Torture their Bodies and Ignore my s : Policy Demand following Terror Attacks

Transcription

1 Torture their Bodies and Ignore my s : Policy Demand following Terror Attacks Marco Giani October 7, 2017 Abstract To assess the causal effect of terrorism on the support for counterterror policies that prioritize security over liberty, I exploit the coincidence that the London Bombing (7/7/2005) and the Orlando shooting (12/6/2016) occurred during the fieldwork period of respectively the British Social Attitudes Survey and the General Social Survey, allowing to construct two equally designed block randomized experiments. Whether terrorism increases the support for security-enhancing policies depends on the specific dimension of civil liberties that is threatened. I find strong, mixed and no evidence that terrorsim increases support for securityenhancing policies that reduce, respectively, human rights, freedom of speech and privacy. London School of Economics and Political Science, Department of Government. m.giani@lse.ac.uk 1

2 1 Introduction Arbitrary arrest, indefinite detention without trial, rendition, suspension of habeas corpus, even torture - who would have thought such things possible?, asked The Economist in Alike total war, terrorism calls for revising the complex balance between the need of enhancing security and that of limiting restriction of civil rights in favor of the former priority. Unlike wars, however, terrororism is potentially undless. For this reason, while the suspension of these fundamental - but not absolute - rights is viewed as necessary, policymakers must weigh competing issues when deciding what and how to implement security policy, balancing these concerns in order to achieve certain broader security objectives without unnecessarily and disproportionately infringing human rights (The RAND Corporation, 2010). Alike experts, few academics doubt that restoring security is top priority under a salient terror threat. Many, however, worry that western democracies might be sacrificing too many civil liberties for nothing else than an apparent security enhancement. Bueno de Mesquita and Dickson (2007) show that counterterror policies that harm the whole aggrieved population, while increasing security in the short-term, are likely to favor proselytism and hence terrorist recruitment. Dragu (2011) shows that the limitations on citizen s privacy often implemented following terror attacks may reflect a policy bias of executive agencies, rather than a credible attempt to reduce the risk of future attacks. Similarly, Dragu (2017) shows that limiting free speech in correspondence to a salient terror threat may be counterproductive, since while constraints on free speech makes the terrorist propaganda more costly, the institution of free speech itself constraints the emergence of terrorism in the first place. These key contributions yield two important conclusions concerning the desirability of counterterror policies. Firstly, they demonstrate that most policy tools typically undertaken to fight against terrorism reduce are liberty-reducing for sure and security-enhancing only with some probability, a finding that the empirical literature supports (Daxecker and Hess, 2013; Benmelec et al., 2015; Daxecker 2017). Observed counterterror policy-making based on repression may not be based on evidence, but rather driven by electoral incentives. To address this possibility, this paper seeks to answer the following question: Does public opinion support security-enhancing and liberty-reducing counterterror policies? A related observation is that whereas security in this context simply refers to the minimization of the likelihood of future attacks, the meaning of liberty is diversified. When counterterror policies that allow for extrajudicial practices against the aggrieved population, liberty refers to the respect of the human rights of terrorist suspects. Similarly, when counterterror policies that allow for spying on or phone conversations, liberty means privacy. Finally, when counterterror policies that ban demonstrations and the free expression of any belief upon security concerns, liberty means freedom of speech. One main contribution of this paper is to tackle the following question: Is public opinion equally willing to sacrifice human rights, privacy and free speech? 2

3 I address the highlighted questions by analyzing and comparing the outcomes of two separate but equally designed natural experiments. I exploit the fact that two major terror episodes, the London Bombing on July 7, 2005 and the Orlando Shooting on June 12, 2016, fall inside the fieldwork period of, respectively, the British Social Attitude Survey (GSAS) and the US General social survey (GSS). By relying on pre-treatment non-parametric matching techniques in the spirit of Iacus et al. (2012), I mimic a block randomized experiments and compute sample average treatment effects on the treated within the fifteen days following the two attacks. Both surveys exceptionally include valid items about the support for counterterror policies that are intended to minimize the risk of future attacks at the cost of harming individual freedom on three dimensions: human rights of terrorist suspects (e.g arbitrary detention and torture), privacy (e.g tapping phone and conversations) and free speech (banning free speech or demonstrations). I find that both the UK and the US public, while reluctant to mandate the government to intrude privacy, increase support for security enhancing policies that harm human rights and free speech. While I am not aware of any other paper that share my premises and focus, the literature on 9/11 does provide some evidence in favor of the hypothesis that public opinion backs security-enhancing and libertyreducing counterterror policies (see e.g Davis and Silver, 2004; Huddy et al., 2005; Heterington and Suhay, 2011). The main advantage of my design with respect to the aforementioned studies is its high internal validity. By focusing on the very short run effect of terror attacks on public opinion, as well as by appropriately pruning data so as to mimic a block randomized experiment, I minimize attribution issues. The ability to apply the same design to two comparable attacks increase the external validity of my outcomes. A similar design is used by Finseraas (2011) and Legewie (2013), who however study the effect of terrorism on the preferences for immigration and focus on the transnational effects of a single attack. My design parallels the one in Giani (2017), where again the main focus is placed on the relationship between terrorism and immigration in France following the Charlie Hebdo shooting. In contrast to these studies that only focus on one case, I compare two different but comparable events. The London Bombing and the Orlando shooting caused respectively 56 and 50 casualties and happened under a leftwing incumbent in two countries that share similar core values. 2 Empirical analysis 2.1 Data I use the London Bombing, which occurred on July 7, 2005, and the Orlando shooting, which occurred on June 12, 2016, as instrument for an exogenous increase in terror threat. These attacks share a very close magnitude of casualties, 56 in the former and 50 in the latter case. Both of them occurred under a left-wing 3

4 incumbent and far apart from episodes of similar importance. 1 The former episode falls inside the fieldwork period of the British Social Attitude Survey (BSAS) whereas the latter falls inside the fieldwork period of the US General Social Survey. Crucially, both surveys exceptionally include suitable survey items concerning support for counterterror policies that are supposed to enhance security at the expenses of civil rights and privacy. Survey items are the ideal ones, and were chosen precisely to tackle the research question I am investigating without expecting of a major terror attack to hit during the data collection period. 2 Of course, these items are not exactly equal. There is, however, a substantial area of overlap that allows to check the consistency of outcomes across two equally designed case studies. Moreover, available answers range from full disagreement (1) to full agreement (4) in both the UK and the US survey, favoring the comparison of outcomes. The 2005 round of the British BSAS incorporates a set of survey items that explicitly refer to policies that minimize the risk of future attack. One further advantage of these items is that wording effects are taken care of by a detailed pre and post formulation. Survey items include specific questions on the support for extrajudicial practices, including arbitrary detention and torture, limited privacy including stricter and phone control, and censorship, including limits on demonstrations and free speech. In a similar manner, the 2016 round of the American GSS incorporates novel survey items that no previous round. These include again extrajudicial practices that harm the human rights of suspects, and limited privacy. The Appendix details each of these survey items for both countries. 2.2 Identification issues Figure 1 plots the daily distribution for each case study. I am going to focus on the effect of terror attacks on public opinion in the very short run. Control and treated units may, in both cases, be unbalanced on socioeconomic covariates. This would, in turn, jeopardize the validity of my results. Following the application of Giani (2017) based on the seminal contribution of Iacus et al. (2012), I control for both ex ante imbalance, by relying on non-parametric pre-treatment matching using coarsened exact matching (Blackwell et al., 2009), and ex-post, by adding controls to the OLS regression that estimates the (sample) treatment effect on the treated. Control variables include income (0-5 in the UK and 0-10 in the US), education (0-7), sex (0-1), age, squared age, whether the respondent lives with children (0-1) and whether she has immigration background (0-1). Finally, accounting for the fact that responses to terror attacks might be mitigated by the incumbent s partisanship (Merolla and Zechmeister, 2013), I add a dummy taking value one if the respondent voted Labour or Obama during the 2001 election in the UK or the 2012 election in the US. Pre-treatment matching 1 Of course, this is not to deny that other contextual variables, including the role of media (Gadarian, 2010) and more generally the framing of the attack (Haider-Markel, et al., 2006), might be key in driving public opinion responses. 2 Detailed survey items on the support for counterterror policies that entail a security vs. liberty trade-off are not available in any previous round of the GSS. SImilarly, they are not available in other surveys, including the European Social Survey or the World Value Survey. 4

5 (a) UK, On July 7, four suicide bombers of Al-Quaeda denonated three bombs in public transports, killing 56. (b) US, On June 12, a member of Isil opened fire at a gay nightclub in Orlando, killing 50. Figure 1: Data collection around terror attacks. The red spike is the date at which each attack took place. reveals that the UK (US) control and treated groups are unbalanced on education, age and whether the respondent has children living at home (age only). I match control and treated units exactly on education, while coarsening age in five years interval before matching them. As such, I prune 10 control and 17 treated units in the UK, whereas I match all units in the US. The low number of ouliers show that control and treated units share close covariates in both cases. 2.3 Estimation I compute the sample average treatment effect (SATT) according to three different model specifications: a (i) naif model, which only test the mean-difference through an OLS regression that does not control for imbalance; a (ii) standard model, which controls for ex post sample imbalance by adding controls to the regression, but does not apply pre-treatment matching and a (iii) full model, which applies matching weights in order to prune bad matches, controlling for both ex ante and ex post imbalance. I focus on the very short run of 15 days after the attack. In order to assess the dynamic path of the hypothesized effects of terrorism on attitudes, the full model is also estimated after 30 days. By focusing on the very short term, I assess public reaction when the threat is salient and prior to the policy implementation stage. As such, attribution issues are minimized. On the other hand, my approach leaves the researcher with an amount of data that makes it impossible to focus on some research questions, such as the mitigating role of ideological (Kossowska et al., 2011) or psychological (Stevens and Vaughan-Williams 2016) orientations that scholars in the same field have assessed following other approaches. 5

6 Interval of time 15 days 15 days 15 days 30 days Model specification Full Standard Naif Full Arbitrary Detention (1-4) SE (.108) (.098) (.102) (.101) Deny trial (1-4) SE (.113) (.119) (.122) (.105) Tag suspect (1-4) SE (.083) (.081) (.085) (.074) Torture Suspect (1-4) SE (.092) (.087) (.089) (.084) Compulsory ID (1-4) SE (.116) (.112) (.113) (.106) Phone/ surveillance (1-4) SE (.113) (.119) (.122) (.105) Ban Protest (1-4) SE (.108) (.100) (.103) (.102) Ban Free Speech (1-4) SE (.108) (.102) (.103) (.101) N. Obs Bold: significant at.05. Outputs estimated by OLS. Controls include income, education, age, squared age, sex, a dummy equal to one if the unit has children, immigration background, voted Obama in 2012,egion fixed effects. Table 1: Effect of London Bombing on support for counterterror policies: full model and robustness checks. 3 Results Full results for the UK case study are provided in Table 1. Public support for counterterror policies that harm the human rights of terrorist suspects increased following the London Bombing. According to the full model specification, support for arbitrary detention and torture on terror suspects increase respectively by 7, 2% and 7, 1%. Similarly, the UK public backs restrictions on freedom of speech, while it does not increase support for banning demonstrations. On the other hand, there is no change in the support for controlling or phone conversation. All other items display non-significant increases. Apart from a policy that makes ID compulsory for adult, my outcomes are robust to each model specification and time interval. Full results for the US case study are provided in Table 2. I confirm mild support for H 1 ; outcomes are, overall, close to those found for the UK case. When the counterterror policy tool are street patrolling and arbitrary detention, that move in a very correlated manner, the increase in policy support is respectively 6, 2% and 6, 5% in the full model after 15 days and is significant at.05. Instead, I find a non-significant result of terror attacks on support for phone, video and surveillance. In the two latter cases, the sign is negative and non-significant. Finally, support for government collection of information about nationals or foreigners without their knowledge does not change following terror attacks. 6

7 Interval of time 15 days 15 days 15 days 30 days Model Specification Full Standard Naif Full Street patrolling (1-4) SE (.103) (.102) (.100) (.091) Arbitrary Detention (1-4) SE (.106) (.106) (.101) (.090) N. Obs Phone Surveillance (1-4) SE (.102) (.101) (.100) (.090) N. Obs Video Surveillance (1-4) SE (.098) (.097) (.092) (.088) Surveillance (1-4) SE (.093) (.092) (.089) (.082) N. Obs Spy Nationals (0-10) SE (.106) (.105) (.099) (.094) Spy Foreigners (0-10) SE (.101) (.101) (.095) (.088) N. Obs Bold: significant at.05. Outputs estimated by OLS. Controls include income, education, age, squared age, sex, a dummy equal to one if the unit has children, immigration background, voted Obama in 2012,egion fixed effects. Table 2: Effect of Orlando Shooting on support for counterterror policies: full model and robustness checks. 4 Conclusion Counterterror policies that harm the human rights of terrorist suspects and curtails free speech match popular support, thereby potentially increasing the chances of re-election. Instead, counterterror policies that harm privacy do not match policy demand. These conclusions help refining the boundaries of support for counterterror policy initiatives. They are based on two comparable case studies in which a terror event is studied as a natural experiment. Two cases are clearly not enough to draw strong conclusions, but the norm in experimental and quasi experimental studies is just one case. In this sense, my outcomes have larger external validity than previous works sharing a similar design. That public opinion be willing to sacrifice core liberties for presumably enhanced security is still a hypothesis than a stylized fact. Given its centrality in the public debate and in actual policy-making, and given the unlucky increase in terror episodes, future research should replicate and expand this analysis. Then latter is limited in several extents. Focusing on the very short run minimizes attribution issues, yet it also cuts observations. As such, further interesting research questions, such as the interaction of socioeconomic or political identities with the treatment, would have to be carried out with little statistical power. Furthermore, since the geographical information about respondents is scarce, I cannot test hypotheses that differentiate respondents as a function of their exposure to the attack, as in Davis and Silver (2004). Finally, to the extent that the causal challenge cannot be completely overcome, I abstract from the important role of political leaders (Merolla and Zechmeister, 2009) and the exposure to threat (Davis and Silver, 2004) in 7

8 driving public opinion. With its advantages and disadvantages, my causal inference approach complements other approaches such as lab or survey experiments, cross country and qualititative analyses. References [1] Benmelec, Efraim, Claude Berrebi, and Esteban F. Klor Counter-Suicide Terrorism: Evidence from House Demolitions. Journal of Politics 77 (1): [2] Blackwell, Matthew, Stefano Iacus, Gary King, and Giuseppe Porro CEM: Coarsened Exact Matching in Stata, The Stata Journal 9: [3] Bueno de Mesquita, E., and Dickson, E.S., The Propaganda of the Deed: Terrorism, Counterterrorism, and Mobilization, American Journal of Political Science 51(2): [4] Daxecker, Ursula E. and Michael L. Hess Repression Hurts: Coercive Government Responses and the Demise of Terrorist Campaigns. British Journal of Political Science 43 (3): [5] Daxecker, Ursula E Dirty Hands: Government Torture and Terrorism, Journal of Conflict Resolution 61(6): [6] Davis, Darren W. and Brian D. Silver Civil Liberties vs. Security: Public Opinion in the Context of the Terrorist Attacks on America, American Journal of Political Science 48(1): [7] Dragu, Tiberiu Is There a Trade-off Between Security and Liberty? Executive Bias, Privacy Protections, and Terrorism Prevention, American Political Science Review 105(1): [8] Dragu, Tiberiu The Moral Hazard of Terrorism Prevention, Journal of Politics 79(1): [9] Finseraas, Henning, Niklas Jakobsson and Andreas Kotsadam Did the Murder of Theo van Gogh Change Europeans Immigration Policy Preferences? Kyklos 64(3): [10] Gadarian, Shana K., The Politics of Threat: How Terrorism News Shapes Foreign Policy Attitudes, Journal of Politics 72(2): [11] General Social Survey, [12] Giani, Marco Don t Ban the Muslims: Public reaction to Terror Threat, WP. Accessed here: marcogiani.altervista.org/wp-content/uploads/2017/08/mergednew.pdf. [13] Haider-Markel, Donald P., Joslyn; Mark R. and Mohammad Tarek Al-Baghal Can We Frame the Terrorist Threat? Issue Frames, the Perception of Threat, and Opinions on Counterterrorism Policies, Terrorism and Political Violence 18(4). 8

9 [14] Hetherington, Marc J. and Elizabeth Suhay Authoritarianism, Threat, and Americans Support for the War on Terror, American Journal of Political Science 55(3): [15] Huddy, Leonie, Stanley Feldman, Charles Taber and Gallya Lahav Threat, Anxiety, and Support of Antiterrorism Policies, American Journal of Political Science 49 (3) : [16] Iacus, Stefano, Gary King and Giuseppe Porro Causal Inference Without Balance Checking: Coarsened Exact Matching, Political Analysis 20: [17] Kossowska, Małgorzata, Trejtowicz, Mariusz, de Lemus, Soledad, Bukowski, Marcin, Van Hiel, Alain and Robin Goodwin Relationships between Right-wing Authoritarianism, Terrorism Threat, and Attitudes towards Restrictions of Civil rights: A Comparison among four European Countries, British Journal of Political Science 102(2): [18] Legewie, Joscha Terrorist Events and Attitudes toward Immigrants: A Natural Experiment, American Journal of Sociology 118(5): [19] Merolla Jennifer L. and Elizabeth J. Zechmeister Terrorist Threat, Leadership, and the Vote: Evidence from Three Experiments, Political Behavior 31(4): [20] Merolla Jennifer L. and Elizabeth J. Zechmeister Evaluating Political Leaders in Times of Terror and Economic Threat: The Conditioning Influence of Politician Partisanship, Journal of Politics 75(3): [21] Mondak, Jeffery J. and Jon Hurwitz Examining the Terror Exception: Terrorism and Commitments to Civil Liberties, Public Opinion Quarterly (76)2: [22] Robinson, Neil, Potoglou, Dimitris, Woo Kim, Chong, Burge, Peter, Warnes, Richard Security, At What Cost? Quantifying people s trade-offs across liberty, privacy and security, Technical report RAND Europe. Accessed here: rand.org/content/dam/rand/pubs/technical_reports/2010/rand_tr664.pdf. [23] Stevens, Daniel and Vaughan-Williams, Nick Citizens and Security Threats: Issues, Perceptions and Consequences beyond the National Frame. British Journal of Political Science 46(1): [24] The Economist, Civil liberties under threat: The real price of freedom. September 20, Accessed here ( 9

10 Appendix: Survey Items British Survey of Social Attitudes (UK, 2005). Extrajudicial practices: A number of measures have been suggested as ways of tackling the threat of terrorism in Britain. Some people oppose these because they think they reduce people s freedom too much. Others think that the reduction in freedom is a price worth paying. For each of the measures I mention, please say which of the views on this card comes closest to your own: (i) Allowing the police to detain people for more than a week or so without charge if the police suspect them of involvement in terrorism; (ii) denying the right to a trial by jury to people charged with a terrorist-related crime; (iii) putting people suspected of involvement with terrorism under special rules, which would mean they could be electronically tagged, prevented from going to certain places, or prevented from leaving their homes at certain times; (iv) Torturing people held in British jails who are suspected of involvement in terrorism to get information from them, if this is the only way this information can be obtained. Limited privacy: A number of measures have been suggested as ways of tackling the threat of terrorism in Britain. Some people oppose these because they think they reduce people s freedom too much. Others think that the reduction in freedom is a price worth paying. For each of the measures I mention, please say which of the views on this card comes closest to your own: (i) Following people suspected of involvement with terrorism, tapping their phones and opening their mail; (ii) having compulsory identity cards for all adults. Free speech: A number of measures have been suggested as ways of tackling the threat of terrorism in Britain. Some people oppose these because they think they reduce people s freedom too much. Others think that the reduction in freedom is a price worth paying. For each of the measures I mention, please say which of the views on this card comes closest to your own: (i) Banning certain peaceful protests and demonstrations; (ii) banning certain people from saying whatever they want in public. General Social Survey (US, 2016). Extrajudicial practices: Suppose the government suspected that a terrorist act was about to happen. Do you think the authorities should have the right to: (i) Detain people for as long as they want without putting them on trial; (ii) stop and search people in the street at random. Limited privacy: Suppose the government suspected that a terrorist act was about to happen. Do you think the authorities should have the right to: (i) tap people s telephone conversations. I have some questions about civil liberties and public security. Do you think that the American government should or should not have the right to do the following: (ii) Keep people under video surveillance in public areas; (iii) Monitor s and any other information exchanged on the Internet. Some people think that governments should have the right to take certain measures in the name of national security. Others disagree. Do you think that the American government should or should not have the right to do the following? (iv) Collect information about anyone living in America without their knowledge; (v) Collect information about anyone living in other countries without their knowledge. (vi) Here is a scale from 0 to 10 where 0 is "All government information should be publicly available, even if this means a risk to public security" and 10 is "public security should be given priority, even if this means limiting access to government information". 10