ON ETHNIC CONFLICT AND THE ORIGINS OF TRANSNATIONAL TERRORISM

Abstract

Using the ITERATE dataset, we explore the origins of transnational terrorist activity, from 1982 through 1997, in 118 countries. We model terrorism, not as a function of a nation’s ethnic, religious or linguistic fractionalization, but as an independent measure of perceived ethnic tensions. When we control for institutional quality, evidence that political rights and civil liberties mitigate the terrorism‐producing effects of ethnic tensions exists only since 1990. Economic freedoms, on the other hand, robustly reduce the number of terrorist attacks originating in ethnically tense societies.

Keywords:

• Ethnic tensions
• Transnational terrorism
• Political rights
• Economic rights
• Cold War

JEL Codes:

• D74

ACKNOWLEDGMENT

The authors benefitted from the comments and suggestions of our discussant, Lisa Chauvet, and other participants at a session of the First World Congress of the Public Choice Societies in Amsterdam, including Art Carden, Roger Congleton, Bernie Grofman and Erich Weede. Scott Beaulier, who discussed our paper at the 2007 meetings of the Public Choice Society, added considerable value to the paper, as did audience members Todd Nesbit, Russ Sobel and Sasha Tomic. The thoughtful written comments of Dann Arce, John Warner and the journal’s two anonymous referees were especially helpful in sharpening our arguments and analysis. As is customary, though, the authors accept full responsibility for any remaining errors of omission or commission.

Notes

¹ Quoted in Hardin (1999: 276).

² See Shughart (2006), Fearon (2006: 856) and, for an in‐depth treatment of the ‘exceptional mismatch between ethnic identities and political structures’ in certain parts of the world, Ferguson (2006). In contrast to Shughart (2006), Horowitz ([1985] 2000: 75–76) claims that, in Africa at least, ‘the boundary‐drawing process frequently took ethnic interests into account, and boundaries often were redrawn later by colonial powers in response to ethnic demands.’ He goes on to argue, however, that ‘the colonies were artificial, not because their boundaries were indifferent to their ethnic composition, but because they were, on the average, many times larger than the political systems they displaced or encapsulated’, thus greatly expanding the range of ethnic group interactions (and points of potential friction), most notably in the case of national elections: ‘many groups encountered each other for the first time during colonial rule’ (Horowitz [1985] 2000; 98).

³ ‘Terrorism is theater’, according to Brian Jenkins (quoted in Coll, 2004: 139). Alternative definitions are supplied by, for example, the US Department of State (1997).

⁴ A euphemistic neologism for ‘ethnic conflict’ delighted in by the Nazis and more effective in maintaining control of a heterogeneous state than overt propaganda, according to philologist Victor Klemperer (Ferguson, 2006: 455).

⁵ ‘Ethnicity is connected to birth and blood, but not absolutely so’ (Horowitz [1985] 2000: 51–52). Cautioning ‘against the acceptance of the figment of the pigment’, Horowitz argues in favor of ‘an inclusive conception of ethnicity that embraces differences identified by color, language, religion, or some other attribute of common origin’. Indeed, ‘it is not the attribute that makes the group, but the group and group differences that make the attribute important’ (Horowitz [1985] 2000: 46, 50).

⁶ The right conditions follow from ‘Hamilton’s Rule … [which] states that, other things equal, evolutionary selection will lead a Donor organism D to aid a recipient organism R if the cost‐benefit ratio C _D /b _R is less than their relatedness r _DR …’, where ‘cost C _D and benefit b _R are measured in increments to the ‘fitness’ (i.e., reproductive survival) of Donor and Recipient, respectively’ (Hirshleifer, 2001: 21; emphasis in original).

⁷ ‘Among the most important needs met by ethnicity is the need for … emotional support and reciprocal help, and for mediation and dispute resolution …’ (Horowitz [1985] 2000: 81).

⁸ Specifically, in societies where one ethnic group comprises 45–90% of the population there is a 28% chance of civil conflict. However, when ethnic groups are more evenly distributed the probability of conflict falls to 3%. It turns out, though, that Africa contains many of the world’s most ethnically diverse nation‐states – Uganda tops the list; Kenya ranks third (Alesina et al., 2003). Fearon (2006: 854) likewise notes that ‘Sub‐Saharan Africa stands out as the only region in which fewer than half the countries have an ethnic majority group.’ These observations tend to undermine the argument that violence is more likely where one ethnic group is dominant. On the other hand, Collier and Hoeffler (2004) report evidence suggesting that civil war is more likely to erupt in such countries.

⁹ Hirshleifer’s (2001: 45) ‘paradox of power’ implies that in any conflict over resources disadvantaged groups will fight harder than better‐endowed groups, whose members have more to lose and, hence, more to gain from cooperation with the other side.

¹⁰ Others have drawn an analogy between terrorism and hate crimes since the definition of such crimes includes attacks by one religious or ethnic group on another religious or ethnic groups (see, for example, Hamm, 1998; Kressel, 1996).

¹¹ Frey and Luechinger (2003: 247) offer ‘three specific strategies: visits to other countries, principal witness programmes, and formal contact, discussion processes and access to normal political participation.’ Anderton and Carter (2005) present a more rigorous analysis of the tradeoffs that emphasizes the importance of underlying cross‐price elasticities, expenditure shares and income elasticities in evaluating the effectiveness of Frey and Luechinger’s ‘benevolence’ strategy versus a deterrence strategy.

¹² Anecdotally, Collier and Hoeffler (2004) point out that during the Russian civil war the desertion rates for both contending armies (the Reds and the Whites) were about ten times higher during the summer than in winter – the income foregone by the peasant rebels was much higher when fields needed to be plowed and harvests gathered.

¹³ John le Carré (2008: 56–57) supplies a fictional account of the paid‐rider option circa 9/11:

Okay, [Hamburg] harbored a few Islamist terrorists, and a trio of them had gone off and blown up the Twin Towers and the Pentagon. So what? It was what they’d come here to do, and they’d done it. Problem solved. They’d struck at the heart of the Great Satan, and they’d killed themselves in the process. We were their lauchpad for Christ’s sake, not their target! Why should we worry? So we lit candles for the poor Americans. And we prayed for the poor Americans. And we showed them a lot of free solidarity. … And when the Iraq War came along, and we good Germans stayed aloof from it, that made us even more immune. Madrid happened. Okay. London happened. Okay. But no Berlin, no Munich, no Hamburg. We were too … immune for any of it. (Emphasis in original)

¹⁴ As one of many other examples, Sikh militants based in the Canadian province of British Columbia blew up an Indian airliner on June 23, 1985, in an effort to pressure the Indian government into ceding them an independent state carved out of India.

¹⁵ Part of the problem confronting scholars is that ethnic and linguistic fractionalization data typically are gathered from census returns, which are subject both to manipulation ‘since there is an element of self‐definition in ethnic affiliation’, and to tampering by groups that stand to gain either by inflating their own numbers or deflating those of others. In consequence, ‘disputes over census results in ethnically divided societies are common’ (Horowitz [1985] 2000: 195). Fearon (2003: 198) nevertheless suggests using survey‐based methods for obtaining information on ethnic identities, cautioning, however, that responses may vary depending on external factors. During episodes of poor economic performance, for example, hitherto unimportant ethnic divisions may become relevant; the same respondents may report less ethnic distinctiveness when the economy is doing well.

¹⁶ According to the World Bank’s WDI database, Botswana’s GDP grew by 8.2% in 2000, and by 6.2% in 2006.

¹⁷ Alesina et al. (2003) provide some additional examples of these problems. An index based on language would not capture any polarization between blacks and whites in the United States. An index based on a static definition of ethnicity would not capture the five different groups into which the Oromo of Ethiopia have split as a result of exogamy and migration.

¹⁸ Further support for Keefer and Knack’s (2002) conclusion is provided by Collier and Hoeffler (2004: 581), who find that fractionalization affects the likelihood of ethnic conflict nonlinearly: the potential for conflict at first increases with ethnic fractionalization, but then falls as fractionalization continues to rise.

¹⁹ The additive index used by Keefer and Knack (2002) includes, in addition to EXPROPRIATION and REPUDIATION, the ICRG’s ratings for ‘corruption’, ‘bureaucratic quality’ and ‘rule of law’. Keefer and Knack report that these three other components are highly correlated with one another, with correlation coefficients ranging from 0.66 to 0.88. They find the index itself to be highly reliable, producing a Cronbach’s alpha of 0.94.

²⁰ GDP PER CAPITA and ETHNIC TENSIONS are uncorrelated contemporaneously, although as expected from studies such as that of Easterly (2000), last period’s ETHNIC TENSIONS is highly correlated with current GDP PER CAPITA.

²¹ Terrorist acts in the current year are even less likely to influence ethnic tensions and property rights two years earlier. We therefore ran all variants of the regressions reported below entering two lags of the explanatory variables rather than one and found that the results were unchanged. Given that our estimates seem to be robust to lag structure, we formally report those obtained with the right‐hand side variables lagged one period.

²² The case of Canada’s home‐grown jihadis is well‐known. On June 2–3, 2006, 18 mostly Muslim Canadian men were arrested in a Toronto suburb on charges of plotting to bomb the Canadian Parliament and storm the offices of the Canadian Broadcasting Corporation.

²³ This argument is not sensitive to the choice of panel estimation techniques. A random‐effects approach on Model 6 (see Model 6.1 in Table XIII(a)) shows that ECONOMIC FREEDOM is the only explanatory variable with a statistically significant impact on TOTAL ATTACKS.