[Cramer, Christopher is an . "Homo economicus goes to war: methodological individualism, rational choice and the political economy of war." World development 30.11 (2002): 1845-1864. https://www.sciencedirect.com/science/article/abs/pii/S0305750X02001201]
In The Periodic Table, Primo Levi (1986) finds in one of his tales the following moral: that one must distrust the almost-the-same... the practically identical, the approximate, the or-even, all surrogates, and all patchwork. The differences can be small, but they can lead to radically different consequences, like a railroad’s switch points; the chemist’s trade consists in good part in being aware of these differences, knowing them close up, and foreseeing their effects. And not only the chemist’s trade. If there is anything substantial in the rational choice economist’s models of violent conflict, it will emerge through empirical testing. But, where this testing has been done it has been, inevitably, through the use of proxies. What matters, then, is whether the difference between a proxy and the reality to which it is meant to be tagged makes a difference: whether their signals might divert the train of analytical and policy-making attention onto a siding. Do the proxies actually measure what they purport to measure? There are two dimensions to answering this question: are the data adequate and are the proxies conceptually or substantially equivalent to the thing they are meant to represent? 16 The most widely known use of proxy measurements for rational choice models of violent conflict or civil wars currently is in the work of Collier and Hoeffler. An early version (Collier & Hoeffler, 1996) posited that the utility of choosing rebellion (Uw) was a function of the probability of victory (p) and the gains to rebels upon victory (T), the potential for government defence spending (D), the expected duration of the conflict (M) and the co-ordination costs of mobilizing for rebellion (C), such that: 17 Uw ¼ fpðDÞ:T ; M; Cg The proxies chosen for the incentive for rebellion aimed to capture potential government military expenditure and revenue-raising capacity through three variables: income per capita, the natural resource base, and the degree of inequality in a society. Hence, poorer countries are more prone to conflict because, with income per capita at a low level, there is insufficient tax-raising capacity to deter rebellion. The use of inequality, measured in Gini coefficients, as a proxy for potential government defence spending is especially interesting. The reasoning adopted in Collier and Hoeffler (1996) is that high inequality reflects the presence of an elite liable to encourage the government to raise taxation of elite wealth if this is temporarily necessary to repel rebellion and thereby preserve this unequal status quo. The gains to rebellion conditional upon victory are assessed via the proxy of the natural resource endowment, which has uncertain consequences given that resource endowment also affects the government’s potential defence spending. The proxy for the costs of rebel co-ordination is the ‘‘ethno-linguistic fragmentation index’’ (with complete homogeneity scoring zero and maximum fragmentation scoring 100). The proposal here was that the cost of co-ordination is... a quadratic function of the degree of cultural fractionalization, initially decreasing in it as potential rebels are distinguished from governments supporters, and then increasing in it as potential rebels are distinguished from each other (p. 5). Collier and Hoeffler (1998) amend this model, retaining its core hypothesis and some of the proxies but combining ethno-linguistic with religious fractionalization and dropping inequality completely from the model. The purpose of econometric analysis using proxies to quantify indirectly the proposed variables in the models is, of course, to achieve predictive (probabilistic) power and Collier and Hoeffler (1998) generates implications such as: ‘‘For example, a highly fractionalized society such as Uganda would be about 40% safer than a homogeneous society, controlling for other characteristics.’’ Collier and Hoeffler (1999) and Collier (2000) engage more directly in the greed versus grievance question, separating the two motivations and selecting proxies for each. Greed or loot seeking is proxied by the share of primary commodities in total exports––because this represents the ease of instant taxation by looting––and by the proportion of young males in the total population combined with average years of schooling (the latter supposed to capture the availability of economic opportunities other than violence and looting in an economy). Grievance or justice-seeking as a cause of civil war is captured by the ‘‘almost-the-same’’ or ‘‘practically identical’’ proxies of economic growth in the five years prior to a conflict, scores of repression and restricted political rights, inequality (reappearing not as a proxy for potential government revenue but for grievance), and ethno-linguistic fragmentation. A little confusingly, some of these proxies, e.g., dependence on primary commodities, are switched in Collier and Hoeffler (2001) to stand not for motive (i.e. greed) but for opportunity. 18 The use of these proxies does seem to be patchwork. The argument that high degrees of inequality generate a greater capacity for taxation of capital is derived from Alesina and Rodrik (1994). Their argument, however, had nothing to do with the self-preservation instincts of the wealthy but rather proposed that where there is high inequality there will be democratic pressure, exercised by the median voter, for higher capital taxation, which in turn would harm growth prospects as this taxation would reduce incentives to invest. The inequality/taxation proxy for potential government military expenditure and hence for the incentive/disincentive for rebellion is also flawed in two other respects. First, it makes something of a naive assumption about the willingness of elites to be taxed: it is not clear that even in democratic Latin American states, for example, high inequality has made a substantial dent, through effective tax increases, in the Gini coefficient (Szekely & Hilgert, 1999). Tax evasion remains extremely widespread in many such countries. In countries threatened by conflict capital flight also has to be expected. 19 Second, the financing of war is rather more complex than is conveyed by this notion of readily flexible tax rates. Most wars have been financed to significant degrees, historically, by borrowing and by inflationary money printing (Galbraith, 1975). Certainly tax increases have been critical over the long-term but where this has been so, for example, during the consolidation of most European states, this has tended to involve hard-won innovations in fiscal institutions rather than quick fix appeals to the coffers of the rich (Tilly, 1992; Ferguson, 2001). The commitment by the wealthy, or by capitalists, to supporting the funding of war has been more complex, varied and politically contested than can be captured by a proxy like this. Confidence in these proxies must be undermined, also, by the way in which one proxy, inequality, is used across models: first it is taken as almost-the-same as the self-preservation instinct of elites and hence of the fiscal prospects for war finance; then it is dropped as not the same as anything; then it is reintroduced as almost-the-same as the source of injustice and hence potential grievance-fuelled conflict. Ethno-linguistic fragmentation is used fairly constantly as a proxy for co-ordination costs and so for the likelihood of collective action. For this to be effective as a proxy, to be ‘‘practically identical’’ to the actual possibilities of and constraints on mobilization for collective action, two conditions must be met. The first is that ethnicity is the only significant category of collective identity and affiliation, and that it operates in the same way across all societies such that more, or less, of it (fragmentation) will have much the same consequences irrespective of geographical and historical context. The second condition is that there are no other effective sources of mobilization. Evidence suggests that neither condition holds. Ethnicity has played a very secondary role at most in many conflicts and the intensity of ethnic animosity does not seem to vary with differences in the index of fragmentation. Its significance has been highly particular and very different across, for example, Somalia, Rwanda, Sierra Leone, Liberia and Angola. Likewise, ethnicity may have played a role in Central American conflicts but it has hardly been a dominant cause in many, and clearly its role has varied between El Salvador, Mexico and Nicaragua. In Yugoslavia ethnicity only came to matter politically in a distinctly ‘‘modern’’ way, that is, as an outcome of economic, administrative, and political policies and experiences from the 1950s onward (Schierup, 1992, 1993); and then in a cumulative, modular or mimetic way drawing on real and imagined pasts in a way very similar to the spread of nationalism as an ‘‘imagined community’’ (Anderson, 1991). What matters more, then, is surely how ethnicity or race or whatever form collective identity takes does or does not come to play a significant role in the origin and feasibility of conflict (Turton, 1997). 20 Meanwhile, regarding the second condition, it is increasingly clear that in apparently obviously ideological conflicts such as the Zimbabwean liberation war and in apparently clearly ‘‘ethnic’’ conflicts such as Rwanda, there is a range of sources of mobilization, and motivation, enabling violent collective action. Primary among these is force itself. Coercion, whether subtle or brutal, is a common constant of most contemporary conflicts. This in itself suggests that, however neat the ‘‘fit’’ with some quadratic function of the role of ethnic fractionalization or indeed with the availability of instantly ‘‘taxable’’ primary commodities, such incentives and organizing principles are typically insufficient to make war work, either for rebels or incumbent governments. Meanwhile, there is evidence of a range of agendas of social conflict that affect people’s readiness to fight or logistically to support civil conflict, aside from straightforward profitability at the margin. These include gender and generational conflict and highly personal or interpersonal relations of rivalry and envy. A further difficulty with the proxies used in these models is that it is far from obvious that the proxy is being assigned to the appropriate object. Preponderance in exports of primary commodities might indicate the availability of lootable goods and so make violence more profitable at the margin than the dreary grind of underemployment and poverty. But in doing this it might just as well be a proxy for failed policy, missing economic dynamism, a probable shortage of consumer goods and imports and widespread grievance or dissatisfaction with this predicament. There is the further problem that the category ‘‘primary commodities’’ is absurdly broad. Open cast or alluvial diamonds might be easy to loot (for soldiers) but not deep-mined diamonds, or aluminum, oil, etc. Other primary commodities like cotton or cocoa do not generate profit levels associated with conflict commodities. Others still, such as coffee, have often been associated with conflict but not in a direct way through lootability. One approach that has added a much needed detail to the analysis of commodities in conflict––and that finds still that commodities are subsidiary to political dynamics––is the political geography of Le Billon (2001). Similarly, a low average for years of schooling might represent the lack of decent opportunities for (nonconflictual) gainful employment and therefore reveal a low opportunity cost of conflict: however, it might just as well reflect or be directly a source of social anger. 21 Evidence for this is especially sharp for Sierra Leone, where poor educational attainment has, reportedly, fuelled grievances among young age cohorts who have appealed for improved resource allocations to schools, in a country with a particularly strong educational heritage in sub-Saharan Africa (Richards, 1996). Since loot has been a central feature of the Sierra Leone conflict, within the framework of the orthodox economic models of conflict this presents a paradox of apparently coexisting grievance and greed. 22 One way to resolve the paradox is to appeal to the ‘‘false consciousness’’ argument, as Collier (2000) does by arguing that the ‘‘narrative of grievance’’ has ‘‘no informational content.’’ 23 Another is to reject the categorical distinction between greed and grievance as false. In this case it becomes clear that greed and grievance do not just coexist in the origin of conflict, but that they might be internally related to one another. 24 Put differently, where does ‘‘greed,’’ or loot-seeking, or aggressive goods and profit acquisition behavior come from? Is it not feasible that greed emerges from grievance? 25 If Kosovo Albanians want better material conditions (and even if some of them emerge as wartime entrepreneurs, profiteers, etc.) is this a manifestation of the aggregated greed of individual Kosovo Albanians or of a historical and relational grievance given particular historical and political opportunities? 26 How big is the difference and might the two combine in the mobilization and decision making of individuals? Another example of the ambiguous mingling of greed and grievance as a motivation for violence is given in the widespread political violence during the 1990s in Congo-Brazzaville, chiefly by a range of militias with different political affiliations. Looting in Brazzaville in 1997 was known as ‘‘slaughtering the pig’’ or, referring to the Elf Aquitaine-controlled oil field of N’kossa, ‘‘N’kossa, everyone gets his share,’’ both capturing a criticism of the failure by the political elite to redistribute the wealth of the country. As one militiaman put it, when accused of theft: You call that stealing? When they incite us to kill, they call it human folly, then afterwards they drink cham pagne together. That is called national unity. While this is going on, we get nothing. Have you ever seen any of the national leaders losing one of their own children in this war? Why is it always us? This has gone beyond a joke. We are going to loot all their houses and tomorrow, when we rebuild the country, perhaps we will finally have a job to go to (quoted in Bazenguissa-Ganga, 1999, p. 48). 27 The claims of political leaders or the statements of individual rebels or soldiers are bound to be partial and subjective, but surely it would be absurd to regard them as having no informational content? As Keen (2001) argues, a more useful focus for research than the separation of greed and grievance is the way in which they interact, most notably how the greedy manipulate the grievances of others. Further, to complicate matters, it is far from obvious that a given conflict, e.g., a ‘‘civil war,’’ is a single phenomenon whose quantifiable aggregates clearly ‘‘reveal’’ individual preferences. For example, participants in a conflict may take part through varying degrees of coercion or voluntary choice and may represent a range of agendas of differing passions and interests that overlap in the overall conflict. Kriger (1992) shows how many different forms of conflict were fought under the mantel of the liberation war in Zimbabwe. Among these conflicts, many women were more or less coercively mobilized behind the Zanu-PF war effort but still used the war to advance their own gender-related conflicts. Similarly, part of the motivation for young males joining the war was not necessarily grand hopes of relief from racial discrimination, or immediate material gains, but escape from the oppressive hierarchy of rural life dominated by male elders. Each of these levels or types of conflict subsumed in the greater war contains its own specific history of social relations, and its own complex mixture of direct interest in material gain with a push for changes in the rules or customs of social relationships. This degree of complexity has not, to date, been captured by neoclassical theory, nor by the proxies used in applications of the theory. The false consciousness argument is based on a rigid conceptual distinction between greed and grievance, it is sustained by the conceit that the proxies used in applications of the models are genuinely ‘‘almost-the-same’’ as their objects, and lastly it depends on the reliability of the data explored to test the model. But, there are hugely significant empirical weaknesses with most, probably all of the proxies employed. Inequality data are notoriously frail, especially for use in cross-country comparisons (and a fortiori for sub-Saharan Africa, where many recent conflicts have taken place). 28 Production data emphasizing the preponderance of primary commodities are prone to errors, thanks to unrecorded manufacturing activity and the elision in some data sets (particularly concerning mining) between manufacturing and primary commodity activities (Riddell, 1990; Yeats, 1990). Data on educational attainment have their weaknesses, but beyond the crude figures there are problems in assuming a direct, universal and predictable relationship between educational attainment and viable ‘‘economic opportunities,’’ e.g., in employment (Bennell, 1996; Pritchett, 1997). The relationship depends, among other things, on the availability of agricultural wage labor (rarely captured effectively in rural survey data) and on the particular set of policies adopted by governments. Finally, the ethno-linguistic fragmentation index is far from foolproof (McIlwham, 1998). Needless to say, data observations for countries already affected by conflict are even more unreliable, when many goods are traded illegally and statistical services do not have the same reach as in peacetime (Cramer & Weeks, 2000). The use of conceptually and empirically vague proxies is an instance of the more general problem of the empirical weakness, arbitrary prioritization of particular variables, and manipulation of technique over and above the search for ‘‘truth’’ in econometrics. Mainstream economists and econometric practitioners such as Mayer (1993) and Leamer (1983) have often cautioned against the frequent, almost institutionalized abuse of technique. More radical critics, such as Lawson (1997), have argued for the more fundamental failure as an analytical device of econometrics.