What causes terrorism?

Transcription

1 What causes terrorism? Tim Krieger Daniel Meierrieks y Abstract We provide an overview of empirical evidence on the determinants of terrorism, in particular focusing on the origins and targets of transnational terrorism. We also assemble several broad theoretical families that relate terrorism to, e.g., economic, political and institutional and demographic factors. We provide a critical discussion of the existing empirical evidence and refer to a number of areas of future research before describing some modest policy implications. Keywords: Determinants of Terrorism, Transnational Terrorism, Counter-Terrorism Policy JEL Classi cation: D74, H56, N40 University of Paderborn, Department of Economics, Warburger Straße 100, Paderborn, Germany. Ph.: +49-(0) , fax: +49-(0) , tim.krieger@notes.uni-paderborn.de. y University of Paderborn, Department of Economics, Warburger Straße 100, Paderborn, Germany. Ph.: +49-(0) , daniel.meierrieks@notes.uni-paderborn.de 1

2 1 Introduction Since the attacks on the US on September 11, 2001, politicians and the public have tried to bring to light the factors that incite terror. The then US President George W. Bush (2002) argued that there is a vicious circle of disenfranchisement, state failure and terror, stating that because [...] persistent poverty and oppression can lead to hopelessness and despair [...] failed states can become havens for terror. The media and public were highly receptive to lines of argumentation that linked terror to socio-economic and political underdevelopment. Accordingly, policy measures have been advocated that aim at alleviating underdevelopment, e.g., by reducing socio-economic strain or political instability. For instance, the then German Chancellor Gerhard Schröder (2003) argued that [...] to address the root causes of terrorism and insecurity [...] we must ensure social and material but also cultural security. Unveiling the causes of terrorism and deriving sound policy advice is important because terrorism is costly to a ected countries. Even if the immediate costs of terrorism are marginal, its indirect political and economic costs may be substantial. For instance, terrorism may reduce governmental stability (Gassebner et al. 2008). Terrorism may also negatively impact trade and capital ows and thus economic development (Nitsch and Schumacher 2004; Crain and Crain 2006; Abadie and Gardeazabal 2008; Mirza and Verdier 2008) and may result in a loss of individual and collective life satisfaction (Frey et al. 2009). Given popular discourse and the unquestioned costs of terrorism, we provide a survey of empirical studies on the determinants of terrorism. In Section 2 we discuss the basic theoretical ideas underlying an economic analysis of terrorism. In Section 3 we review the empirical literature on terrorism determinants. We focus on large-sample 2

3 studies with long time horizons to assess long-run global trends. We also present complementary evidence where relevant, recognizing that case- or region-speci c studies and micro evidence may be important to revealing speci c dynamics of terrorism. For this review we focus on the sources (terrorists countries of origin) and targets of transnational terrorism, as the majority of empirical studies have focused on this form of terrorism. 1 Our discussion should help to clarify which broad hypotheses are most suited to explaining transnational terrorist activity and why no empirical mainstream has emerged so far. In Section 4 we sum up our ndings, provide some cautious policy implications and hint at avenues of future research. 2 Explaining the causes of terrorism There is no universally accepted de nition of terrorism. It is commonly de ned as the deliberate use of violence and intimidation directed at a large audience to coerce a community (government) into conceding politically or ideologically motivated demands. The main tactical (short-run) goals of terrorism are (1) gaining publicity and media attention, (2) destabilizing polity and (3) damaging economies (e.g., Tavares 2004). Among the long-run goals of terrorism is a redistribution of power, in uence and wealth (e.g., Frey and Luechinger 2004). Terrorist organizations must have goals that are not enforceable in the ordinary political process and for whose implementation their members are willing to use force. Tactical terrorist behavior (e.g., assassinations, hostage-takings) then serves to achieve these strategic goals, making violence a means to meet more abstract objectives. Individually, terrorists must exhibit certain character traits that enable them to carry out terrorist actions. Organizationally, the dominance of group leaders, group dynamics 3

4 and other (psychological) factors also in uence terrorist behavior (cf. Victoro 2005). Hence, popular opinion often links terrorism to irrationality or insanity. However, such an argumentation is too simple, perhaps even wrong. An economic view of terrorism instead assumes that terrorists are rational, so the average terrorist behaves more or less as a homo economicus (Caplan 2006). As rational actors terrorists act violently to maximize their utility, given certain bene ts, costs and constraints that are linked to these actions (e.g., Sandler and Enders 2004). 2 The utility-maximizing level of terrorism is the level at which the marginal costs equal the marginal bene ts of terrorism. Bene ts from terrorism arise from obtaining the tactical and strategical terrorism goals. The costs of terrorism are, e.g., linked to the use of resources and to the opportunity costs of violent behavior (e.g., Frey and Luechinger 2004). 3 Aggregate country-speci c factors impact the terrorists cost-bene t matrices and thus their behavior. Such determinants may raise (lower) the price and/or the opportunity costs of terrorism, causing a decline (increase) in terrorist activity. Using this idea, several schools of thought have put forward speci c global hypotheses (GH ) that emphasize the importance of certain factors in determining terrorism. Table 1 relates the global hypotheses to speci c determinants and gives examples of how these determinants are measured empirically. The global hypotheses stress the role of economic deprivation (GH1), socio-economic and demographic change and resulting strain (GH2), the political order (GH3) and political transformation (GH4), identity (GH5), the global economic and political order (GH6) and temporal and spatial contagion (GH7) in explaining terrorist activity. 4 Next, we describe the global hypotheses in detail, speci cally with respect to the causal links running from certain determinants to terrorist activity. We keep this discussion rather short as we will also discuss the general hypotheses in Section 3. Further, the reader should be aware that the hypotheses may be of di erent relevance 4

5 to the source and target countries of terrorism. In the following we hint at this, but delegate a more exhaustive discussion to Section 3. Table 1 about here 2.1 Economic deprivation Some scholars suggest that terrorism is rooted in economic deprivation, i.e., in poverty and within-country inequality. Gurr (1970) puts forward the idea of relative deprivation, where violence is generated when there is a discrepancy between what individuals think they deserve and what they actually receive through the economic (distributive) process. Poor structural economic conditions create frustration, which in turn makes violence more likely. This link from economic deprivation to terrorism should matter to the source countries of terrorism. For instance, terrorist organizations should nd it easier (less costly) to recruit frustrated followers or to receive funding from supporters when economic deprivation prevails. The lack of non-violent economic activities may also ll the ranks of terrorist organizations by lowering the opportunity costs of violence. Thus, relative economic deprivation is expected to lead to more terrorism by inciting frustration and lowering the (opportunity) costs of violence. With respect to the target countries of terrorism, economic success may attract attacks when economic deprivation is assessed globally (poor vs. rich countries). 2.2 Socio-economic and demographic strain Other scholars argue that terrorism is fostered by the process of modernization. Modernization encompasses, inter alia, economic change (e.g., economic growth), new forms of communication and lifestyles (shift from agricultural to urban societies) and new ideas (e.g., Western ideology). These factors may create grievances associated with socio-economic and 5

6 demographic strain (Robison et al. 2006). For instance, economic growth may be associated with a restructuring of labor markets, creating grievances among modernization losers who become unemployed due to economic change. As another example, modern forms of communication may challenge traditional elements of a society, generating social con ict (Robison et al. 2006), while terrorist organizations may use modern means of communication to disseminate their opinions more e ectively (Ross 1993). Most grievances are generated during the transition from a traditional to a modern society (Ross 1993). Here, terrorists are able to capitalize on the grievances of modernization losers linked to economic dissatisfaction, new forms of alienated living or other challenges to traditional societal patterns, thus making recruitment, nancing or other forms of support more likely. This mechanism is particularly relevant to the source countries of terrorism. Generally, modernization is linked to socio-economic and demographic changes which may feed through to con ict. These changes are di cult to capture in empirical analyses. Therefore, researchers often resort to speci c socio-economic (e.g., low educational attainment) and demographic (e.g., population growth) factors to indicate the impact of modernization on terrorism. 2.3 Political and institutional order The political and institutional order is also argued to matter to terrorism. There is an ongoing academic debate on whether a certain political system is more prepared to deal with terrorism. Democratic regimes can o er non-violent means of voicing dissent but are unable to realize hard counter-terrorism measures due to an obligation to civil liberties (Li 2005). This may make terrorism production (country of origin perspective) less likely but may increase the probability of terrorist attacks (target country perspective). Autocratic regimes can capitalize on their capability of repression 6

7 which may at the same time generate grievances linked to political disenfranchisement. A low level of political openness may make the genesis of terrorism more likely but lessens the likelihood of terrorist attacks. In general, there is no consensus on which political regime can fend o terrorism most e ectively. Some authors suggest that semi-open societies (partial democracies) are most prone to terrorism because they cannot capitalize on the advantages of either pure regime, suggesting a non-linear link between political institutions and terrorism. Regardless of the exact regime type, government strength (e.g., military or police power), structure (centralized vs. decentralized), policies (e.g., welfare policies) and ideological a liation (e.g., left-wing vs. right-wing) may also in uence terrorists calculi. For instance, a large-scale government may make it more di cult for societal groups to pursue rent-seeking, making it more attractive to gain rents through violence (Kirk 1983). In any case, the speci city of various institutions is expected to enter the terrorists calculi in various ways. For instance, government policies and characteristics (which ensue from the political and institutional order) may also matter to terrorist activity. 2.4 Political transformation and instability Political transformation and instability are also named as causes of terrorism, in particular in popular discourse. The main idea is that political change may create political vacuums which terrorist groups use to push their agendas. Such vacuums are attractive as radical groups are less likely to be challenged by an instable, thus weak government (e.g., measured by a regime stability variable), making terrorism a less costly venture. Also, an individual may nd it more attractive to join or support a radical organization because there are few non-violent alternatives (meaning low opportunity costs of violence) but 7

8 high payo s from terrorist success (meaning increased bene ts of violence). Instable or failed states may even serve as schools of international terrorism, where in phases of domestic instability (e.g., civil war) individuals gain an education in violence that they can also use for internationalized terrorist campaigns (Campos and Gassebner 2009). Political transformation may generally amplify terrorist behavior, where this process in uences both the production of terrorism and terrorists target decisions. 2.5 Identity and cultural clash Famously, Huntington (1996) states that civilizational clash may also result in violence. When groups exhibit di erent identities (e.g., di erent religions or ethnicities), this may lead to more con ict either between di erent groups within a country or between di erent country groups organized along civilizational lines (e.g., Islamic countries versus the West). 5 For terrorist groups, it should be easier (less costly) to muster support against antagonistic identity groups (identi ed, e.g., by ethnic or religious fractionalization measures). This holds even more when terrorists build on identity-related ideologies that stress the supremacy of their respective identity (e.g., representing the only true faith ). Such a world view eliminates moral constraints and strengthens an organization s cohesion, making terrorism less costly and more e ective (Bernholz 2006). The abstract con ict between world views also becomes manifest in realpolitik, where groups with di erent identities pursue di erent (often diametrically opposed) policies. Such behavior may, e.g., be represented in rent-seeking or other forms of socio-economic and political interaction between groups with di erent identities (Arce and Sandler 2003). Terrorism is used by the inferior group not only as a means to voice their world view but also to shift (material) outcomes in their favor. Identity (and opposition to other identities) works as a bond facilitating, e.g., terrorist recruitment and nancial support. When terrorists 8

9 succeed, related payo s are particularly high due to the enforcement of a claim to the absolute (Bernholz 2006). 2.6 Global economic and political order The global political and economic order (globalization) may also matter to terrorism, where terrorism driven by global factors is more likely to be of an international nature. Economic integration (e.g., measured by trade openness), foreign policy (e.g., indicated by political proximity to the West) and alliance structures (e.g., indicated by membership in alliances) may play a role when developing grievances and resistance against globalization. Traditionalist or disenfranchised segments of a society may use violence to counter foreign dominance (i.e., Western supremacy) and global modernization (Lizardo and Bergesen 2004). If individuals are incited by an existing global order that is perceived as unfair, it should be easier for terrorist organizations to nd support by building on related grievances in the source countries of terrorism. However, the targets of terrorism may also be chosen in response to the existing global order, especially if this order is perceived as unjust from the perpetrators perspective. For instance, a con ict between a government and an opposing group may be exported to a foreign ally of the government (Addison and Murshed 2005). Terrorism may also be used as a foreign policy tool to ght antagonistic world views, e.g., as observed during the Cold War (O Brien 1996). However, political and economic globalization does not necessarily automatically generate violence. For instance, if political integration ameliorates the international distribution of wealth and power by means of cooperation or if economic integration through trade gains also bene ts the poor, it may reduce terrorist support as globalization grievances are reduced. That is, while a globalization-terrorism nexus may very well exist, its exact mechanics remain disputed. Also, it is di cult to capture in cross-country analyses. 9

10 Hence, empirical analyses have often resorted to straightforward (but rough) measures of the global order, e.g., trade openness or political proximity to the US. 2.7 Contagion Contagion may be another factor explaining terrorism. The main idea is that terrorism exhibits a strong self-energizing nature with respect to both time and space (Midlarsky et al. 1980). Temporal contagion means that past terrorism bears new terrorism within one country (empirically implemented by lagged terrorism variables). For a terrorist organization it is more bene cial to run a terrorist campaign because this increases the bene ts from terrorist activity (e.g., through increased media attention). Spatial contagion means that if one country su ers from terrorism, it is also likely to infect neighboring countries(e.g., indicated by the distance to terrorism). Terrorist activity in one country may lead to imitational behavior in neighboring countries. Emerging terrorist groups may capitalize on the experience of existing groups in adjacent countries. Additionally, when terrorist organizations cooperate they may reduce their costs (e.g., by sharing know-how and weapons) and increase the payo (e.g., by joint terrorist actions). Temporal and spatial proximity to terrorism may thus in uence the cost-bene t considerations of terrorists in ways that may promote the generation of violence and the likelihood of a certain country being targeted. 6 3 Evidence on the determinants of terrorism In this section, we review the evidence on the determinants of transnational terrorism that tests the previously discussed global hypotheses. Due to the dyadic nature of this kind of terrorism, we rst consider studies that center on factors that may impact 10

11 the production of terrorism in certain countries (country of origin perspective). Then, we review the evidence regarding factors that make a country a more or less likely target of transnational terror (location of attack or target perspective). We center on cross-sectional and panel studies investigating underlying hypotheses on a global scale. Here, we acknowledge that comparing empirical studies involves some non-trivial methodological problems as the empirical set-ups of the studies di er, potentially leading to (arti cially) con icting evidence. Generally, this may be due to (1) di erences in the dependent variables. As shown by Drakos (2009), terrorism measures constructed from di erent sources (e.g., ITERATE, MIPT-RAND) may correlate but nevertheless report distinct terrorism patterns, potentially as a consequence of underlying under-reporting biases (Drakos and Gofas 2006b). Also, dependent variables may be used on di erent time horizons, so con icting evidence may rather point to changes in the nature of terrorism. (2) The empirical analyses use di erent explanatory variables. The diversity of results may to some extent be due to the econometric context. This leaves open the possibility of omitted variable biases when models are speci ed incompletely. Furthermore, the mere labeling of variables may hide the fact that di erent studies use di erent variables (data sources) to measure certain relationships. (3) The econometric methodology of the analyses di ers. While most studies rely on negative binomial models, others employ, e.g., OLS or gravity model estimations. The dissimilarity of estimation techniques may also contribute to the dissimilarity of empirical results. In order to tackle these problems, in the following we discuss the empirical set-ups when appropriate. This approach allows us to identify whether the evidence is truly con icting or rather methodology-dependent. 11

12 3.1 Origins of transnational terrorism Economic deprivation One controversial subject is whether terrorism originates in poor economic conditions. Evidence on this issue is con icting. On the one hand, the ndings of Blomberg and Hess (2008a), Azam and Delacroix (2006), Lai (2007) and Azam and Thelen (2008) suggest that successful economic development (measured in real GDP per capita) signi cantly reduces the genesis of terrorism. On the other hand, empirical results by Krueger and Maleckova (2003), Kurrild-Klitgaard et al. (2006), Basuchoudhary and Shughart (2010), Plümper and Neumayer (2010), Krueger and Laitin (2008) and Freytag et al. (2008) generally point at only weak links between economic development and terrorism. Furthermore, Kurrild-Klitgaard et al. (2006) nd no signi cant links between income poverty, income inequality and terrorism. In other words, there is only limited evidence to support the hypothesis that economic deprivation causes terrorism (GH1). As we shall see later, poor economic conditions matter less to terrorism once it is controlled for institutional and political factors. The notion that economic deprivation is not a strong catalyst of terrorism is also supported by the micro evidence of Krueger and Maleckova (2003) and Berrebi (2007). Neither study nds a negative relationship between wealth and terrorism. A study using aggregated data by Piazza (2007) for the Middle East for the period 1972 to 2003 returns similar ndings. At best, we may hypothesize that a combination of poor economic and institutional conditions may matter to the genesis of terror, e.g., when poverty is accompanied by low opportunities to escape it (as poor institutions constrain political and economic participation). However, such an assumption only stresses that poverty and inequality do not translate into terrorism of their own accord. 12

13 3.1.2 Socio-economic and demographic strain The idea that modernization-related socio-economic and demographic strain leads to transnational terrorism is di cult to operationalize in empirical studies. For lack of better proxies, rough measures such as economic growth, population size and education are used to indicate modernization-related strain. However, they may also constitute elusive indicators. Given these preliminary remarks, we nd little convincing support for the idea that various social strains cause terrorism (GH2). First, Kurrild-Klitgaard et al. (2006) and Krueger and Laitin (2008) detect no signi cant connection between economic growth and terrorism. In other words, the little evidence there is does not support the notion that economic change translates into violence from those who lose out on it. Second, many studies nd that more populous countries are more likely producers of transnational terrorism (Krueger and Maleckova 2003; Burgoon 2006; Lai 2007; Plümper and Neumayer 2010; Freytag et al. 2008; Piazza 2008b; Krueger and Laitin 2008). Only in Azam and Delacroix (2006) and Azam and Thelen (2008) is population size insigni cant. Such ndings may indicate that demographic strain fosters con ict as population size comes with higher demographic stress, so terrorism costs are reduced. Arguably though, population size is a poor proxy for demographic stress. The correlation between population size and terrorism production may simply indicate that terrorism is more likely in larger countries, so there is no independent mechanism behind this correlation. In order to investigate the true e ect of modernization-related demographic strain on terrorism, it would be necessary to use proxies that more directly measure underlying concepts (e.g., youth burdens). Third, evidence for education is diverse and needs future research, as we cannot con rm 13

14 that terrorism is in uenced by education. 7 Azam and Thelen (2008) suggest that higher secondary school enrollment discourages terrorism production. Freytag et al. (2008) nd a signi cant e ect of the average years of schooling on terrorism which varies with di erent analytical scopes. In Krueger and Maleckova (2003), the e ect of the illiteracy rates on terrorism is negative but insigni cant. Kurrild-Klitgaard et al. (2006) nd that higher levels of overall education even tend to encourage terrorism, albeit insigni cantly. The diversity of results may result from the use of di erent indicators of education, i.e. education measures Political and institutional order As previously noted, several studies nd that economic deprivation leads to terrorism. However, most of these studies do not properly control for institutional settings. Once institutions are taken into consideration, the terror-amplifying e ect of poor economic conditions becomes less important. Accounting for economic and institutional development, several studies nd that more liberal and democratic countries are signi cantly less likely to produce transnational terrorism (Krueger and Maleckova 2003; Burgoon 2006; Kurrild-Klitgaard et al. 2006; Plümper and Neumayer 2010; Krueger and Laitin 2008; Piazza 2008b). 8 Basuchoudhary and Shughart (2010) nd that more liberal political institutions have tended to discourage transnational terrorism production only since the end of the Cold War. Only the ndings of Eubank and Weinberg (2001) and Lai (2007) suggest that more democratic countries generate substantially more terrorist activity, arguing that autocratic regimes are better prepared to suppress opposition. These ndings indicate that high economic development may work only as a proxy for institutional quality but that there is no independent causal e ect of income on terrorism. 14

15 The role of the broader institutional order is also stressed by Basuchoudhary and Shughart (2010) and Lai (2007), who nd that higher levels of economic freedom (e.g., in the form of secure property rights) reduce the production of terrorism. However, Kurrild-Klitgaard et al. (2006) control for the impact of a variety of economic institutions and nd no signi cant association between them and the genesis of terrorism. Regardless of the exact type of institutional order, there may also be e ects of government activity on terrorism. Burgoon (2006) nds that welfare policies signi cantly discourage terrorist activity. 9 Freytag et al. (2008) point in a similar direction, nding that government expenditures are to some extent associated with a reduction in terrorism. The composition and capacity of governments is also found to be associated with terrorism. When governments are more leftist or when they are more capable, the terrorism production in these countries increases (Burgoon 2006). In general, the production of transnational terrorism seems to be rooted in domestic political and institutional orders (GH3). Once institutional factors are taken into account, poor economic conditions do not matter strongly to terrorism. 10 Current evidence suggests that the advent of liberal political institutions reduces terrorism since, e.g., fringe groups are more likely to articulate their demands non-violently (Ross 1993). Democratic institutions may thus reduce con ict by o ering cost-e cient non-violent con ict resolution. There is considerably less evidence linking liberal societies to more terrorism production, e.g., by granting certain rights to their citizens that facilitate the terrorists access to communication, the media or other equipment (Ross 1993). Furthermore, evidence alludes to the importance of the broader institutional order in explaining terrorism. Economic institutions and government policies may increase the price of terrorism, e.g., by enabling better access to economic resources and opportunities. The results also indicate that governmental activity may cause more terrorism when 15

16 governments are powerful, as this produces hostility against them which increases the payo of terrorism (Sandler and Lapan 1988). That said, the evidence of the e ect of institutions on terrorism production is by no means resolved, opening avenues of future research Political transformation and instability Few empirical studies have investigated the relationship between political instability and terrorism (GH4). The ndings of Lai (2007) and Piazza (2008b) indicate that political instability (indicated by incidences of civil war and state failure) goes along with more terrorism production. By contrast, Eubank and Weinberg (2001) nd that stable democratic countries are the most likely breeding grounds of terrorism, whereas Piazza (2008b) nds that regime durability by itself has no signi cant impact on terrorism production. As the evidence is limited and contradictory, we cannot conclusively state whether political instability drives terrorism, e.g., as political instability teaches violent behavior, making it less costly (Campos and Gassebner 2009). For the Middle East Piazza (2007) nds that state failure and political instability are linked to increased terrorism production, indicating that a future investigation of such links on a global scale is promising Identity and cultural clash Identity con icts may also be linked to the genesis of transnational terrorism (GH5). With respect to ethnic identity con ict, Kurrild-Klitgaard et al. (2006) nd no signi cant association between ethnic or linguistic fractionalization and terrorism generation. Piazza (2008b) nds that countries with more homogeneous populations generate less terrorism. With respect to religious identity con ict, Blomberg and Hess (2008a) show that more 16

17 religiously fractionalized countries tend to generate less transnational terrorism. However, Krueger and Maleckova (2003), Kurrild-Klitgaard et al. (2006) and Krueger and Laitin (2008) nd no signi cant linkages between religious factors and terrorism. In general, the studies nd rather weak links between identity factors and terrorism, yet this does not necessarily negate the underlying hypothesis. Rather, it may be a measurement issue, as all studies above measure identity con ict via an invariant concentration index (e.g., ethnic or religious fractionalization) or via a variable indicating simple proportions (e.g., share of Muslim population). Such measures do not necessarily say much about actual tensions that are rooted in ethnic or religious diversity. Only Basuchoudhary and Shughart (2010) consider the genesis of transnational terrorism as a function of a country s (varying) degree of ethnic tensions. Measuring identity con ict in this manner, they nd that a higher level of ethnic tension is indeed associated with a higher level of terrorism production. This may imply that the identity hypothesis on con ict roots (GH5) is valid. It is possible that other studies have not arrived at a similar conclusion because they have failed to operationalize the identity-terrorism nexus properly Global economic and political order Some studies stress the role of international factors in in uencing terrorism production. The general idea is that terrorism develops in response to the global economic and political order. Even while it seems intuitive to assume that unfair global conditions morph into terrorism, it is di cult to empirically capture linkages from globalization to terrorism. As for the global economic order, trade openness (the ratio of exports and imports to GDP) is often used to measure the degree of and exposure to economic integration. 17

18 Evidence by Kurrild-Klitgaard et al. (2006) and Blomberg and Hess (2008a) indicates that higher levels of economic integration signi cantly reduce the generation of terrorism. Freytag et al. (2008) nd that higher levels of trade openness may produce more terror in more-developed countries, while openness is still found to reduce terrorism in less-developed economies. Only Burgoon (2006) nds a positive relationship between trade openness and terrorism, yet it is not signi cant. These ndings indicate that economic integration works salutary rather than threatening, decreasing terrorism support and thus its production. The terror-dampening e ects of economic integration (e.g., an erosion of funding of terrorist organizations through increased global competition) seem to dominate possible terror-enhancing e ects (cf. Mirza and Verdier 2008). However, using measures that better di erentiate between the terror-dampening and terror-amplifying e ects of economic integration may help to further disentangle the potentially complex relationship between terrorism and the global economic order. As for the global political order, Azam and Delacroix (2006) and Azam and Thelen (2008) nd that receiving foreign aid decreases a country s likelihood of producing terrorism. Blomberg and Hess (2008a) nd that political integration (WTO membership) also reduces this likelihood. These ndings indicate that international political cooperation may reduce transnational terrorism production (cf. Sandler 2005). However, other international political factors may increase terrorism production. Lai (2007) and Piazza (2008b) nd that involvement in interstate war is signi cantly associated with more terrorism, whereas Burgoon (2006) nds a negative but insigni cant relationship. Lai (2007) detects a strong and positive linkage between a country that is a state sponsor of terrorism and the likelihood of it actually generating terrorism. Plümper and Neumayer (2010) nd that membership of international alliances may increase the ow of transnational terrorism between two countries, particularly when there is a major power di erential 18

19 between them. In general, these ndings suggest that terrorism production signi cantly interacts with international political factors. 11 The direction of in uence seems to be dependent upon the exact impetus of related policies. While cooperative international policies may help to reduce terrorism, hostile and exclusive ones may propel it Contagion Lastly, we want to review whether spatial and temporal proximity also in uences the generation of terrorism (GH7). Investigating spatial contagion, Lai (2007), Plümper and Neumayer (2010) and Blomberg and Hess (2008a) nd that proximity to other countries su ering from terrorism makes the production of terrorism on own soil substantially more likely. Considering temporal contagion, Lai (2007) shows that past incidences of terrorism predict terrorist activity in present time periods, hinting at the persistence of terrorism. This nding is also supported by the studies of Enders and Sandler (2005) and Piazza (2007) who point towards the autoregressive and infectious nature of transnational terrorism. The sparse empirical evidence thus o ers support for the contagion theory of terror production (GH7). 3.2 Targets of transnational terrorism Economic deprivation The mainstream nding on the link from economic deprivation to the target decisions of transnational terrorists (GH1) is that terrorist attacks are associated with higher levels of economic development (Eyerman 1998; Tavares 2004; Blomberg et al. 2004; Kurrild-Klitgaard et al. 2006; Plümper and Neumayer 2010; Blomberg and Hess 2008a, 2008b; Krueger and Laitin 2008). This is consistent with the theory by Blomberg et al. (2004) who argue that fringe groups resort to terrorism in rich countries (as 19

20 the most cost-e ective way of violent dissent) while they resort to open rebellion in poorer economies. However, some studies do not nd signi cant links between economic development and terrorist attacks (Piazza 2006; Sambanis 2008; Dreher and Fischer 2010; Campos and Gassebner 2009; Dreher and Gassebner 2008; Piazza 2008a). This sheds doubt on the mainstream notion that favorable economic conditions attract terrorist strikes, in particular as variables besides per capita income (e.g., income inequality) are normally not robustly linked to terrorism (Li 2005; Kurrild-Klitgaard et al. 2006; Piazza 2006; Koch and Cranmer 2007; Piazza 2008b). Even more, Li and Schaub (2004), Li (2005) and Braithwaite and Li (2007) nd that high levels of economic development make it less likely for a country to become a target of transnational terrorism. So, while the mainstream notion is that transnational attacks are more likely in richer countries, some ndings suggest that caution is advised. Variables indicating economic development may correlate with the true causes of terrorism. For instance, Li (2005), Piazza (2006) and Kurrild-Klitgaard et al. (2006) control for the impact of economic development and institutional factors at the same time, nding that only the latter factors matter to terrorism. This suggests that economic development is associated with high institutional quality (e.g., e cient democratic institutions), where only institutions have a causal impact on terrorist attacks. 12 This leads back to the broader question of whether economic conditions or institutional surroundings attract terrorist strikes. We shall discuss this issue in more detail below Socio-economic and demographic strain Next, we review the evidence linking modernization-related strain to terrorist attacks (GH2). Here, little evidence links economic change (e.g., measured by GDP growth) to terrorist strikes (Drakos and Gofas 2006a; Kurrild-Klitgaard et al. 2006; Piazza 2006; 20

21 Campos and Gassebner 2009; Krueger and Laitin 2008). Only in Blomberg et al. (2004) are times of recession (when coupled with democratic institutions) associated with an increased vulnerability to transnational attacks. By contrast, Tavares (2004) nds that higher growth rates are linked to more terrorist attacks in some speci cations. Overall, these ndings do not suggest that economic change translates into higher terrorist activity. As a corollary, these ndings further indicate that economic success exerts no (substantial) independent e ect of on terrorist attacks. Turning to population factors, most studies (as with the origins of transnational terrorism) con rm that population size is a strong and positive predictor of transnational attacks (Braithwaite and Li 2007; Burgoon 2006; Campos and Gassebner 2009; Dreher and Fischer 2010; Dreher and Gassebner 2008; Koch and Cranmer 2007; Krueger and Laitin 2008; Li 2005; Li and Schaub 2004; Piazza 2006, 2008a, 2008b; Plümper and Neumayer 2010; Sambanis 2008), with Eyerman (1998) o ering complementary but insigni cant results. It is unclear whether such a correlation can be genuinely attributed to underlying causal mechanisms, given that population size as a poor proxy for such mechanisms. The correlation may simply reveal that attacks are more likely in more populated countries. Thus, it is more promising to use indicators of truly modernization-related demographic factors to capture related e ects. Here, Tavares (2004) demonstrates a terror-enhancing e ect of youth burdens. Tavares (2004) and Campos and Gassebner (2009) nd that high urbanization rates are associated with more terrorist strikes. Drakos and Gofas (2006a) show that high population density makes terrorist strikes more likely. That is, when using more re ned measures of demographic change empirical studies reveal that terrorist strikes are indeed more common when demographic strain is present. Considering the in uence of education on terrorist strikes, the ndings of Drakos and Gofas (2006a), Kurrild-Klitgaard et al. (2006) and Blomberg and Hess (2008b) do not 21

22 indicate that education is signi cantly associated with transnational terrorist attacks. However, all three studies use di erent measures to indicate educational levels (e.g., secondary school enrollment). Only in Tavares (2004) is a higher illiteracy rate among adult males to some extent linked to a reduction of terrorist activity. As discussed above, the diversity of results may indicate a methodological problem (of measuring education), as the e ect of education on terrorist attacks varies with the proxies used. On theoretical grounds, the weak links between education and the targets of terrorism also appear plausible. Most theoretical considerations approach the education-terrorism link from a country of origin perspective. From a target country perspective, such causal mechanism seems to be rather di cult to construct. 13 Overall, we nd some evidence for the modernization hypothesis (GH2), in particular when demographic strain associated with modernization (indicated by youth burdens or urbanization) is linked to terrorist attacks. Only marginal evidence is found linking economic change and education to terrorist strikes. As argued before, modernization-related strains are often operationalized by measures such as economic growth, population size and education. These proxies may be considered as improper indicators of such strain, especially as they may also be connected to the economic deprivation (e.g., economic growth) hypothesis Political and institutional order The mainstream nding considering the relationship between the political and institutional order and the probability of a terrorist attack (GH3) is that transnational terrorist attacks are associated with a liberal and democratic political system (Blomberg et al. 2004; Blomberg and Hess 2008a, 2008b; Braithwaite and Li 2007; Burgoon 2006; Eubank and Weinberg 2001; Li and Schaub 2004; Weinberg and Eubank 1998; Piazza 2008b; 22

23 Campos and Gassebner 2009). However, some results come to the opposite conclusion, nding that terrorist attacks co-occur with political repression (Piazza 2006; Eyerman 1998; Krueger and Laitin 2008). Other studies nd no signi cant or consistent links between political institutions and the likelihood of terrorist strikes (Drakos and Gofas 2006a; Koch and Cranmer 2007; Piazza 2008a; Dreher and Fischer 2010; Plümper and Neumayer 2010; Sambanis 2008; Tavares 2004; Dreher and Gassebner 2008). Hypothesizing a non-linear relationship between politics and the vulnerability to terror (Kurrild-Klitgaard et al. 2006) may explain the contradictory results outlined above. That is, semi-open polities are most prone to terrorist strikes, presumably as they cannot defend themselves against terrorism by e ectively using democratic or repressive tools against terrorism. Also, various components of the political order may interact di erently with terrorism. For instance, democratic participation may decrease the likelihood of attacks, whereas government constraints associated with a democratic order (which may constrain counter-terrorism actions) may make terrorist attacks easier (Li 2005). Such complex interactions may also account for the di erences in results. Apart from the (unsettled) discussion of which political system is most prone to terrorist strikes, a look at the evidence regarding the broader institutional order yields further interesting results. Higher levels of economic freedom are not associated with fewer terrorist strikes (Kurrild-Klitgaard et al. 2006; Piazza 2008a), which seems to con rm our previous notion pointing at the primacy of institutional over economic factors when explaining terrorist attacks. This notion is supported by further studies. First, once studies simultaneously control for government capacity 14 and economic development (Li and Schaub 2004; Li 2005; Koch and Cranmer 2007) there is no independent e ect of economic factors on terrorist attacks. Richer states are potentially more capable, meaning that it is then more cost-e cient to use terrorism than to engage in a full-scale 23

24 civil war (Blomberg et al. 2004). Second, there is evidence that government activities such as decentralization (Dreher and Fischer 2010) and social policies (Burgoon 2006) discourage terrorism. Third, government ideology a liation also matters. Interestingly, while Burgoon (2006) nds that left-wing governments tend to more successfully discourage terrorist attacks, Koch and Cranmer (2007) nd opposing results. In general, this evidence indicates that a closer look at economic and institutional environments may help to further unveil the mechanics of economics, institutions and terrorism. To sum up, the political order (GH3) may in uence the attack decisions of transnational terrorists, potentially in non-monotonic ways and through various channels. Independent of the formal political order related factors such as government capabilities, policies and ideological a liation also seem to matter. Given that economic and institutional development are correlated, most studies indicate that countries are primarily not attacked because they are rich but because they exhibit certain institutional traits; once these traits are controlled for, the e ect of economic factors on terrorist attacks becomes rather marginal Political transformation and instability Terrorist attacks may be more common in politically instable environments (GH4). Some studies nd that political stability (indicated by regime stability) discourages terrorist attacks (Li 2005; Campos and Gassebner 2009; Piazza 2008b). Similarly, political instability (indicated by, e.g., state failure, civil war, riots and strikes) encourages terrorist attacks (Weinberg and Eubank 1998; Piazza 2008a, 2008b; Campos and Gassebner 2009). 15 That is, not only the level of political freedom but also its change over time matters to terrorist strikes. Accordingly, Eyerman (1998) nds that while established democracies are less likely to experience terrorist strikes than autocratic regimes, new 24

25 democracies are most prone to attacks. Overall, the small body of literature on the relationship between political transformation and terrorist attacks suggests that the former attracts the latter Identity and cultural clash Considering the hypothesis linking identity con ict to terrorist attacks (GH5), some studies indeed nd that ethnic and/or linguistic diversity may encourage terrorist attacks signi cantly (Tavares 2004; Piazza 2006; Piazza 2008b). Others come to less consistent and signi cant ndings (Blomberg and Hess 2008b; Drakos and Gofas 2006a; Sambanis 2008). Kurrild-Klitgaard et al. (2006) nd that linguistic fractionalization drives terrorism substantially, while ethnic fractionalization does not. In general, these ndings o er some evidence for the idea that identity divides tend to encourage terrorist attacks. However, as discussed before most studies try to capture identity con ict by invariant measures that do not necessarily re ect underlying con icts over power and resources along ethnic lines. In fact, Drakos and Gofas (2006a) use a variant indicator of ethnic repression (from the Minority-at-risk project) and do not nd that this variable signi cantly drives terrorist attacks. This suggests that more care should be taken when indicating true identity con ict and its concomitants. 17 Turning to religious factors Piazza (2006) nds that religious diversity encourages terrorism but other studies come to inconclusive and insigni cant results (Blomberg and Hess 2008a, 2008b; Tavares 2004; Krueger and Laitin 2008; Kurrild-Klitgaard et al. 2006). Again, there are methodological reservations about measuring religious identity. Overall, such measurement issues do not allow us to conclude whether religious factors (as well as ethnic ones) drive terrorist attacks. 25

26 3.2.6 Global economic and political order Similar to the ndings on the origins of transnational terrorism, the results suggest that some components of the global economic and political order (GH6) matter to transnational attacks. Considering economic integration, Blomberg and Hess (2008a, 2008b) nd that higher levels of trade openness coincide with an increased probability of attack, whereas Drakos and Gofas (2006a) and Kurrild-Klitgaard et al. (2006) nd exactly the opposite relationship. Koch and Cranmer (2007) nd that the e ect of integration is dependent upon trade structure. That is, when a country trades with poor countries it is particularly likely to be targeted by transnational terrorism. The latter measure may more appropriately indicate grievances from an unfair trade order than the simple trade ratio. However, Braithwaite and Li (2007), Burgoon (2006) and Campos and Gassebner (2009) nd no statistically signi cant e ect of openness on terrorist attacks. Also, Li and Schaub (2004) who control for the e ect of several integration proxies (e.g., FDI) on terrorism detect no signi cant relationship and suggest that the e ect of integration on terrorism, if any, may be indirect. In general, these results do not suggest a strong link between economic integration and the patterns of terrorist attacks. Considering global political factors and terrorist attacks, we again di erentiate between its positive and negative e ects. On the one hand, Blomberg and Hess (2008a) nd that international cooperation (participation in international organizations) discourages transnational attacks. Campos and Gassebner (2009) nd no strong link between foreign aid and transnational terrorist attacks. These ndings seem to indicate, weakly, that peaceful political integration may discourage terrorist activity. On the other hand, confrontation and crises may impact terrorism. Drakos and Gofas (2006a) and Piazza (2008b) nd that international disputes foster transnational terrorist attacks. 26

27 Several other studies fail to nd evidence of strong links between international con icts (interstate war and other disputes) and terrorism (Li 2005; Braithwaite and Li 2007; Koch and Cranmer 2007). Interestingly, the ndings of Dreher and Gassebner (2008), Dreher and Fischer (2010) and Campos and Gassebner (2009) indicate that political proximity to the US is accompanied by a signi cant increase in the likelihood of being targeted by transnational terrorism. This nding ts the results of Plümper and Neumayer (2010) which suggest that political bonds (military alliances) and power di erentials between allies are associated with increased in ows of transnational terrorism. That is, it does not seem to be international con ict per se that attracts transnational terrorist attacks. Rather, speci c components of foreign policy seem to matter, particularly political proximity to the US. Interestingly, the sparse evidence so far suggests that economic integration matters less to terrorist attacks than international politics. In a broader sense, this is consistent with our previous notion that terrorist attacks are driven more by institutional (GH3) than economic conditions (GH1) and shaped more by political (GH4) than by socio-economic transformation (GH2) Contagion Strong support is found for the contagion hypothesis (GH7). The ndings of Plümper and Neumayer (2010) and Blomberg and Hess (2008b) suggest that distance between countries (spatial contagion) decreases the likelihood of terrorist attacks. Related to this, Braithwaite and Li (2007) report that terrorist attacks are more likely when countries are located in terrorism hot spots, i.e., in places where terrorism abounds. Similarly, Drakos and Gofas (2006a) nd that terrorist attacks are determined by location in certain world regions. With respect to temporal contagion, Li (2004), Li and Schaub (2004), Drakos and Gofas (2006a), Braithwaite and Li (2007) and Koch and Cranmer (2007) report 27