Doing Well out of War Paul Collier The World Bank April 10, 1999 Paper prepared for Conference on Economic Agendas in Civil Wars, London, April 26- 27, 1999. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author. They do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. Doing Well out of War Paul Collier Introduction The discourse on conflict tends to be dominated by group grievances beneath which inter-group hatreds lurk, often traced back through history. I have investigated statistically the global pattern of large-scale civil conflict since 1965, expecting to find a close relationship between measures of these hatreds and grievances and the incidence of conflict. Instead, I found that economic agendas appear to be central to understanding why civil wars get going. Conflicts are far more likely to be caused by economic opportunities than by grievance. If economic agendas are driving conflict, then it is likely that some groups are benefiting from conflict and that these groups therefore have some interest in initiating and sustaining it. Civil wars create economic opportunities for a minority of actors even as they destroy them for the majority. I consider which groups benefit, and what the international community can do to reduce their power. Economic agendas as causes of conflict A useful conceptual distinction in understanding the motivation for civil war is that between greed and grievance. At one extreme rebellions might arise because the rebels aspire to wealth through capturing resources extra-legally. At the other extreme they might arise because rebels aspire to rid the nation, or the group of people with which they identify, of an unjust regime. These two motivations obviously imply radically different types of policy intervention if the international community wishes to promote the prospects of peace. The most obvious way of discovering what motivates people is to ask them. However, here we immediately encounter a problem. Those rebel organisations which are sufficiently successful to get noticed are unlikely to be so naïve as to admit to greed as a motive. Successful rebel organisations place considerable emphasis upon good public relations with the international community. Narratives of grievance play much better with this community than narratives of greed. A narrative of grievance is not only much more functional externally, it is also more satisfying personally: rebel leaders may readily be persuaded by their own propaganda. Further, an accentuated sense of grievance may be functional internally for the rebel organization. The organization has to recruit: indeed, its success depends upon it. As the organization gets larger, the material benefits which it can offer its additional members is likely to diminish. By playing upon a sense of grievance, the organization may therefore be able to get additional recruits more cheaply. Hence, even where the rationale at the top of the organization is essentially greed, the actual discourse may be entirely dominated by grievance. I should emphasize that I do not mean to be cynical. I am not arguing that rebels necessary deceive either others or themselves in explaining their motivation in terms of grievance. Rather, I am simply arguing that since both greed-motivated and grievance-motivated rebel organisations will embed their behaviour in a narrative of grievance, the observation of 1 that narrative provides no informational content to the researcher as to the true motivation for rebellion. To discover the truth we need a different research approach. The approach I take, which is the conventional one in social science, is to infer motivation from patterns of observed behaviour. If someone says `I don't like chocolates', but keeps on eating them, we infer that she really likes them, and the question of why she says the opposite is then usually relegated to being of secondary importance. I try to determine patterns in the origins of civil war, distinguishing between those causal factors which are broadly consistent with an economic motivation, and those which are more consistent with grievance. I then try to predict whether each country has a civil war during each five-year period from 1960 to 1995 in terms of the values of the causal factors at the beginning of each period. For example, I try to predict whether Kenya had a civil war during the period 1970-74 in terms of its characteristics as at 1970. This approach only becomes reasonably robust if the coverage is large and comprehensive. I therefore follow current research practice in opting for global coverage, only dropping countries where there is too little data. I first describe the proxies which I use to capture the notion of an economic agenda. The most important one is the importance of exports of primary commodities. I measure this as the share of primary commodity exports in GDP. Primary commodity exports are likely to be a good proxy for the availability of `lootable' resources. We know that they are by far the most heavily taxed component of GDP in developing countries, and the reason for this is that they are the most easily taxed component. Primary commodity production does not depend upon complex and delicate networks of information and transactions as with manufacturing. It can also be highly profitable because it is based on the exploitation of idiosyncratic natural endowments rather than the more competitive level playing fields of manufacturing. Thus, production can survive predatory taxation. Yet for export it is dependent upon long trade routes, usually originating from rural locations. This makes it easy for an organised military force to impose predatory taxation by targeting these trade routes. These factors apply equally to rebel organisations as to governments. Rebels, too, can impose predatory taxation on primary commodities as long as they can either interrupt some point in the trade route or menace an isolated, and difficult to protect, point of production. For rebels, primary commodities have one further advantage over other sources of taxation which does not apply to governments. Sometimes, taxation can be much higher if it is levied in kind: the rebels directly extract a proportion of the production, rather than cash. This is particularly likely to apply where production is conducted by poor households rather than by large firms, and the households are themselves cash-scarce because they can only command a small fraction of the international value of their production. If rebels receive taxes in kind, they will need to be able to dispose of the output. Because rebel organisations are extra-legal, the disposal of output on international markets potentially poses problems. The more identifiable is the original source of the output, the deeper will be the discount below the international price. Primary 2 commodities have the considerable advantage to rebel organisations that they are generic rather than branded products, and so their origin is much more difficult to determine. The discount from reliance upon extra-legal marketing channels can therefore be much smaller. While primary commodities are thus a good proxy for the lootable resources which greed-motivated rebels would seek to capture, there are other factors likely to matter for an economic agenda. The most important other factor is likely to be the cost of attracting recruits to the rebellion. Overwhelmingly, the people who join rebellions are young men. Hence, other things equal, we might expect that the proportion of young men in a society, say those between ages 15-24, would be a factor influencing the feasibility of rebellion: the greater the proportion of young men, the easier would it be to recruit rebels. Relatedly, the willingness of young men to join a rebellion might be influenced by their other income-earning opportunities. If young men face only the option of poverty they might be more inclined to join a rebellion than if they have better opportunities. I proxy these income-earning opportunities by the amount of education in the society: the average number of years of education which the population has received. In developing countries this education will have been disproportionately supplied to young men, so that differences in the average educational endowment between societies will reflect much larger differences in the educational endowments of young males. It might seem to some non-economists that considerations of alternative income-earning opportunities do not enter into the decision process of potential recruits to rebellions. I will therefore give an example of where such considerations were hugely important. The largest civil war of the twentieth century was the Russian civil war of 1919-20. Both the Red and the White armies were essentially scratch, rebel armies, since the Tsarist army had collapsed. For both these rebel armies recruitment and desertion were huge problems. Between them they lost four million men to desertion. Thus, the desertion rate is large enough to be a social rather than just an idiosyncratic phenomenon. The desertion rate was ten times higher in summer than in winter (Figes, 1996). The reason for this was obviously that both armies were composed of peasants, and during the summer peasants had much higher income-earning opportunities, notably the harvest, than in the winter. To summarise, my measures of economic agendas will be primary commodities, the proportion of young men in the society, and the endowment of education. There are of course many other potential economic agendas in conflict, such as suppliers of armaments and opportunities for bureaucratic corruption. However, most of these are difficult to measure in a comparative way and so preclude the sort of analysis I undertake here. I now contrast these economic factors with those which proxy grievance. Rebel narratives of grievance are focused around one or other of four factors. One, which probably attracts the most horrified fascination from western media, is the expression of raw ethnic or religious hatred. While such narratives may contain a sub-text of specific economic or social grievances, sometimes these refer to very remote time periods, or may appear to be merely illustrations or even pretexts for a deeper hatred. For example, this might seem to be the most obvious interpretation of the Serb attack on the population of Kosovo. I measure the tendency to such raw grievances by the extent to which the society 3 is fractionalised by ethnicity and by religion. Specifically, I use indices constructed from historical work by anthropologists, which show the probability that any two randomly drawn people from the society are from different ethnic, and different religious groups. I also multiply the two indices, which gives a measure of potential cross-cutting cleavages: societies which are highly fractionalised by both ethnicity and religion will thus get the highest scores on this combined index. Of course, ethnic and religious identities are not given, fixed phenomena, but social constructions. However, they are rather slow to change. I measure them as of 1965 and attempt to explain conflict over the ensuing thirty years and over such a period they have probably changed little. A second important narrative of grievance is focused on economic inequality. The grievance might refer either to unequal incomes or to unequal ownership of assets. For example, some of the conflicts in Central America are commonly attributed to one or other of these types of inequality. Both of these are now objectively measurable for most societies, although my measure of asset inequality is confined to the ownership of land. However, in low-income countries, land is the major single asset, and so inequalities in its ownership should be a good proxy for overall asset inequality. A third narrative of grievance is focused on a lack of political rights. If the government is autocratic and repressive then people will have a natural and justifiable desire to overthrow it in the pursuit of democracy. For example, the uprising in Romania in 1989 is usually seen as a demand for democracy. Political scientists have now carefully classified political regimes according to the degree of political rights and I use the one on which most political scientists now base their analyses (the `Polity III' data set). A final narrative of grievance focuses on government economic incompetence. If a government is seen to inflict sufficient economic misery on its population then it may face an uprising. The successful National Resistance Movement rebellion in Uganda in the early 1980s is often seen as being motivated by despair at gross economic mismanagement by successive regimes. I proxy such economic performance by the rate of growth of per capita income in the preceding five years. Other things equal, an economy which had experienced rapid decline might be more prone to rebellion than one which had experienced rapid growth and so offered hope. I will now describe the results. The purpose of this paper is to present results to people who are not necessarily familiar with (or interested in) modern social science research methods. I will simply note that the method used is a `probit' model which predicts the occurrence of civil war in terms of these underlying factors. The results from this analysis tell the researcher both how important each factor appears to be, and how much confidence we can place in that appearance. The results are reported formally in Collier and Hoeffler (1999). The results overwhelmingly point to the importance of economic agendas as opposed to grievance. Indeed, the grievance factors are so unimportant or perverse that there must be a reason for it, and I go on to explain why, I think, grievance-based explanations of civil 4 war are so seriously wrong. First, however, I describe the evidence on the importance of economic agendas. The presence of primary commodity exports massively increases the risks of civil conflict. Specifically, other things equal, a country which is heavily dependent upon primary commodity exports, with a quarter of its national income coming from them, has a risk of conflict four times greater than one without primary commodity exports. The result is also highly significant statistically, meaning that there is only a very small chance that it is a statistical fluke. The presence of a high proportion of young men in a society also increases the risk of conflict, whereas the greater the educational endowment the lower is the risk. Education is relatively more important than the proportion of young men. For example, if we double the proportion of young men its effect can be offset by increasing the average educational endowment by around two months. Each year of education reduces the risk of conflict by around 20%. Thus, some societies are much more prone to conflict than others, simply because they offer more inviting economic prospects for rebellion. The risk factors multiply up. A country with large natural resources, many young men and little education is very much more at risk of conflict than one with opposite characteristics. Before drawing out the policy implications, I will turn to the results on grievance. The only result which supports the grievance approach to conflict is that a prior period of rapid economic decline increases the risk of conflict. Each 5% of annual growth rate has about the same effect as a year of education for the population in reducing the risk of conflict. Thus, a society in which the economy is growing by 5% is around 40% safer than one which is declining by 5%, other things equal. Presumably, growth gives hope, while rapid decline may galvanize people into action. Inequality, whether measured in terms of income or land ownership, has no effect on the risk of conflict according to the data. This is, of course, surprising given the attention which inequality has received as an explanation of conflict. The results cannot, however, be lightly dismissed. For example, the measures of inequality have proved to be significant in explaining economic growth and so are evidently not so noisy as to lack explanatory power. Nor is our result dependent upon a particular specification. Hoeffler and I have experimented with well over a hundred variants of our core specification and in none of these is inequality a significant cause of conflict. (By contrast, primary commodity exports are always significant). Political repression has ambiguous effects on the risk of conflict. A society which is fully democratic is safer than one which is only partially democratic. However, severe political repression yields a lower risk of conflict than partial democracy. These effects are of moderate size and only weakly significant: a fully democratic country has a risk of conflict about 60% lower than the most dangerously partially democratic societies. In related work, Hegre et al (1999) investigate the effects of political transition. They find that the transition from one type of political regime, such as repression, to another, such as partial democracy, itself temporarily increases the risk of conflict. However, they find 5 that the increased risk fades quite rapidly. One year after the change, three quarters of the risks generated by political transition have evaporated. The most surprising result for those who emphasise grievance as the cause of conflict concerns ethnic and religious fractionalization. We find that such fractionalization is significant in changing the risk of conflict. The effect is most pronounced and significant when we measure social fractionalization as the combination of ethnic and religious divisions, that is the potential cross-cutting fractionalization created by multiplying the two underlying indices. Thus measured, ethnic and religious fractionalization significantly reduces the risk of conflict. Fractionalized societies are safer than homogenous societies. For example, a highly fractionalized society such as Uganda would be about 40% safer than a homogenous society, controlling for other characteristics. The grievance theory of conflict thus finds surprisingly little empirical support. Inequality does not seem to matter, while political repression and ethnic and religious divisions have precisely the opposite of their predicted effects. Why might this be the case? I think that the reason that the grievance theory is so at variance with the actual pattern of conflict is that it misses the importance of what social scientists call the `collective action problem'. Justice, revenge, and relief from grievance are `public goods' and so are subject to the problem of free-riding. If I am consumed with grievance against the government, I may well prefer to rebel than to continue to suffer its continuation. However, whether the government gets overthrown does not depend upon whether I personally join the rebellion. Individually, my preferred choice might be that others fight the rebellion, while I benefit from the justice which their rebellion achieves. This standard free-rider problem will often be enough to prevent the possibility of grievance- motivated rebellions. However, it is compounded by two other problems. In order for a rebellion to achieve justice it probably needs to achieve military victory. For this it needs to be large. Small rebellions face all the costs and risks of punishment without much prospect of achieving justice. Hence, grievance-motivated potential rebels will be much more willing to join large rebellions than small ones. Obviously, however, rebellions have to start small before they can become large. It is quite possible that many people would be willing to join a large rebellion but that nevertheless it does not occur, because only few people are willing to join a small rebellion and so it does not scale up. Social scientists think of this as a coordination problem. The final problem is that rebels have to fight before they achieve justice. The rebel leader may promise to assuage grievances, but once he has won he may have an incentive to behave much like the current government. More generally, the rebel leader has a much stronger incentive to promise things than he has subsequently to deliver them. Because potential recruits can recognize this problem they may not be able to trust the rebel leader and so may decide not to join the rebellion even though it promises relief from grievances. Social scientists term such a phenomenon a time-consistency problem. Between them, the free-rider, coordination and time consistency problems pose formidable obstacles to rebellions motivated purely by grievance. How might a rebel 6 leader overcome them? All societies face collective action problems of a great many varieties. Many are not overcome, or are overcome by the function being taken over by government, supported by taxation and enforcement powers. However, where they are overcome less formally, in a way which could be pertinent for a rebellion, the usual way is through what we now term `social capital', that is, the trust generated by participation in informal or formal groupings of people into networks, clubs and societies. Through such interactions people learn to set each decision in the context of past and future decisions about other matters: I better not free-ride now because other people didn't free ride last time, and if I do, they might free ride next time. Thus, a rebel leader might seek to overcome the collective action problem by drawing upon existing social capital. This, I think, is why ethnic and religious fractionalization reduces rather than increases the risk of rebellion. Social capital usually does not span ethnic and religious divides. Thus, in highly fractionalized societies it is much harder to mobilize large numbers of people than in homogenous societies. It may only be possible to mobilize the people within a particular ethnic-cum-religious group, but if this is only a small part of the national population, the prospects of victory are poor and so the prospect of assuaging grievance are poor. Grievance-motivated rebellions by small minorities are liable to be quixotic. The pattern of rebellion is sufficiently strongly related to the proxies for greed, and sufficiently negatively related to ethnic and religious fractionalization, to suggest that most rebellion is not quixotic. The remaining strategy for a rebel leader is to rely upon greed. Greed-motivated rebellion does not face any of the collective action problems of grievance-motivated rebellion. There is no free-rider problem because the benefits of the rebellion can be confined to those who participate in it. There is no coordination problem because the rebellion does not need to be so large as to be victorious nationally in order to gain spasmodic control of some territory and so be predatory on the export trade in primary commodities. There is no time-consistency problem because if rebellions are able to cream off some of the rent from primary commodity exports during the rebellion then rebel recruits can be paid during the conflict rather than be dependent upon promises. Hence, we might expect that those grievance-motivated rebellions which actually take hold do so by combining some material pay-off with the grievance. We see this in many rebellions. For example, in Colombia groups which began as grievance-based organizations (of the political extreme left and extreme right) have evolved into drug baronies. To conclude this section, rebellions based purely on grievance face such severe collective action problems that the basic theories of social science would predict that they are unlikely to occur, and the empirical evidence supports this prediction. Societies indeed differ markedly in the underlying objective causes of grievance. We can reasonably expect that a society which is fractionalized into many ethnic and religious groups, with high income and asset inequality, and which has a government which represses political rights, will have many more grievances than a homogenous, equal democracy. Yet this does not translate into a higher risk of conflict. I suggest that what it does produce is a high-pitched discourse or narrative of grievance. There is a disconnect between these narratives and action. Even in apparently highly- charged ethno-religious conflicts such as the former Yugoslavia, there were apparently 7 cases of one side renting tanks from the other side! Such behavior could not occur if the objective of conflict was simply to harm the opposing ethnic or religious group, but it can be explicable if there are economic advantages to the control of territory. To understand action we have to shift our focus from the discourse to the economic agenda. For the reasons I discussed above, this economic agenda will be concealed. The true cause of much civil war is not the loud discourse of grievance, but the silent force of greed. Who Gains during Conflict? Civil wars inflict very high costs on an economy. I estimate that on average during civil wars the economy as a whole declines by around 2.2% per annum relative to its underlying growth path (Collier, 1999). This may seem a small number, but it implies that after a decade of war a society will have an income 20% lower than it would otherwise have been. Despite these overall losses, civil wars create some opportunities for profit which are not available during peace. These fall into four groups. First, during civil war life tends to become less predictable. As a result, people shorten their time horizons, or equivalently, discount the future more heavily. This changes the calculus of opportunistic behavior. In normal circumstances people tend not to be opportunistic in business relationships because such behavior damages their reputations and so makes it more difficult for them to reach agreement on deals in the future. The less predictable is the future, or the more peculiar are current circumstances, the less worthwhile is it to sacrifice current opportunities for profit in order to maintain reputation. Hence, civil war societies tend to become opportunistic. This will affect business practices, so that some firms will thrive through sharp practices, while others become their victims. Profit rates will therefore become more dispersed and increase for the opportunistic. Secondly, there is likely to be an increase in criminality. Governments reduce expenditure on the police during conflict as they increase spending on the military. As a result, the risks of punishment for criminal behavior decline. The main economic activity of criminals is theft, and this reduces asset-holding through two routes. An increase in theft makes assets less attractive. Hence, households will tend to run their assets down or shift them out of the country. For example, a common phenomenon is civil wars is for the livestock herd to decline quite drastically. Further, the criminals themselves face an even more acute asset-holding problem than their potential victims. If a criminal accumulates assets through theft, he lacks good title, and so his possession is insecure. A likely response is to shift the assets out of the country, either directly, as when stolen cattle are moved over the border, or indirectly, as when their value is first converted into some other asset (Collier and Gunning, 1995). Thirdly, during civil war markets become disrupted. In normal circumstances the main force keeping marketing margins down, and indeed profits more generally, is competition. If there is good information and easy entry into trading, marketing margins 8 will be driven down to the point at which traders earn no higher incomes than they would in any other activity. Civil wars make information much more costly and particular. Further, they make entry into the activity much more difficult. Existing traders may be able to resort to illegal means to discourage entry, and as opportunism becomes more rife, viable trading will contract to those relationships which can still be trusted. Thus, during civil war competition tends to break down. Trade becomes increasingly monopolistic and so marketing margins increase. Of course, during conflict the actual volume of transactions will decrease, but if margins are initially narrow and widen sufficiently, then the profits from trade can actually increase. Fourthly, the scope for rent-seeking predation on trade increases for rebels and may even increase for government officials, as their actions become less open to scrutiny. Indeed, in some instances the very distinction between rebels and government can become blurred: government soldiers by day become rebels by night. The rebels are not rebelling against the government at all, they are simply taking off their uniforms in order to reduce detection and thereby increase the opportunities which their official weapons provide for predation. In the limit, if such rent-seeking becomes too competitive then it can kill trade off. Imagine that primary commodities must be transported from their point of production to the coast. If at many points along the road each locally powerful rebel, off-duty army officer, or official exacts a charge in an uncoordinated way, then the combined extractions can be so high that they make exporting unprofitable: the competitive predation simply kills the activity. Thus, sufficiently decentralized greed-motivated rebellions tend to kill off the economic goose and so die out. If there is no trade there is no loot. To prevent this, a rebel movement will try to create a monopoly of predation and for this it must generate a monopoly of rebel violence. This may be why a very common characteristic of rebel movements is that they go through a phase in which considerable military effort is expended on fighting other rebel groups. To be economically successful, a rebel group does not need to defeat the government, but it does need to replace the government monopoly of violence with a rebel-government duopoly of violence. Fully competitive rebellions will not normally be profitable except in the short run. This suggests that even when a country collapses into anarchy, such a state of affairs will seldom persist. There will be strong economic forces creating sufficiently large units of power that the primary commodity export trade will not be killed off. Rebellions in which no group can impose its authority may thus fade out. Evidence for this is that the duration of rebellion, as opposed to the risk of its occurrence, is prolonged if the society consists of two ethnic groups (Collier, Hoeffler and Sodersbom., 1999). Both ethnically homogenous and ethnically highly fractionalized societies have shorter conflicts. When there are two ethnic groups, probably one being the government and the other the rebels, the rebel organization has the best chance of imposing a cohesive monopoly on rebellion. The implications of the above are that various identifiable groups will `do well out of the war'. They are opportunistic businessmen, criminals, traders and the rebel organizations themselves. The rebels will do well through predation on primary commodity exports, traders will do well through widened margins on the goods they sell to consumers, 9 criminals will do well through theft, and opportunistic businessmen will do well at the expense of those businesses which are constrained to honest conduct. If some people do well out of civil war they may not be particularly concerned to restore peace. While they have increased incomes, all other groups will suffer sharply declining incomes and so have a strong interest in peace. Overall, the losers lose more than the winners gain, so that potentially there is scope for a mutually beneficial peace settlement. However, there are reasons to expect that it will be very difficult to achieve peace through a settlement in which all these groups are confident of being better off. There are two major problems. The first is that even if a settlement can be found in which all groups are better off, it is unlikely that the settlement can be trusted. Settlements face the `time-consistency' problem discussed above in the different context of whether potential rebels can trust their leader. The application to a peace settlement is as follows. Usually, a settlement will involve some military disbandment of rebel forces. As a result, the balance of military advantage is likely to shift to the government. As a result, the government will have an incentive to promise ex ante, things which it will not have an incentive to adhere to ex post. Since shrewd rebels can see this problem, they may rationally decide to decline a peace settlement which would ostensibly benefit them. The second major problem is that it is not realistically possible to construct a settlement in which all of the four groups who benefit from civil war are bought off. For example, although the offer of modest financial incentives to the leadership of Renamo proved feasible, and may have been critical in ending the conflict in Mozambique, it is morally and politically much harder to offer drug barons the large financial incentives which would be needed to switch their interest from the perpetuation of conflict to the conclusion of a peace settlement. Hence, although the costs of war appear to offer the potential for mutually beneficial peace settlements, in practice peace will depend upon those groups which gain from peace being more influential than those which gain from continued war. The relative power of economic interest groups is the classic question posed by modern political economy. The literature tells us that small, cohesive groups will be disproportionately influential. Unfortunately, because most people lose from war, the pro-peace group faces a massive free-rider problem in lobbying for peace. By contrast, because the beneficiaries of war are a much smaller group, some of whom gain very large amounts, the free-rider problem of the pro-war lobby is very much less severe. An implication is that peace may sometimes prove illusive because the small groups which have an economic interest in sustaining or reviving conflict are disproportionately influential. Because the private interests of these groups are very much against the public interest, their true agendas will be actively concealed. Thus, the true motivations for the perpetuation of conflict are normally unobserved, not simply because they get crowded out by the discourse of grievance, but because they will be kept secret. If such interest groups cannot be bought off, then they have to be overcome. Interventions which reduce their profits from conflict can work both to reduce their incentives for conflict and, perhaps more importantly, to reduce their capacity to influence decisions. 10 One test of these ideas is whether conflict becomes more likely as a result of previous conflict. If grievance is the main driver of conflict, then for sure a powerful impetus to grievance will be previous conflict. Conflicts leave a legacy of atrocities crying out for revenge. By contrast, the greed-based approach to conflict would argue that it is the underlying economic conditions which create the risk of conflict. Some societies will have repeated conflicts, not because of the cumulative legacy of the desire for vengeance, but because war is profitable for some groups. Although the evidence is only preliminary, at present it supports the latter interpretation. Once we allow for the risk factors described above, countries which have had a conflict are not more likely to have a further conflict than countries which have been conflict-free. To the extent that this is correct, it is good news for the international community. It implies that conflicts are not deeply intractable in the sense that they are driven by historical loyalties. The loyalties of local communities may indeed be so determined, and the observed discourse may reflect these loyalties, but there is a wide gulf between this and actual large-scale conflict. If only the international community can change the economic incentives for conflict, it can substantially reduce their incidence, even in societies riven by long-standing hatreds. Reducing the incentives for conflict How can international policy reduce the economic incentives for conflict? For this we must work through the list of causes of conflict and determine where there is scope for intervention. Recall that the most powerful single driver of the risk of conflict is for an economy to have a high proportion of primary commodity exports. This gives the international community some opportunity for risk reduction. Most of the international markets for primary commodities are highly centralized, with a small number of key intermediaries. The most extreme case of this is probably the diamonds trade. One reason for centralization is that there are almost always questions of product quality: primary commodities are not completely standardized. To the extent that it is possible to curtail the sales of primary commodities which are financing conflict, the prospects for peace are increased. For example, diamond exports from Sierra Leone probably account for the high incidence of conflict in that country. Many of these exports originate in highly informal marketing channels but find their way onto world markets. Of course, some markets, notably those for narcotics, are illegal throughout their entire chain, making them uncompetitive and thus providing very high profits to traders. In most markets, however, the task may be to prevent illegitimate supplies from gaining access to legitimate channels. This will drive down the incomes of the illegitimate acquisition of the commodities and thereby reduce the incentive to contest the control of primary exports. A further way in which the international community can reduce the risks generated by primary commodity exports is to assist in the diversification of the economies of those societies which are most at risk. The instrument for this is development assistance. Obviously, substantial development assistance is usually only feasible during peacetime conditions, so that its role is preventative, whereas the control of marketing channels may 11 also be able to influence the incentives for settling current conflicts. Whether aid programs can succeed in diversifying an economy depends both on the underlying comparative advantage of the country and on its absorptive capacity for aid. A well- located country without major natural resources such as Mozambique has a better chance of export diversification than a landlocked country with large natural resources. If an economy has a high absorptive capacity for aid then development assistance can reduce the risk of conflict not just through increasing diversification but through reducing poverty and increasing the growth rate. Recall that both poverty and economic decline increases risks. The absorptive capacity of an economy for aid depends primarily upon the economic policies which governments adopt. Recent work using the World Bank's scoring system for twenty different aspects of policy finds that those developing countries with average policy scores have an absorptive capacity for aid about double that in countries with fairly poor scores (Collier and Dollar, 1999). Hence, if governments adopt policy environments which are highly discouraging for economic activity, there is rather little that donors can do to offset these effects through large aid flows. However, where governments adopt policies which are more conducive to growth then donors can do a great deal to accelerate the process of development and thereby reduce the risks of conflict. The limitations of aid, in that it cannot offset the effects of highly damaging policies, should not blind us to the considerable contribution which aid can make to enhancing peace in most environments. I have suggested that during conflict marketing margins tend to widen, creating some lucrative monopolistic trading opportunities and giving these traders an incentive for conflict continuation. To the extent possible, policy should therefore be focused on making markets as competitive as possible. Competition will reduce profits to normal levels and reduce the attraction of conflict for wartime traders. Agencies of the international community, broadly defined, are themselves often major purchasers during conflicts. If their own purchasing practices are insufficiently cost-conscious, they will become a source for super-normal profits. The international community (though obviously not the World Bank) may also increase the incentives for peace through political actions. First, the time-consistency problem surrounding peace settlements implies that there is a role for external guarantors of the settlement terms. The incentives for settlement maintenance may range from the military, through the diplomatic, to the financial. Secondly, the above analysis implies that full democratic rights are an effective means of reducing the risk of conflict, and that, although political transition temporarily increases risks, these risks do not persist for long. Indeed, a slow transition from repression may be dangerous because it implies a long period of partial democracy, during which the risks are at their peak. There may therefore be a role for assisting countries during a brief phase of rapid transition to democracy. Finally, were the world to be composed of small, ethnically and religiously homogenous states, the statistical evidence suggests that it would have a much higher incidence of civil war. I have already discussed the result that ethnic and religious fractionalization 12 actually make states safer rather than more dangerous, so that ethnic cleansing is not only repellent in itself but would result in more dangerous political entities. A result I have not yet described is that large states are proportionately much safer than small states. The risk of civil conflict occurring somewhere on the territory of one large state is approximately one third lower than if the same territory is divided into two identical states. Thus, the political forces for self-determination of small, ethnically or religiously homogenous groups may not be benign. Economic policies in post-conflict conditions Finally, I consider briefly economic policy priorities in post-conflict societies. Such societies need to reduce the underlying risks of conflict. This will involve the same policies which are appropriate in conflict prevention, such as diversification and poverty reduction. However, there are also some economic legacies from conflict: particular interest groups which develop during the conflict and which have little interest in peace. These interests need to be weakened as rapidly as possible. A civil war society tends to favor the opportunistic and the criminal, and to permit the encroachment of monopoly. These characteristics persist after the conflict has ended. Yet the groups who benefit from these characteristics have an interest in perpetuating wartime conditions. One approach is therefore to weaken these groups as rapidly as possible by reducing their profits. Market integration can be promoted by deregulation, improved transport, and improved market information, for example, by means of better communications. In post-conflict Uganda, when the government deregulated the transport of coffee, the road haulage industry expanded and this new entry into the sector broke the road haulage cartel which had informally operated during the conflict. As a result, road haulage rates approximately halved and so rural produce markets in turn became more competitive. In the process, the politics of conflict probably changed. A larger, more competitive transport and trading sector which has made investments which depend upon the continuation of peace, is a strongly pro-peace lobby. The former interest of a small cartel enjoying monopoly profits has disappeared. Opportunism thrives on conflict. For example, in Uganda a trader who purchased mattresses on credit from the local manufacturer to sell in the North claimed that his purchases had been stolen by rebels. The manufacturer suspected that this claim is false, but could not prove the contrary and so had to accept the default of the trader. The opportunistic trader thus has an interest in unrest. A firm can guard against such opportunism by improving its information. If the manufacturer had a better network of contacts in the North, or a better network of information from other suppliers to the trader, it would be more difficult for the trader to be opportunistic. But information networks are costly. Particularly where the telephone system is poor, and where newspaper circulation is low, information is expensive and so limited. The Ugandan government has indirectly reduced opportunism post-conflict by encouraging cell phones, radio and a free press. 13 Crime thrives on low detection and poor justice systems. The rehabilitation of the police and the courts are thus post-conflict priorities, partly to ease problems of contract enforcement. There is also a need to professionalise the army. As discussed, sometimes during conflict the government army will itself be an important source of crime and predation, so that it will have little interest in peace. For this reason, demobilization may not be as problematic as is commonly feared. Ill-paid government soldiers may be less of a threat once disarmed, disbanded and dispersed to their farms than when they are together in barracks. The Ugandan demobilization actually reduced crime rates despite the widespread fear that it would do the opposite (Collier, 1994). Conclusion Discussion of civil conflict is dominated by the narrative of grievance. Hence, policy towards conflict tends to be focused upon on the one hand assuaging perceived grievances, and on the other, attempting to reconcile populations which have deep-rooted hatreds. The evidence on the causes of conflict does not really support this interpretation. The objective factors which might contribute to grievance, such as income and asset inequality, ethnic and religious divisions, and political repression do not seem to increase the risks of conflict. Indeed, to the extent that they have any effect it is to make societies safer. I do not wish to imply that the parties to a conflict do not hold grievances and historical hatreds, and it is indeed sensible to attempt to reduce them. However, the evidence on the causes of conflict points to economic factors as the main drivers of conflict. The combination of large exports of primary commodities, low education, a high proportion of young men and economic decline between them drastically increase risks. Greed seems more important than grievance. Although societies as a whole suffer economically from civil war, some small identifiable groups do well out of it. They thus have an interest in the initiation, perpetuation and renewal of conflict. Naturally, these interests tend to remain low profile. Hence, the discourse of grievance is much louder than that of greed, even if it is less significant. Policy intervention should, however, focus rather more than in the past on these economic agendas. Effective policy should reduce both the economic incentives for rebellion, and the economic power of the groups which tend to gain from the continuation of social disorder. The restriction of access to international commodity markets for illegitimate exports from countries in conflict, and the targeting of development assistance to high-risk countries not currently in conflict, are both feasible strategies for the international community. 14 References Collier, P. (1994) `Demobilisation and Insecurity', Journal of International Development. Collier, P. (1999) `On the Economic Consequences of Civil War', Oxford Economic Papers 51, 168-83. Collier, P. and D. Dollar (1999) `Aid Allocation and Poverty Reduction', mimeo, DECRG, World Bank. Collier, P. and J.W. Gunning (1995) `War, Peace and Private Portfolios', World Development. Collier, P. and A. Hoeffler (1999) `Justice-Seeking and Loot-Seeking in Civil War', mimeo, DECRG, World Bank. Collier, P. A. Hoeffler and M. Sodersbom (1999) `On the Duration of Civil War', mimeo, DECRG, World Bank. Figes, O. (1996) A People's Tragedy, Pilmico. Hegre, H., T. Ellingsen, N.P. Gleditsch, and S. Gates (1999), `Towards a Democratic Civil Peace?' mimeo, International Peace Research Institute, Oslo. 15