¹ White, Robert W., ‘From Peaceful Protest to Guerrilla War: Micromobilization of the Provisional Irish Republican Army’, American Journal of Sociology, 94 (1989), 1277–302. p. 1292.CrossRefGoogle Scholar

² LaFree, GaryDugan, Laura and Korte, Raven, ‘The Impact of British Counterterrorist Strategies on Political Violence in Northern Ireland: Comparing Deterrence and Backlash Models’, Criminology, 47 (2009), 17–45CrossRefGoogle Scholar

Sanchez-Cuenca, Ignacio and Calle, Luis de la, ‘Domestic Terrorism: The Hidden Side of Political Violence’, Annual Review of Political Science, 12 (2009), 31–49CrossRefGoogle Scholar

³ White, ‘From Peaceful Protest to Guerrilla War’.

⁴ Guelke, Adrian, ‘The Northern Ireland Peace Process and the War Against Terrorism: Conflicting Conceptions?’, Government and Opposition, 42 (2007)CrossRefGoogle Scholar

⁵ Jones, Seth G. and Libicki, Martin C., How Terrorist Groups End: Lessons for Countering Al Qa'ida (Santa Monica: Rand Corporation, 2008)Google Scholar

⁶ Blomberg, S. BrockEngel, Rozlyn C. and Sawyer, Reid, ‘On the Duration and Sustainability of Transnational Terrorist Organizations’, Journal of Conflict Resolution, 54 (2009), 303–30 Google Scholar

Blomberg, S. BrockGaibulloev, Khusrav and Sandler, Todd, ‘Terrorist Group Survival: Ideology, Tactics, and Base of Operations,’ Public Choice, 149 (2011)CrossRefGoogle Scholar

Cronin, Audrey Kurth, How Terrorism Ends: Understanding the Decline and Demise of Terrorist Campaigns (Princeton: Princeton University Press, 2009)Google Scholar

⁷ Only studies by Jones and Libicki and Blomberg, Gaibulloev and Sandler account for group-level characteristics. Yet Jones and Libicki's regression analysis may be biased because their empirical analysis includes only groups that ended their campaigns within their time frame. Blomberg et al. is the only systematic study that controls for group characteristics. Jones and Libicki, How Terrorist Groups End; Blomberg, Gaibulloev and Sandler, ‘Terrorist Group Survival’.

⁸ Blomberg, Gaibulloev and Sandler, ‘Terrorist Group Survival’.

⁹ Jones and Libicki, How Terrorist Groups End; Blomberg, Gaibulloev and Sandler, ‘Terrorist Group Survival’.

¹⁰ Jones and Libicki, How Terrorist Groups End.

¹¹ Blomberg, Engel and Sawyer, ‘On the Duration and Sustainability of Transnational Terrorist Organizations’; Jones and Libicki, How Terrorist Groups End; Young and Dugan, ‘Why Do Terrorist Groups Endure?’.

¹² Abrahms, Max, ‘Why Democracies Make Superior Counterterrorists’, Security Studies, 16 (2007), 223–53 CrossRefGoogle Scholar

Walsh, James I. and Piazza, James A., ‘Why Respecting Physical Integrity Rights Reduces Terrorism’, Comparative Political Studies, 43 (2010)CrossRefGoogle Scholar

Davenport, Christian, ‘State Repression and Political Order’, Annual Review of Political Science, 10 (2007), 1–23CrossRefGoogle Scholar

¹³ Sanchez-Cuenca, and de la Calle, , ‘Domestic Terrorism’; LaFree, Dugan and Korte, ‘The Impact of British Counterterrorist Strategies on Political Violence in Northern Ireland’; Donatella Della Porta, ‘Leaving Underground Organizations: A Sociological Analysis of the Italian Case’, in Tore Bjorgo and John Horgan, eds, Leaving Terrorism Behind: Individual and Collective Disengagement (New York: Routledge, 2009), pp. 66–88Google Scholar

Parker, Tom, ‘Fighting an Antaean Enemy: How Democratic States Unintentionally Sustain the Terrorist Movements They Oppose’, Terrorism and Political Violence, 19 (2007), 155–79CrossRefGoogle Scholar

¹⁴ Rashwan, Diaa, ‘The Renunciation of Violence by Egyptian Jihadi Organizations’, in Bjorgo and Horgan, Leaving Terrorism Behind, pp. 113–33Google Scholar

Handelman, Howard, ‘Labor-Industrial Conflict and the Collapse of Uruguayan Democracy’, Journal of Interamerican Studies and World Affairs, 23 (1981), 371–94 CrossRefGoogle Scholar

Lewis, Paul H., Authoritarian Regimes in Latin America: Dictators, Despots, and Tyrants (Oxford, UK: Rowman & Littlefield, 2006)Google Scholar

Testas, Abdelaziz, ‘Determinants of Terrorism in the Muslim World: An Empirical Cross-Sectional Analysis’, Terrorism and Political Violence, 16 (2004)CrossRefGoogle Scholar

¹⁵ We evaluate the possible implications of our argument on terrorist group activity in the robustness test and find support for our main contention when using terrorist events as the dependent variable. We believe that a focus on duration is more helpful, since a temporary increase or decline in attacks as a result of repression cannot tell us whether government coercion contributed to the eventual decline of terrorist groups.

¹⁶ Mesquita, Ethan Bueno de and Dickson, Eric S., ‘The Propaganda of the Deed: Terrorism, Counterterrorism, and Mobilization’, American Journal of Political Science, 51 (2007)CrossRefGoogle Scholar

Kydd, Andrew and Walter, Barbara F., ‘The Strategies of Terrorism’, International Security, 31 (2006), 49–79CrossRefGoogle Scholar

Rosendorff, Peter B. and Sandler, Todd, ‘Too Much of A Good Thing? The Proactive Response Dilemma’, Journal of Conflict Resolution 48 (2004)CrossRefGoogle Scholar

¹⁷ We borrow the terms backlash effects and deterrence effects from LaFree, Dugan and Korte, ‘The Impact of British Counterterrorist Strategies on Political Violence in Northern Ireland’.

¹⁸ Crenshaw, Martha, ‘How Terrorism Declines’, Terrorism and Political Violence, 3 (1991), 69–87CrossRefGoogle Scholar

¹⁹ Mason, T. David and Krane, Dale A., ‘The Political Economy of Death Squads: Toward a Theory of the Impact of State-Sanctioned Terror’, International Studies Quarterly, 33 (1989), 175–98, p. 74.CrossRefGoogle Scholar

²⁰ Gibney, MarkCornett, L. and Wood, Reed, ‘Political Terror Scale 1976–2006’, available at http://www.politicalterrorscale.org/Google Scholar

Cingranelli, David L. and Richards, David L., ‘The Cingranelli-Richards (CIRI) Human Rights Dataset’, available at http://ciri.binghamton.edu/Google Scholar

²¹ Rosendorff and Sandler, ‘Too Much of A Good Thing?’, p. 658.

²² The argument raises the question of why democracies would use repressive strategies in the first place. Research by Bueno de Mesquita and Kydd and Walter suggests one possible explanation. The authors argue that democracies may have to resort to more public and less discriminatory practices when engaging in counterterrorism. Democratic governments are under pressure to ‘do something’ in response to terrorist violence, and less visible strategies can create the perception that the government is not protecting the population. See Ethan Bueno de Mesquita, ‘The Quality of Terror’, American Journal of Political Science, 49 (2005), 515–30; Kydd and Walter, ‘The Strategies of Terrorism’.

²³ Posner, Richard, Countering Terrorism: Blurred Focus, Halting Step (Lanham, Md.: Rowman & Littlefield, 2007)Google Scholar

Tyler, Tom R.Schulhofer, Stephen and Huq, Aziz Z., ‘Legitimacy and Deterrence Effects in Counterterrorism Policing: A Study of Muslim Americans’, Law & Society Review, 44 (2009), 365–402CrossRefGoogle Scholar

²⁴ Donohue, Laura K., The Cost of Counterterrorism: Power, Politics, and Liberty (Cambridge: Cambridge University Press, 2008)CrossRefGoogle Scholar

²⁵ Tyler, Schulhofer and Huq, ‘Legitimacy and Deterrence Effects in Counterterrorism Policing’, p. 380.

²⁶ Huq, Aziz Z.Tyler, Tom R. and Schulhofer, Stephen J., ‘Why Does the Public Cooperate With Law Enforcement? The Influence of the Targets and Purposes of Policing’, Psychology, Public Policy, and Law, 17 (2011), 419–50, pp. 429–31.CrossRefGoogle Scholar

²⁷ LaFree, Gary and Ackerman, Gary, ‘The Empirical Study of Terrorism: Social and Legal Research’, Annual Review of Law and Social Science, 5 (2009), 347–74 CrossRefGoogle Scholar

²⁸ Abrahms, ‘Why Democracies Make Superior Counterterrorists’; LaFree and Ackerman, ‘The Empirical Study of Terrorism’; Walsh and Piazza, ‘Why Respecting Physical Integrity Rights Reduces Terrorism’.

²⁹ See Sanchez-Cuenca and de la Calle, ‘Domestic Terrorism’, p. 42; LaFree, Dugan and Korte, ‘The Impact of British Counterterrorist Strategies on Political Violence in Northern Ireland’.

³⁰ Della Porta, ‘Leaving Underground Organizations’, p. 84.

³¹ The Israeli response to terrorist activity is widely seen as more aggressive than most modern democracies. See Parker, ‘Fighting an Antaean Enemy’, p. 163.

³² Iqbal, Zaryab and Zorn, Christopher, ‘Sic Semper Tyrannis? Power, Repression, and Assassination Since the Second World War’, Journal of Politics, 68 (2006), 489–501CrossRefGoogle Scholar

³³ Walsh and Piazza, ‘Why Respecting Physical Integrity Rights Reduces Terrorism’, p. 559.

³⁴ Walsh and Piazza, ‘Why Respecting Physical Integrity Rights Reduces Terrorism’.

³⁵ Our focus on authoritarian regimes as a single category neglects potential variations in the importance of legitimacy considerations across different types of authoritarian regimes. While developing an argument that takes these differences seriously is beyond of the scope of this article, the robustness tests include an exploratory analysis.

³⁶ Tankebe, Justice, ‘Public Co-operation With the Police in Ghana: Does Procedural Fairness Matter?’’ Criminology, 47 (2009), 1265–93 CrossRefGoogle Scholar

Lambert, Eric G., Jiang, Shanhe, Khondaker, Mahfuzul I., Elechi, O. OkoBaker, David N. and Tucker, Kasey A., ‘Policing Views From Around the Globe: An Exploratory Study of the Views of College Students From Bangladesh, Canada, Nigeria, and the United States’, International Criminal Justice Review, 20 (2010), 22–47CrossRefGoogle Scholar

³⁷ Pape, Robert A., Dying to Win: The Strategic Logic of Suicide Terrorism (New York: Random House, 2005)Google Scholar

³⁸ Shabad, Goldie and Ramo, Francisco José Llera, ‘Political Violence in a Democratic State: Basque Terrorism in Spain’, in Martha Crenshaw, ed., Terrorism in Context (University Park: Pennsylvania State University Press, 1995), pp. 410–73 Google Scholar

³⁹ We are not arguing that non-democratic leaders are completely insulated from public opinion and electoral concerns. However, while elections are now held in the vast majority of countries, non-democratic regimes frequently limit or prohibit the participation of opposition candidates and incumbents often manipulate the electoral process.

⁴⁰ Cases of group success are empirically rare. Of all groups that ended their campaigns in our data, only 6.6 per cent of cases resulted in a group victory.

⁴¹ Moreover, splintering could be the result of competition among different groups within the same state, and thus be the result of groups competing over material resources and popular support rather than the government's actions. See Kydd and Walter, ‘The Strategies of Terrorism’. The proliferation of terrorist groups in Pakistan is an example of such a dynamic.

⁴² Jones and Libicki provide a list of all terrorist organizations and group attributes in their appendix. The data cover 648 terrorist groups between 1968 and 2006 and provide start and end dates for all groups, including those that did not terminate their campaigns by 2006. The RAND-MIPT data used to create their list of terrorist groups is available at http://www.rand.org/nsrd/projects/terrorism-incidents.html. Jones and Libicki include a few terrorist groups in colonies that were active before the country's independence, such as the Armed Revolutionary Action group in Mozambique. We excluded such groups from the analysis until states achieved independence. Data limitations on the repression variable limit our time frame to the 1976–2006 period. Jones and Libicki, How Terrorist Groups End, pp. 142–86.

⁴³ Jones and Libicki identify a single host country for 430 of 539 groups in our sample. See Jones and Libicki, How Terrorist Groups End. For the remaining 109 groups, two or more states are identified as host countries. To identify the primary host for these groups, we consulted the list of terrorist groups provided by the Terrorism Knowledge Base (TKB), which formed the basis of the data on terrorist groups included in RAND-MIPT. The TKB provides a narrative of each group's evolution and frequently refers to the country in which groups organize the majority of their operations. The TKB data are available online at http://www.start.umd.edu/start/data_collections/tops. For groups without additional information, we consulted the RAND-MIPT data to determine the country in which groups carried out the majority of their attacks.

⁴⁴ Hoffman, Bruce, ‘The Confluence of International and Domestic Trends in Terrorism’, Terrorism and Political Violence, 9 (1997), 1–15Google Scholar

⁴⁵ Moreover, it is possible that such groups simultaneously engage in domestic attacks, since the two types of attacks are not mutually exclusive. The frequency with which groups target both domestically and across borders should ultimately be an empirical question, but we are not aware of research that addresses this question.

⁴⁶ While Jones and Libicki's empirical analysis excludes ongoing campaigns, their appendix provides information on groups that have not terminated. Jones and Libicki, How Terrorist Groups End.

⁴⁷ Jones and Libicki, How Terrorist Groups End, p. 5.

⁴⁸ Jones and Libicki's categories are very similar to ours, but the authors further separate government victory into ‘victory through policing’ and ‘victory through military force’. Since terrorist groups that end because of policing and military force represent an achievement of the government's goals, we collapse these two categories. Jones and Libicki, How Terrorist Groups End.

⁴⁹ Jones and Libicki, How Terrorist Groups End, pp. 142–85.

⁵⁰ Gibney, Cornett and Wood, ‘Political Terror Scale 1976–2006’.

⁵¹ A score of 1 indicates that a country is under a secure rule of law, people are not imprisoned for their views, torture is rare or exceptional, and political murders are extremely rare. Countries that have a limited amount of imprisonment for nonviolent political activity were rated as a 2. Those with extensive political imprisonment, common executions or political murders and brutality, and unlimited detention for political views receive a score of 3 on the PTS. A rating of 4 indicates that a country violates the political and civil rights of a large portion of the population and that murder, disappearances and torture are a common part of life, especially for those who interest themselves in politics and ideas. Finally, a rating of 5 indicates that terror has expanded to the entire population of the country, and the leaders are not limited in the way they pursue personal or ideological goals. The PTS data provide scores for both Amnesty International and US State Department rankings; we average the two scores, meaning that the repression variable can take empirical values of 1, 1.5, 2, 2.5 and so forth.

⁵² Monty G. Marshall and Keith Jaggers, ‘Polity IV Project: Political Regime Characteristics and Transitions, 1800–2007’, available at http://www.systemicpeace.org/polity/polity4.htm.

⁵³ Of all countries in the data, 68 per cent are democracies with polity scores of 6 or higher.

⁵⁴ One might object that findings based on this interaction are the result of a small number of cases, since repression and democracy likely co-vary. While we do not dispute that democracies, on average, have lower levels of repression than non-democratic regimes, numerous case studies document the use of coercive strategies in democracies. Moreover, a cross-tabulation of the repression and polity variables showed that 30 per cent of all democracies with polity scores of 6 or greater score 4 or higher on the repression variable. When only highly democratic countries (polity scores of 8 or greater) are included, the percentage of democracies engaging in such levels of repression decreases to 26 per cent. Although research by Davenport and Armstrong suggests that democracies exhibit lower levels of repression only when high levels of democracy are reached, one-fourth of highly democratic states in our data apply high levels of repression. Conversely, 13 per cent of non-democratic regimes have repression levels lower than 3. Christian Davenport and David A. Armstrong, ‘Democracy and the Violation of Human Rights’, American Journal of Political Science, 48 (2004), 538–54.

⁵⁵ Jones and Libicki, How Terrorist Groups End.

⁵⁶ This variable does not change over time. While it would be ideal to use a time-varying covariate, no such data are available.

⁵⁷ Jones and Libicki, How Terrorist Groups End.

⁵⁸ Nationalist groups were most represented in our sample, representing 38.2 per cent of cases, followed by leftist groups with 33.4 per cent and religious groups with 23.2 per cent; right-wing groups represented only 5 per cent of all cases.

⁵⁹ The variable ranges from 1–6, where lower values indicate that the group's goals are more limited, and greater values represent more extensive goals. Groups are coded as 1 if their goal is to maintain the status quo, 2 if their goal is policy change, 3 if their primary goal is regime change, 4 if their goal is territorial change, 5 if their goal is revolution and 6 if their goal is empire. Since group goals could be correlated with group ideology and such multicollinearity could mask the statistical significance of covariates, we specified separate models excluding the group goal and ideology variables, respectively. The results remained similar to the ones presented here. In addition, correlation matrices for the variables did not show correlations greater than 0.16.

⁶⁰ Alan Heston, Robert Summers and Bettina Aten, ‘Penn World Table Version 6.3: Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania’, available at http://pwt.econ.upenn.edu/php_site/pwt_index.php. The natural log was taken because of high skewness in the data.

⁶¹ Data are available at http://publications.worldbank.org/WDI/. The variable was logged in order to achieve a normal distribution.

⁶² A Hausman test showed that the outcomes are independent of each other. Results for a multinomial logit model were similar.

⁶³ Box-Steffensmeier, Janet and Jones, Bradford S., Event History Modeling (Cambridge: Cambridge University Press, 2004)CrossRefGoogle Scholar

⁶⁴ Tests for non-proportionality produced different results for different models. The group size variable showed time-varying effects in all empirical models that include group characteristics. In addition to group size, the variables measuring democracy, population size, religious groups, nationalist groups and whether a group has a foreign presence showed signs of non-proportionality in some (but not all) models. We include interactions between the logarithm of time and the respective variables for time-varying variables in all models in which tests showed evidence of non-proportionality for respective variables. In order to preserve space, hazard ratios for time-interactions are not presented.

⁶⁵ Box-Steffensmeier and Jones, Event History Modeling, p. 136.

⁶⁶ In competing risk models, subhazard ratios are presented; they represent the effect of covariates on the hazard of the four terrorist campaign outcomes.

⁶⁷ We also analysed models with standard errors clustered on the group's host country. Results were similar to the ones presented here.

⁶⁸ Results for other distributional forms – such as the exponential, log-logistic and Gompertz distribution –were similar to the ones reported.

⁶⁹ We cannot include time-interactions in this model because the command used to create the figure does not allow for their inclusion.

⁷⁰ Cingranelli and Richards, ‘The Cingranelli-Richards (CIRI) Human Rights Dataset’.

⁷¹ Asal, Victor and Rethemeyer, R. Karl, ‘The Nature of the Beast: Organizational Structures and the Lethality of Terrorist Attacks’, Journal of Politics, 70 (2008), 437–49 CrossRefGoogle Scholar

LaFree, Gary and Dugan, Laura, ‘Introducing the Global Terrorism Database’, Terrorism and Political Violence, 19 (2007), 181–204CrossRefGoogle Scholar

⁷² Cronin, How Terrorism Ends, p. 133.