secession itself extremely unlikely.

DISCUSSION AND CONCLUSION

Our overall results help explain why scholarship in thisarea has produced many seemingly conflicting claims.Even when, as in the experiments reported here, keyvariables (such as international involvement, relativeeconomic position of the potentially secessionist re-gion, and distinctiveness of group identity) are heldconstant, the data often suggest nonlinear patterns. Onthe other hand, our findings also encourage support for

222


the analyses of researchers such as Hechter and Kohli.These scholars discern trade-offs that may be expectedin choosing between repressive policies and policiesthat emphasize responsiveness and representation. Theformer are likely to produce short-termquiescence andlonger termthreatsof secessionismandsecession,whilethe latter are likely to lead to broader and noisiermobi-lizations but substantially lower threats of secessionismand, particularly, of secession. In this regard, it is worthreviewing the results of these experiments in terms ofthe three outcome variables of interest: extent of eth-nopolitical mobilization, amount of secessionism, andfrequency of secession.

Ethnopolitical Mobilization

Analysts who have argued or implied that repression,domination, or control can work (Adam 1971; Fearonand Laitin 2003; Lustick 1980) receive strong supportinsofar as limiting the amount of ethnopolitical mo-bilization is concerned. Indeed, rates of mobilizationdecrease steadily as the size of the control apparatusincreases. Relatedly, the wider mobilizations producedby increasing either the representativeness of the ex-isting regime bureaucracy or the size of a representa-tive semiautonomous bureaucratic apparatus supportat least parts of the arguments of scholars who warn ofthe energizing effects of democratization for patternsof nationalist or ethnopolitical mobilization (Snyder2000).To be sure, our model does not distinguish between

violent and nonviolent forms of mobilization. If oneimagines this mobilization as violent, then regardlessof the longer-term consequences, regimes may find itadvisable to engage in repressive or control policies,avoiding responsiveness or representativeness in thegovernance of a regionally concentrated, potentiallysecessionistminority, if only to avoid violent confronta-tions. However, if one imagines that secessionist strug-gles, per se, are more likely to be violent than others,then to judge the attractiveness or necessity of suchpolicies a regime would also have to consider their ef-fect on the impetus given among those mobilized toengage in secessionist activity.

Secessionist Activity

Repression, i.e., strengthening a rigid and unrespon-sive regime bureaucracy, may decreasemobilization bythe latently secessionist identity, but it does not sig-nificantly decrease (nor increase) secessionist activitywithin that community. There is no strong support inour experiments, in other words, for a direct relation-ship between increased repression and the robustnessof secessionist movements. Similarly, increasing the re-sponsiveness of the authority structure did not substan-tially alter secessionist activity.In contrast, increasing representativeness encour-

aged public participation by the potentially secession-ist, regionally concentrated minority and did in factdecrease the secessionist activity. Particularly strongeffects in this direction followed the creation of semiau-

tonomous governing structures. Low levels of power-sharing, that is, small or moderate increases in the rep-resentativeness of the existing bureaucracy, have nosignificant effect on the amount of secessionist activ-ity. But with the regime bureaucracy in the less well-integrated region fully or nearly fully staffed by rep-resentatives of the latently secessionist minority, weregistered significant decreases in secessionist activ-ity. Of particular interest, when the regime maintainedits existing bureaucratic array and then added even asmall-sized bureaucracy staffed by representatives ofthe regionally dominant identity, secessionist activityby this group decreased quite sharply.These findings lend support to the body of scholar-

ship referred to above that has argued in favor of self-administration, limited autonomy, or other schemesto create settings for the exercise of public power byrepresentatives of potentially secessionist groups as ameans of preventing the development of secessioniststruggles. Ready to accept higher levels of ethnopoliti-cal mobilization, these authors imagine that with semi-autonomous institutions to channel that mobilization,secessionist struggles can be averted. The mechanismsthat are understood to produce this combination of ef-fects for these authors vary and are not always clearlyspecified. An advantage of the agent-based simulationapproach is that we can understand how this seeminglyodd combination of effects arises from the simple as-sumptions made about the conditions widely believedto contribute to secessionism.As the proportion of the population that publicly

identifies with identity 10 in SE increases, larger num-bers of those who come to activate on that identity alsohave within their repertoires the identities they werepreviously activated on, including the original govern-ing identity 5 or other identities that may become dom-inant in Beita. As these agents are integrated into theethnopoliticallymobilized group, a larger proportionofthe entire group comes to contain within their individ-ual repertoires the dominant identity in Beita—usuallyidentity 5. This reduces the likelihood of secessionistactivity. Moreover, as identity 10 becomes the prevail-ing activated identity in SE, the number of agents pub-licly expressing this identity, and who find themselvesin neighborhoods inhabited predominantly by otheragents also activated on identity 10, increases. Withlower tension levels secessionist activity is also likely todecrease. Representative institutions, even if not fullyautonomous, thus seem to inhibit secessionism. Theydo so by reducing the overall amount of alienation orpolarization between regime-dominant and regionallydominant identity groups and by reducing the likeli-hood that individuals identifying with the regionallydominant group will be in tension with the majority ofthose with whom they have direct contact.

Secession

Increasing the size of a repressive bureaucracy did sig-nificantly reduce the frequency of secession itself, butonly when the regime was willing to invest the re-sources necessary to double (or nearly so) the size of

223


the original bureaucracy in the SE region. Although in-creasing the responsiveness of the existing bureaucracydid not decrease the frequency of secession, significantdecreases were achieved by increasing power-sharingor by erecting semiautonomous institutions that com-plemented the existing regime identity-dominated bu-reaucracy.There are two reasons why—from the point of view

of the regime—sharing power or building semiau-tonomous institutions for the regionally concentratedand disaffected minority group may be a more at-tractive approach than repression. First, significant de-creases in secession frequency were achieved with rel-atively smaller-scale increases in representativenesscompared to the scaleof increases in the repressionnec-essary to produce comparable decreases in secessionfrequency (See Table 2). Second, it may reasonably beassumed that under conditions of both power-sharingand creation of a subsidiary bureaucracy fully staffedby agents activated on identity 10, revenues from taxa-tion would be more dependable and more robust thanunder the repression condition. On the other side ofthe political ledger, of course, may be the regime’sneed to accept, in the absence of greatly increased re-pression, higher levels of ethnopolitical mobilizationby the potentially disaffected minority—increases thatare indeed associated with maintaining a more rep-resentative bureaucracy. It is striking, however, thateven a relatively small investment in the developmentof minority self-governance through semiautonomousinstitutions reduced the probability that even the largermobilizations associated with this condition resulted ina break from the state.Looking across all of our conditions and dependent

measures it appears that the broadest and most com-pelling finding emerging from our data is that explana-tions for variation in amounts of ethnopolitical mobi-lization, even by members of communities that seemprimed for secessionism, cannot be expected to cor-respond to explanations for patterns in the variationof amounts of secessionism or outright secession. Thisresult not only helps explain the difficulty scholars havehad making sense of cumulative findings, but also jus-tifies the search for techniques of investigation permit-ting more precision and control in the examination ofthe effects of specific variables on patterns of outcomes.In part it was the search for such techniques, by schol-ars who emphasize process-tracing in small-N studiesand by those who search for their insights through theconstruction and refinement of large databases, thatinspired our effort to deploy agent-based computersimulation modeling to this problem.Brief consideration of several real-world cases in

terms of our findings illustrates what may be our mostimportant conclusion—that noncoercive policies avail-able for the reduction of secessionism and secessionmay work only at the “cost” of a state accepting a sig-nificantly larger role in the public political arena forpolitical expressions of historically “out-group” identi-ties. For what Rogers Brubaker has called “nationaliz-ing states,” featuring regimes dedicated to the exclusiveexpression of a particular identity and to serving the in-

terests of those attached to it, this costmaybeperceivedas too high (Brubaker 1996). Such states will be ratherlikely to face a choice between extended and extensiveregimes of repression or the emergence of robust andpossibly successful secessionist movements.Canada’s extraordinarily generous response to the

potent threat of Quebecois separatism included chang-ing the country’s flag and transforming its political cul-ture from an “Anglo” state to a truly binational state.The extent to which French-speaking and Quebecois-identifying inhabitants of Quebec were granted thekindof semiautononous and representative institutionswhose effects we have sought to investigate is reflectedin the palpable discomfortmanyEnglish speakers havefelt in Montreal and other cities and towns in Quebecand the significant out-migration of these Canadiansfrom the region. On the other hand, another effect ofthis real policy change by the Canadian regime hasbeen to create conditions in which Quebecois sepa-ratists have repeatedly failed to gain majority supportfor referenda on secession itself.In July 2003 the French government sponsored a

referendum in Corsica on a new statute that wouldsignificantly expand the power and prestige of semiau-tonomouspolitical institutions inCorsica.Thismeasurewas explicitly justified as a means to combat Corsicanseparatism, which had produced chronic violence andprotest against French rule of the island.Adding dramato the referendum was a French police announce-ment days before of the arrest of Ivan Colonna, aCorsican separatist leader charged with assassinatinga French official. The referendum failed by a narrowmargin. Though the overwhelming majority of non-Corsican French on the island opposed it, the statutehad been supported by many mainstream Corsican na-tionalists and by the French government. Observerssubsequently suggested that the timing of the arrest ofthe separatist leader, reflecting a successful exercise in“repression,” had shifted what would have been many“Yes” votes among Corsican nationalists into the “No”column.This episode would appear to illustrate several of

our findings with Beita. First, the French governmentdoes appear to have concluded that strengthening rep-resentative and semiautonomous institutions would bea more attractive and expedient way to reduce seces-sionist pressures in Corsica than continued repression.Support of most Corsican nationalists for the referen-dum, at least prior to Colonna’s arrest, would seem toindicate theirpreference fora regimewithinwhich theiridentity would be expressed more vigorously and theirinterests honored more substantially than a concertedand direct struggle to separate from France. The strongopposition of the “French” inhabitants of Corsica isalso understandable as an expression of their realiza-tion that the implication of combating secessionism viaautonomy andCorsican representativeness rather thana centralized and repressive order enforced from Pariswould be an unacceptable “indigenization” of the is-land’s political culture and public space.Finally, consider the case of India in its relation-

ship with Kashmir—its only Muslim majority state.

224


Aside from relatively meaningless symbolic frame-works adopted in the late 1940s, India has generallyrefused to recognize the ethnopolitical distinctivenessof Kashmir, where a nationalist movement with strongsecessionist overtones has been active for 50 years de-manded,with varying degrees of intensity and violence.Elsewhere within India’s borders the state has empha-sized a federal structure that acknowledges and givesexpression to a host of regional cultural and ethnopo-litical or linguistic variants. But whether dominated bythe secularist Congress Party or, more recently, by theHindutva-oriented BJP, the regime in New Delhi hasalways firmly opposed Kashmiri demands and empha-sized rigorous repression and direct rule of the territoryby non-Kashmiris.Our finding that under such conditions repression

can, if rigorous enough, prevent robust public ethnop-olitical mobilizations, but only at great cost and with-out eliminating the threat of secessionism, is well il-lustrated in this case. Although support for separatistsfrom Pakistan and other external sources is clearly sig-nificant and lies outside the terms of the experimentsreported here with Beita, the combination of Kashmiriexclusion from the public arena, continued high levelsof separatist violence, and high levels of costly regimerepression is quite consistent with our findings regard-ing the trade-offs likely to prevail between repressionand representative, semiautonomous or responsive in-stitutions, as well as between preservation of the ex-clusiveness of a public political space and the stabi-lizing consequences of increasing the responsiveness,autonomy, and/or representativeness of governing in-stitutions.As there are limits to other approaches, so too are

there limits to agent-basedmodeling simulation in gen-eral, and PS-I in particular. We as yet have no reliablemethod for portraying a simulation of Beita of a par-ticular length as corresponding to a particular lengthof time in a country comparable to Beita in the realworld. We are in the process of integrating other keyvariables of interest to scholars in this area into Beitaand our study of its dynamics, but as we do so we runthe risk of creating, in our virtual space, the intractablyentangled and complex causal relationships that makeit difficult to isolate effects in the real world. The vari-able appearance of violence as an aspect of regimerepression or of ethnopolitical mobilization and of se-cessionist struggles is of great interest but is not readilycaptured in a computationalmodel such asBeita, basedas it is on a cellular automata design. It is our hope,however, that having demonstrated the usefulness ofa user-friendly platform such as PS-I, researchers willadd this technique to their methodological quiver,whether as an “idea pump” or as a method for ex-ploring the logic of their arguments independent of theaccidents of available data and confounding circum-stances that may be associated with particular cases ordatabases.Finally, having considered the results of our inves-

tigations with Beita, having interpreted them as in-terventions in the scholarly debate over relationshipsamong ethnopolitical mobilization, secessionism, and

various institutional responses, and having illustratedtheir application in a few real world cases, it is worthreturning to the issue of “empirical validity.” Froma Lakatosian (1970) perspective in which science ad-vances via the competition of technically incommen-surable research programs, the real test of a scientificenterprise is whether the results of its work articulatewith the results of work done in other programs to forman interesting and progressive conversation within thescholarly community. In this context we take satisfac-tion in the extent to which the findings drawn fromour experiments with Beita yield nuanced answers toquestions being asked on the cutting edge of the field,questions of great interest to researchers using radicallydifferent techniques, and questions politiciansmust an-swer to alleviate the real suffering that ignorance ofprocesses associated with ethnopolitical conflict helpsproduce.

APPENDIX: BEITA TECHNICAL DETAILS,MODEL SPECIFICATION, AND NOTATION

This appendix presents the BEITA simulation environmentin greater detail. It does not provide general informationabout the PS-I toolkit, nor does it present technical detailsregarding all of its simulation capabilities.26 We wish to em-phasize, however, that the templates, scripts, and softwarenecessary to replicate our findings exactly are available toall researchers at the URL listed in the Acknowledgments.At least as important, however, is that researchers who maywish to explore the robustness of our findings to variationin particular parameter settings or coding rules may do soquite straightforwardly and systematically by adjusting thosesettings in the templates (snapshots) provided and observingthe consequences when those adjusted templates are used togenerate batches of 100 histories.The Appendix is divided into three sections. The first sec-

tion addresses issues related to the definition of agents inBEITA and provides more details regarding the distributionof agents of different types in the landscape. In the secondsection we present the rules for repertoire and identity trans-formation in BEITA. The third section discusses the condi-tions for the emergence of border cells.

Agents in BEITA

BEITA is designed to approximate the characteristics of amulticultural, multiethnic, multiregional state. The landscapeis a square grid (66× 66) and contains 4,356 cells. At the be-ginning of eachBeita history 4,096 of these cells are inhabitedby agents. The remaining 260 “border cells” form an externaland unchanging boundary along its four straight edges. (seeFigure 1).BEITA contains agents from different types or agent

classes. Agents belonging to different classes can be distin-guished visually by different icons. Agent behavior is af-fected by differences they may have in four parameters: (a)influence—the effect of one agent on other adjacent agents;(b) sight radius—thesizeof theagent’s “social”neighborhoodor the number of agents in its proximity that it surveys before

26 Such information has been provided elsewhere. See Dergachev2003 and Lustick 2002.

225


TABLE A1. Agent Classes in BEITASight Graphic Icon

Influence Radius Inactive Immutable (Figure 1)Basic agent 1 1 False False NoneInfluential 2 2 1 False False CrossInfluential 3 3 1 False False SpiralInfluential 4 4 1 False False CircleBorder cell 0 0 True True Black color

implementing rules governing identity activation and reper-toire composition; (c) inactive—whether an agent affectsother agents at all; and (d) immutable—whether an agentcan change its activated identity, i.e., whether it can be aff-ected at all. Combinations of values on these parameters pro-duce the five agent classes in BEITA (Table A1). Note thatonly a few of the many agent classes producible in PS-I areincluded in Beita.In addition to the parameters discussed above, each agent

in BEITA has its own unique location and a particular arrayof identities, with one identity activated at any time. Eachagent’s identity repertoire is a subset of the total number ofidentities available in BEITA. (There is a total array of 20identities present in Beita—its “spectrum.”) But at any par-ticular time step, only the identities within an agent’s identityrepertoire are available to that agent to be rotated into acti-vation on any particular time step. Identities in the spectrumare each assigned a number and a corresponding color. Eachagent’s activated identity is viewable via this color coding (seeFigure 1). Table A2 offers a summary of authority structuresand patterns of activation for agents in BEITA (Baseline).

TABLE A2. Summary of Authority Structures and Activation Patterns in BEITA (Baseline)Region Agent Class Repertoire Avg. Repertoire Size Influence NumberNW High-echelon RO {[Regime]} 3.00 4 4

Mid-echelon RO {[Regime], regional, parochial} 5.00 3 6Low-echelon RO {[Regime], regional, parochial} 5.02 2 81RA {[Regime], regional} 5.06 1 490RA {[Regime], other} 6.00 1 52PA {Regime, [other]} 6.02 1 391

NE & SW High-echelon RO {[Regime], regional} 3.50 4 8Mid-echelon RO {[Regime], regional, parochial} 5.00 3 10Low-echelon RO {[Regime], regional, parochial} 6.00 2 135RA {[Regime], regional, parochial} 6.40 1 622High-echelon RegO {Regime, [regional]} 4.00 4 1Mid-echelon RegO {Regime, [regional], parochial} 4.60 3 5Low-echelon RegO {Regime, [regional], parochial} 6.00 2 33RegA {Regime, [regional], parochial} 6.47 1 831RegA {[Regional], parochial} 6.00 1 43PA {Regime, regional, [parochial]} 6.00 1 403

SE High-echelon RO {[Regime]} 3 4 1Mid-echelon RO {[Regime]} 3 3 3Low-echelon RO {[Regime]} 3 2 54RA {[Regime], 16, 10, parochial} 5.61 1 113Strong DRM influential {[10], parochial} 3 4 2Mid DRM influential {[10], parochial} 3 3 1Low DRM influential {[10], parochial} 3 2 1DRM {Regime, [10], parochial} 4.67 1 93FRM {Regime, [16], parochial} 4.37 1 129PA {Regime, 10, 16, [parochial]} 5.38 1 627

Note: Brackets, active identity; curly braces, identities in repertoire; RO, regime official; RA, agent activated on one of the regime identities;PA, agent activated on one of the parochial identities; RegA, agent activated on one of the regional identities; RegO, regional officials;DRM, disgruntled regional minority; FRM, favored regional minority; other, in the NW region some agents are activated/subscribed toidentities associated with one of the other regions.

Repertoire and Identity Transformationin BEITA

Once BEITA begins to move forward in time agents beginto interact, guided by a set of simple, but precise algorithmsthat govern their adaptive responses to information avail-able from their local neighborhoods and from bias signals.These rules are designed to operationalize the main tenets ofconstructivist identity theory and to remain consistent withsocial psychological analyses of individual and group behav-ior, including social identity theory and self-categorizationtheory (Tajfel and Turner 1986; Turner et al. 1987; Turner andReynolds 2001). Under specified conditions the evaluation ofthe information may result in changing the agent’s activationor the composition of an agent’s identity repertoire.In general, agents act in a boundedly rational fashion,

adapting to their world by monitoring biases and align-ing their activated identities with the identity most promi-nently activated in their neighborhoods. For every mutableagent, the algorithms controlling these calculations take intoaccount identities available within the agent’s repertoire,

226


identities activated within its neighborhood, its currently ac-tivated identity, biases currently assigned to different identi-ties, and the influence levels of agents in its neighborhood,including itself.Each mutable agent updates its activation on every other

time step. On each updating time step each agent counts allactivated identities in its neighborhood (including its ownidentity), taking into account each agent’s influence, as wellas the biases assigned to the identities. This operation returnsan “identity weight,” or “count,” for each activated identityin an agent’s neighborhood.

counti =∑

activated[A]=iA∈ neighborhood

Influence[A]+ biasi .

In addition to calculating the relative weight of each iden-tity in its neighborhood, agents must also identify candidateidentities that could be discarded, rotated, swappedout, or ac-quired.A discard candidate is the identity with the smallest count

in an agent’s repertoire (including the active identity):

discard candidate = argmini∈repertoire

counti .

A swapout candidate is the identity with the smallest countamong the agents’ nonactive identities (i.e., all excluding theactive identity):

swapout candidate = argmini∈repertoirei �=activated

counti .

A rotate candidate is the identity with the highest countamong all the subscribed identities:

rotate candidate = argmaxi∈repertoire

counti .

Finally, an acquire candidate must not be part of the sub-scribed set and is the onewith the highest count in the identityspectrum:

acquire candidate = argmaxi /∈repertoire

counti .

As time progresses the repertoire of active agents evolvesaccording to the following rules. In this example the triggeringthresholds for each operation are those assigned in BEITAto basic agents.

1. If the count of the activated identity equals or is larger thanthe count for any other identity, the agent’s repertoire andactivated identity stay the same.

2. If the count for the rotate candidate is larger than or equalstwo, the agent activates on the rotate candidate.

3. If a discard candidate and an acquire candidate have beendefined and the count for the acquire candidate is equalto or larger than seven, the agent discards the candidateidentity and activates the acquired identity.

4. If both swapout and acquire identities have been de-fined, and the count of the acquire identity is equal toor larger than five, the acquire candidate replaces theswapout candidate in the agent’s repertoire.

new repertoire =

countactivated ≥max counti , old repertoire.countrotate candidate ≥ 2, activate(rotate candidate).countacquire candidate ≥ 7, activate(acquire candidate),

unsubscribe(discard candidate).countswapout candidate ≥ 5, subscribe(acquire candidate),

unsubscribe(swapout candidate).otherwise old repertoire.

Agent Transformation in BEITA

In PS-I one can set conditions governing the transformationof an agent from one agent class to another. In BEITA weimplemented a set of conditions that would allow the trans-formation of basic-agents into border cells. To be eligible forborder transformation (bt) an agent must meet a set of threeconditions.First, the activated identity must be defined as an oppo-

sitional identity (OI). Oppositional identities are a subset ofidentities that are not the dominant identity (DI). DI is theidentity, at any particular time, with a plurality ofBeita agentsactivated on it. It is computed from the distribution of acti-vated identities:

DI = argmax activatedi .

To qualify as an OI, no more than 20% of the agentsactivated on that identity can have the DI identity in theirrepertoires:

OIi ={1 opposition counti > 0.8 · activated,0 otherwise.

Second, the activated identity agent must be defined as asubordinated identity (SI). This condition is met in any timestep by identities activated by at least 10% of the agents inBEITA, excluding the DI.

SIi ={1 activatedi > 0.1 · active agent,0 otherwise.

Third, no individual agent meeting the first and secondconditions can transform into a border cell unless it is expe-riencing a high level of tension (a value of at least three).Tension is calculated for each agent by summing the num-ber of agents in its neighborhood with an activated identitydifferent from its own.

tension[A]=∑

B∈neighborhood[A]activated(B)!= activated(A)

1.

As indicated before, all three conditions (OI, SI, and ten-sion) must be met before an agent becomes eligible to trans-form into a border cell.

eligible for bt[A]= SIactivated(A) ∩ OIactivated(A) ∩ (tension[A]≥ 3).During an updating time step an eligible agent has a 20%

probability of actually transforming into a border cell.

REFERENCES

Adam, Heribert. 1971. Modernizing Racial Domination: SouthAfrica’s Political Dynamics. Berkeley: University of CaliforniaPress.

Aronoff,Myron J. 1998. “The Politics of Collective Identity.”Reviewsin Anthropology 27 (March): 71–85.

Axelrod, Robert. 1997. The Complexity of Cooperation: Agent-Based Models of Competition and Collaboration. Princeton, NJ:Princeton University Press.

Axelrod, Robert, and Ross A. Hammond. 2003. “The Evolution ofEthnocentric Behavior.” Presented at the Midwest Political Sci-ence Convention, Chicago.

Barth, Frederik. 1959. Political Leadership among Swat Pathans.London: Athlone Press, University of London.

227


Bhavnani, Ravi, and David Backer. 2000. “Localized Ethnic Con-flict and Genocide: Accounting for Differences in Rwanda andBurundi.” Journal of Conflict Resolution 44 (June): 283–306.

Brass, Paul R. 1974. Language, Religion and Politics in North India.London: Cambridge University Press.

Brass, Paul R.1980. “Ethnic Groups and Nationalities: The Forma-tion, Persistence, and Transformation of Ethnic Identities OverTime.” InEthnicDiversity andConflict inEasternEurope, ed. PeterF. Sugar. Santa Barbara, CA: ABC Clio, 1–68.

Brown, David. 1998. “Why Is the Nation-State so Vulnerable to Eth-nic Nationalism?” Nations and Nationalism 4 (January): 1–15.

Brubaker, Rogers. 1996.Nationalism Reframed: Nationhood and theNationalQuestion in theNewEurope. Cambridge:CambridgeUni-versity Press.

Carley, Patricia. 1997. U.S. Responses to Self-Determination Move-ments: Strategies for Nonviolent Outcomes and Alternatives to Se-cession.Washington, DC: United States Institute of Peace.

Cederman, Lars-Erik. 1997. Emergent Actors in World Politics: HowStates andNations Develop andDissolve. Princeton, NJ: PrincetonUniversity Press.

Cederman, Lars-Erik. 2001. “Agent-BasedModeling in Political Sci-ence.” Political Methodologist 10 (Fall): 16–22.

Cederman, Lars-Erik. 2002. “Endogenizing Geopolitical Bound-aries with Agent-Based Modeling.” Proceedings of the NationalAcademy of Science U.S.A. 99 (May): 7296–7303.

Chandra, Kanchan. 2001. “Introduction: Constructivist Findings andTheirNon-Incorporation.”APSA-CP:Newsletter of theOrganizedSection in Comparative Politics of the American Political ScienceAssociation 12 (Winter): 7–11.

Collier, Paul. 2000. “Economic Causes of Civil Conflict and TheirImplications for Policy.”World Bank. http://www. worldbank. org/research/conflict/papers/civilconflict. pdf (June 15).

Connor,Walker. 1998.Ethnonationalism: TheQuest for Understand-ing. Princeton, NJ: Princeton University Press.

Cornell, Svante E. 2002. “Autonomy as a Source of Conflict: Cau-casian Conflicts in Theoretical Perspective.” World Politics 54(January): 245–76.

Crawford, Beverly. 1998. “TheCauses of Cultural Conflict:Assessingthe Evidence.” In The Myth of “Ethnic Conflict:” Politics, Eco-nomics, and “Cultural” Violence, ed. Beverly Crawford and Ron-nie D. Lipshutz. International and Area Studies, Research Series,No. 98. Berkeley: University of California, 513–61.

Danspeckgruber, Wolfgang, ed. 1997. Self-Determination and Self-Administration: A Sourcebook. London: Lynne Rienner.

Dergachev, Vladimir. 2003. “Description of Sample Models Dis-tributed with Ps-I.” http://ps-i.sourceforge.net/manual.pdf

Doyle, Michael W., and Nicholas Sambanis. 2000. “Interna-tional Peacebuilding: A Theoretical and Quantitative Analy-sis.”American Political Science Review 94 (December): 779–801.

Eley, Geoff, and Ronald Grigor Suny. 1996. “Introduction: From theMoment of Social History to the Work of Cultural Representa-tion.” InBecomingNational: AReader, ed. Geoff Eley andRonaldGrigor Suny. New York: Oxford University Press, 3–37.

Epstein, Joshua M., John D. Steinbruner, and Miles T. Parker. 2001.“Modeling Civil Violence: An Agent-Based Computational Ap-proach.”CenteronSocial andEconomicDynamicsWorkingPaperNo. 20. Brookings Institution.

Fearon, James D. 1998. “Commitment Problems and the Spreadof Ethnic Conflict.” In International Spread of Ethnic Conflict:Fear, Diffusion, and Escalation, ed. David A. Lake and DonaldRothchild. Princeton, NJ: Princeton University Press, 108–26.

Fearon, James D., and David D. Laitin. 2003. “Ethnicity, Insurgency,and Civil War.” American Political Science Review 97 (February):75–90.

Freeman, Michael. 1999. “The Right to Self-Determination in Inter-national Politics: Six Theories in Search of a Policy.” Review ofInternational Studies 25 (July): 355–70.

Hale, Henry E. 2000. “The Parade of Sovereignties: Testing Theo-ries of Secession in the Soviet Setting.” British Journal of PoliticalScience 30 (January): 31–56.

Hardin, Russell. 1995. “Self-Interest, Group Identity.” In National-ism and Rationality, ed. Albert Breton, Gianluigi Galeotti, PierreSalmon, andRonaldWintrope. Cambridge: CambridgeUniversityPress, 14–42.

Hechter, Michael. 1975. Internal Colonialism: The Celtic Fringe in

British National Development, 1536–1966.Berkeley: University ofCalifornia Press.

Hechter, Michael. 2000. “Nationalism and Rationality.” Studies inComparative International Development 35 (Spring): 3–19.

Hechter,Michael, andDinaOkamoto. 2001. “PoliticalConsequencesofMinorityGroup Formation.”Annual Review of Political Science4 (June): 189–215.

Hirschmann, Albert O. 1970. Exit Voice and Loyalty: Responsesto Decline in Firms, Organizations, and States. Cambridge, MA:Harvard University Press.

Hoffmann, Matthew J. 2003. “Entrepreneurs and Norm Dynamics:An Agent-Based Model of the Norm Life Cycle.” Presented atPenn’sAgent-BasedModelingLaboratory (PAMLA). http://www.polisci.upenn.edu/abir/ private/Pamla/Hoffmann norms.doc.

Hoppe, Hans-Hermann. 1998. “TheWestern State as Paradigm.” So-ciety 35 (July/August): 21–25.

Horowitz, Donald L. 1985.EthnicGroups in Conflict.Berkeley: Uni-versity of California Press.

Horowitz,DonaldL. 1997. “Self-Determination: Politics, Philosophy,and Law.” InEthnicity and Group Rights, ed. Ian Shapiro andWillKymlicka. New York: New York University Press, 421–63.

Johnson,GaryR.1997. “TheArchitectureofEthnic Identity.”Politicsand the Life Sciences 16 (September): 257–63.

Kaufmann, Chaim D. 1998. “When All Else Fails: Ethnic PopulationTransfers and Partitions in the Twentieth Century.” InternationalSecurity 23 (Fall): 120–56.

King, Gary, and Lanche Zeng. 2001. “Improving Forecasts of StateFailure.”World Politics 53 (July): 623–58.

Kohli, Atul. 1997. “The Bell Curve of Ethnic Politics: The Riseand Decline of Self-Determination Movements in India.” InSelf-Determination and Self-Administration: A Sourcebook, ed.Wolfgang Danspeckgruber. Boulder, CO: Lynne Rienner, 309–36.

Kowert, Paul, and Jeffrey Legro. 1996. “Norms, Identity, and TheirLimits: A Theoretical Reprise.” In The Culture of National Secu-rity: Norms and Identity in World Politics, ed. Peter J. Katzenstein.New York: Columbia University Press, 451–97.

Kumar, Radha. 1997. “The Troubled History of Partition.” ForeignAffairs 76 (January/February): 22–34.

Laitin, David D. 1998. Identity in Formation: The Russian-SpeakingPopulations in the Near Abroad. Ithaca, NY: Cornell UniversityPress.

Laitin, David D. 2000. “What Is a Language Community.”AmericanJournal of Political Science 44 (January): 142–55.

Laitin, David D. 2001. “Secessionist Rebellion in the Former SovietUnion.” Comparative Political Studies 34 (October): 839–86.

Laitin, DavidD., andDaniel Posner. 2001. “The Implications of Con-structivism for Constructing Ethnic Franctionalization Indices.”APSA-CP: Newsletter of the Organized Section in ComparativePolitics of the American Political Science Association 12 (Winter):13–17.

Lakatos, Imre. 1970. “Falsification and theMethodology of ScientificResearch Programmes.” In Criticism and the Growth of Knowl-edge, ed. Imre Lakatos andAlanMusgrave. NewYork: CambridgeUniversity Press, 91–196.

Lapidoth, Ruth. 1996. Autonomy: Flexible Solutions to Ethnic Con-flicts.Washington, DC: United States Institute of Peace Press.

Laptev, Alexander A. “Modeling of Sociogenetic Processes,” http://www.univer.omsk.su/MEP.

Lijphart, Arend. 1977.Democracy in Plural Societies: AComparativeExploration. New Haven, CT: Yale University Press.

Lijphart, Arend. 1985. Power-Sharing in South Africa. Berkeley: In-stitute of International Studies, University of California Press.

Lustick, Ian S. 1979. ”Stability in Deeply Divided Societies: Conso-ciationalism vs. Control.”World Politics 31 (April): 325–44.

Lustick, Ian S. 1980. Arabs in the Jewish State: Israel’s Control of aNational Minority. Austin: University of Texas Press.

Lustick, Ian S. 1996. “History, Historiography, and Political Science:Multiple Historical Records and the Problem of Selection Bias.”American Political Science Review 90 (September): 605–18.

Lustick, Ian S. 2000. “Agent-Based Modeling of Collective Identity:Testing Constructivist Theory.” Journal of Artificial Societies andSocial Simulations 3 (January) http://jasss.soc.surrey.ac.uk/3/1/1.html.

Lustick, Ian S. 2002. “PS-I: A User-Friendly Agent-Based ModelingPlatform for Testing Theories of Political Identity and Political

228


Stability.” Journal of Artificial Societies and Social Simulations 5(June): http://jasss.soc.surrey.ac.uk/5/3/7.html.

Lustick, Ian S., with Dan Miodownik. 2000. “Deliberative Democ-racyandPublicDiscourse:TheAgentBasedArgumentRepertoireModel.” Complexity 5 (March): 13–30.

Lustick, Ian S., and Dan Miodownik. 2002. “The Institutionalizationof Identity: MicroAdaptation,Macro Effects, and Collective Con-sequences.” Studies in Comparative International Development 37(Summer): 24–53.

Macy,MichaelW., andRobertWiller. 2002. “FromActors to Factors:Computational Sociology and Agent-Based Modeling” AnnualReview of Sociology 28 (August): 134–66.

McGarry, John, and Brendan O’Leary. 1993. “The Macro-PoliticalRegulation of Ethnic Conflict.” In The Politics of Ethnic ConflictRegulation, ed. John McGarry and Brendan O’Leary. London,Routledge, 1–40.

McGarry, John, and Brendan O’Leary. 1994. “The Political Regu-lation of National and Ethnic Conflict.” Parliamentary Affairs 47(January): 94–115.

Mozaffar, Shaheen, and James R. Scarritt. 2000. “Why TerritorialAutonomy Is Not a Viable Option for Managing Ethnic Con-flict in African Plural Societies.” In Identity and Territorial Auton-omy in Plural Societies, ed. William Safran. London: Frank Cass,230–53.

Nagel, Joane. 1994. “Constructing Ethnicity: Creating and Recreat-ing Ethnic Identity and Culture.” Social Problems 41 (February):152–76.

O’Leary, Brendan, Ian S. Lustick, and Thomas Callaghy, eds. 2001.Right-Sizing the State: The Politics of Moving Borders. Oxford:Oxford University Press.

Reilly, Ben, and Andrew Reynolds. 1999. Electoral Systems andConflict in Divided Societies.Washington, DC: National AcademyPress.

Roeder,PhilipG. 1991. “SovietFederalismandEthnicMobilization.”World Politics 43 (January): 196–232.

Roeder, Philip G. 2003. “Clash of Civilizations and Escalation ofDomestic Ethnopolitical Conflicts.” Comparative Political Studies36 (June): 509–41.

Saideman, Stephen M. 1998. “Is Pandora’s Box Half Empty or HalfFull? The Limited Virulence of Secessionism and the DomesticSources of Disintegration.” In The International Spread of EthnicConflict: Fear, Diffusion, and Escalation, ed. David A. Lake andDonaldRothchild. Princeton,NJ: PrincetonUniversity Press, 127–50.

Sambanis, Nicholas. 2000. “Partition as a Solution to EthnicWar: AnEmpirical Critique of the Theoretical Literature.” World Politics52 (July): 437–83.

Scarritt, James R., Susan M. McMillan, and Shaheen Mozaffar.2001. “The Interaction between Democracy and EthnopoliticalProtest and Rebellion in Africa.” Comparative Political Studies 34(September): 800–27.

Smith,Anthony,D. 1981.TheEthnicRevival.Cambridge:CambridgeUniversity Press.

Smith, Anthony, D. 1986. The Ethnic Origins of Nations. Oxford:Blackwell.

Snyder, Jack. 2000. From Voting to Violence: Democratization andNationalist Conflict. New York: Norton.

Spencer,Metta. 1998. “Introduction.” In Separatism:Democracy andDisintegration, ed.Metta Spencer. London:Rowman&Littlefield,1–6.

Srbljinovic Armano, Penzar Drazen, Rodik Petra, and KardovKruno. 2003. “An Agent-Based Model of Ethnic Mobilisation.”Journal of Artificial Societies and Social Simulation 6 (January):http://jasss.soc.surrey.ac.uk/6/1/1.html.

Tajfel, Henri, and John C. Turner. 1986. “The Social Identity Theoryof Intergroup Behavior. In The Psychology of Intergroup Rela-tions, ed. StephenWorchel &WilliamG.Austin. Chicago: Nelson–Hall, 7–24.

Tullberg, Jan, and Birgitta S. Tullberg. 1997. “Separation orUnity?AModel for Solving Ethnic Conflicts.” Politics and the Life Sciences16 (September): 237–49.

Turner, JohnC., andKatherine J.Reynolds. 2001. “TheSocial IdentityPerspective in Intergroup Relations: Theories, Themes and Con-troversies.” In Blackwell Handbook of Social Psychology: Inter-group Processes, ed. Rupert J. Brown&SamuelGaertner. Oxford,UK: Blackwell, 133–52.

Turner, John C., Michael A. Hogg, Penelope J. Oakes, Stephen D.Reicher, andMargaret S.Wetherell. 1987.Rediscovering the SocialGroup: A Self-Categorization Theory. Oxford: Blackwell.

van der Veen, A. Maurits. 2002. “The Emergence of European Iden-tity:Modeling International Experiences.” Presented at the annualmeeting of the American Political Science Association, Boston.

Verdery, Katherine. 1991. National Ideology Under Socialism: Iden-tity and Cultural Politics in Ceausescu’s Romania. Berkeley: Uni-versity of California Press.

Wilkinson, Steven Ian. 2000. “India, Consociational Theory, andEth-nic Violence.” Asian Survey 40 (September/October): 767–91.

Wilkinson, Steven Ian. 2001. “Constructivist Assumptions and Eth-nic Violence.” APSA-CP: Newsletter of the Organized Section inComparative Politics of the American Political Science Association12 (Winter): 17–20.

Yiftachel,Oren. (2000). “Ethnocracy and ItsDiscontents:Minorities,Protest and the Israeli Polity.” Critical Inquiry 26 (Summer): 725–56.

229

Documents

Secessionism in Multicultural States: Does Sharing …othello.alma.edu/~cartrite/abm/lusmioeid.pdfSecessionism in Multicultural States: Does Sharing Power Prevent ... (Cederman 2001;MacyandWiller2002).7