Centre Publications

Community Conflict: Policy and Possibilities

by Donald L. Horowitz
Published by the University of Ulster, Coleraine 1990
(Out of Print)
Paperback 13 pp £1.50

Out of Print

This material is copyright of the Centre for the Study of Conflict and the author(s) and is included on the CAIN web site with the permission of the publisher. Reproduction or redistribution for commercial purposes is not permitted.

All violent conflicts have general as well as specific characteristics. The general characteristics operate, to a greater or lesser extent, in all plural societies. The specific characteristics are distinctive to each conflict and arise from its peculiar history, institutions and internal relationships. Almost inevitably, when one conflict is examined in detail, there is a tendency to narrow the focus in the search for explanations. This can lead, and in Northern Ireland has lead, to an exaggeration of the unique features of the conflict and a corresponding reluctance to stand back and consider broader approaches and issues.

The aim of this series of Occasional Papers, issued by the Centre for the Study of Conflict, is to redress this balance. It builds on the body of research carried out on the Northern Ireland conflict but is not a research series.

Our main hope is that the series will broaden the context in which the Irish conflict is considered, and will encourage comparative and innovative approaches to ethnic conflict in general.

John Darby, Director
April, 1990.

Return to publication contents

Donald L. Horowitz holds the Charles S. Murphy Chair in the School of Law, and is also Professor of Political Science, at Duke University, North Carolina. He has carried out field research in Guyana and Trinidad, Malaysia, Sri Lanka, Nigeria and Ghana. He has written extensively on aspects of ethnicity and ethnic conflict, and his Ethnic Groups in Conflict (University of California Press, 1985) is widely acknowledged as the most authoritative work in the field.

Return to publication contents

Northern Ireland is an intractable conflict by Western standards. In some key ways, it resembles the severely divided societies of Asia and Africa more than it does the less severely divided multiethnic societies of the West, such as Belgium, Canada, Switzerland, or the Netherlands.

Later on, when I speak of some techniques that seem to reduce ethnic conflict, and here I include religious and linguistic conflict, I shall refer to what seems to work in those severely-divided societies of Asia and Africa. I shall do that deliberately because, if a technique has an effect there, it might have an effect here. In the European cases, by contrast, there is an element of circularity: we cannot quite be sure whether the Western cases are conflicts that are moderate because they have effectively been controlled or whether they are effectively controlled because they are moderate conflicts to begin with. This is the problem with advocates of "consociational democracy" on the model of Arend Lijphart. So it is far more sensible to compare only techniques that have an effect in severely divided societies.

I said there were some resemblances between Northern Ireland and severely divided societies of Asia and Africa, or, to put it another way, Northern Ireland is something of a deviant case in Western terms. How?

In ethnically divided societies of the West, there is, first, an important overarching level of national identity. A survey of Switzerland found that, in spite of ethnic differences, about half of all respondents identified themselves as "Swiss." In France, only twenty-five percent of French Basques called themselves "Basques;" twenty percent called themselves French, and the remaining 55 percent responded with "Basque-French" or "French-Basque". These are not findings that could be obtained in Nigeria, or Malaysia or Sri Lanka. And, of course, the very terms Unionist and Nationalist reveal that the nature of a supra-ethnic national identity is the central issue in Northern Ireland as well.

Second, not only is there an overarching level of national identity but there are also alternative identities at the same sub-state level as ethnic identity. Belgium, for all its Fleming-Walloon differences, also has religious and class differences that compete for attention with ethnic differences. Switzerland has linguistic, class, religious and cantonal differences. Canada has class, regional, and religious conflicts in addition to Anglophone-Franco-phone conflict. These differences show up in party organisations, voting behaviour, and the structure of divisive issues. In multiethnic societies of Asia and Africa, wherever free elections prevail, parties tend to be organised closely along ethnic lines. In Western Europe and North America, they do not. In Belgium, the three main parties have long bridged the Fleming-Walloon cleavage, and one of them, with its distinctly non-ethnic perspective - the Liberal Party - actually increased its support as ethnic issues gained in prominence. Ethnic issues never wholly pre-empted others in Belgium. Rather, a "triple issue spectrum" - language, class, and religion - characterises political debate, with each issue changing the alignment of forces and to some extent neutralising the others. In Switzerland, language is no stronger a predictor of party preference than is class, though it is somewhat stronger than religion. Ethnicity in the West typically does not displace all other forms of group difference. But one can hardly say that about Northern Ireland, where religion and ethnicity, after all, are not sources of alternating affiliations but are coterminous, where regional loyalties do not significantly crosscut ethno-religious differences, and where class has been alleged to reinforce those ascriptive differences. So here, too, Northern Ireland resembles Malaysia, Sri Lanka, and Nigeria more than Belgium, Canada and Switzerland, and this is reflected in the ethno-religious basis of party politics.

Third, there is inferential evidence that intensity of ethnic conflict is lower in the West than in Asia and Africa. Surveys show that French and Italian minorities in Switzerland are actually more satisfied with government and politics than German-speaking majority. Such phenomena are incompatible with strong ascriptive loyalties. There is much less inter-ethnic violence in Western countries as well. Once again, Northern Ireland is an exception - although it is not quite on all fours with Asia and Africa here; because intense mass hostility usually produces not just terrorism, but also face-to-face ethnic riots and mass killings, as in Sri Lanka, in 1983,. And in the most severely divided societies ascriptively based parties usually drive out completely the multiethnic parties located in the centre; but in Northern Ireland in the 1970s, the Alliance party pretty generally held its own, with about ten percent of vote.

Fourth, there is the external factor. Some of the most severely divided societies are located on fault lines between two worlds: the Arab and the African, as in Mauritania, Chad, and Sudan; the Christian and the Muslim, as in Cyprus and Lebanon; the Dravidian and the so-called "Aryan," as in Sri Lanka; the Malay world and the Chinese world, as in Malaysia. In all of these cases, there are external pulls, external influences and especially external fears, such as the fear of the Sinhalese that the Sri Lankan Tamils are really the cutting edge of the invasion of the 50 million Tamils in India - and there is therefore a fear of being swamped by the people across the border. Northern Ireland, too, may lie between the Celtic World and the Anglo-Saxon world; and the fear of swamping is not far from the surface. But, more to the point, the double minority issue is as familiar in Sri Lanka or Mauritania as in Northern Ireland.

Now that I have made Northern Ireland part of the Third World, at least in terms of some characteristics, let me enumerate a few of the obstacles to accommodation in such societies, some of the things it is futile to hope for, and some of the things it is by no means idle to hope for.

First of all, in such societies, many of the claims are zero-sum. That is because relative group advantage is the thing in issue. And not just advantage in the material sense but in the symbolic sense, because symbols tell us whose country this is, who is at home, who belongs, and who is worthy. Symbolic demands, if they reflect claims to the distribution of group worth and group legitimacy in the territory, are difficult to compromise.

Moreover, in such societies, we should not project our own good intentions onto policymakers - often they, too, are participants in the conflict and have hostile feelings toward members of other groups. Even where they do not, they know that their followers do, and they know that to retain their ethnic following, they must pursue the conflict. For most politicians, most of the time, it is more rewarding to pursue the conflict than to pursue accommodation. Again, under conditions of free elections the politician who pursues a strategy of inter-ethnic accommodation will usually lose more followers of his own group than he could conceivably hope to attract from across the ethnic divide.

But buried here is a prescription. It has two sides to it - negative and positive:

First, do not aim too high. Do not try to make people love each other or at least do not put all your eggs in that basket. Do not try for national unity or national integration or nation-building. There is no short way to get there from here. It took hundreds of years to produce integrated national entities where they exist, and severely divided societies like Northern Ireland cannot wait for that. For them, the short-run is crucial.

Second, if much of the conflict revolves around politics, politicians, and their pursuit of group advantage, but politicians have no incentive to be accommodative, let us provide them with such incentives. Let us structure the situation they face so that it is rewarding for them to act in an accommodative fashion, so as to counter and overcome the political incentives to behave in an ethnically exclusive fashion.

In other words, the object of policy is not to surmount or obliterate or ignore ethnic differences, and not necessarily, in the short run at least, to reduce feelings of hostility, but to reduce the level of overt ethnic conflict behaviour.

The aim, in other words, is to make moderation pay. There is a market and there are market incentives in party politics and in leadership. Parties and leaders who do not supply the requisite hostility or promote ethnically exclusive policies demanded by the electoral market will find themselves without buyers for the product they do supply. Party A, representing Group A, can quickly be displaced by Party A1, representing the same group, as soon a Party A is perceived as being too soft or as selling Out group interests. Strategies of accommodation involve intervening in this market to change the incentive structure. The currency in this market consists of votes and support.

Now there are two main underlying mechanisms for doing this. The first involves instituting devices that have the effect of heightening incentives for interethnic co-operation on the part of politicians. The second involves devices that heighten intraethnic differences and therefore lessen interethnic conflict, in two ways: (a) by making it relatively less important or (b) by increasing incentives on the part of fractions of ethnic groups to reach out across ethnic boundaries and co-operate with fractions of other ethnic groups. These latter two tendencies are related, because a fraction of a group may have more need to reach out across ethnic lines.

Take the following situation, which is familiar in an array of ethnically divided societies, including Northern Ireland. Suppose there are two groups, A and B. A is sixty percent, B is forty percent of the population. Each has one party. Party A has a majority in sixty percent of the single-member constituencies in first-past-the-post elections. Elections are held: Party A wins sixty percent of the seats and forms the government in perpetuity. Party B and Group B are excluded from power permanently. As you can imagine, this is a situation fraught with conflict: riots, secessionist violence (if groups are territorially concentrated), and coup attempts (if group B predominates in the officer corps). What to do? Be careful, because if you split Group A into two equal parts, Party A with thirty percent and Party A1 with thirty percent, then Party B, with forty percent, may win a majority of seats on a mere plurality of votes by repeated victories in three-way contests. The result will be an illegitimate government, because a minority has, only because of plurality electoral rules, captured the state machinery.

That is not to say that fragmenting the support of the Party as of this world is always a bad idea - only that it is not always a good idea. Let me show you a case where it was a good idea: Nigeria's second republic.

Nigeria's first Republic, 1960-66, consisted of three main regions, with a dominant group and a dominant party in each, all of them fighting to control the entire state. When one of the groups - the Yoruba - split into two, one Yoruba fraction aligning with the Hausa-Fulani of the North and the other aligning with the Ibo of the East, there was bifurcation and a sixty-forty type election, followed by a coup, by riots, another coup, more riots, and a war of secession.

In 1978, when Nigeria went back to civilian rule, there was a keen desire to avoid a recurrence of this syndrome. One thing the Nigerians wanted to avoid was having whoever controlled parliament control the whole state. So they opted for a presidential system and a separation of powers. But they went further. By then there were not three main regions but nineteen states. The old North was divided now into ten states, with Hausa-Fulani dominant in only five or six. How, then, to elect the president? If the president were to be elected by a plurality, the Hausa-Fulani, the largest group, could win the presidency. The Nigerians wanted the president to be a pan-ethnic figure. So they hit on the idea of plurality plus distribution. The winning candidate had to have largest number of votes and no less than twenty-five percent of the vote in no fewer than two-thirds of the then-nineteen states. That meant that no two or three major groups could capture the presidency alone. A presidential candidate, to be successful, would have to reach out broadly across ethnic lines for votes. And Shehu Shagari, who won the election, did exactly that - he gained one-third of the popular vote and narrowly won twenty-five percent in two-thirds of the states. And he behaved as a pan-ethnic figure in office. Quite simply, he wanted to be re-elected, and he understood how the electoral incentives were structured.

Now there are several points here: The first involves the division of territory. Changing the state boundaries in a federal system prevented the Hausa-Fulani from controlling the whole North by having a majority in a single Northern parliament. This partition of the North led to a flourishing of previously dormant opposition parties in the North. And this in turn heightened the electoral incentives for presidential candidates from the North to reach out across ethnic lines for votes, because such candidates started out with a smaller northern electoral base. The effects of slicing up the territory interacted with the effects of the electoral innovations I have previously described.

Second, the president was ethnically conciliatory because he depended on the votes of members of ethnic groups besides his own. The only way to get such votes and keep them is to accommodate the wishes, the claims, the sentiments of those other ethnic groups.

Third, this mechanism is not dependent on the existence of a presidential system. We can just as easily make legislators in a parliamentary system dependent on votes of members of ethnic groups other than their own. I will show you some ways to do this a bit later.

Fourth, the pooling of popular votes of members of otherwise antagonistic ethnic groups is far more likely to produce accommodative political behaviour than is the mere pooling of seats of parties representing more than one ethnic group to form a government. Where Party A and Party B, representing groups A and B, pool just their parliamentary seats, representing fifty percent plus 1 of all seats in the house, in order to form a government, where each party is dependent only on the votes of members of its own group, this produces only a coalition of convenience. Indeed, the partners in the coalition will often be those with literally no electoral overlap - for it they were rivals for some marginal voters, party officials of each would resist the formation of the coalition. So, often such arrangements are coalitions of opposites, and they quickly turn inconvenient. The stronger partner tries to induce aisle crossings, secure a majority, and push the other party out of the government. In any case, such coalitions (and there are many instances - Punjab before 1966, Benin, Uganda, Nigeria from 1960 to 1964) tend to dissolve acrimoniously over divisive ethnic policy issues.

This is much less likely where parties of different ethnic groups have pooled seats and popular votes - where, for example, in constituencies each cannot win it has told its voters to vote for a party of the other ethnic group. Of course, in a divided society it cannot do that unless it can assure its voters that, compared to the alternatives, this party is moderate and accommodating on ethnic issues of concern to its voters. Therefore, an arrangement based on vote pooling is a coalition of the moderate middle and not of the extremes, as coalitions of convenience tend to be.

Fifth, notice, too, that this vote pooling will not be necessary unless there are more than two parties; for if it is A:B-60:40, the occasion for accommodation will not arise. So there needs to be some intraethnic division, or else a plurality of parties that has not reduced itself to bifurcation, with several groups - A, B, C and D - for vote pooling to work.

Sixth, and before I point out how to induce vote pooling by techniques other than Nigerian distribution, let me make a crucial point for Northern Ireland. When people who study Northern Ireland talk of electoral innovation in the service of accommodation, they often say that Northern Ireland has had an appropriate electoral reform, because it had proportional representation in the 1970s. They are therefore discouraged about the prospects for intergroup accommodation through the electoral process. Their discouragement, however, is unwarranted, for proportional representation does not necessarily produce vote pooling or coalitions committed to accommodation.

It is true that PR creates incentives to proliferate parties, because any party - or, in some systems, any party over a minimum vote threshold - can secure representation in parliament. All else equal, the larger the number of parties, the less the chance that any one party can form a government alone. Consequently, the majority governments that are fostered by the first-past-the-post electoral system used in Great Britain and the United States are less likely under proportional representation. If, under PR, no party has a majority, there will be a need to form a coalition.

Nevertheless, the coalition will generally result from a mere pooling of seats, and not votes. Under the more extreme list systems of proportional representation, there is no occasion even to contemplate vote pooling. Each party will receive the same fraction of seats as it received of votes. Parties will seek to maximise the support of their own groups and will not jeopardise these by making appeals across group lines. The same is not necessarily true under the different form of proportional representation - the single transferable vote - used in the Irish Republic. STV is a preferential system. The voter may cast a ballot for more than one candidate for the same seat, in order of the voter's preference. Vote pooling across ethnic lines is therefore possible. The single transferable vote was utilised in the Northern Ireland elections of 1973, the elections that determined the future turned out to be bleak. One crucial reason that the power-sharing government did not succeed is that STV induced no substantial vote pooling.

The incentives to vote pooling are much too weak under STV to induce vote pooling across ethnic lines. The reason is that, under STV, it is easy to win a seat. If there are four seats in a constituency, about one-fourth of the vote will suffice to secure election for a candidate. Under these rules, candidates (and parties) will simply appeal to members of their own group for votes, rather than undertake the riskier course of making reciprocal agreements to pool votes across group lines. Exactly this is what happened in Northern Ireland in 1973. There are other systems of preferential voting - particularly, those with a majority threshold for election - that provide much stronger incentives to intergroup vote pooling, because the second and third preferences of voters will be crucial to victory. Without those strong incentives, which are not present in most PR systems, coalitions tend to be coalitions of convenience, put together merely to gain a majority of seats, rather than coalitions committed to compromise. To put the point very sharply, Northern Ireland has certainly not had an accommodative electoral reform.

The most reliable way to make interethnic moderation pay is to make politicians (not all, but some) reciprocally dependent on the marginal votes of members of groups other than their own. And the electoral system can do this despite the wishes of the politicians, by making vote pooling profitable.

How? There are several ways, so I shall illustrate one way rather quickly and two at greater length, because I can then also demonstrate the impact of such techniques by a paired comparison.

The quick one is Lebanon - a severely divided society which had more than thirty years of peace because of intricate accommodative arrangements. And, I should add, that peace did not break down because of any fault in what I am about to describe. It broke down despite these arrangements, because powerful forces overtook them.

Here are the electoral arrangements. All elected positions were assigned by sectarian group. The President was required to be a Maronite, the Prime Minister a Sunni, the Speaker of the House a Shiite. Likewise, virtually all parliamentary seats were reserved. So it was Maronite contesting versus Maronite, Sunni versus Sunni, etc. Immediately, there was intraethnic competition among politicians and a concomitant reduction of interethnic competition. But every voter voted for all the candidates running in the constituency. The typical parliamentary constituency was multi-member, and each of the seats was reserved to a different group, e.g., one Druze, one Sunni, one Shiite, one Maronite. This led to the creation of competing interethnic tickets, because a Sunni candidate, relying only on the votes of his own group, could easily be defeated by a smart Sunni rival who added Druze, Shiite and Maronite votes to his Sunni votes. So there was intraethnic competition and interethnic co-operation and moderation, because a Maronite candidate could not persuade his supporters to vote for a Sunni with whom he was aligned unless he could assure them that this Sunni was one who understood Maronite concerns. So this was a system with powerful tendencies toward conciliation. I cite it, not because it is apt for Northern Ireland, but only to show how a clever system can be designed for this purpose in a severely divided society.

Now to the longer comparison, which shows that divided societies that do not design such systems can end up in a worse condition than even more severely divided societies that do. The comparison is Sri Lanka versus Malaysia.

What I shall show is that Sri Lanka had the easier problem but political institutions that exacerbated it; Malaysia had the harder problem but institutions that ameliorated it - and the difference is cast in terms of incentives for politicians to behave moderately.

At independence, anyone forecasting the ethnic future of the two countries would have predicted far more difficulty for Malaysia than for Sri Lanka. Relative group proportions, conceptions of group legitimacy, recent political events, the relations of elites of the various groups, and the political culture of the two countries all suggested a Sri Lankan advantage.

1. Numbers. The Sri Lankan Tamils were a mere eleven percent of the population at independence. The Malaysian Chinese were about one-third and the Indians about ten percent of the Malaysian population, so the non-Malay total was about half the Malaysian population.

2. Indigenousness. The Sri Lankan Tamils arrived, on the average, 1,000 years ago. The Chinese and Indians, on the other hand, were relatively recent migrants to Malaysia. The Sri Lankan Tamils were citizens. The Chinese and Indians, by and large, were not. The Tamils were legitimate participants in the political system and were early participants in the national movement. The Chinese were not accepted as legitimate participants in Malaysian politics. Segments of the Malay press were advocating the return of the Chinese to China. The Malays were regarded as "sons of soil," having priority in the country. The Sinhalese did not use that term or make such distinctions.

3. Events Around Independence. Recent events were unconducive to peaceful ethnic relations in Malaya. During World War II, Chinese guerrillas, fighting the Japanese, had fought Malay villagers who resisted their exactions of food. After the war, the guerrillas emerged from the jungle, proclaimed the abolition of Malay sultanates, and purported to annex Malaya to China. Until the British completed the reoccupation of Malaya, there were bloodbaths up and down the peninsula. Thereafter, Chinese guerrillas returned to the jungle to fight the British and the largely-Malay armed forces in the emergency of 1948-60. These battles had the character of ethnic hostilities, undermining the Chinese position in the country. In Ceylon, the Tamils were well represented in the Ceylon Defence Force during the war and in the Ceylon Army after independence. Tamil leaders had asked for ethnically balanced representation in parliament, but the British had rejected that. Independence nevertheless found the Tamils with ministerial portfolios.

4. Elites. Malay and Chinese elites were divided by the structure of educational institutions in colonial and post-colonial Malaya. Sinhalese and Tamil elites were brought together by the educational system in Ceylon. The Malay top elite were segregated in the Malay College at Kuala Kangsar, which was modelled on the British public school. No comparable monoethnic institution existed in Ceylon. The result was that Malay and Chinese leaders were not, at first, on intimate terms, whereas Sinhalese and Tamil leaders frequently knew each other well. There was, in Ceylon, a genuinely intercommunal elite, sharing many common values. The same description would certainly not hold for Malayan elite at independence.

5. Interethnic Diplomacy and Political Culture. Malay politicians were quite discriminating and cautious about whom they would deal with. Some Malay newspapers were urging "no diplomacy with the Chinese." The Ceylonese, by contrast, had a bargaining political culture. No agreement was automatically foreclosed. Tamil parties dealt with several Sinhalese parties, and vice versa. "What are your terms?" was a phrase frequently heard. Party discussions often revolved around whether a better deal could not be obtained from some other party than the one with which negotiations were being conducted. For interethnic negotiations, it is reasonable to assume that such a bargaining political culture is more advantageous than one that put a premium on personal relations, was hesitant to deal at arm's length, and had a set of unwritten rules governing interethnic negotiations.

So, on all of these grounds, Sri Lanka started out with considerable advantages. But where are they now? Despite those favourable conditions, Sri Lanka is in the midst of an ugly ethnic war - or rather two wars. Despite Malaysia's unfavourable conditions, Malaysia is at peace. The last serious episode of ethnic violence was in May 1969. This contrast is not fortuitous. Malaysia has had the more difficult problem, but it has also had better conflict management.

The most important contrast between Malaysian and Sri Lankan ethnic politics has been the role of interethnic political coalitions, based on vote pooling, in the two countries. The dominant parties in Sri Lanka have all been ethnically based, whereas the dominant force in Malaysia has been a permanent interethnic coalition of ethnically based parties -namely, the Alliance and its successor, the National Front.

By the mid-1950s, practically all Sri Lankan Tamils had abandoned Sinhalese parties like the UNP and moved either to the Tamil Congress or the Federal Party, leaving the UNP, SLFP, and the various Left parties to the Sinhalese. Ever since, Sri Lanka's party system has revolved around the competition of the two main Sinhalese parties for Sinhalese votes and the two main Tamil parties for Tamil votes until the two Tamil parties merged in 1972. The dynamics of intraethnic competition, especially for the Sinhalese vote, has pushed the parties toward meeting ethnic demands and limited their leeway to make concessions across ethnic lines.

The rise of the SLFP as the main competitor to the UNP in the 1950s went hand in hand with appeals to Sinhalese ethnic sentiment. After the landslide victory of the SLEP-led coalition in 1956, a "Sinhalese Only" language act was passed, and Tamil civil servants were discriminated against on linguistic grounds. Rebuffed at the polls, the UNP responded by becoming as anti-Tamil as the SLFP. When S.W.R.D. Bandaranaike attempted to cool Sinhalese-Tamil tension down by a compromise agreement with the Federal Party in 1957, the UNP campaigned against it, and the compromise was abandoned. From 1960 to 1965, when Mrs Bandaranaike was in power, there was a further acceleration of favouritism toward Sinhalese Buddhists. The UNP-led coalition of 1965 to 1968 made some concessions to redress Tamil grievances, but these came to an abrupt halt when the SLFP opposed them and UNP backbenchers feared losing their seats to SLFP candidates if they went along. Interethnic compromise was strictly limited by intraethnic electoral competition between the two main Sinhalese parties. This is a phenomenon not entirely unknown in Northern Ireland.

Mrs Bandaranaike's second regime, from 1970 to 1977, was characterised by a virulent anti-Tamil strain. In 1970 a new constitution was promulgated. By its terms Buddhism was accorded a "foremost place" in the country. The constitution utterly ignored the Tamil presence in the country. Around the same time, a scheme to "standardise marks" was implemented. Its effect was to reduce the marks received by Tamil students on university entrance exams, thereby depriving large numbers of Tamil students of the university education for which they were plainly more qualified than many of the Sinhalese students who were admitted in their place. An entire half generation of recruits for Tamil separatist organisations was thereby created, and the seeds were planted for guerrilla warfare.

Underlying this process of bidding and outbidding for the Sinhalese vote was an electoral system that translated small swings in popular votes into large swings in seats. The system was first-past-the-post in mainly single-member, largely homogeneous constituencies. With multiparty competition in the Sinhalese South, it was often possible to win a parliamentary majority on a plurality of thirty to forty percent of the popular vote. In every parliamentary election from 1952 to 1970 - six times, in fact - there was alternation in office between the SLFP and the UNP. In the South, the vast majority of constituencies was overwhelmingly Sinhalese in composition. As a result, parties derived rich rewards from appealing to Sinhalese ethnic sentiment and opposing proposals to conciliate the Tamils.

The combination of (1) largely homogeneous constituencies, (2) plurality elections in mainly single-member constituencies, and (3) a competitive party configuration for power and two plausible contenders for nearly every seat created a system that was exceedingly sensitive to Sinhalese opinion and inhospitable to interethnic accommodation.

Several of these conditions were later altered. In 1978, the UNP government promulgated a new constitution. A separately-elected presidency was instituted. The president is now elected by a system of preferential voting that accords weight to voters' second choices in a way that they had not been weighted in plurality elections for parliament. To be elected president, a candidate needs a majority. Under preferential voting, if no candidate receives a majority of first preference ballots, second preferences of voters whose first choices are not among the top two candidates are then counted. Under this system, Tamil second preferences for president will be reallocated as if they were first preferences whenever the Tamil candidate is not among the top two, as, of course, he never will be. So Tamil second preferences could provide the margin of victory, and prudent presidential candidates could hardly ignore Tamil interests under such conditions. Since presidential candidates would make deals with Tamil parties to pool votes, and Tamil parties would only make such deals with Sinhalese candidates who were moderate on Tamil issues, candidates will sort themselves out with respect to their position on Tamil issues.

By the time this system was put into effect, however, separatist violence had begun in earnest, and Sinhalese and Tamil opinion had so polarised that, in the short term at least, no electoral system could foster moderation. In the two presidential elections held thus far, civil war conditions, including a Tamil boycott of the 1989 elections, have prevented the conciliatory features of this ingenious system from having an effect.

Very few conditions were different in Malaysia, and yet the results have been dramatically different. Malaysia also had first-past-the-post in single-member constituencies, and also party competition revolving around attention to mutually exclusive ethnic claims. Unlike Sri Lanka, however, interethnic compromise has also had a claim on party attention; and moderation, as well as extremism, has brought political rewards.

Three differences changed the balance of incentives.

1. The first was fortuitous. Malaysians began working on interethnic accommodation early. They had had a bitter taste of violence during and after World War II and did not wait until accumulated grievances had reached the threshold of widespread violence. But electoral incentives also played a big part, as I shall suggest in a moment.

2. The second condition is the existence of many more ethnically heterogeneous parliamentary constituencies in Malaysia than in Sri Lanka. In Malaysia, the electorate in the 1964 elections was thirty-eight percent Chinese and eight percent Indian. Forty percent of the parliamentary constituencies had Chinese pluralities and in toto non-Malay majorities. Another twenty percent had an electorate at least thirty percent Chinese. In Sri Lanka, only eleven percent of parliamentary constituencies had a Sri Lankan Tamil plurality, and in all but one of these Tamils comprised the vast majority of the constituency. In eighty-one percent of all constituencies, Tamils were less that ten percent of the electorate. In Malaysia, by contrast, Chinese comprised less than ten percent of the electorate in only eighteen percent of all constituencies - a graphic illustration of how much more heterogeneous Malaysian constituencies were. These contrasts occurred not just because the Chinese were more numerous than the Ceylon Tamils but because the Tamils and the Sinhalese are much more regionally concentrated. So in overwhelmingly Sinhalese constituencies, there simply was no electoral constraint on taking anti-Tamil positions. The one or two percent Tamil voters could hardly offer anything in return for the moderation of a Sinhalese candidate. Overall there were many more Sinhalese votes to be had by being extreme than there were Tamil votes to be had by being moderate. And likewise in the country as a whole: no Tamil party could help a Sinhalese party to win a hotly contested Sinhalese-majority constituency by contributing marginal Tamil votes. And vice versa for Sinhalese help for Tamil candidates. So interethnic coalitions based on vote pooling were very unlikely, since vote pooling could not really occur, given the nature of the territorial constituencies.

Malaysia's heterogeneous constituencies made ethnic calculations more complex. In many constituencies, Chinese voters could punish Malay extremists and reward moderates. There were not always more Malay votes to be gained than Chinese votes to be lost by taking extreme positions. By the same token, Chinese and Malay parties could pool votes profitably at the constituency level and come out ahead. Where there were more Malays than Chinese in a constituency, a Chinese party could urge its supporters to vote for a friendly Malay candidate, and vice versa where there were more Chinese than Malays in a constituency. Parties might still evolve along wholly ethnic lines, but - especially if there were more than one party per ethnic group - there would be countervailing incentives fostering interethnic coalition.

It is important to note that the structure of constituencies is not given: it is not a function of mere demography and geography. If ethnic groups are geographically concentrated and heterogeneous constituencies are regarded as desirable, it is possible to do what the Nigerians did and what the Sri Lankans later did - namely, make the whole country one large heterogeneous constituency for the purpose of electing a president. And, to turn the conventional wisdom around, perhaps the greatest utility of a presidential system in a divided society has nothing to do with the separation of powers. Rather, it is that it permits the institution of electoral devices that can guarantee vote pooling across ethnic lines.

The third difference between Malaysia and Sri Lanka follows from the first two. An interethnic coalition was formed before independence, and it occupied the centre of the ethnic spectrum. It was formed, not out of "statesmanship," not out of goodwill or tolerant attitudes - these were not always present - but out of selfish electoral calculations. The two parties that came together were in danger of losing a crucial election before independence that would undermine all their subsequent claims in negotiations for independence - the town council elections, where non-Malay votes were especially important. Once the coalition was formed for these elections and made permanent, other parties, all of them ethnically-based, took extreme positions on the flanks, locking the coalition in the centre. With Malay and Chinese votes split, the centre coalition could not compete with the extreme parties in taking extreme positions, and there were often more votes to be had across ethnic lines by being conciliatory on ethnic issues than there were within one's own group by being extreme. Had the interethnic coalition not been formed, the configuration of Malaysian politics would resemble Sri Lankan politics, and Malaysian ethnic relations would be in the same parlous state - or, very likely, even worse.

I cannot go any more deeply into all this, but let me be bold and point five morals for Northern Ireland:

1. In a severely divided society, it is futile to devote too much attention to trying to make people love and respect each other. It is more fruitful to try to encourage decent behaviour by politicians, whatever they may feel in their hearts.

2. The most enduring basis for interethnic conciliation in a severely-divided society, where communal allegiances are politicised, lies in an electoral system that creates some incentives rewarding moderation, even though other incentives continue to reward extremism.

3. An electoral reform without regard to all the functions of elections in severely divided societies - apart from the function of merely providing representation in accordance with numbers - will not help, may make things worse, and may convince people, incorrectly, that nothing works. Everything has not been tried in Northern Ireland.

4. A plan for "power sharing" from the top down, without the appropriate under-lying electoral incentives for power sharing will fail, and it will reinforce the false notion that a substantial reform has been implemented without success. I want to suggest that this is exactly the sort of "power sharing" Northern Ireland has attempted.

5. The challenge these problems pose to human ingenuity is a bit like the old academic joke about the three disciplines:
   Economics is about how people make choices.
   Sociology is about how there are really no choices to be made.
   Politics is about how people make them anyway.

For conflict in severely divided societies, there are no feasible ways out, but we need to find them anyway.

Return to publication contents