2. METODOLOGÍA
2.6 Instalación y configuración de los servicios que conforman al subsistema
2.6.3 Bloque de tratamiento de datos
In addition to the primary independent variable, I include other variables to ac- count for alternative explanations of group longevity. This significantly bolsters the strength of the conclusions I can draw from the analyses. By including other variables into the equation I can be more confident that it is in fact the nature of a group split and the consistency or inconsistency of internal preferences that are affecting group duration and not something else.
First, I include dummy variable indicators representing different group identities. An organization’s identity and goals might affect their longevity by providing a means to unite existing members while also enabling group leaders to reach a particular pool of recruits. The group identities I account for are: nationalist-separatist, communist- socialist, religious, anarchist, or leftist. In this system, groups can belong to several different identities at once so an organization can be both nationalist-separatist and
14. For a discussion see Ashworth et al., “Design, inference, and the strategic logic of suicide terrorism.”
communist-socialist, for example. Scholars have found that group identities can affect organizational endurance through a number of mechanisms. With regards to militant groups, however, the precise theoretical link between identity and survival is rather underdeveloped. For instance, Blomberg et al (2011) argue that “Terrorist ideology may also play a role in a groups success; however, this influence is at bottom really an empirical question that can be answered only by seeing how other ideologies fare against religious groups.15” Nonetheless, it is plausible that religious organizations, for instance, can use their shared faith to enhance cooperation and unify the group in a way that bolsters their survivability.16
Second, I include variables that are aimed at capturing particular state character- istics. Existing literature finds strong theoretical and empirical links between regime type and both group proliferation and states’ capacity to fight militants. Some ar- gue that terrorism is more likely to proliferate in democracies since the institutional arrangement incentives conflict between competing actors and interests.17 However, with regards to group duration, there are several competing hypotheses. Young and Dugan argue that because “democratic societies offer institutional recourse for ag- grieved individuals, people have formal mechanisms for resolving their anger towards the state”,18 and the resulting terror groups should be short lived. Although this would seemingly suggest that militant groups would also form less often in democ- racies—which is not the case—it nonetheless implies that violent nonstate actors
15. S. Brock Blomberg, Khusrav Gaibulloev, and Todd Sandler, “Terrorist group survival: ideology, tactics, and base of operations,”Public Choice 149, nos. 3-4 (October 2011): 450.
16. Richard Sosis and Candace Alcorta, “Militants and martyrs: Evolutionary perspectives on religion and terrorism,” Natural security: A Darwinian approach to a dangerous world, 2008, 105– 124.
17. Erica Chenoweth, “The Inadvertent Effects of Democracy on Terrorist Group Proliferation” (Doctoral Dissertation, University of Colorado, 2007); Erica Chenoweth, “Democratic competition and terrorist activity,” Journal of Politics 72, no. 1 (2010): 16–30; Quan Li, “Does Democracy Promote or Reduce Transnational Terrorist Incidents?,” Journal of Conflict Resolution 49, no. 2 (April 2005): 278–297; William Lee Eubank and Leonard Weinberg, “Does democracy encourage terrorism?,”Terrorism and Political Violence6, no. 4 (1994): 417–435.
within democratic countries have additional pathways of abandoning their violent tactics, making them more short-lived on average.
Scholars also debate whether or not democracies make better19 or worse20 coun- terinsurgents. Those in the positive camp might argue that the armed forces of demo- cratic nations are superior and they should therefore be more effective against all enemies, even nonstate actors. On the other side of the debate, those who argue that democracies are worse counterinsurgents might contend that democratic publics are averse to long, costly, casualty-prone wars and as a result they are less dedicated and worse off than equally-equipped autocratic regimes. Either way, it is theoreti- cally plausible for regime type to impact group longevity in meaningful ways so I include dummy variables for autocracy and democracy. Similar to the motivation be- hind regime type, I also include the regime’s durability—the total number of years since the most recent regime change.21Regime durability could be connected to orga- nizational longevity through several mechanisms but most significantly, more durable regimes might be better prepared to combat nonstate actors. It is likely that more stable states have increasingly developed armed forces, more capital to fund their military, and more developed police forces to quickly detect extremist elements. Con- sequently, greater regime durability should correlate with decreased group longevity. Also at the state level I include measures of a country’s GDP per capita and its population. GDP might influence militant survivability in several ways. Some argue that poor economic performance and low development are correlated with militant
19. Dan Reiter and Allan C. Stam, “Democracy and battlefield military effectiveness,”Journal of Conflict Resolution 42, no. 3 (1998): 259–277; Dan Reiter and Allan C. Stam, Democracies at war (Princeton University Press, 2002).
20. As pointed out by Lyall (2010), this argument rests on the assumption that democratic publics are particularly casualty-averse and sensitive to large-scale attacks on civilians. Ethan Bueno de Mesquita, “Politics and the suboptimal provision of counterterror,”International Organization 61, no. 01 (2007): 9–36; John Mueller,Policy and opinion in the Gulf War (AAPOR, 1994).
21. Marshall, Jaggers, and Gurr, Polity IV Project: Political Regime Characteristics and Transi- tions, 1800-2011, http://www.systemicpeace.org/polity/polity4.htm, February 2013.
activity and violent extremism.22These environments create grievances that motivate individuals to take up arms, they provide fodder to recruit new members, and they also lower the cost of joining the militant group when there are fewer opportunities for employment. These mechanisms are expected to work in the opposite direction as well: greater economic prosperity should reduce popular grievances and subsequently the support for militant organizations, raising the likelihood that they forgo their violent activities. According to this logic, GDP per capita should be negatively corre- lated with militant organizational survival. However, a higher GDP per capita could also be correlated with decreased group longevity since more developed states (with higher GDPs per capita) are expected to be increasingly able to successfully com- bat violent extremism. These states should have more money and more resources to devote to counterterrorism and policing, increasing the odds that groups are quickly defeated. Similarly, Fearon and Laitin (2003) argue that states with higher GDPs will correlate with greater road penetration, ultimately meaning that rural areas are more easily within the reach of the central government. This is similar to the pro- posed effect of regime durability though there is still some variation between the two. Finally, Blomberg et al (2011) conceptualize GDP in another way: as they write, “GDP provides a target-rich environment,” and it “may also provide more skilled recruits,” ultimately making groups more resilient. Though this view is in the mi- nority it nonetheless provides a compelling explanation for why militant groups in more prosperous countries could actually persevere. Data for GDP per capita and population come from the Penn World Tables.
The final variable related state characteristics that I also included is a logged estimate of the percent of a country’s terrain that is mountainous. Scholars have used
22. James A. Piazza, “Poverty, minority economic discrimination, and domestic terrorism,”Journal of Peace Research 48, no. 3 (May 2011): 339–353; Andreas Freytag et al., “The origins of terrorism: Cross-country estimates of socio-economic determinants of terrorism,”European Journal of Political Economy 27 (2011): S5–S16; Bueno de Mesquita,“Conciliation, Counterterrorism, and Patterns of
this measure in recent years as a means of proxying for the conditions that favor insurgencies though the logic should equally apply to terrorists and other types of militants just as well.23 Rough terrain provides terrorists, guerrillas, and insurgents a means of evading detection and a space to more easily train and plan operations. States will find it difficult to police and monitor these areas, so when they exist in high quantities they should bolster the survivability of nonstate actors.
Third, group competition might make it harder and less likely for individual groups to survive. A more competitive environment could lead to a group’s early demise since resources and recruits are more difficult to come by, and also because direct intergroup competition might weaken personal commitment and cause individuals to rethink their dedication to the group. I approximate these dynamics with several measures: first, with the number of active groups in a country-year, second, with whether or not a militant organization is the “top dog” in a single year, and third, whether or not the Cold War is ongoing.
The number of active militant group in a given country-year is intended to capture the possibility of group outbidding—a dynamic that is theorized to occur when multi- ple nonstate actors exist in the same environment. Since most resources are zero-sum, groups are necessarily in competition with one another for recruits, popular support, etc. Bloom (2005) was the first to theorize the effects of outbidding and she argues that groups will compete both with more and more severe violence to “gain credibil- ity and win the public relations campaign.”24 The outbidding phenomenon is most commonly used to explain the level of violence in a particular country but Young and Dugan extend this logic to group survival as well. As they write, “While a competitive environment may encourage terrorism, it also likely dampens group survival as other organisations drain the pool of potential recruits. Thus, similar to interest groups
23. James D. Fearon and David D. Laitin, “Ethnicity, insurgency, and civil war,”American political science review 97, no. 1 (2003): 75–90.
operating in competitive environments, some will succeed and some will fail.25” The presence of multiple groups should correlate with outbidding, and outbidding should then produce resource scarcity which drives down the ability of groups to survive.
Young and Dugan also introduce the idea that being a “top dog”—the most active organization in a country—will correlate with group resiliency. Using an analogy to firms in the marketplace, groups that already have a strong base of support should be more established and consequently more resilient than others, making it less likely that they die out in any particular year. While there are a few problems with this logic—for instance, does being the top dog also attract the most government attention which can actually lower the odds of survival?—the theoretical link between top dog status and group survival, at least in a single year, seems plausible and they find strong supporting evidence in their own research.
4
Empirical Results I: Organizational Survival
and Factional-Multidimensional Schisms
Table 4 displays the results from the first analysis of organization longevity that includes important group and country characteristics and whether or not a group was a factional or multidimensional splinter. The next analysis will disaggregate factional splits into the precise reasons causing the schism but this is a baseline test of my model. Each successive model in this table adds additional variables to the equation until Model 5, which includes everything.
It is important to note what these coefficients mean since they are different from ordinary least squares. Since the dependent variable is the time until group failure the model is estimating the hazard rate—or in other words the likelihood that a group fails in a given year. A coefficient greater than one means that the hazard rate
increases and a group is more likely to fail.Coefficients below one imply that a group is less likely to fail and instead more likely to survive.
Model 1 is the most basic including only a categorical variable indicating whether a group is a nonsplinter, factional splinter, or multidimensional splinter, as well as year and country fixed effects. The results suggest that there is a significant differ- ence in longevity between these groups. The significant and positive coefficient on multidimensional splinters suggests that they are much more likely to fail in a given year than nonsplinter groups, while the coefficient below one on factional splinters implies they areless likely to fail. Both of these estimates are statistically significant at the 10% level and they are significant in relation to each other as well (Wald test, p=.000). This provides evidence of the baseline differences between groups according to how they form, conforming to the initial expectations of my theory. However, more covariates are needed to truly understand if this relationship is robust.
More variables are added in Models 2 and 3. Model 2 adds dummy variables for different group identities and goals: whether or not a group is nationalist-separatist, communist-socialist, religious, anarchist, or leftist. Model 3, on the other hand, adds country characteristics to the equation: whether a group is operating in an autocracy or democracy,26 the durability of the current regime, logged GDP per capita, logged population, and the logged percentage of mountainous terrain. The coefficients be- low one on communist-socialist groups implies that these organizations last longer than others—the omitted categories such as environmental, racist, right-wing, anti- globalization, etc. On the other hand, the model finds that leftist groups fail at a much greater rate.
With regards to country characteristics, there is mixed evidence that groups in au- tocracies tend to last longer, lowering the hazard of failure in any given year. Although
26. The omitted category is thus anocracies, or countries that are a mix of autocratic and demo- cratic structures but do not fall entirely into either camp.
Table 5.2. Cox Proportional Hazards Model of Organizational Longevity
(1) (2) (3) (4) (5)
Baseline Group Country Environment Full Factional Splinter 0.682∗ 0.625∗∗ 0.673∗ 0.684∗ 0.622∗∗ (0.143) (0.133) (0.143) (0.143) (0.134) Nonfactional Splinter 1.619∗ 1.917∗∗ 1.586∗ 1.726∗∗ 1.908∗∗ (0.454) (0.625) (0.440) (0.476) (0.581) Nationalist-Separatist 0.803 0.808 (0.142) (0.140) Communist-Socialist 0.572∗∗∗ 0.585∗∗ (0.120) (0.123) Religious 0.855 0.878 (0.197) (0.201) Anarchist 0.644 0.648 (0.393) (0.391) Leftist 2.929∗∗∗ 2.826∗∗∗ (0.745) (0.718) Autocracy 0.417∗ 0.498 (0.186) (0.242) Democracy 1.031 1.161 (0.281) (0.329) Regime Durability 0.989 0.991 (0.009) (0.009)
GDP Per Capita (log) 0.939 0.794
(0.326) (0.296) Population (log) 1.401 1.390 (0.304) (0.305) Pcnt. Mountainous (log) 1.308 0.940 (0.958) (0.711) Terrorist Competitors 1.019 1.014 (0.015) (0.015) Lead Organization 0.915 0.972 (0.156) (0.164) Observations 2912 2912 2912 2912 2912
Standard errors in parentheses (clustered by group). Coefficients reported as hazard ratios. Country and year fixed effects included in every model.
the coefficient estimate is nearly similar in Models 3 and 5, it is only significant in the former. Neither the democracy dummy, the measure of regime durability, nor any other country-level factor produces a meaningful effect. This implies that one cannot reject the null hypotheses that there is no relationship between group duration and regime durability, GDP per capita, population size, and the proportion of rough ter- rain. In terms of the primary independent variables, however, there continues to be a strong relationship between factional/multidimensional formation and the likelihood of failure.
Model 4 introduces variables relating to a militant group’s operating environment: in other words, certain conditions that affect how it conducts itself. This focuses on the number of other active militant groups in its home state and whether or not it is the lead organization in a single year (conducting more than 50% of attacks). These variables fail to generate any meaningful effect and their inclusion barely changes other estimates in the model.
Finally, Model 5 includes every variable from the first four analyses, representing the most complete model yet. Several things stand out: first, the dummy indicators for communist-socialist and leftist identities maintain their significant effects, with communist-socialist groups lasting longer and leftist dying out quicker. The dummy for autocratic countries, however, loses significance and is no longer meaningful at the 10% level. And, notably, the estimates for factional and multidimensional dummies actually increase in significance to under 5%. As before, groups forming factionally have a much lower hazard of failure in any given year, and groups forming multidi- mensionally experience a much higher hazard and overall, a lower chance of survival. These results strongly support my hypothesis that factional splinters tend to be more durable than are similar multidimensional splinter groups. The dummy indi- cators for groups that form factionally and multidimensionally surprising reach sta- tistical significance in every model despite which variables are introduced. This is
Figure 5.2. Coefficient plot: duration of factional and multidimensional splinters.
testament to the robustness of this relationship between the consistency of internal preferences and the ability for militant organizations to endure. It is also interesting that both indicators are significant in reference to nonsplinter organizations and not just to each other. This suggests that the effects of variation in splinter formation are not only limited to differences between splinters, subtly influencing their ability to survive, but rather that it produces a larger, more general effect that is noticeable even amongst the broader population of militant organizations.