In what ways can systems of democratic government be structured to help harmonize relations among groups in multiethnic societies? Many policymakers and scholars alike believe that broadly inclusive government, or power sharing, is essential to successful conflict management in societies beset by severe ethnic conflicts. Broadly interpreted, power-sharing political systems are those that foster governing coalitions inclusive of most, if not all, major mobilized ethnic groups in society. Decision making is based on a broadconsensus that transcends groups through coalitions that are widely inclusive. Consensus or near-consensus decision making is differentiated from simple majoritarian forms of democracy, in which decisions are taken for the entire society based on the preferences of a minimum winning majority (Rae 1969).
The term "power sharing" has been defined by scholars such as Arend Lijphart as a set of principles that, when operationalized through practices and institutions, provide every significant identity group in a society representation and decision-making abilities on common issues and a degree of autonomy over issues of importance to the group. Lijphart's principles of power sharing-known as "consociational democracy" (derived from the Latin term consociatio, meaning to associate in an alliance)-differentiate coalescent democracy from majoritarian democracy (Lijphart, 1977a: 25).
Scholars differ over whether the consociational power-sharing approach-in which groups are represented as groups (usually through ethnically exclusive political parties), in essence as building blocks of a common society-leads to better relations among ethnic groups in multiethnic societies than practices that seek to foster political organizations that transcend ethnic group differences, or an integrative (or pluralist) approach. The integrative approach sees as ideal the creation of pre-election coalitions among ethnic parties or (less commonly) the creation of broad multiethnic parties based on interests that transcend ethnic identities, such as region or common economic interests. Traditionally, pluralism also relies on the forces of economic interaction to help create social cleavages that crosscut ascriptive identity. The preeminent example of this crosscutting pattern of democracy is the multiethnic United States, best described in Seymour Martin Lipset's 1960 book, Political Man; India is another oft-cited example (Brass 1990).
In addition, prominent scholars of the politics of multiethnic societies differ over the scope of the term "power sharing." Some, such as Lijphart, argue that the consociational approach to power sharing encompasses a wide variety of practices and instances. Others, such as Donald Horowitz, argue that the consociational approach is more narrow in meaning and that many cases of consociational democracy cited by consociationalists (such as Malaysia or Lebanon) are not in fact consociational but integrative. Lijphart contends that the integrative approach is essentially majoritarian in nature, and that integrative mechanisms encourage majority representatives to behave moderately and with sensitivity toward minorities, which are still excluded from real political power.
Thus, the power-sharing debate revolves around the following central question: Which broad approach best manages conflict, one that essentially sees ethnic groups as building blocks of national politics in multiethnic states, or one that purposefully encourages the formation of political blocs across group lines? Consociationalists suggest that conflict management isbest promoted by accommodation among ethnic group leaders representative of their communities through cooperative problem-solving in post-election coalitions. Critics of the consociational approach-such as Horowitz (1985)-argue that institutions and practices that create incentives for the formation of pre-election coalitions and that encourage intragroup competition, rather than intergroup competition, better reduce the likelihood of violent conflict. Ideally, integrative mechanisms would lead to multiethnic parties or organizations that transcend narrow communal interests.
A central theme of this paper is that the concept of "power sharing" encompasses both consociational power sharing and integrative power sharing. Both the consociational and integrative approaches to ethnic conflict management seek to promote governing coalitions that are broadly inclusive of all ethnic groups in a deeply divided multiethnic society-the hallmark of power sharing-but advocates of these approaches sharply disagree over when and how such coalitions are formed, and which specific institutions and practices better manage ethnic conflict. For this reason, power sharing should be interpreted as encompassing both distinct approaches, keeping in mind that there are different types of institutions and practices to promote democratic ethnic conflict management which can be assembled and arranged in many different ways.
Scholars of comparative politics tend to agree that simple forms of majoritarian government contain special problems for multiethnic societies. Minorities in particular in such societies do not equate democracy with freedom or participation, but with the structured dominance of adversarial majority groups. Permanent minorities such as Tamils in Sri Lanka, Catholics in Northern Ireland, and whites in South Africa have feared the consequences of electoral competition, especially when the expected consequence of majority victory is discrimination against them. For minority groups, losing an election is not simply a matter of losing office, but of losing the means for protecting the survival of the group.1
Majoritarian democracy is typified by the Westminster system of small single-member districts with first-past-the-post (plurality) electoral rules; the party (or parties in coalition governments) with a majority of the seats forms the government while other parties remain in loyal opposition. Analytically, there are three problems with simple majoritarian democracy in multiethnic societies: the possibility of permanent exclusion of minority group-based political parties, the lack of "floating" voters whose preferences are formed on other-than-ascriptive criteria such as class,2 and the pervasiveness of radical outbidding on divisive ethnic issues. Although simple majority rule may be "fairest" from a theoretical point of view (Rae 1969), the scholarly consensus recognizes the principle's limitations in multiethnic societies.3 (Horowitz [1993: 30] also demonstrates how a procedurally free and fair election can lead to equally exclusive minority rule.)
Advocates of power sharing in divided societies agree on the dangers of majoritarianism, citing the potential distortions in vote-to-seat outcomes, the inability of geographically dispersed minority parties to achieve representation, and-in the context of an ethnic party system-the likelihood that a single ethnic group or coalition of ethnic groups will govern exclusively and to the detriment of others.4 Lijphart, the most indefatigable critic of majoritarian and plurality electoral rules for divided societies (and indeed in other democracies), identifies the core problem when he refers to the potential for "majority dictatorship" (1985: 102). Horowitz concurs, arguing in his seminal work, Ethnic Groups in Conflict, that under conditions of simple majority rule:
Ethnic parties developed, majorities took power, and minorities took shelter. It was a fearful situation, in which the prospect of minority exclusion from government, underpinned by ethnic voting, was potentially permanent . . . Civil violence, military coups, and the advent of single party regimes can all be traced to this problem of inclusion-exclusion (1985: 629).
Simply put, simple majority rule results in minimum winning coalitions that tend to exclude a significant minority; when minority preferences are intense, and when there is little chance of their becoming a majority, a recipe for conflict exists. Simple majoritarianism in a deeply divided society leads to zero-sum politics (Welsh 1993). Without an assurance that the electoral system will not lead to permanent exclusion, why should a minority group that perceives a threatening environment be willing to accept the inherent risks of electoral competition?
Rejection of majoritarian democracy does not mean a rejection of democratic values. What distinguishes advocates of majoritarianism from advocates of coalescent democracy, or power sharing, is belief in the prospects for "political engineering" (Sartori 1968) to mitigate conflicts in divided societies. That is, the rules of the political game can be structured to institutionalize moderation on divisive ethnic themes, to contain the destructive tendencies, and to preempt the centrifugal thrust created by ethnic politics. There is no assertion that deft political engineering can prevent or eradicate deep enmities, but appropriate institutions can nudge the political system in the direction of reduced conflict and greater governmental accountability. The common assumption is that choices about the basic rules of the game affect its outcomes. Horowitz writes: "Where there is some determination to play by the rules, the rules can restructure the system so the game itself changes" (1985: 601).
The central question of political engineering is: In multiethnic societies, which kinds of institutions and practices create an incentive structure for ethnic groups to mediate their differences through the legitimate institutions of a common democratic state? Alternatively, how can the incentive system bestructured to reward and reinforce political leaders who are moderate with respect to divisive ethnic themes and to persuade citizens to support moderation, bargaining, and reciprocity among ethnic groups?
As highlighted above, there are two distinct approaches to constructing conflict-ameliorating democratic institutions in deeply divided societies: the consociational model most associated with Lijphart (1968, 1969, 1977a, 1977b, 1985) and what I term the integrative approach associated with Horowitz (1985, 1990, 1991, 1993). These approaches are summarized in Table 1. Although dichotomizing these approaches may be a too-simplistic description of their advocates' views, I do so here in order to highlight the differences. The former approach places greater faith in assurances for minority group protection, whereas the latter places greater emphasis on the role of incentives in encouraging interethnic cooperation. What unites them is the belief in coalescent democracy as an alternative to the adverse effects of majoritarianism and the assumptions that support a rejection of majoritarian practices. "Coalescent" decision making is argued to be a better prescription for the ills that plague deeply divided societies than the adversarial pattern associated with majoritarian democracy.5
Consociationalism, above all, relies on elite (political leaders) cooperation as the principal mechanism of successful conflict management in deeply divided societies.6 Consociationalists suggest that even if there are deep communal differences, overarching elite cooperation is a necessary and sufficient condition to assuage conflict.7¦ In the consociational approach, elites directly represent various societal segments and act to forge political ties at the center. This is the case in many of the consociational democracies-Belgium, the Netherlands, Switzerland, Malaysia (1955-69), and Lebanon (1943-1975)-that have been considered successful experiences by these theorists.8
Table 1. Approaches to Power Sharing| Consociational | Integrative | |
| Characteristics | Elites cooperate after elections to form multiethnic coalitions and manage conflict; groups are autonomous; minorities are protected. | Parties encouraged to create coalitions before elections, creating broadly inclusive but majoritarian governments. |
| Principles | Broad-based or "grand" coalitions, minority veto, proportionality in allocation of civil services positions and public funds, group autonomy. Dispersion and devolution of power, promotion of intraethnic competition, inducements for interethnic cooperation, policies to encourage alternative social alignments, managed distribution of resources. | Institutions and practices to promote these principles Parliamentary government, proportional reservation of seats, PR electoral system. Federalism, vote pooling, electoral systems, president elected by "super-majority." |
| Strengths of the approach | Provides groups firm guarantees for the protection of their interests. | Provides politicians with incentives for moderation -- "coalitions of commitment." |
| Weaknesses | "Coalitions of convenience." Elites may pursue conflict rather than try to reduce it; communal groups may not defer to their leaders; system relies on constraints against immoderate politics. | Lack of whole-country empirical examples of working systems; assumption that politicians respond to incentives and citizens will vote for parties not based on their own group. |
According to Lijphart, consociationalism relies on four principles: broad-based or "grand" coalition executives; minority veto; proportionality in the allocation of civil service positions and public funds; and group autonomy. Lijphart argues persistently that the institutions that give life to these principles must be specially adapted to the society they are to serve, and they cannot be implemented and expected to work singularly. Lijphart also identifies a number of conditions that are favorable to the successful operation of consociational democracy: popular deference to elites, "a multiple balance of power, small size of the country involved, overarching loyalties, segmental isolation, prior traditions of elite accommodation, and-although much more weakly and ambiguously-the presence of cross-cutting cleavages" (1977a: 54).
Power sharing in the executive in a grand coalition, or a variant thereof, ensures that the minority is not permanently excluded from political power.9 Parliamentary systems are argued to be more conducive to the creation of inclusive governing coalitions. In grand coalitions, political elites-representing the various segments of society-thrash out their differences in an effort to reach consensus, but public contestation among them is limited. The common denominator, and most important feature, is that decision making takes place consensually at the top among elites representing underlying social segments (Lijphart 1977a: 31-6).
The second feature of consociationalism is the mutual or minority veto, through which each segment is given "a guarantee that it will not be outvoted by the majority when its vital interests are at stake" (Lijphart 1977b: 118). Through the mutual veto, the majority's ability to rule is qualified by "negative minority rule" (Lijphart 1977a: 36). The minority veto is at the heart of the concrete assurances of consociationalism. The veto provides an ironclad guarantee of political protection to each segment on issues related to its vital interests. While the minority veto gives minorities the right to prevent action by others on the most sensitive issues, such as language, cultural, or education rights, it also serves a more important overriding goal. Like the Calhounian "concurrent majority," it invests each segment with thepower of protecting itself (Lijphart 1977a: 37).
In every sphere of political life, the principle of proportionality lies behind consociational practices. Proportionality is introduced at every level of government decision making (central, regional, and local) to give minority groups power, participation, and influence commensurate with their overall size in society. The principle is manifested in two ways. First, through the electoral system proportional representation (PR) is used to faithfully translate the demographic strength of the segments into commensurate representation in parliament; parties are awarded seats in parliament in direct proportion to votes garnered in an election. Second, the allocation of resources by the state-including the appointment of civil servants and public spending-should be doled out according to the proportionality principle.
Either through territorial federalism or "corporate federalism" (nonterritorial autonomy), consociationalism provides internal autonomy for all groups who want it by devolving decision-making authority to the segments. Lijphart draws distinctions between those issues that concern the common interest and those that primarily concern the segments; on the former, decisions are made by consensus, otherwise decision-making power is delegated to the segments. The basic principle underlying communal autonomy is "rule by the minority over itself in the area of the minority's exclusive concern" (Lijphart 1977a: 41). An important feature of the call for entrenched group rights on certain issues is the principle of "voluntary affiliation." Group identification should not be predefined or determined; instead, the segments of society would be able to define themselves through the proportional electoral system (Lijphart 1995).
Lijphart consistently asserts that consociationalism is the only viable option for democracy in divided societies: "For many plural societies of the non-Western world, therefore, the realistic choice is not between the British [majoritarian] model of democracy and the consociational model, but between consociational democracy and no democracy at all" (1977a: 238).
Consociationalism is not without its critics, of course. Below I will address criticisms of some of the specific practices (as opposed to principles) that are raised in response to the policy recommendations that flow from consociationalism, but here I would like to raise three broad criticisms of the consociational approach in order to better highlight the differences between it and the integrative approach. They are the reliance on elite accommodation and the problem of elite-initiated conflict; the reification of ethnic identity; and the tendency toward antidemocratic and inefficient decision making. (A fourth broad criticism raised by Horowitz, that consociationalism relies on constraints, not incentives, is outlined below in the discussion ofthe integrative approach.)
Consociationalists have been criticized for the assertion that elites can effectively regulate conflict in divided societies. As the Anglo-Irish Accord of 1985 demonstrates-as does the 1990 failure of the Meech Lake Accord agreed to by Canada's provincial leaders in 1987-even though political elites may agree on a formula for accommodation, peace cannot endure without grassroots backing. In Northern Ireland, the 1985 Anglo-Irish power-sharing agreement was reached without the inclusion of local Ulster, unionist, Protestant involvement; this constituency perceived the agreement, negotiated by the United Kingdom on its behalf, as a step toward a unified Ireland. As Rose (1990: 148) has suggested, "exclusion from the deliberations was regarded as part of a deliberate British plan to `sell out' the Protestant majority." Moreover, George Tsebelis (1990 b) suggests that consociational institutions may provide incentives for politicians to foment conflict along group lines in order to bolster their own bargaining position vis-à-vis other groups at the political center-what he terms "elite-initiated conflict."
In a similar vein, Steven Burg is critical of consociationalism as the "ultimate form of elite manipulation and control," reducing the accountability of political leaders to their communities. He writes:
There is mounting evidence that consociational arrangements (power sharing and mutual veto) encourage elites to rule in opposition to mass beliefs. Why did Yugoslavia and Czechoslovakia fall apart? It was not because of . . . interethnic hatreds at the mass level. It was because of elite mobilization of latent nationalisms, and because the structural characteristics of each system included power sharing and mutual veto, enabling [for example] the Slovene regional leadership and the Slovak regional leadership to paralyze their respective federal governments. . . . The mass electorates did not want their respective countries to break up until elites had pushed these conflicts beyond the point of no return.10¦
Critics also assert that consociationalism serves to maintain, legitimize, and strengthen segmental claims against the state, reinforcing and entrenching ethnicity in the political system. By "freezing" group boundaries in the political system, for example through statutory reservation of offices for specific group representatives, a consociational power-sharing system is said to be an undynamic model for conflict management (Barry 1975). Providing structural guarantees for communities, for example through a minority veto, can provide systemic incentives for maintaining the rigidity of the segments. Finally, consociational institutions are arguably antidemocratic because they can stifle vigorous opposition politics. For example, the absence of an opposition party in a grand coalition may detract from the accountability of the government. For example, Marc Chernick (1991) suggests that the exclusive nature of Colombia's "National Front" government between 1958 and 1974 led to the emergence of insurgentgroups in subsequent years.
In response to these criticisms, Lijphart refers to the consociational arrangement not as an institutional blueprint, but as a set of principles to which certain institutions-such as a proportional representation electoral system-are naturally suited. For example, he replies to the charge that consociationalism can "freeze" or rigidify segments by arguing that the PR electoral system allows the segments to "define themselves." While the institutions of consociational decision making vary, its advocates argue, the principles are rediscovered time and time again as societies seek solutions to the existence of intense ethnic politics and methods to harness ethnicity for constructive purposes.11
In contrast with the consociational model, Horowitz (1985: 597-600) proposes a typology of five mechanisms aimed at reducing ethnic conflict: dispersions of power, often territorial, which "proliferate points of power so as to take the heat off of a single focal point"; devolution of power and reservation of offices on an ethnic basis in an effort to foster intraethnic competition at the local level; inducements for interethnic cooperation, such as electoral laws that effectively promote pre-election electoral coalitions through vote pooling; policies to encourage alternative social alignments, such as social class or territory, by placing political emphasis on crosscutting cleavages; and reducing disparities between groups through managed distribution of resources.
Horowitz's prescriptions for conflict-regulating institutions for divided societies overlap those of Lijphart in certain respects: both advocate federalism, for example, and assert the importance of proportionality and ethnic balance. Yet Horowitz is an indefatigable critic of the consociational model for two important reasons (1985: 568-76; 1991: 137-45). First, he argues, is the problem of "elite-initiated conflict" that Tsebelis and Burg also identify. "There is no reason to think automatically," Horowitz writes, "that elites will use their leadership position to reduce rather than pursue conflict" (1991: 141). Consociationalism overestimates deference by communal groups to their leaders and underestimates the power and role of popular dissatisfaction with intergroup compromise.
Second, consociational institutions rely on constraints against immoderate politics, such as the mutual or minority veto, as opposed to incentives for moderation (1991: 154-60). Horowitz argues that political institutions should encourage or induce integration across communal divides. For effective democratic governance in a divided society, moderates must be rewarded, extremists sanctioned. The aim is to engineer a centripetal spin to the political system by providing electoral incentives for broad-based moderation by political leaders and disincentives for extremist outbidding (1985: 601-52). This differentiates Horowitz's prescriptions from those of consociationalism in two important respects.
First, the key to any successful democratic political system in divided societies is to provide demonstrable incentives for politicians to appeal beyond their own communal segments for support. The only assumption is this: politicians will do whatever they need to do to get elected; they are rational electoral actors (Horowitz 1991: 261). When politicians are rewarded electorally for moderation, they will temper their rhetoric and actions. Given this premise, the political system can be engineered essentially to encourage intergroup cooperation as a prerequisite for electoral success. Horowitz contends that incentives are better than consociational constraints (like the mutual veto) because they offer reasons for politicians and divided groups to behave moderately, rather than obstacles aimed at preventing them from pursuing hegemonic, defeat-the-other aims.
The second difference is a concern with constituency-based moderation rather than reliance on political leaders as the engines of moderation. The solution is to design the electoral system so that leaders must appeal to underlying moderate sentiments in the electorate and shun the forces of extremism in order to win elections. Office-seekers, by appealing to the most moderate sentiments of the electorate, maximize moderation at both the elite and popular levels. Looking for the basis of consent at the constituency level allows politicians to make the kinds of compromises they must make at the center if the divided society is to be truly democratic and stable. The key to constituency-based moderation is the electoral system. To safeguard minority interests, according to Horowitz, the system should make the votes of minority members count. Minorities should have more than representation, they should have influence. What institutions and practices are argued to have these effects?
Dramatic devolution of power can serve four important purposes in divided societies, according to Horowitz (1985: 601). First, it can combine with the electoral system to encourage the party proliferation that is conducive to intersegmental compromise and coalition building. Second, politics at the regional and local levels can serve as training grounds for politics at the center: political leaders can form intergroup ties at the constituency level before they contest higher-stakes issues at the level of central government. Third, federalism disperses conflict at the center by resolving some issues at subtier levels, and may in communally homogeneous federal states promote cleavages within groups. Finally, it creates difficulties for any parties hoping to get a hegemonic grip on the entire country; capturing all of the provincial states would be a difficult task. For example, the adoption of federalism at the time of democratization in Spain is an instructive example of successful ethnic conflict management through devolution (Horowitz 1985: 623; Share 1986). "Federalism can either exacerbate or mitigate ethnic conflict," he writes, ". . . much depends on the number of components, the number of states, boundaries, and the ethnic composition" (Horowitz 1985: 603).
To Horowitz, divided societies need electoral systems that fragment support of one or more ethnic groups, especially ethnic majorities; induce interethnic bargaining; encourage the formation of multiethnic coalitions; produce fluidity and a multipolar balance; and produce proportional outcomes. Three types of electoral systems can achieve these aims: a subsequent preference voting system (preferably among such systems, alternative voting or AV); mixed lists with a common voters roll; and single-member districts in multiethnic constituencies. In each instance, the purpose is to promote vote pooling by candidates or parties across ethnic lines. Although electoral systems and conflict management will be more thoroughly discussed below, a brief introduction here will highlight the differences between the consociational and integrative approaches.
Why are electoral systems that provide for vote pooling superior for divided societies, in Horowitz's view? The logic is this: in order to win, politicians must seek to obtain the second- or third-preference votes of those who would not ordinarily vote for them (presumably because they do not represent the voter's community). In order to gain second- or third-preference votes, leaders must behave moderately toward other communal groups. Outbidding on ethnic themes will inevitably occur, Horowitz agrees, but so too will moderation. In response to the incentive structure of the electoral system, most politicians will vie to appear the most moderate-they will compete with one another to define and occupy the political center. Centripetal forces will override centrifugal ones. The critical difference between the consociational approach to electoral systems and Horowitz's is thus the formation of electoral coalitions by constituents as they specify their second or third preferences beyond their own narrow group interests. Horowitz cites the system established by the Sri Lankan constitution of 1978, and the electoral politics of the Indian state of Kerala, where four major ethnic blocs share power in a fluid system of changing coalitions and alliances, as examples of successful interethnic vote pooling (1993).
A presidency, argues Horowitz, if elected directly on the basis of a super-majority distributional formula or a subsequent preference voting method, is a less exclusive institution than parliamentarism. Presidentialism is argued to have two important advantages in divided societies. First, if a president is elected with an electoral system that requires broadly distributed support, an executive who has the broadest possible national appeal can be elected. A strong, statesmanlike, moderate president-forced to appeal to the least common denominator of electoral sentiments-can serve a unifying, nation-building role (Horowitz, 1990b). Second, a strong executive would be able to push legislation through a divided parliament. If strong but benevolent leadership is required, to make tough economic decisions or redress historical injustices, for example, a strong president is desirable. Anexample of such a presidential system, according to Horowitz, is Nigeria's (1985: 636).
Horowitz's broad approach to ethnic conflict management-"the political incentive structure is one package," he writes (1985: 651)-has also encountered criticism, considered here, as have the specific conflict-regulating practices (considered below). There are four interrelated concerns: a paucity of empirical examples of the system at work; the questionable assumptions that politicians will respond to the incentive system for moderation if it exists, and that voters will be willing to vote for parties not based in their own group; and the contention that the electoral systems Horowitz advocates are essentially majoritarian in nature. Like criticisms of consociationalism, these concerns go beyond simple conflict-regulating mechanisms and are rooted in basic beliefs about the fluidity and malleability of ethnic identity and representation.
At the heart of the difference between consociational and integrative approaches to power sharing are the nature and formation of multiethnic coalitions. In the consociational approach, coalitions are formed after an election by elites who realize that exclusive decision making will make the society ungovernable or who are compelled to do so by prior constitutional arrangements that are based on the same reasoning. In an integrative power-sharing system, coalitions are formed prior to an election-either as a coalition of parties in pre-election pacts (vote pooling) or by a party with a broad multiethnic candidate slate. Consociational arrangements formed after elections, Horowitz contends, are fragile and tenuous "coalitions of convenience" as opposed to firm and enduring "coalitions of commitment" (1985: 365-95).
A central argument of this paper is that in the above-outlined debate, neither approach can be said to be the best in all circumstances. Rather, the two approaches should be seen in contingent terms, and in terms of a spectrum of options from the most consociational to the most integrative. The appropriate question is: Under what conditions is the consociational approach likely to mitigate conflict, and under what conditions is an integrative approach likely to produce success? The challenge is not to develop a singular model of conflict-regulating practices, but rather a menu of conflict-regulating practices from which policymakers can choose and adapt to the intricacies and challenges of successfully regulating any given ethnic conflict.
Eric Nordlinger's seminal 1972 study, Conflict Regulation in Divided Societies, identified six political methods and practices, the presence of which account for successful conflict regulation in societies with deep ethnic fissures: stable governing coalitions, the principle of proportionality, mutual veto, purposive depoliticization, compromises on key issues, and concessions by conflict groups. These practices generally reflect the consociational vein of thinking about democracy in divided societies, relying principally on elite accommodation, populardeference to elites, group solidarity, and legitimate group representation. Given recent experiences with power sharing, both successful and unsuccessful, and criticisms lodged against consociational and integrative approaches, it is possible to refine, build upon, and expand Nordlinger's typology to encompass an amended and enlarged typology of conflict-regulating practices, including integrative practices. It is useful to consider power-sharing practices in terms of three sets of variables that apply to both approaches: territorial divisions of power, decision rules, and defining state/ethnic relations. The first two deal with constitutional structure, whereas the third relates to public policies.
When the territorial dimensions of ethnicity are strong, practices that configure the territorial division of power are exceptionally important. The range of options within these types of practices is between partition (separation) and a centralized unitary state (the ultimate form of sharing), with a variety of options such as confederal, federal, and semiconfederal systems in between. It is widely agreed that the territorial division of power can serve a variety of purposes beyond simple devolution, including providing economic mechanisms for affecting the distribution of resources and political mechanisms for reducing the stakes of conflict at the center.
Among the variety of methods for dividing territory, federalism is the most extensively analyzed for its potential conflict-regulating effects. Federalism, it should be noted, can be structured for either consociational or integrative purposes. Indeed, the opportunities for innovation are so extensive that federalism can be structured to serve both ends within a given state. In multiethnic societies, the test of federalism is the degree to which territorial units coincide with or are parallel with communal boundaries.12 By promoting intraethnic conflict within each subnational territorial unit, federalism can potentially create incentives for interethnic cooperation, encourage alignments along nonethnic interests, and be fiscally structured to level socioeconomic disparities. Devolution of power can give minorities some degree of power when it is unlikely that they would ever achieve majority status at the center.13
Federalism implies a division of power based on mutual consent. The national or central government is bestowed with a defined area of authority and the territorial units are provided degrees of autonomy, with both tiers of government enjoying some limited coordinated powers-what the Europeans call "competences."14 The hallmark of federalism is that neither the center nor the regions can amend the arrangement without mutual consent (Wheare 1964: xviii). It is through federal structures that the principle of unity through diversity, an essential norm of democracy in divided societies, can best be realized; this is especially true with respect to the protection of minorities through a panoply of options such as grants of autonomy, indigenous rights and semisovereign ancestral lands, and recognition of limitedterritorial self-determination (Hannum 1990).
Like other power-sharing practices, federalism is highly flexible. As new territorial units are created, so too are opportunities for innovative practice. Six of India's states, for example, have adopted the use of ombudsmen; many have complex language policies to meet local ethnic demands. It is for this reason that federalism offers innumerable opportunities for addressing the absolute complexity of ethnic demands in deeply divided societies within a democratic framework. For complex multiethnic states in the developing world, as Richard Sklar (1987: 698) writes, "federalist futures are democratic images."
The crux of decision making within a democracy is the threshold of consensus required for some members of the society to take decisions that apply to all members of society (Rae 1969). Thus, practices that establish decision-making rules for executive and legislative decisions by elites, and decision-making rules for the selection of representatives (the electoral system) by the electorate, are critical. Options range from minimum winning decision-rules (for example, a plurality) to complete consensus.
There is a long-standing debate between advocates of parliamentary government and advocates of presidential executives as to which system is more inherently stable and inclusive.15 In a typical parliamentary system, the executive is drawn from the ranks of parliament and ultimately dependent on its members for its continued governance. The leader of the largest party in the parliament serves as prime minister and the cabinet is drawn from the majority party or parties in the case of coalition governments. In grand coalitions, sometimes known as governments of national unity or a "national front," all significant political parties are represented in a ruling coalition and participate in executive decision making. Cabinet posts are doled out carefully and with an eye toward a balance of power commensurate with each party's electoral strength, creating a plural executive. Decisions are made by consensus, with each segment exercising a mutual veto.
Parliamentary structures are attractive options for divided societies because they may be structured to facilitate the inclusion of many groups, including minorities, at the highest levels of government, for instance in a broadly representative cabinet. Juan Linz (1990: 52) argues that parliamentary systems "are more conducive to stable democracy" than presidential systems and that this "applies especially to nations with deep political cleavages and numerous political parties." Parliamentarism, he contends, allows for many shades of possible political outcomes. When parliamentarism is combined with a variable term (the government can be forced to resign when majority parliamentary support is withdrawn), it is, in sum, a highly flexible arrangement. Presidential systems-with executive branch authority highly centralized in one individual, usually but not always directly elected-are by definition more exclusive than parliamentary systems, particularly if the president is unambiguously identifiable as a member of any one community orinterest.
The problem with a parliamentary-chosen executive, according to Horowitz (1990b: 73-9), is that in typical parliamentary systems whichever party or coalition of parties has a bare majority in the legislature can choose an executive without regard to the preferences of the minority; parliamentary-chosen executives can fall in the winner-take-all, government-versus-opposition pattern of politics which serves to further divide an already divided society. A minimum winning coalition usually forms the government. Instead, Horowitz believes that, combined with a strict separation of powers with the legislature, a separately elected presidency can proliferate points of power at the center, allowing some parties to win sometimes, and others to win at other times. Dispersing power in a political system through a system of checks, balances, and divided responsibility serves to lower the stakes of control for any particular institution or office (the separation of powers doctrine). If the stakes are very high for any particular office or number of seats, conflict between winners and losers will rise commensurately.
For Horowitz, given conditions in divided societies, a nationally elected president with exceptionally broad support is more likely to have conflict-reducing effects than its parliamentary counterpart.16 A presidential system can be constructed so as to ensure that, in order to be elected, candidates present themselves as conciliatory broad-based leaders.17
A broad-minded individual fairly pursuing aims of national integration can serve a symbolically important conflict-reducing role. South Africa's Nelson Mandela, with broad public support and an ethos of national reconciliation and moderation, is Horowitz's ideal president-in essence, standing above the ethnic fray. Although South Africa technically has a parliamentary system (the president is indirectly elected and the system essentially allows for the majority party in parliament to elect its candidate president), Mandela's behavior as a president is exemplary of the type of role a president can play.
Broad-based executives are more easily created if the principle of proportionality is operational. Perhaps the best-known example of proportional appointments in the executive was the Lebanese National Pact of 1943, which mandated top posts for representatives of the Christian (Maronite), Druze, Shi'a, Sunni, and other communities. Seats may also be reserved in legislatures. In Indonesia, Chinese, Europeans, and Arabs are awarded nine, six, and three seats in parliament, respectively, if as many are not duly elected.18 For the first ten years of postindependence Zimbabwe, 20 of the 100 seats in parliament were reserved for whites, even though they constituted 3 percent of the country's population. Nigeria's 1989 constitution stipulates that if the president comes from the North, the majority party leader must come from the South; Tanzania has similar arrangements. Like many of the conflict-regulating practices, proportional appointments can be handled informally.
Problems with proportional communal representation arise when the structure is inflexible to demographic changes in theunderlying population. At least one factor in the collapse in 1975 of the Lebanese National Pact of 1943 was the demographic shift as the Christian majority population became a minority. The deterioration of Lebanon into one of the world's intractable civil wars between 1976 and 1990 attests to the dangers of such rigidity.19 Whether the National Pact can be labeled a success (for mediating conflict for more than 30 years), or a failure (because it degenerated into civil war) is an open question. In many ways it was both.20
An appropriate electoral system in a multiethnic society is possibly the most important mechanism through which parties in conflict can adopt a democratic conflict-regulating practice. This is true because, as Giovanni Sartori has written, electoral systems are "the most specific manipulative element of politics" (1968: 273).21 The debate over which electoral system is best is complicated by the fact that electoral system design can be a very technical matter, and the outcomes that flow from a specific choice are highly dependent on unknowns such as the spatial distribution of votes, shifting party alignments and interparty pacts, voting behavior, ballot design, and myriad other variables.22 Finally, electoral system choices inevitably involve trade-offs among values such as legitimacy, simplicity, accountability, and proportionality.
For multiethnic societies, the central issue is whether a majoritarian system (or the less demanding plurality systems) are best or some type of proportional representation (PR) system is best. Underlying the debate is the clear understanding that electoral system choice, over time, has a strong effect on the type of party system that emerges (Lijphart 1994a); plurality and majoritarian systems tend to produce two-party systems, whereas PR usually leads to a fragmented multiparty system (Duverger 1964). It is in this regard that the policy-relevant differences between the consociational and integrative approaches to power sharing are most acute.
Advocates of consociationalism are strong believers in simple systems of PR that allow all parties, presumably but not necessarily ethnically based, representation in government commensurate with their share of the electorate. PR is not a single system but instead a broad set of electoral systems that seek to minimize differences between votes and seats in parliamentary elections (Taagepera and Shugart 1989: 24-5).23 By directly translating the number of votes won to the proportion of seats, PR systems may be more suited to providing incentives for inclusion of minority parties (Dahl 1989: 156-62).24
Opting for PR incurs the risk that the system may in fact provide incentives for social fragmentation. While PR may produce proportionality and allow for the politics of ethnic inclusion, the prospect of at least some presence in parliament (and of a parliamentary salary) can make it attractive for politicians to factionalize into narrowly based, exclusive parties, a phenomenon that Sartori has termed "polarized pluralism."25 The proliferationof parties during the continually troubled Weimar Republic led postwar constitutional designers in Germany to adopt a threshold for representation (5%), a now-common practice in PR systems. In circumstances of fragile minimum winning coalitions, such as in Israel, small parties can force onto the agenda narrow interests or highly volatile ethnic issues by threatening to withdraw from majority coalitions.
Horowitz, on the other hand, sees advantages to PR under certain specific conditions (1985: 628-653), but the purposes of electoral system design are different from those proposed by Lijphart. In addition to achieving proportionality and reducing disproportionate vote-to-seat ratios, Horowitz suggests that the effects of the electoral system should also be to fragment support for ethnic parties; induce an ethnic group, particularly a majority group, to behave moderately; preserve fluidity and multipolar balance to prevent exclusion; and, most importantly, encourage pre-election coalitions of parties across ethnic cleavages or, if possible, the creation of broad multiethnic parties (1985: 632).
Although Horowitz agrees that list PR can be structured to meet these goals, he identifies problems with simple list system PR, namely the strong role played by "party bosses" and lack of a constituency link by representatives; a lack of accountability; the potential incentives contained in the system for the proliferation of ethnic parties and ethnic outbidding; and the lack of incentives for cross-group integration. He and other integrationists (e.g., Lardeyret 1991) argue that the electoral system should root accommodation across group lines at both elite and popular levels in society and should seek to integrate groups into large multiethnic parties. On the other hand, large multiethnic parties are criticized for being a mask for ethnic dominance, in which the votes of some minorities are taken for granted given the electorate's choices (for example, black votes for the Democratic Party in the United States).
To meet the integrative aims he sets out, Horowitz favors vote pooling as a mechanism for inducing moderation; vote-pooling systems provide opportunity for voters to cast not only their first preference votes, but subsequent preference votes as well. In providing examples of vote pooling, Horowitz favorably cites (1985; 1991) the alternative vote (AV, where parties can agree to pool votes) or the single transferable vote (STV, where candidates can make vote-pooling agreements); second- and third-preference votes can be transferred among candidates or parties in a complex computation that produces overall winners.26 The major precondition for a successful vote-pooling framework is sufficient party proliferation, large heterogeneous constituencies, and conditions that make vote pooling profitable: that is, when moderation by political leaders causes them to gain more second- and third-preference votes than the first-preference votes, they lose by appearing "soft" on communal interests.27
Neither Nigeria nor Sri Lanka, the principal cases that Horowitz relies on to make the argument for the salutary effects of vote pooling, have had their electoral systems fully tested, and other examples such as Kerala (in India) are of limited comparative value given the idiosyncratic context of regional politics withinoverall national politics in India. Nevertheless, these examples do show that innovative solutions are conceivable to create an electoral system that provides incentives for intergroup moderation. Crawford Young writes that "In Malaysia, the semi-consociational management of racial divisions has probably been facilitated by the operation of the plurality system which in this setting provided incentives for cooperation within the Malay-dominated ruling alliances, whose electoral majority is magnified by the Westminster model" (1994: 12).28
The problem, of course, is the difficulty of winning support across group boundaries in situations of deep intergroup conflict. Nordlinger anticipates the problem when he writes that "even if party leaders were to mitigate their positions on the less salient issues, members of the opposing conflict group are not likely to change their party attachments on the basis of a secondary issue" (1972: 102). Horowitz notes that the only assumption to the successful operation of a vote-pooling framework "is that voters will vote for candidates other than those of their own group if advised to do so by leaders of the ethnic party they support (and usually only for second preferences). There is abundant evidence from Malaysia that they do this regularly, and it does not depend on the malleability of ethnic identity or even a softening of conflict."29 In situations where the population is highly illiterate or innumerate, however, these systems may simply be too complicated, especially when severe conflict places a premium on clearly legitimate and simply understood election results.
Electoral systems are highly flexible and can be pieced together in many ways to be appropriate for specific conditions. For example, given large heterogeneous electoral districts, it may be possible to create a simple PR list system that gives parties incentives to put up multiethnic slates-an integrative practice-to maximize support. When large multimember constituencies are ethnically diverse and no single group dominates, a party seeking to maximize its vote share would want to appeal as broadly as possible and to moderate on ethnic themes during an election campaign. In this manner, list PR can achieve integrative aims by providing incentives for coalitions of different variations within a single party, as was the case with the choice for PR in South Africa's first all-race elections (Sisk 1995a). In a subsequent analysis of the April 1994 vote in South Africa, I suggest that parties did in fact respond to the incentives embedded in the electoral system to moderate their campaign rhetoric. Moreover, the top two vote-getters (the ANC and the National Party) deliberately structured their candidate lists to appear racially inclusive (Sisk 1995b).
Thus, PR was a seemingly appropriate choice for this transitional or "founding" election, and may be a good choice for other elections that culminate a negotiated transition from deep conflict. Yet the South African system will likely be changed because large multimember district PR systems arguably lack accountability. For that reason, it is likely that South Africa will adopt a hybrid system, similar to Germany's combination of national PR plus majoritarian single-member districts, a systemwhich has contributed to stable electoral outcomes in the postwar era in that country (Kaase 1986). In the current deliberations over a more permanent electoral system for South Africa, there is a keen awareness that the system should try to maximize the values of proportionality, incentives for moderation, and representivity and accountability.
Practices that define relationships between ethnic groups and the state are an essential dimension of power-sharing practices. At one end of a spectrum are practices that do not identify or name specific groups or specify group rights or preferences in ethnic terms; on the other are those that enumerate, recognize, and confer special rights or preferences to distinct ethnic groups. Economic policies, the allocation of public funds, education and language policy, the delineation of rights and duties (especially whether group rights are entrenched and, if so, their formulations), citizenship, and procedures to administer justice all are critical components of successful ethnic conflict management.
Crawford Young, in summarizing the findings of a United Nations Research Institute for Social Development study of public policies in ethnically diverse societies, writes:
Few if any state policies will be absolutely neutral in their distributive effects among ethnic groups. What matters, then, is whether the ethnic distributive effect is widely perceived as a product of deliberate bias towards those groups with favored access to the state, and whether offsetting or compensating policies are undertaken to redress imbalance created by a given policy (Young 1992: 21).
Thus, conflict-regulating practices in divided societies have a significant "political economy" component.30 If a source of ethnic conflict is the maldistribution of resources, what sorts of practices can lead to the appropriate redistribution of wealth or income? Kenneth McRae (1974) refers to a host of options as "fiscal equalization devices"; that is, measures that promote the equitable distribution of society's resources. Such practices might include directed public policies, labor market policies, and differential access to certain resources. But, as Esman (1994: 239) notes, "excursion into economic variables demonstrates how difficult it is to explain or predict the effects of economic trends on ethnic-based conflict."
When ethnic groups have been historically disadvantaged or discriminated against, provisions have often been included-for example, in the United States, Sri Lanka, South Africa, India, or Pakistan-to acknowledge problems of discriminatory inequality. Often constitutional provisions do not spell out with specificity the public policy strategies to ameliorate the effects of past discrimination or comparative inequality, but make the important symbolic statement professing it the aim of the polity to seek redistribution as a goal. Enshrining the objective of encouraging the uplifting of those who may be disadvantaged can help chart thesubsequent course of politics toward that end.
Often the answer to such vexing problems is not to specify their solution in constitutional terms, but to set up new institutions and procedures to which all groups can subscribe that address equalization measures in collaborative, problem-solving forums. South Africa's new land restitution commission and court, whose establishment has served to defuse and constructively channel tensions, is an example worth considering for replication in other conflicts in which land disputes are a central cause of continued hostilities.
One important practice often ignored by power-sharing theorists, or at least downplayed, is the role of human rights (particularly the careful balancing of individual and group rights) and the role of judicial institutions in ethnic conflict mitigation. This is understandable because courts are weak in most developing states. Group rights, providing for multiple official languages, own-language education, special statutory protection for named groups, access to broadcast and media, and special religious practices and customs are legal protections that may give groups comfort that their cultural identity is secure within a multiethnic framework. The self-management of community institutions and associations, or nonterritorial self-determination (sometimes known as "corporate," as opposed to territorial, federalism), may be a sufficiently reassuring practice that groups will not seek other types of privileges or special representation.
A critical issue in divided societies is the careful balancing of individual and group rights and their adjudication. The protection of individual rights-and increasingly the protection of minority group rights-has traditionally been perceived as the telos of modern constitutionalism: the principle of garantisme (Sartori 1962: 85ff) against arbitrary rule. In multiethnic societies the issue has been whether there should be fully articulated rights to address minority group concerns. Such claims often take the form of demands for linguistic freedoms, particularly the right to education in one's mother tongue, rights to establish cultural and religious institutions or associations, protection against discrimination, and in some cases a right to self-determination. Some international law scholars assert strongly the case for further codification of minority rights; Hannum (1989: 19) writes: "The essential philosophical underpinnings of human rights include the right to be and to live in community with other members of one's own group."
From the discussion above, it is clear that any menu of power-sharing practices will be quite broad. Practices need to fit together into a grand puzzle that carefully meets a divided society's particular needs. With this in mind, however, it is possible to discern ten distinct practices that embody either a consociational or an integrative approach to power sharing. As mentioned above, some practices can serve either consociational or integrative aims depending on how they are conceived and structured. Table 2 provides a summary.
Consociational Practices: Territorial Divisions of Power
1. Granting Autonomy and Creating Confederal Arrangements. Territorially concentrated ethnic groups, particularly minority groups, can be accommodated through grants of autonomy or joint decision-making via agreements reached between the rump government and the autonomous units over issues such as economic and foreign relations and regional commerce. A critical variable is the degree of economic interdependence, the structure of fiscal relations, and the balance of dependency.
2. Creating a Polycommunal Federation. Territorially concentrated ethnic groups can also be accommodated in a polycommunal federation, that is, through "ethnic federalism." Ethnic federations require more extensive interaction than confederations between central and regional govrnments, and the allocation of powers between them is invariably a difficult, and ongoing, balancing act. Management of the economy and the distribution of commonly held resources (e.g., water, mineral rights) are critically important. Other thorny issues in constructing ethnic federations are boundary delimitation; the structure of security; the containment of secessionist tendencies; relations between subunits and foreign governments and international organizations; disparities across region or state in the adjudication of law, language, and education policy; and-perhaps most important-the status of minorities and majorities within any given region.
Table 2. Conflict-Regulating Practices
| Consociational Approach | Integrative Approach | |
| Territorial divisions of power | Granting autonomy and creating confederal arrangements. | Creating a polycommunal federation. Creating a mixed or noncommunal federal structure. Establishing a single inclusive unitary state. |
| Decision-making rules | Adopting proportional representation and consensus rules in executive, legislative, and administrative decision making. Adopting a highly proportional electoral system. | Adopting majoritarian but integrated executive, legislative, and administrative decision making. Adopting a semimajoritarian or semiproportional electoral system. |
| State/ethnic relations | Acknowledging group rights or corporate federalism. | Adopting "ethnicity-blind" public policies. |
3. Adopting Proportional Representation and Consensus Rules in Executive, Legislative, and Administrative Decision Making. When all groups demand an influential role in decision making at various levels, proportional representation in the executive (through grand coalitions), legislative (through the minority veto), and administrative (through appointments) can serve conflict-mitigating aims. The principle underlying suchpractices is that of consensus decision making. Such practices work best when there are few clearly identifiable groups and where there has been a history of exclusion of disadvantaged groups. Problems may arise when consensus is impossible or difficult to achieve and any one group's leaders perceive it as advantageous to act as a "spoiler" (withdrawing from participation and violently opposing accommodating groups). The most critical question is whether positions are reserved on a predetermined group basis or whether positions are reserved on the basis of electoral outcomes.
4. Adopting a Highly Proportional Electoral System. PR electoral systems are useful for defining group boundaries where they are ambiguous or where an ethnic party competes against multiethnic parties. Moreover, these electoral systems (particularly list PR) can serve as the basis for determining the relative weight of various groups in terms of proportional representation in executive, legislative, and administrative arenas, especially when census data is inaccurate, suspicious, or absent. A critical issue is whether a simple PR system is expected to fragment the party system over time and what the implications of such fragmentation may be. A second issue is the appreciation that PR systems may not mitigate the effects of majority domination when the majority bloc is sufficiently cohesive.
5. Acknowledging Group Rights or Corporate Federalism. By giving groups autonomy over issues that concern them most, such as language rights, state-financed own-language education, protection for cultural or religious activities, and access to customary law, ethnic groups can feel sufficiently protected to participate in a common polity without fear that their identity will be subsumed in the overarching national ethos. Group rights may also amount to group preferences, particularly for historically disadvantaged groups. When an ethnic group is not territorially concentrated, nonterritorial or corporate federalism can be introduced through the structure of group rights. However, group rights (and especially preferences) can be difficult to adjudicate and can precipitate demands by other groups which may be more difficult to accommodate and which may provoke a backlash by nonpreference groups. Group rights and principles of equality are in constant tension.
6. Creating a Mixed or Noncommunal Federal Structure. When groups are not territorially concentrated, or when the aim of federalism is to promote intra-group cleavages and foster alignments across groups, a mixed or polycommunal federal approach may serve conflict-mitigating aims. Mixed federations are appropriate when one or two ethnic groups are mobilized,aggrieved, and territorially concentrated, but where other groups are more integrated; those territorially concentrated groups can be given special status or recognition while preserving a nonethnic hue for the remainder of the polity. Noncommunal federalism can be especially appropriate when groups are not territorially concentrated or where significant minority communities will reside in all of the subunits. Dangers occur in mixed federations when special status is conferred on one territory but not others, and in noncommunal federations when some groups strive for greater territorial autonomy and such autonomy is not forthcoming.
7. Establishing a Single Inclusive Unitary State. Under some situations, it may be best to avoid the territorial division of power and work at the central and local government level to develop conflict-regulating practices. In small states, states where economic integration is especially high, and where grants of autonomy will lead to violent secessionist attempts, it may be best to centralize power but equitably exercise it. Unitary states are useful when a well-integrated elite exists and when efficient decision making is a premium, for example to implement difficult structural adjustment economic policies.
8. Adopting Majoritarian but Integrated Executive, Legislative, and Administrative Decision Making. When political elites are integrated and when cleavages crosscut ethnic divisions, majoritarian decision making in the executive, legislative, and administrative fields is more efficient than reliance on consensus. Moreover, there are a host of integrative options that are more demanding than simple majority rule but less demanding than consensus. The use of special majorities on key issues in both cabinet and legislative institutions (for example, through innovative parliamentary rules) can serve this aim without reliance on grand coalitions or minority vetoes for named groups. Inclusive, legitimate, and authoritative arbiters of conflict such as broadly accepted commissions and judicial bodies are good examples of integrated administrative decision-making practices.
9. Adopting a Semimajoritarian or Semiproportional Electoral System. As a conflict-regulating practice, semimajoritarian electoral systems have traditionally been difficult to implement, but they nevertheless hold great promise. Electoral systems that are essentially majoritarian but that may, under certain conditions, have proportional effects (such as AV) or that are proportional but still have a majoritarian element (such as STV) advance conflict regulation by providing concrete incentives for candidate moderation across group lines. The principal problem with such systems is their complexity (either for voters or for understanding vote-to-seat formulae). However, sophisticated and complex electoral systems like AV or STV may be effectively used inelite decision making, such as parliamentary elections for prime ministers or presidents.
If semimajoritarian or semiproportional systems can facilitate vote pooling, their conflict-regulating effects can be realized. In some instances, when group mutual security is of less particular concern, plurality or simple majoritarian systems are expected to be the most integrative electoral system. It should also be kept in mind that majoritarian and proportional systems can be combined, such as Germany's double ballot system. Like PR, a critical issue is how the electoral system relates to the territorial division of power.
10. Adopting "Ethnicity-Blind" Public Policies. When there is no particular pattern of historical inequality or no clear economic base to group mobilization, an "ethnicity-blind" public policy approach may serve integrative aims. By protecting the rights of a group through essentially individual rights (of association and nondiscrimination), group preferences can be avoided and the principle of equality can emerge as paramount. Practices such as adopting a lingua franca and eliminating ethnic traits in the overarching national identity are examples of such conflict-regulating practices.
Policymakers considering power sharing as a potential solution to the problems of multiethnic societies deservedly ask straightforward questions about experiences with various forms of democratic practice in multiethnic societies. Under what conditions does power sharing work and under what conditions does it fail? Is elite power sharing antidemocratic? Under what conditions do power-sharing systems entrench group identities and collapse into violent conflict, and when do they lead over time to more integrative and majoritarian patterns of democracy?
Naturally, there are no simple answers to these questions, but some conclusions can be drawn. A necessary condition for the mitigation of conflict in multiethnic societies is the existence, or creation, of a centrist core of moderates-both elites and more broadly in civil society-that adheres to rules and norms of pragmatic coexistence with other groups. The commitment to pragmatic coexistence must be broad (accepted by most disputants) and deep (accepted by political elites, their organizations, and constituencies in civil society).
When a sufficiently cohesive core of moderates does exist, power sharing is a viable means of democratic governance. Power sharing describes a wide variety of conflict-regulating practices, and each power-sharing system has its own peculiar characteristics. There is no single, transportable model of power sharing, but rather a broad menu of conflict-regulating practices, institutions, and mechanisms. Whether a consociational or integrative power-sharing approach is "best" is highly conditionalon the structure and dynamics of a given society and is ultimately a matter for the society itself to decide.
However, conditional generalizations can be made about successful and unsuccessful power-sharing practices. Power sharing is successful in managing ethnic group tensions when:
1. In other types of voting, governments do not countenance plebiscites or referenda on secessionist claims because of fear of the consequences of determining the popular will by simple majority rule. The government of India has not allowed the implementation of the United Nations General Assembly resolutions of 1947 promising a plebiscite on the territorial dispensation of Kashmir because of the likelihood that a majority of Muslims would opt for accession to Pakistan, or, more recently, for independence. The trigger of the onset of the war in Bosnia was the move in February 1992 by the predominantly Muslim and Croat government to hold a referendum on independence from former Yugoslavia, the outcome of which would have been determined by simple majority rule; the referendum was boycotted by the Bosnian Serbs.
2. The absence of floating voters is a critical assumption for antimajoritarians, one that I will return to below in the section on integrative electoral systems. In defense of the assumption, Nordlinger writes:
Orthodox democratic theory, which presupposes alternating or shifting majorities, is certainly not applicable to dividedsocieties. Given the conflict's intensity, individuals belonging to a particular segment or conflict group will be adamantly and emotionally attached to "their" political party or parties. And in this kind of electoral arena party leaders rarely seek to broaden their appeals in order to win the support of intractable individuals belonging to the opposing segment (1972: 34-5).
3. Vernon Bogdanor, writing in the Basil Blackwell Encyclopedia of Political Institutions, summarizes the scholarly consensus: "Where . . . the basic cleavage is ethnic, religious, territorial, or tribal, the plurality method will emphasize concentrations of support, and it will tend to emphasize territorial cleavages at the expense of socio-economic ones. Plurality and majority methods will work less successfully in deeply divided or plural societies than in homogenous ones" (1987: 195). This consensus is by no means unanimous, however, a point that I will return to below.
4. See Lijphart (1977: 25-8, 114-18) and Horowitz (1985: 629-30) for critiques of majoritarianism in deeply divided societies. Horowitz (1985: 333-64) provides a lengthy analysis of the acute problems of ethnic party systems.
5. The coalescent/adversarial distinction is Lijphart's (1977a: 25).
6. The literature on consociational democracy is well developed. While the groundbreaking work is Lijphart's, particularly his book Democracy in Plural Societies (1977a), many other scholars have contributed to the approach. First, consociationalism has its antecedents in the earlier work of Lijphart (1968), which termed the approach "the politics of accommodation." Landmark works in the school include Daalder (1971), McRae (1974), and Pappalardo (1981). Several scholars, led by Jürg Steiner, have sought to extend the consociational approach to a broader framework of decision making in coalitions. See Steiner's articles (1981a; 1981b). Other scholars, such as Lembruch and Schmitter, have related consociational theory to the corporatist model, arguing that these approaches are complementary; their views are best stated in Lembruch and Schmitter (1979). Lijphart (1985) catalogs and responds to critics of the consociational approach.
7. Eric Nordlinger (1972: 73) goes so far as to argue that elites "alone can initiate, work out and implement conflict-regulating practices, therefore they alone can make direct and positive contributions to conflict-regulating outcomes."
8. Horowitz does not consider the Malaysian and Lebanese experiences to have been consociational, arguing that "neither was a grand coalition, neither had a proportional electoral system, and neither had a minority veto. In neither case did the ruling coalition ever represent the segments as such, only parts of the segments. In fact, none of the (only) four cases cited by Lijphart in his 1977 book as a consociational regime in a severely divided society was one." Personal communication with the author, 25 August 1995. For more on Horowitz's views on this issue, see his book, Ethnic Groups in Conflict (1985: 575-6).
9. Grand coalitions can occur either in the cabinet or parliamentary systems, in "grand councils," or as a grand coalition of a president and senior executives in presidential systems (Lijphart 1977b: 118).
10. Personal communication with the author, 30 August 1995.
11. For instance, despite the failure of power sharing in Cyprus in the early 1960s, power sharing appears to be the only solution to that conflict as evidenced by the substance of the UN secretary-general's 1992 "Set of Ideas" for Cyprus.
12. Boundary demarcation is inevitably a critical and sensitive issue in creating or amending federal systems. On criteria for determining "just" boundaries, see Walker and Stern (1993: 4).
13. An example is the conflict-mitigating effects that South Africa's choice for regionalism has had. In the 1994 election, the primarily white National Party, supported by a majority of the so-called "Colored" community (mulattoes), won a province, as did the primarily Zulu-based Inkatha Freedom Party.
14. For ten "yardsticks" to measure federal-subfederal relations, see Duchacek (1987: 188-276).
15. Lijphart does argue (1977a: 33) that types of presidentialism and consociationalism are compatible, although the type of decision-making structure he advocates (grand coalition, as above) is clearly different from the broad-based presidential model advocated by Horowitz.
16. Federal Republic of Nigeria, The Constitution of the Federal Republic of Nigeria Promulgation Decree (1989), Article 131 (1) (a).
17. The provisions of the 1979 and 1989 constitutions of Nigeria illustrate. In Nigeria's Second Republic (1979 constitution), election to the office of the presidency required the candidate to take due notice of the "federal character" of the state; in order to win, the candidate was required to garner a plurality of votes nationwide and at least 25 percent of the votes in 13 of Nigeria's then 19 states. In the Third Republic's constitution, promulgated by the Babangida military government (following a constitutional constituent assembly) in 1989, an even more stringent requirement was built into the constitution to require a "super majority" for election to the presidency: "He has to have not less than one-third of the votes cast at the election in each of at least two-thirds of all the states in the Federation." In the event a single candidate does not meet this test in the first round of balloting, a runoff election is required. It was however, a dispute over the 1983 presidential election that contributed in part (another cause was corruption in the civil service) to the downfall of the Second Republic (1979-83).
18. For a thorough discussion of the communal proportional representation principle, see Duchacek (1987: 108ff).
19. See Samia (forthcoming) and Maila (1992) for a review of the terms of the Taif Agreement, the power-sharing arrangements agreed in the context of the experiences with earlier Lebanese efforts to manage confessional differences, and the implications for conflict resolution. Samia concludes that the agreement shouldnot be considered a final statement of inter-group relations in Lebanon, but that "amending, reviewing, reconsidering, and rewriting some [of its key terms] to recreate balance and equity is a virtue."
20. It is significant that in the 1990 Lebanese constitution the formulas for representation were changed (seats in the National Assembly are allocated on a half-Moslem, half-Christian basis with future subgroup sectarian and regional representation) and that the 1989 Taif Accords that led to the new constitution stipulate that an electoral law is eventually to be designed which wholly removes inflexible guarantees for sectarian representation. Joseph Maila notes that the Taif Accord "recognizes that confessionalism is a regulating principle of political society by claiming that power cannot be legal if it contravenes the `Pact of Co-existence' or the `desire to live together' (al aysh al-mushtarik)" (1992: 17).
21. See Lijphart (1994: 139- 52) for the effects of political engineering on the degree of proportionality, the party system, and the nature of majority victory.
22. Moreover, formal theorists have shown that no single system is arguably the "fairest" because of the "paradox of voting"-with the same set of voter preferences, different systems yield different winners (Arrow 1963).
23. See the essays in Lijphart and Grofman (1986: especially parts 1 and 2) for a discussion of the varieties of PR and PR list systems.
24. The benefits of simple PR (and especially list-system PR in which parties offer candidate slates) include a more precise vote-to-seat ratio; the lack of "wasted votes"; the ability of ethnic groups to "define themselves" and to achieve representation by their own leaders in legislative and executive institutions (Lijphart 1990: 10); an immunity to gerrymandering (given sufficiently high district magnitude); and the likelihood that over time fragmentation of the party system will result in coalition governments.
25. Examples of polarized pluralism include Italy (before recent changes in its electoral system) and Israel.
26. See Taagepera and Shugart (1989) for a thorough discussion of alternative preference voting systems. Lijphart (1991) counters that systems such as AV, particularly, are no better and sometimes worse than simple majoritarian electoral systems and that vote pooling can be achieved in many variants of list PR, primarily through the system of apparentement (in which parties can link their candidate lists).
27. The last condition is critical. If the constituencies are too divided-if there exists no sentiment for accommodation in the electorate-vote pooling cannot establish it.
28. Lijphart notes that "Malaysia has managed as a power-sharing system in spite of, not because of, the plurality system. Some form of PR would have been more straightforward." Personal communication with the author, 25 August 1995.
29. Personal communication with the author, 21 August 1995.
30. See Samarasinghe and Coughlan, Economic Dimensions of Ethnic Conflict (1991).
Bogdanor, Vernon, ed. 1987. The Blackwell Encyclopedia of Political Institutions. New York: Basil Blackwell.
Brass, Paul. 1990. The Politics of India Since Independence. Cambridge: Cambridge University Press.
Chernick, Marc. W. 1991. "Insurgency and Negotiations: Defining the Boundaries of the Political Regime in Colombia." Ph.D. dissertation, Columbia University.
Daalder, Hans. 1971. "On Building Consociational Nations: The Cases of the Netherlands and Switzerland." Legislative Studies Quarterly 3 (2): 11-25.
Daalder, Hans. 1974. "The Consociational Democracy Theme." World Politics 26 (3): 604-621.
Dahl, Robert A. 1989. Democracy and its Critics. New Haven: Yale University Press.
Diamond, Larry. 1990. "Three Paradoxes of Democracy." Journal of Democracy 1 (3) 48-60.
Duchacek, Ivo. 1987. Comparative Federalism: The Territorial Dimension of Politics. Lanham, MD: University Press of America.
Duverger, Maurice. 1964. Political Parties. Trans. Barbara and Robert North. New York: John Wiley & Sons.
Esman, Milton J. 1994. Ethnic Politics. Ithaca, NY: Cornell University Press.
Hannum, Hurst. 1990. Autonomy, Sovereignty, and Self-Determination. Philadelphia, PA: University of Pennsylvania Press.
Hannum, Hurst. 1989. "The Limits of Sovereignty and Majority Rule: Minorities, Indigenous Peoples and the Rights to Autonomy." In New Directions in Human Rights, Ellen Lutz, Hurst Hannum, and Kathryn J. Burke, eds. Philadelphia, PA: University of Pennsylvania Press.
Horowitz, Donald L. 1993. "Democracy in Divided Societies." Journal of Democracy 4 (4): 18-38.
Horowitz, Donald L. 1991. A Democratic South Africa? Constitutional Engineering in a Divided Society. Berkeley: University of California Press.
Horowitz, Donald L. 1990a. "Making Moderation Pay." In Conflict and Peacemaking in Multiethnic Societies, Joseph Montville, ed. Lexington, Massachusetts: Lexington Books.
Horowitz, Donald L. 1990b. "Comparing Democratic Systems." Journal of Democracy 1 (4): 73-79.
Horowitz, Donald. 1985. Ethnic Groups in Conflict. Berkeley and Los Angeles: University of California Press.
Kaase, Max. 1986. "Personalized Proportional Representation: The `Model' of the West German Electoral System." In Choosing an Electoral System: Issues and Alternatives, Arend Lijphart and Bernard Grofman, eds. New York: Praeger.
Lardeyret, Guy. 1991. "The Problem with PR." Journal of Democracy 2 (3): 30-35.
Lijphart, Arend. 1995. "Self Determination versus Pre-Determination of Ethnic Minorities in Power-sharing Systems." In The Rights of Minority Cultures, Will Kymlicka, ed. Oxford: Oxford University Press.
Lijphart, Arend. 1994. Electoral Systems and Party Systems: A Study of Twenty-Seven Democracies, Oxford: Oxford University Press.
Lijphart, Arend. 1994b. The Puzzle of Indian Democracy: A Reinterpretation. RGICS Paper No 18. New Delhi: Rajiv Gandhi Institute for Contemporary Studies.
Lijphart, Arend. 1992. "Constitutional Choices in Poland, Hungary and Czechoslovakia, 1989-1991." Journal of Theoretical Politics 4 (2): 207-224.
Lijphart, Arend. 1991. "The Alternative Vote: A Realistic Alternative for South Africa?" Politikon 18 (2): 91-101.
Lijphart, Arend. 1990a. "The Political Consequences of Electoral Laws, 1945-85." American Political Science Review 84: 481-96.
Lijphart, Arend. 1990b. "Electoral Systems, Party Systems and Conflict Management in Segmented Societies." In Critical Choices for South Africa: An Agenda for the 1990s, Robert Schrire, ed. Cape Town: Oxford University Press.
Lijphart, Arend. 1989. "The Ethnic Factor and Democratic Constitution-Making in South Africa." In Edmond J. Keller and Louis A. Picard, eds. South Africa in Southern Africa. Boulder, CO: Lynne Rienner Publishers.
Lijphart, Arend. 1987. "Choosing an Electoral System for Democratic Elections in South Africa: An Evaluation of the Principal Options." University of Cape Town Institute for the Study of Public Policy, Critical Choices for South African Society, occasional paper.
Lijphart, Arend. 1985. Power Sharing in South Africa. Policy Papers in International Affairs #24. Berkeley, California: Institute of International Studies, University of California, Berkeley.
Lijphart, Arend. 1977a. Democracy in Plural Societies. New Haven: Yale University Press.
Lijphart, Arend. 1977b. "Majority Rule versus Consociationalism in Deeply Divided Societies." Politikon 4 (December): 113-126.
Lijphart, Arend. 1969. "Consociational Democracy." World Politics 4 (January): 207-25.
Lijphart, Arend. 1968. The Politics of Accommodation: Pluralism and Democracy in the Netherlands. Berkeley: University of California Press.
Lijphart, Arend, and Bernard Grofman, eds. 1986. Choosing an Electoral System: Issues and Alternatives. New York: Praeger.
Linz, Juan J. 1990. "The Perils of Presidentialism." Journal of Democracy 1 (1): 72-84.
Lipset, Seymour Martin. 1960. Political Man: The Social Bases of Politics. New York: Doubleday.
Maila, Joseph. 1992. Prospects for Lebanon: The Document of National Understanding. Oxford: Centre for Lebanese Studies.
McRae, Kenneth, ed. 1974. Consociational Democracy: Political Accommodation in Segmented Societies. Toronto: McLelland and Stewart.
Nordlinger, Eric A. 1972. Conflict Regulation in Divided Societies. Cambridge, MA: Center for International Affairs, Harvard University.
Pappalardo, Adriano. 1981. "The Conditions for Consociational Democracy: A Logical and Empirical Critique." European Journal of Political Research 8 (4): 365-390.
Przeworski, Adam. 1991. Democracy and the Market. Cambridge: Cambridge University Press.
Rae, Douglas W. 1969. "Decision-Rules and Individual Values inConstitutional Choice." American Political Science Review 63: 40- 56.
Rose, Richard. 1990. "Northern Ireland: The Irreducible Conflict." In Joseph Montville, ed. Conflict and Peacemaking in Multiethnic Societies. Lexington, MA: Lexington Books.
Samia, Elie. "The Taif Agreement: Implications for Conflict Resolution in Lebanon." In Acknowledgement, Forgiveness and Reconciliation: Lessons from Lebanon, George Irani and Laurie Kings, eds. Gainesville, FL: University Presses of Florida, forthcoming.
Sartori, Giovanni. 1968. "Political Development and Political Engineering." Public Policy No. 17. (John D. Montgomery and Alfred O. Hirschmann, eds.). Cambridge, Massachusetts: Harvard University Press.
Sartori, Giovanni. 1966. "European Political Parties: The Case of Polarized Pluralism." In Political Parties and Political Development, Joseph La Palombara and Myron Weiner, eds. Princeton: Princeton University Press.
Sartori, Giovanni. 1962. "Constitutionalism: A Preliminary Discussion." American Political Science Review 56: 853-864.
Samarasinghe, S. W. R. de A., and Reed Coughlan. 1991. Economic Dimensions of Ethnic Conflict. New York: St. Martin's Press.
Share, Donald. 1986. The Making of Spanish Democracy. New York: Praeger.
Sisk, Timothy D. 1995a. Democratization in South Africa: The Elusive Social Contract. Princeton, NJ: Princeton University Press.
Sisk, Timothy D. 1995b. "Electoral System Choice in South Africa: Implications for Intergroup Moderation." Nationalism and Ethnic Politics, Vol. 1 (2): 178-204.
Sisk, Timothy D. 1993a. "The Violence-Negotiation Nexus: South Africa in Transition and the Politics of Uncertainty." Negotiation Journal 9 (1): 77-94.
Sklar, Richard. 1987. "Developmental Democracy." Comparative Studies in Society and History 29: 688-714.
Steiner, Jürg. 1981b. "Research Strategies Beyond Consociational Theory." Journal of Politics (November): 1241-1250.
Steiner, Jürg. 1974. Amicable Agreement versus Majority Rule: Conflict Resolution in Switzerland. Chapel Hill, NC: University of North Carolina Press.
Taagerpera, Rein, and Matthew Soberg Shugart. 1989. Seats and Votes: The Effects and Determinants of Electoral Systems. New Haven: Yale University Press.
Tsebelis, George. 1990a. Nested Games: Rational Choice in Comparative Politics. Berkeley, California: University of California Press.
Tsebelis, George. 1990b. "Elite Interaction and Constitution Building in Consociational Democracies." Journal of Theoretical Politics 2 (1): 5-29.
Walker, Lee, and Paul C. Stern, eds. 1993. Balancing and Sharing Power in Multiethnic Societies: Summary of a Workshop. Washington, D.C.: National Academy Press.
Welsh, David. 1993. "Domestic Politics and Ethnic Conflict." In Ethnic Conflict and International Security, Michael Brown, ed. Princeton, NJ: Princeton University Press.
Wheare, Sir Kenneth. 1964. Modern Constitutions. London: Oxford University Press.
Young, Crawford. 1995. "Ethnic Diversity and Public Policy: An Overview." United Nations Institute for Social Development, draft occasional paper.