Each drone attack amounts to even more terrorist attacks in the future
Anouk S. Rigterink, 18, Rigterink S., Anouk. (Postdoctoral Research Fellow at the University of Oxford and has Ph.D. at the London School of Economics and Political Science) “The Wane of Command∗: Evidence on drone strikes and control within terrorist organizations.” Department of Economics: University of Oxford. 10-30-18. Accessed 7-1-19. (Pembrokehill-MLT)
Targeting terrorist leaders has become a commonly used US counterterrorism policy since 9/11. According to the US National Strategy for Counterterrorism 2018, targeting key terrorists remains the number one priority action. An underlying goal of this policy is to undermine control within terrorist organizations: the ability of individuals higher up the hierarchy – or more central to the organization – to determine what others in the organization do. This policy is primarily implemented using armed drones, Unmanned Aerial Vehicles, which can surveil and kill targets. The US now owns hundreds of these armed drones, and 28 other countries have acquired weaponized drones in the last ten years. Targeted killing of terrorist leaders is also referred to as “cutting off the head of the snake”, implying that if one does so, the body will die. The snake analogy portrays the terrorist group as a single organism: a unitary actor. Many theoretical models similarly considering terrorist groups as unitary actors straightforwardly predict that targeting terrorist leaders, and undermining control within terrorist organizations, decreases terrorism (Sandler and Arce, 2003; Sandler and Siqueira, 2006; Powell, 2007; Bandyopadhyay and Sandler, 2011). However, if we consider terrorist organizations as non-unitary actors, predictions are less straightforward. Models that consider terrorist groups as organizations subject to collective action and principal-agent problems, which I will call problems of control for short, suggest that undermining control within terrorist groups may not be effective, or may even backfire. This leaves an empirical question, which this paper addresses: how does counterterrorism which undermines control within terrorist organizations affect terrorism? Specifically, this paper investigates how drone strikes killing terrorist leaders affect terrorist attacks. To investigate this empirically, this paper exploits a natural experiment provided by drone strikes ‘hitting’ and ‘missing’ terrorist leaders in the Federally Administered Tribal Areas (FATA) of Pakistan, which is inspired by Jones and Olken (2009). I construct a new dataset of drone attempts on terrorist leaders’ lives, executing several cross-checks to safeguard data quality. This dataset captures variation across time and terrorist organizations. I argue that conditional on a drone attempt to kill a terrorist leader, drone ‘hits’ and ‘misses’ are quasirandom. Drone hits and misses are not statistically significantly different from one another on an extensive range of characteristics, including pre-trends in terrorist violence. Narratives of why drones miss also suggest that misses are largely driven by chance. This enables a difference-in-difference design, investigating changes in attacks by a terrorist organization before and after a drone hit on its leader, compared to before and after a miss. Results indicate that a drone hit on a terrorist organization’s leader is associated with an increase in the number of terrorist attacks by this organization, compared to a miss. Estimates indicate that this increase amounts to 29 additional terrorist attacks in total across the world in the six months after a drone hit, an increase of 43% compared to the six months after a miss. This result cannot be explained by terrorist organizations ‘speeding up’ the timing of attacks, or by a decrease in the lethality of attacks. Results are robust to estimating the main test statistic using randomization inference, including leader-fixed effects, alternative aggregations of terrorist groups, controlling for Pakistani military action and peace agreements between the terrorist group and the Pakistani government, and using alternative econometric specifications. Several game-theoretical models provide explanations for the results obtained. A first family of models considers problems of control. A principal-agent model by Shapiro (2013) suggests that leader killing, by increasing the costs for the leader to control his operatives, could lead to increased terrorist violence if operatives have a greater preference for (indiscriminate) violence than the leader. Modelling terrorist violence as a public good to terrorists, Enders and Jindapon (2010) suggest that terrorist networks strategically decrease the number of network connections in response to targeted leader killing, leading individual nodes to increase their efforts to commit attacks as they can no longer free-ride on the efforts of others in the network. Other families of models provide alternative explanations of why targeted leader killing may increase terrorist violence: terrorist groups may respond to a loss in capacity by substituting a few high-capacity attacks with many low-capacity attacks, pro-active counterterrorism policies that result in civilian casualties may create backlash if they spur terrorist recruitment, and terrorist organizations may commit more terrorist attacks after their leader has been killed to signal strength. Further analysis suggests that problems of control explain the main result better than alternative theoretical mechanisms. The observed increase in terrorist attacks is primarily driven by attacks that leaders of terrorist organizations do not favour: attacks on civilian and private targets – which terrorist leaders repeatedly instruct their followers to avoid – and attacks by its members that the organization does not publicly claim responsibility for. The observed effect is also stronger for terrorist organizations and leaders that rely more strongly on central control. Furthermore, a drone hit is associated with proxies for network breakdown: an increase in group splintering and infighting. Simultaneously however, results suggest that a drone strike on one group’s leader is associated by an increase in terrorist attacks by other groups, affiliates in its network. A considerable number of existing papers investigate empirically the impact of targeted killing of leaders of terrorist organizations, yet results are mixed. Some authors conclude that targeted leader killing is effective, either because it speeds up the decline of terrorist organizations (Price, 2012) or diminishes the number or intensity of terrorist attacks (Jaeger and Paserman, 2009; Johnston, 2012). Others conclude that it has no effect (Jordan, 2009; Mannes, 2008; Hafez and Hatfield, 2006), or even an adverse effect – either on terrorist organizations in general or on particular categories of terrorist organizations (Kaplan et al., 2006; Jordan, 2009; Mannes, 2008; Abrahms and Potter, 2015; Abrahms and Mierau, 2017).