Assignment Week 6: Quantitative Article Reviews

Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=ftpv20

Terrorism and Political Violence

ISSN: 0954-6553 (Print) 1556-1836 (Online) Journal homepage: https://www.tandfonline.com/loi/ftpv20

Repression and Terrorism: A Cross-National Empirical Analysis of Types of Repression and Domestic Terrorism

James A. Piazza

To cite this article: James A. Piazza (2017) Repression and Terrorism: A Cross-National Empirical Analysis of Types of Repression and Domestic Terrorism, Terrorism and Political Violence, 29:1, 102-118, DOI: 10.1080/09546553.2014.994061

To link to this article: https://doi.org/10.1080/09546553.2014.994061

Published online: 23 Feb 2015.

Submit your article to this journal

Article views: 4330

View related articles

View Crossmark data

Citing articles: 21 View citing articles

Repression and Terrorism: A Cross-National Empirical Analysis of Types of Repression

and Domestic Terrorism

JAMES A. PIAZZA

Department of Political Science, The Pennsylvania State University, University Park, Pennsylvania, USA

While some scholars have theorized that repression reduces terrorism because it raises the costs of participating in terrorist activity by dissidents, others argue that repression stimulates terrorism by either closing off nonviolent avenues for express- ing dissent or by provoking or sharpening grievances within a population. This study investigates these contradictory sets of expectations by considering whether or not different specific types of repression yield different effects on patterns of terrorism in 149 countries for the period 1981 to 2006. By assessing the impact of nine specific types of repression on domestic terrorism, the study produces some interesting findings: while, as expected, forms of repression that close off nonviolent avenues of dissent and boost group grievances increase the amount of domestic terrorism a country faces, types of repression that raise the costs of terrorist activity have no discernible suppressing effect on terrorism.

Keywords dissent, domestic terrorism, regime type, restriction of freedoms, state repression

What is the relationship between state repression and terrorist activity within countries? Can repression both suppress and stimulate terrorism? The burgeoning body of work on regime type and terrorism gives some clues about these questions. Investigation of the structural determinants of terrorism has most consistently found that democratic regimes experience more terrorist activity than nondemocratic or illiberal regimes.1 The common explanation for this finding limits itself to the obser- vation that the executive limitations and preservation of individual rights that are part and parcel of democratic rule provide a more hospitable environment for terror- ists than is found in illiberal regimes.2 Democracies extend civil liberties to citizens, place restrictions on policing, extend due process and rights of the accused to arrestees, and tolerate a free media. All of these elements make it easier for terrorist movements to form, to plan and conduct attacks, to claim credit for attacks via media, and to protect their terrorist network if members are arrested. The overall conclusion is that the same ingredients that make a democracy nurturing of civil society and individual freedoms make it vulnerable to terrorism. Conversely, regimes

James A. Piazza is Associate Professor in the Department of Political Science at The Pennsylvania State University.

Address correspondence to James A. Piazza, Department of Political Science, The Pennsylvania State University, University Park, PA 16802, USA. E-mail: [email protected]

Terrorism and Political Violence, 29:102–118, 2017 Copyright # Taylor & Francis Group, LLC ISSN: 0954-6553 print=1556-1836 online DOI: 10.1080/09546553.2014.994061

102

that are able to more easily and widely employ repression, such as dictatorships, should see reduced terrorist activity. Indeed, it is this line of reasoning that prompted commentators after the 9=11 attacks to suggest that countries reduce their vulner- ability to terrorism by curtailing or redefining citizens’ rights.3

The above findings would seem to close the book on the relationship between regime type and terrorism. However, research by scholars focusing more specifically on regime institutional features has substantially complicated the picture. For example, there is evidence that terrorist activity is not uniformly high within liberal polities and that different institutional elements of liberal democratic rule have dif- ferent effects on the level of terrorism a country experiences. The framework for this observation was first established by Eyerman,

4 who summarized theoretical wisdom of the time in noting that some institutions within democratic rule, such as citizens’ opportunities to engage in political activity and to play a role in selecting regime lea- ders, could be expected to dampen terrorism by providing a nonviolent outlet for political dissent while other qualities, such as the abundance of easy targets and the presence of a free media to amplify the propaganda impact of terrorist attacks, might make it more frequent. Eyerman described the first quality as the ‘‘accessible system’’ school of thought and the latter as the ‘‘soft target’’ school of thought. Extending Eyerman, one may presume that illiberal regimes, by restricting opportu- nities for free political engagement, might present themselves to dissidents as ‘‘inaccessible systems,’’ thereby incentivizing terrorism while also presenting them- selves as inhospitable or ‘‘hard’’ targets to terrorists by repressing individual rights, thereby reducing terrorism.

Subsequent scholarship has examined the intersection of regime type and regime institutions, and other features as explanatory variables for terrorism. Li

5 first pro- vided direct empirical evidence that, in part, supports the contention that different institutional aspects of liberal democracy, such as political participation in elections and constraints on executive authority, alternately reduce and stimulate terrorism, while Piazza6 found that young democracies are more terrorism-plagued. Work by Aksoy and Carter7 finds that among democracies, regimes with proportional rep- resentation systems and with higher levels of subnational district magnitudes see a more frequent emergence of certain types of terrorist groups. Leveraging broader scholarly trends in comparative politics, other current research has also found that dictatorships—regimes assumed to be inhospitable to terrorist movements and terrorist activity due to their heightened ability to mobilize repression against dissenters—are not uniformly impervious to terrorism. Wilson and Piazza

8 find evi- dence that among authoritarian regimes, military regimes experience substantially more terrorist attacks than do civilian-led, single-party dictatorships. This is due to the ability of such regimes to maximize both coercive and co-optive tools to manage domestic political dissent. Conrad et al.9 produce a corresponding finding in their empirical study of types of dictatorships and terrorism: authoritarian regimes that generate higher audience costs—military, single-party, and dynastic autocracies— experience more terrorism. Finally, Aksoy et al.10 find that dictatorships that tolerate opposition parties within their legislatures see fewer terrorist groups emerge than those that exclude opposition parties. This is because such dictatorships are able to better manage dissent and channel it into controllable avenues.

We can, therefore, see evidence that both regime type and specific intra-regime political institutions and institutional configurations matter as predictors of terrorist activity in countries. Can we extend this to say that regime behaviors vis-à-vis

Repression and Terrorism 103

citizens’ rights—which can be conditioned on or shaped by regime type and regime institutions, but are distinct phenomena—matter as well? Some research suggests, at least preliminarily, that the answer to this question is yes. For example, we do know that regime treatment of citizens in terms of physical integrity rights and regime respect for minority rights both affect patterns of terrorism in countries.11 Coupled with a more general literature showing that degrees of preservation of liberal rights and deployment of repression against political dissent vary considerably among democratic regimes, as well as among dictatorships,12 these scant findings suggest that particular regime behaviors—whether or not citizens are subject to political repression—might predict the circumstances under which terrorism will occur.

This suggests a closer look at repression, across countries, as a predictor of ter- rorism. Regime type is an overly aggregate predictor of terrorism and disaggregating repression into specific manifestations, I argue, is a way to make sense of these observed divergent and complex relationships. In this study, I examine the impact of different types of repression on the amount of domestic terrorist attacks a country sustained for the period 1981 to 2006. I undertake this investigation starting with the observation that repression can take different forms, affecting different aspects of citizens’ political, social, and personal rights. These different forms, I theorize, could yield different impacts on patterns of terrorism in countries. In the next section, I discuss three such theoretical relationships involving different mixes of types of repression that are hypothesized to either suppress or stimulate terrorist attacks. I then, in the subsequent section, test these different relationships using data on nine different types of repression—restriction of freedoms of speech, association, move- ment, religion, political self-determination, access to a free and independent media, physical integrity, labor rights, and minority political and economic rights—as well as aggregated indices of types of repression. I conclude with a discussion of the implications of the findings.

Repression as a Suppressor and Stimulant of Terrorism: Three Theoretical Stories

In investigating repression, in all of its manifestations, as a predictor of terrorist attacks within countries, I identify three theoretical relationships between forms of repression and domestic terrorism. This set of three relationships is intended to be collectively exhaustive, from a theoretical standpoint. In the first relationship, I argue that repression raises the costs for engaging in terrorism by political dissidents, thereby reducing terrorist activity in a country. The second relationship theorizes that repression closes nonviolent avenues for political dissent, which incentivizes engagement in terrorist activity, leading to an increase in terrorism. The third and final theoretical relationship considers repression to be a key ingredient in the forma- tion and aggravation of group grievances, leading to higher rates of terrorism in countries. I discuss each of these relationships in turn.

Repression and Raised Costs of Terrorism

In this first theoretical story, state application of repression to compel citizen political support and to quash dissent produces a poor strategic environment for would-be terrorists, in a manner consistent with previously discussed literature (e.g., Schmid

13). Suppression of citizens’ ability to freely assemble and engage in

104 J. A. Piazza

autonomous political activity inhibits the ability of terrorist groups to form, draw recruits, and plan their activities. Restrictions on media and on free speech further curtail the communication and propaganda efforts of terrorist movements, com- monly assumed to be the raison d’être of terrorism.14 Finally, lack of constraints on police surveillance, arrest, detention, physically punishing interrogation, and even disappearances of dissidents that are hallmarks of repressive states dramatically enhance the counterterrorism advantages of officials. All of these elements are essen- tially the flip side of Eyerman’s15 ‘‘soft target’’ depiction of democratic regimes. While liberal democracies are easy venues for terrorist movements to work within, illiberal regimes are inhospitable environments that suppress terrorist activity by making it more difficult, dangerous, and costly for dissidents to engage in, and less likely to be effective in terms of garnering public attention. This scenario rests heav- ily on the assumption that the decision of dissidents to engage in terrorism is driven by a strategic, rational calculus in which they opt to use the tactic most likely to advance their political objectives: to get attention, to influence an audience, and to secure concessions from their adversaries. State repression, in this scenario, makes the decision to engage in terrorism suboptimal. Therefore, we should observe states employing repression under these conditions, and using specific types of restrictions that raise the costs of terrorism—restriction of citizens’ movement, control over independent citizen association, press censorship and unconstrained policing, deten- tion and interrogation—to experience less terrorist activity.

Repression and Closed Avenues for Dissent

In the second theoretical story, repressive means employed by the state actually incentivize would-be peaceful political dissidents to engage in terrorism. This is a scenario consistent with theoretical work by Crenshaw

16 and DeNardo17 and with some of the empirical findings produced by Li,18 Aksoy and Carter,19 Aksoy et al.,20 Bravo and Dias,21 and Wilson and Piazza.22 In this scenario, state repressive measures that close legal avenues for political dissent and redress of grie- vances incentivizes dissenters to resort to more extreme, extra-legal measures such as terrorism. In contrast to the first scenario, suppression of free speech and inde- pendent press prompts dissidents to engage in demonstrations of violence, like ter- rorism, in order to break through official censorship to call attention to political grievances.

23

This theoretical story, therefore, makes use of the flip side of Eyerman’s24

‘‘accessible system’’ school of thought regarding democracies and terrorism. In conditions under which the political system is ‘‘inaccessible,’’ dissidents are more likely to see value in engaging in political violence and terrorist activity, despite the risks of doing so, than would be the case if legal avenues to engage in dissent were present. Again, this scenario is informed by some empirical findings. Using a sample of Latin American countries, Bravo and Dias25 find that those that respect political, civil, and human rights experience less anti-government terrorist activity. Aksoy et al.26 and Wilson and Piazza27 determine that dictatorships that provide some official opportunity for political dissent, albeit incomplete and managed, are more impervious to terrorism than those that do not. And like the first scenario, it rests on a rational=strategic actor assumption: dissidents engage in terrorism because it is a potentially more profitable course of action relative to working within the system.

Repression and Terrorism 105

Repression and Elevation of Group Grievances

In the third theoretical story, state repression also stimulates terrorism, but via a different route from the ‘‘Closed Avenues’’ story above. Rather than altering the strategic costs and benefits of using terrorism, repression in this scenario alters the overall climate of public approval of the government, thereby affecting the potential scope and effectiveness of terrorist activities. Experience of repression de-legitimizes the state and alienates citizens from government, fostering and strengthening anti-state, anti-status quo popular grievances. Repression creates an environment that is easily exploited by extremists engaged in terrorism, who can more profitably draw support from a sympathetic public, can more easily recruit new members, can more easily turn attacks into propaganda tools, and are less vulnerable to potential backlash normally generated by terrorist attacks.

28 In instances where state use of repression is broad and indiscriminate, affecting dissidents and apolitical bystanders alike, opportunities for extremist movements to exploit public outrage are even greater.29 Furthermore, repressive states may find other states less likely to cooperate on counterterrorism efforts such as sharing of information and extraditing terrorism suspects because such activities violate internationalized norms.30

There are several strands of theoretical and empirical support for this scenario. Borrowing from a theoretical framework for grievance and rebellion developed by Gurr,31 Crenshaw32 and Ross33 demonstrates, using examples from historical case studies of terrorist campaigns, that state oppression is an important precipitant of group grievances that help terrorist groups overcome collective action and other problems standing in the way of recruitment and mobilization of political violence. Research by Piazza34 empirically determines that countries characterized by political and economic discrimination against ethnic minority groups experience significantly more terrorism than countries without minority discrimination. Moreover, noting that some qualitative historical literature observes that state use of repression— particularly human rights abuses—ultimately undermines government counterter- rorism and counterinsurgency efforts by damaging relations with local populations and spurring domestic and transnational political opposition,

35 Walsh and Piazza36

find a positive empirical link between government respect for physical integrity rights and lower levels of terrorism in a cross-national sample. Bravo and Dias37 produce corresponding results for a sample of Latin American countries.

Hypotheses

These three theoretical stories translate into three testable hypotheses:

H1: Forms of repression that contribute to the raising of the costs associated with engaging in terrorism reduce terrorist activity.

H2: Forms of repression that contribute to the closure of peaceful avenues for political dissent or redress increase terrorist activity.

H3: Forms of repression that provoke or exacerbate group grievances increase terrorist activity.

In the study, these hypotheses are tested using sets of repression indicators developed from existing databases on regime attributes and behaviors. These include measures of restriction on citizens’ freedom of movement, both domestically and internationally, restriction on freedom of association and membership in political

106 J. A. Piazza

and social organizations, restriction on electoral self-determination through voting, repression of independent labor unions, workplace organizations, strikes and collec- tive bargaining rights, restriction of free speech by citizens, religious repression and restriction of freedom of conscience, formal and informal discrimination against ethnic minorities, press restriction, and censorship and abuse of citizens’ rights to physical integrity.

To test the first hypothesis, forms of repression most clearly associated with raising the costs of engaging in terrorism by dissidents, or with producing an inhos- pitable or suboptimal strategic environment for using terrorism, are regressed against counts of terrorism. These include: restriction of movement and association, both of which make organizing, planning, and executing terrorism difficult; press censorship, which severely discounts the ability to use terrorist attacks to transmit propaganda or to influence a wider audience; and physical integrity rights abuse, which are associated with the use of torture, indefinite detention and targeted assassination=extrajudicial killing of terrorists, terror suspects, and terrorist group supporters. These types of repression are hypothesized to reduce terrorism.

The second hypothesis is tested using forms of repression reasonably argued to be associated with closing peaceful avenues of dissent, including: restriction of elec- toral self-determination, restriction of free speech, and labor restriction including bans on the right to strike or public protest. Also tested along with these forms of repression is press censorship, which in the context of this hypothesis is used to oper- ationalize reduced opportunity for citizens to express independent, critical, and unauthorized political opinions. These types of repression are hypothesized to increase terrorism.

The third hypothesis is tested using forms of repression that are most closely associated with the provocation and sharpening of grievances in the population, including religious repression and ethnic minority discrimination. Also included in this category is abuse of physical integrity rights, given that Walsh and Piazza

38 the- orize that such abuses often undermine state counterterrorism efforts because they aggrieve and alienate the wider population, thereby hamstringing government efforts to gather intelligence about terrorists and their supporters and to garner community support to fight terrorism.

Analysis

To test the three hypotheses, the study conducts a series of zero-inflated negative binomial estimations using the nine individual measures of repression in 149 coun- tries for the period 1981 to 2006, the full range of years for which I have relatively complete data for all variables. These nine types of repression constitute the main independent variables of the study. Seven of the independent variables are derived from the Cingranelli and Richards (CIRI) Human Rights Data project.39 CIRI pro- vides ordinal scale measures of the status of and protections for various political, civil, social, and physical rights within countries for the period 1981 to 2006. I pro- cess and re-scale many of these to produce a complement of indicators of repression of these rights in countries for the same time period to use in my analyses. These include measures of Physical Integrity rights abuse, restriction of movement, free speech, free association, electoral self-determination, restriction of women’s empow- erment, labor rights repression, and religious repression. To measure abuse of Physi- cal Integrity rights—protections against physical torture, political imprisonment,

Repression and Terrorism 107

extrajudicial killing, or disappearance—I subtract eight from the ‘‘PHYSINT’’ (Physical Integrity) indicator coded by CIRI to produce a scale between 0 and 8, where 8 indicates severe repression of physical integrity of citizens. I measure restric- tion of movement similarly by adding the two CIRI indicators for restriction of dom- estic and foreign movement or travel (‘‘DOMMOV’’ and ‘‘FORMOV’’) and subtracting the sum from 4, producing a scale where 4 indicates severe restriction on citizen movement. Restriction of free speech, free association, electoral self- determination, labor rights, and religious rights are similarly re-scaled in the analysis by subtracting the original CIRI scores (for ‘‘SPEECH,’’ ‘‘ASSOC,’’ ‘‘ELECSD,’’ ‘‘WORKER’’ and ‘‘NEW_RELFREE’’) from 2, producing new scales where 2 indicates severe repression of these rights.

I use data from non-CIRI sources for the remaining two measures of repression. I re-code data from the Freedom House Index of Press Censorship,

40 converting the original measure into a 10-point index where 10 indicates severe restriction of media. I also operationalize repression of minority group rights by combining two indica- tors derived from the Minorities at Risk41 database—political discrimination (‘‘POLDIS’’) and economic discrimination (‘‘ECDIS’’) suffered by minority groups—to construct a singular index scored between 0 and 8, where 8 indicates severe minority discrimination.

The dependent variable is a count of domestic terrorist attacks occurring in a country-year. This count is derived from data from the Global Terrorism Database (GTD) in Enders, Sandler, and Gaibulloev’s

42 study. The types of repression engaged in by states would seem to most impact the amount of domestic terrorist activity—defined as terrorism launched by nationals of a country targeting co-nationals or domestic targets within the boundaries of the country—occurring within a country. So regressing measures of repression to counts of domestic terror- ism is an obvious design element.43 Alternative measures, such as counts of terrorist attacks that include transnational terrorism—defined as attacks by foreigners against domestic targets—would not seem to be clearly affected by country-level attributes such as political repression by a regime. Such attacks are, therefore, excluded from the analysis. Because the dependent variable is a count indicator char- acterized by significant levels of spatial and temporal dispersion, and has a prepon- derance of zero values (67.3% of all observations are zeros), I utilize a zero-inflated negative binomial regression estimation technique.

44 My decision to do this is further buttressed by the results of Vuong tests conducted on all models, the results of which are all significant, indicating that a zero-inflated negative binomial estimation tech- nique is more efficient than a negative binomial technique that pools zero-observations and counts of terrorism.45 Finally, in all models I also calculate robust standard errors clustered on country.

Controls

Included in every model estimation are some standard covariates, frequently found in other cross-sectional time series empirical studies of terrorism.46 Because the inde- pendent variables measure regime behaviors rather than regime type per se, I also include a measure of political regime in each estimation. I use the 21-point Polity score for this. To hold constant level of economic development and distribution of incomes within countries, all specifications include country Human Development indices—which measure gross national income, literacy and life expectancy

108 J. A. Piazza

rates—and national Gini coefficient measures of income inequality. These are both expected to be positive predictors of terrorism given that Piazza47 found that econ- omically developed countries—specifically countries that score highly in terms of human development as measured by the HDI—are more prone to terrorist attacks than poor or developing countries,48 and Eyerman49 and Li50 both find countries with high levels of income inequality experience more terrorism. To consider the impact of state capacity to project military force, some of which may be deployed in a counterterrorism capacity, I also include the CINC index of national capacity from the Correlates of War database. My expectation is that this will be a negative predictor of terrorism. Also included are natural logged measures of national popu- lation and surface area of countries, both of which have been found to positively pre- dict terrorism.

51 The estimations also control for whether or not the country is engaged in an interstate war and is experiencing a civil or intra-state war to hold con- stant other manifestations of violence.52 Because Eyerman53 found that older, mature regimes are less likely to experience terrorism, the study also controls for age of political regime using the ‘‘Durable’’ score from the Polity IV database. Finally, all estimations control for past experience of terrorist attacks.54 All inde- pendent variables in the study are furthermore lagged one period, within country-case, to capture delayed effects and to aid in determining direction of causation.

Descriptive statistics for all variables used in the study are presented in Table 1.

Table 1. Descriptive statistics

Variable Obs. Mean SD Min Max

Domestic terrorist attacks (Enders et. al 2011)55

4,260 7.49 32.99 0 524

Physical integrity rights abuse 3,837 3.15 2.35 0 8 Restriction of movement 3,916 1.12 1.28 0 4 Restriction of association 3,640 .90 .84 0 2 Restriction of electoral

self-determination 3,645 .86 .84 0 2

Press censorship 4,472 5.92 2.84 1 10 Restriction of free speech 3,645 .95 .73 0 2 Labor repression 3,611 1.01 .78 0 2 Religious repression 3,859 .67 .80 0 2 Minority discrimination 3,011 3.28 3.04 0 8 Human development index 4,214 .66 .19 .22 .956 CINC score 3,980 .006 .019 2.73e-06 .190 Gini coefficient 4,228 42.87 9.19 20.7 84.8 (ln) Population 4,228 1.94 1.76 –2.65 7.18 (ln) Area 4,472 11.83 2.14 5.70 16.64 International war dummy 4,472 .01 .12 0 1 Civil war dummy 4,472 .16 .37 0 1 Regime type (polity) 3,810 1.58 7.43 –10 10 Regime durability 4,228 23.04 28.22 0 197

Repression and Terrorism 109

Table 2. Repression and domestic terrorism: Individual measures

[1] [2] [3] [4]

Count 1

Zero- infl

2 Count

1 Zero- infl

2 Count

1 Zero- infl

2 Count

1 Zero- infl

2

Restrict. of movement t-1 .005 (.038)

.115 (.070)

Restrict. of association t-1 .022 (.063)

.129 (.127)

Restrict. of electoral self- determin. t-1

.210��

(.070) .047

(.138) Restrict. of free speech t-1

Labor repression t-1

Religious repression t-1

Minority discrimination t-1

Press censorship t-1

Physical integrity rights abuse t-1

Regime type (polity) t-1

.011�

(.005) –.039���

(.008) .025��

(.007) –.013 (.013)

.025��

(.008) –.015 (.015)

.041���

(.009) –.024 (.016)

Regime durability t-1 –.004 ��

(.001) .005�

(.002) –.004��

(.001) .006�

(.002) –.004��

(.001) .006�

(.002) –.004���

(.001) .006�

(.002) Human Dev. Index t-1 1.815

���

(.305) �2.398���

(.494) 1.569���

(.325) �1.873��

(.578) 1.509���

(.327) –2.012��

(.594) 1.569���

(.327) –1.973��

(.583) CINC score t-1 3.658

�

(2.155) 7.089�

(3.385) .269

(2.419) –.201

(4.426) 1.288

(2.458) 1.719

(4.406) 1.275

(2.463) 1.599

(4.363) Gini coefficient t-1 .038

���

(.006) –.022�

(.009) .033���

(.006) –.015 (.010)

.032���

(.006) –.016 (.010)

.032���

(.006) –.017 (.010)

(ln) Population t-1 .479 ���

(.044) –.248���

(.063) .465���

(.048) –.210��

(.075) .440���

(.049) –.240��

(.079) .450���

(.048) –.220��

(.077) (ln) Area t-1 –.126

���

(.033) –.027 (.045)

–.097��

(.036) –.019 (.052)

–.100��

(.037) –.020 (.053)

–.106��

(.037) –.028 (.053)

International war (COW) t-1

–.008 (.204)

.321 (.385)

.231 (.241)

.544 (.529)

.232 (.241)

.483 (.525)

.313 (.243)

.508 (.518)

Civil war (PRIO) t-1

.875���

(.079) �1.147���

(.189) .947���

(.092) �1.479���

(.265) .934���

(.088) �1.391���

(.263) .958���

(.088) �1.363���

(.258) Terrorist attacks t-1 .018

���

(.001) �1.367���

(.197) .016���

(.001) �1.035���

(.167) .016���

(.001) �1.025���

(.163) .016���

(.001) �1.004���

(.159) Constant –1.182�

(.493) 4.189���

(.780) –1.120�

(.516) 3.013��

(.898) –.932�

(.528) 3.269���

(.932) –1.124�

(.530) 3.389���

(.922) N 4,767 3,338 3,274 3,279 Nonzero obs. 1,657 1,343 1,330 1,331 Zero obs. 3,110 2,039 1,944 1,948 LR v2 1294.36��� 1101.06��� 1062.99��� 1073.75���

Vuong test (z) 12.62��� 9.96��� 10.02��� 9.95���

Y¼Counts of domestic terrorist attacks. All models zero-inflated negative binomial 1Count model, negative binomial. 2Zero-inflated logit (certain zero). �p� .1; ��p� .01; ���p� .000.

110 J. A. Piazza

[5] [6] [7] [8] [9] [10]

Count 1

Zero- infl

2 Count

1 Zero- infl

2 Count

1 Zero- infl

2 Count

1 Zero- infl

2 Count

1 Zero- infl

2 Count

1 Zero- infl

2

–.001 (.075)

.154 (.133)

.091�

(.055) .299��

(.113) .129�

(.052) .124

(.105) .120���

(.016) –.049�

(.030) .024

(.025) –.006 (.045)

.089���

(.024) –.125��

(.045) .023��

(.007) –.018 (.013)

.028���

(.007) –.010 (.012)

.026���

(.007) –.021�

(.012) .027��

(.008) –.010 (.014)

.018�

(.009) –.041��

(.016) .034���

(.007) –.041��

(.012) –.004��

(.001) .006�

(.002) –.005��

(.001) .006�

(.002) –.007���

(.001) .006�

(.003) –.001 (.001)

.004 (.003)

–.004��

(.001) .005��

(.002) –.003�

(.001) .004�

(.002) 1.514���

(.328) –1.983��

(.584) 1.539���

(.330) �2.057���

(.589) 1.944���

(.335) –1.839��

(.601) 1.137��

(.406) –.513 (.732)

1.852���

(.308) �2.411���

(.502) 1.582���

(.328) –2.004��

(.594) 1.129

(2.439) 1.291

(4.400) 1.562

(2.475) –2.848 (5.349)

–7.925��

(2.920) –7.190 (7.234)

2.741 (2.632)

–7.557 (12.180)

3.653�

(2.158) 7.172�

(3.395) 1.523

(2.481) 1.597

(4.408) .032���

(.006) –.016 (.010)

.030���

(.007) –.021�

(.010) .035���

(.007) –.014 (.010)

.028���

(.007) .002

(.012) .039���

(.006) –.022�

(.009) .028���

(.007) –.013 (.010)

.445���

(.048) –.227��

(.077) .429���

(.049) –.223��

(.080) .518���

(.051) –.150�

(.082) .495���

(.056) –.108 (.105)

.475���

(.044) –.248���

(.063) .397���

(.050) –.159�

(.081) –.102��

(.037) –.027 (.053)

–.092�

(.038) –.029 (.054)

–.117��

(.037) –.042 (.054)

-.208���

(.041) –.052 (.066)

–.125���

(.033) –.028 (.045)

–.076�

(.038) –.013 (.054)

.230 (.243)

.484 (.522)

.233 (.244)

.399 (.543)

.161 (.249)

.425 (.562)

.537�

(.274) .858

(.693) –.010 (.204)

.318 (.368)

.296 (.242)

.526 (.546)

.938���

(.091) �1.414���

(.263) .945���

(.089) �1.454���

(.265) .963���

(.089) �1.490���

(.272) .943���

(.100) �1.727���

(.434) .864���

(.080) �1.148���

(.189) .768���

(.099) �1.176���

(.273) .016���

(.001) �1.009���

(.160) .016���

(.001) �1.036���

(.171) .016���

(.001) �1.068���

(.180) .015���

(.001) �1.051���

(.211) .018���

(.001) �1.371���

(.198) .015���

(.001) �1.030���

(.176) –.909�

(.538) 3.250���

(.927) –1.023�

(.532) 3.372���

(.924) –1.349��

(.524) 3.193���

(.915) –.109 (.577)

1.691 (1.105)

–1.366��

(.533) 4.244���

(.870) –1.309�

(.539) 3.404���

(.935) 3,279 3,247 3,326 2,578 4,767 3,245 1,331 1,318 1,308 1,050 1,657 1,326 1,948 1,929 2,018 1,528 3,110 1,919

1064.72��� 1056.82��� 1094.29��� 818.98��� 1295.29��� 1072.58���

9.98��� 9.92��� 9.82��� 7.94��� 12.60��� 9.82���

estimations. All predictors lagged one period.

111

Results

Table 2 presents the results of the estimations regressing the individual measures of repression to terrorist attacks. Because the individual components are correlates of one another, they are run in separate models to avoid problems of multicollinearity.56

Model 1 is a simple baseline model that only includes the covariates. Here we can see that regime type, measured using the Polity score, is a positive and significant predictor of counts of terrorist attacks and is a negative predictor of the likelihood a country will experience no terrorism. This finding is consistent with the mainstay of research showing that democracies, measured as an overall regime type, experience higher rates of terrorism.57 The essence of this finding—for the count model—is reproduced across all of the models in Table 2. It is therefore interesting that in the remaining models, some measures of repression, which operationalize regime behavior rather than regime type, are found to be significant even when holding regime type constant.

The first hypothesis, which expects forms of repression that raise costs for ter- rorist activity and=or worsen the strategic environment for would-be terrorists to reduce terrorist attacks, is tested using the measures of restriction of movement, association, press censorship, and physical integrity rights abuse (Models 2, 3, 9, and 10). However, the results in Table 2 do not yield support. In Models 2, 3, and 9 the repression indicator is not found to be significant for either the count or the zero-inflated logit. The coefficient for the count model for physical integrity rights abuse in Model 10 is significant but positive for the count model, contra expecta- tions, while the coefficient for the zero-inflated logit is significant and negative. (A similar pattern is evident for the other indicators used to test the first hypothesis: coefficients for the count models are positive, rather than negative, but do not meet standard levels of statistical significance.)

The second hypothesis, which anticipates that elements of repression associated with closure of nonviolent opportunities for political dissent boost terrorism, is tested in Models 4, 5, 6, and 9. This hypothesis is partly supported by the results. As hypothesized, restriction of electoral self-determination (Model 4) and labor repression (Model 6) are significant and positive predictors of counts of terrorist attacks—though electoral self-determination is not found to reduce the likelihood that a country experiences zero attacks and labor repression is found to boost both high counts of terrorism and the probability a country experiences zero attacks, sug- gesting a more complex relationship between this manifestation of repression and terrorism.

58 However, restriction of free speech (Model 5) and press censorship (Model 9) are not found to be significant.

The third hypothesis, in which forms of repression that provoke or exacerbate public grievances against the state are expected to increase terrorist activity in coun- tries, is most consistently supported by the results. All three specific repression indi- cators associated with this hypothesis—religious repression (Model 7), minority discrimination (Model 8), and abuse of physical integrity rights of citizens (Model 10)—are found to be significant predictors of counts of terrorist attacks in countries. Moreover, for minority discrimination and physical integrity abuse, the zero-inflated logit coefficient is also significant and signed in the proper direction, suggesting that countries characterized by low levels of these two forms of repression are more likely to experience no terrorism at all.

112 J. A. Piazza

In all 10 of the regression estimation models, these results are found to be robust to the inclusion of some highly significant controls.59 Across all models, regime type, regime durability, level of human=economic development, income inequality mea- sured using the Gini coefficient, national population, the civil war indicator, and counts of previous terrorism are significant predictors of terrorism in the expected direction. Country surface area is also significant in all models, but in the opposite direction expected: negative. Only the national capabilities index (CINC) and the dummy variable for country participation in an international war are not consist- ently significant across models. Inclusion of these control indicators, along with the calculation of country-clustered robust standard errors, help to allay concerns of omitted variable biases in the estimations.

Substantive Impact of Forms of Repression on Terrorism

To assess the substantive effects of the various types of repression examined in the analysis, I calculated marginal effects with all covariates set constant at their means. The results of these for the individual and aggregate measures of repression are graphed in Figure 1.

In the figure, the mean substantive effect on the dependent variable—in this case, counts of terrorism—is graphed along with upper and lower 95% confidence intervals. Figure 1 reveals that of the forms of repression examined, restriction of electoral self-determination results in the highest amount of subsequent terrorism (.95 attacks). On average, for each one-point (out of a 3-point scale) increase in repression of electoral self-determination, countries are found to experience 23.3% more domestic terrorist attacks. Countries featuring repression of labor rights and religious freedom experience on average between .53 and .42 more attacks than

Figure 1. Marginal effects of individual types of repression on domestic terrorist attacks.

Repression and Terrorism 113

countries without such repression. Increasing by one point (again, out of a 3-point scale) labor or religious repression boosts domestic terrorist attacks by 9.5 and 13.9% respectively. Repression of minority rights and physical integrity rights abuses bear a similar substantive effect: countries featuring these types of repression experi- ence, on average, .41 more attacks than countries that do not. A one-point increase (out of an 8-point scale) in minority discrimination increases terrorism by 12.4% and by 10.1% for physical integrity rights abuse. The other forms of repression are not significant.

Conclusion

The study produces several key conclusions. First, the study finds that repression overall is a stimulant, rather than a suppressor, of domestic terrorism within countries: none of the specific types of repression are found to reduce terrorism. This finding, while consistent with some studies of individual predictors of terrorism like human rights abuses,59 largely contradicts the mainstream theoretical expectation in the field of terrorism studies and in the policy community that repression is an effec- tive safeguard against or panacea for terrorism.60 Illiberal states that repress their populations do not inoculate themselves from terrorism via repression, and liberal states should not expect greater protection from terrorism by augmenting the rights of their citizens.

Second, in disaggregating repression into specific forms, the study finds that some manifestations of repression actually increase a country’s vulnerability to attacks. This finding appears to be the product of two processes and a handful of specific types of repression. When repression road-blocks nonviolent alternatives to participation—most notably when it obstructs citizens’ rights to participate in elections and when it deters expressions of labor rights—domestic terrorism increases. Indeed, electoral self-determination is found to be the most substantive precipitant of terrorism in the analysis, prompting almost double the amount of ter- rorism than the other predictors. Discrimination against minorities and physical repression are found to increase domestic terrorism by prompting popular grievances and alienating citizens from the government—a finding that is consistent with pre- vious empirical research. However, religious repression is also found to precipitate terrorism by increasing grievance and alienation, and is found to be a slightly more substantive predictor than minority or physical repression. These findings help to clarify the causal relationship between repression and terrorism—showing it to be conditional on the type of repression—and underscore the importance of disaggrega- tion when contemplating the relationship between regime type and terrorism.

Notes

1. Ana Bela Santos Bravo and Carlos Manuel Mendes Dias, ‘‘An Empirical Analysis of Terrorism: Deprivation, Islamism and Geopolitical Factors,’’ Defence and Peace Economics 17, no. 4 (2006): 329–341; Stephen B. Blomberg and B. Peter Rosendorff, ‘‘A Gravity Model of Globalization, Democracy and Transnational Terrorism,’’ in G. D. Hess, ed., Guns and But- ter (Cambridge: MIT Press, 2007), 25–156; Alex Braithwaite and Quan Li, ‘‘Transnational Terrorism Hot Spots: Identification and Impact Evaluation,’’ Conflict Management and Peace Science 24, no. 4 (2007): 281–296; Axel Dreher and Justina A. V. Fischer, ‘‘Government Decentralization as a Disincentive for Transnational Terror? An Empirical Analysis,’’ Inter- national Economic Review 51, no. 4 (2010): 981–1002; Konstantinos Drakos and Andreas

114 J. A. Piazza

Gofas, ‘‘In Search of the Average Transnational Terrorist Attack Venue,’’ Defence and Peace Economics 17, no. 2 (2006): 73–93; William Lee Eubank and Leonard Weinberg, ‘‘Does Democracy Encourage Terrorism?’’ Terrorism and Political Violence 6, no. 4 (Winter 1994): 417–443; William Lee Eubank and Leonard Weinberg, ‘‘Terrorism and Democracy: What Recent Events Disclose,’’ Terrorism and Political Violence 10, no. 1 (1998): 108–118; William Lee Eubank and Leonard Weinberg, ‘‘Terrorism and Democracy: Perpetrators and Victims,’’ Terrorism and Political Violence 13, no. 1 (Spring 2001): 155–164; Kate Ivanova and Todd Sandler, ‘‘CBRN Incidents: Political Regimes, Perpetrators and Targets,’’ Terrorism and Polit- ical Violence 18, no. 3 (2006): 423–448; Quan Li and Drew Schaub, ‘‘Economic Globalization and Transnational Terrorism: A Pooled Time-Series Analysis,’’ Journal of Conflict Resolution 48, no. 2 (2004): 230–258; James A. Piazza, ‘‘Do Democracy and Free Markets Protect Us from Terrorism?’’ International Politics 45 (2008): 72–91; Burcu Savun and Brian J. Phillips, ‘‘Democracy, Foreign Policy and Terrorism,’’ Journal of Conflict Resolution 53, no. 6 (2009): 878–904; Sarah Jackson Wade and Dan Reiter, ‘‘Does Democracy Matter?’’ Journal of Conflict Resolution 51, no. 2 (2007): 329–348.

2. Alex Schmid, ‘‘Terrorism and Democracy,’’ Terrorism and Political Violence 4, no. 4 (Winter 1992): 14–25.

3. Alan Dershowitz, ‘‘Is an Outright Ban the Best Way to Eliminate or Constrain Torture?’’ in Stuart Gottlieb, ed., Debating Terrorism and Counterterrorism: Conflicting Per- spectives on Causes, Contexts and Responses (Washington, DC: Congressional Quarterly Press, 2009), 320–335; Axel Dreher, Martin Gassebner, and Lars H. R. Siemers, ‘‘Does Terror Threaten Human Rights?: Evidence from Panel Data,’’ CESIFO Working Paper No. 1935 (2007), http://www.econstor.eu/bitstream/10419/25980/1/538079215.PDF; Connor Gearty, ‘‘Terrorism and Human Rights,’’ Government and Opposition 42, no. 3 (2007): 340–362; Paul Hoffman, ‘‘Human Rights and Terrorism,’’ Human Rights Quarterly 26, no. 4 (2004): 932–955; John Yoo, ‘‘Counterterrorism and the Constitution: Does Providing Security Require a Trade-Off with Civil Liberties?’’ in Stuart Gottlieb, ed., Debating Terrorism and Counterterrorism: Conflicting Perspectives on Causes, Contexts and Responses (Washington, DC: Congressional Quarterly Press, 2009), 339–352.

4. Joe Eyerman, ‘‘Terrorism and Democratic States: Soft Targets or Accessible Systems,’’ International Interactions 24, no. 2 (1998): 151–170.

5. Quan Li, ‘‘Does Democracy Produce or Reduce Transnational Terrorist Incidents?’’ Journal of Conflict Resolution 49, no. 2 (2005): 278–297.

6. James A. Piazza, ‘‘Regime Age and Terrorism: Are New Democracies Prone to Terrorism?’’ International Interactions 36, no. 2 (2013): 246–263.

7. Deniz Aksoy and David Carter, ‘‘Electoral Institutions and the Emergence of Terrorist Groups,’’ British Journal of Political Science 44, no. 1 (2014): 181–204.

8. Matthew Wilson and James A. Piazza, ‘‘Autocracies and Terrorism: Conditioning Effects of Authoritarian Regime Type on Terrorist Attacks,’’ American Journal of Political Science 57, no. 4 (2013): 941–955.

9. Courtenay Conrad, Justin Conrad, and Joseph Young, ‘‘Tyrants and Terrorism: Why Some Autocrats are Terrorized While Others Are Not,’’ International Studies Quarterly 58, no. 3 (2014): 539–549.

10. Deniz Aksoy, David Carter, and Joseph Wright, ‘‘Terrorism in Dictatorships,’’ Journal of Politics 74, no. 3 (2012): 810–826.

11. Andreas E. Feldmann and Maiju Perälä, ‘‘Reassessing the Causes of Nongovernmen- tal Terrorism in Latin America,’’ Latin American Politics and Society 46 (2004): 101–132; James A. Piazza, ‘‘Types of Minority Discrimination and Terrorism,’’ Conflict Management and Peace Science 29 (2012): 521–546; James A. Piazza, ‘‘Poverty, Minority Economic Dis- crimination and Terrorism,’’ Journal of Peace Research 48, no. 3 (2011): 339–353; James I. Walsh and James A. Piazza, ‘‘Why Respecting Physical Integrity Rights Reduces Terrorism,’’ Comparative Political Studies 43, no. 4 (2010): 551–577.

12. Christian Davenport, ‘‘State Repression and the Tyrannical Peace,’’ Journal of Peace Research 44, no. 4 (2007): 485–504; Hanne Fjelde, ‘‘Generals, Dictators, and Kings: Authori- tarian Regimes and Civil Conflict, 1973–2004,’’ Conflict Management and Peace Science 27, no. 3 (2010): 195–218; Barbara Geddes, Paradigms and Sandcastles: Theory Building and Research Design in Comparative Politics (Ann Arbor: University of Michigan Press, 2003); Jennifer Gandhi, Political Institutions under Dictatorship (New York: Cambridge University

Repression and Terrorism 115

Press, 2008); Kristian S. Gleditsch and Michael D. Ward, ‘‘Double Take: A Reexamination of Democracy and Autocracy in Modern Polities,’’ Journal of Conflict Resolution 41, no. 3 (1997): 361–383.

13. Schmid, ‘‘Terrorism and Democracy’’ (see note 2 above). 14. Brigitte Nacos, Mass-Mediated Terrorism: The Central Role of the Media in Terrorism

and Counterterrorism (Lanham, MD: Rowman and Littlefield, 2007); Robert Pape, ‘‘The Strategic Logic of Suicide Terrorism,’’ American Political Science Review 97, no. 3 (August 2003): 343–361.

15. Eyerman, ‘‘Terrorism and Democratic States’’ (see note 4 above). 16. Martha Crenshaw, ‘‘The Causes of Terrorism,’’ Comparative Politics 13 (1981):

379–399. 17. James DeNardo, Power in Numbers: The Political Strategy of Protest and Rebellion

(Princeton, NJ: Princeton University Press, 1985). 18. Li, ‘‘Does Democracy Produce or Reduce Transnational Terrorist Incidents?’’ (see

note 5 above). 19. Aksoy and Carter, ‘‘Electoral Institutions and the Emergence of Terrorist Groups’’

(see note 7 above). 20. Aksoy et al., ‘‘Terrorism in Dictatorships’’ (see note 10 above). 21. Bravo and Dias, ‘‘An Empirical Analysis of Terrorism’’ (see note 1 above). 22. Wilson and Piazza, ‘‘Autocracies and Terrorism’’ (see note 8 above). 23. A case example helps to explain this. Domestic political repression and media censor-

ship in Brazil during the 1964–1985 military dictatorship is a commonly cited motive for the 1969 kidnapping of U.S. Ambassador Charles Burk Elbrick by the Left-Wing MR-8 (Revol- utionary Movement 8 October). MR-8 hoped to circumvent government restrictions on media and political dissent by kidnapping a prominent diplomat.

24. Eyerman, ‘‘Terrorism and Democratic States’’ (see note 4 above). 25. Bravo and Dias, ‘‘An Empirical Analysis of Terrorism’’ (see note 1 above). 26. Aksoy et al., ‘‘Terrorism in Dictatorships’’ (see note 10 above). 27. Wilson and Piazza, ‘‘Autocracies and Terrorism’’ (see note 8 above). 28. Kalevi J. Holsti, ‘‘Political Causes of Humanitarian Emergencies,’’ in Wayne

Nafziger, Frances Stewart, and Raimo Vayrynen, eds., War, Hunger and Displacement: The Origins of Humanitarian Emergencies (Oxford: Oxford University Press, 2000), 239–281; Andreas Wimmer, ‘‘Who Owns the State? Understanding Ethnic Conflict in Post-Colonial Societies,’’ Nations and Nationalism 3 (1997): 631–666.

29. Stathis Kalyvas, ‘‘The Paradox of Terrorism in Civil War,’’ The Journal of Ethics 8, no. 1 (2004): 97–138.

30. Martha Finnemore and Kathryn Sikkink, ‘‘International Norm Dynamics and Political Change,’’ International Organization 52 (1998): 887–917.

31. Ted Robert Gurr, Why Men Rebel (Princeton, NJ: Princeton University Press, 1970); Ahmed S. Hashim, Insurgency and Counterinsurgency in Iraq (Ithaca, NY: Cornell University Press, 2006); Ted Robert Gurr, ‘‘Why Minorities Rebel: A Global Analysis of Communal Rebellion and Conflict since 1945,’’ International Political Science Review 14 (1993): 161–201.

32. Crenshaw, ‘‘The Causes of Terrorism’’ (see note 16 above). 33. Jeffery Ian Ross, ‘‘Structural Causes of Oppositional Political Terrorism: A Causal

Model,’’ Journal of Peace Research 30 (1993): 317–329. 34. Piazza, ‘‘Types of Minority Discrimination and Terrorism’’ (see note 11 above);

Piazza, ‘‘Poverty, Minority Economic Discrimination and Terrorism’’ (see note 11 above). 35. Max Abrahms, ‘‘Why Democracies Make Superior Counterterrorists,’’ Security Stu-

dies 16, no. 2 (2007): 223–253; Hashim, Insurgency and Counterinsurgency in Iraq (see note 31 above); Anthony James Joes, Resisting Rebellion: The History and Politics of Counterinsur- gency (Lexington: University of Kentucky Press, 2004); Gil Merom, How Democracies Lose Small Wars (Cambridge: Cambridge University Press, 2003); Edward Rice, Wars of the Third Kind: Conflict in Underdeveloped Countries (Berkeley, CA: University of California Press, 1988).

36. Walsh and Piazza, ‘‘Why Respecting Physical Integrity Rights Reduces Terrorism’’ (see note 11 above).

37. Bravo and Dias, ‘‘An Empirical Analysis of Terrorism’’ (see note 1 above).

116 J. A. Piazza

38. Walsh and Piazza, ‘‘Why Respecting Physical Integrity Rights Reduces Terrorism’’ (see note 11 above).

39. David L. Cingranelli and David L. Richards, CIRI Human Rights Data Project (2010), http://ciri.binghamton.edu/index.asp.

40. Freedom House, Freedom of the Press Data (2012), http://www.freedomhouse.org/ report-types/freedom-press

41. Minorities and Risk Project, ‘‘Minorities at Risk Dataset’’ (College Park, MD: Center for International Development and Conflict Management, 2009), http://www.cidcm. umd.edu/mar

42. Walter Enders, Todd Sandler, and Khusrav Gaibulloev, ‘‘Domestic Versus Transna- tional Terrorism: Data, Decomposition and Dynamics,’’ Journal of Peace Research 48, no. 3 (2011): 319–337.

43. Furthermore, domestic terrorism is a more frequent, widespread, and deadly=damag- ing phenomenon than transnational terrorism. (See Ignacio Sanchez-Cuenca and Luis de la Calle, ‘‘Domestic Terrorism: The Hidden Side of Political Violence,’’ Annual Review of Polit- ical Science 12 [2009]: 31–49; Enders et al., ‘‘Domestic Versus Transnational Terrorism’’ [see note 42 above]). Enders et al. provide a thorough discussion of the operational difference between domestic and transnational terrorism.

44. The zero-inflated technique facilitates the separate investigation of predictors of counts of terrorist attacks and the likelihood that a country will experience zero terrorist attacks. In theory, these can bear different relationships. Patrick T. Brandt, John T. Williams, Benjamin O. Fordham, and Brian Pollins, ‘‘Dynamic Models for Persistent Event Count Time Series,’’ American Journal of Political Science 44, no. 4 (2000): 823–843; Adrian Colin Cameron and P. K. Trivedi, Regression Analysis of Count Data (Cambridge: Cambridge University Press, 1998); Gary King, ‘‘Statistical Models for Political Science Event Counts: Bias in Conventional Procedures and Evidence for the Exponential Poisson Regression Model,’’ American Journal of Political Science 32, no. 3 (1988): 838–863.

45. The results of the Vuong tests are reported in the main results for all models. W. H. Greene, ‘‘Accounting for Excess Zeros and Sample Selection in Poisson and Negative Binomial Regression Models,’’ Working Paper, Stern School of Business, New York Univer- sity (1994), EC-94-10; J. S. Long, Regression Models for Categorical and Limited Dependent Variables (Thousand Oaks, CA: Sage, 1997); Q. Vuong, ‘‘Likelihood Ratio Tests for Model Selection and Non-Nested Hypotheses,’’ Econometrica 57 (1989): 307–333.

46. See Martin Gassebner and Simon Luechinger, ‘‘Lock, Stock, and Barrel: A Compre- hensive Assessment of the Determinants of Terror,’’ Public Choice 149, nos. 3–4 (2011): 235–261.

47. Piazza, ‘‘Poverty, Minority Economic Discrimination and Terrorism’’ (see note 11 above).

48. Though I expect HDI to be a positive predictor of terrorism, I do note the body of complicated empirical findings regarding poverty and terrorism. For example, Braithwaite and Li, ‘‘Transnational Terrorism Hotspots’’ (see note 1 above) and Li, ‘‘Does Democracy Produce or Reduce Transnational Terrorist Incidents?’’ (see note 5 above) find poverty to boost terrorism in countries, while Abadie, Blomberg and Hess, and Enders, Hoover and Sandler find a more complex and nonlinear relationship between poverty and terrorism: Alberto Abadie, ‘‘Poverty, Political Freedom, and the Roots of Terrorism,’’ American Econ- omic Review 96, no. 2 (2006): 50–56; S. Brock Blomberg and Greogory Hess, ‘‘From (No) But- ter to Guns? Understanding the Economic Role in Transnational Terrorism,’’ in Keefer and Loayza, eds., Terrorism, Economic Development, and Political Openness (Cambridge, MA: Cambridge University Press, 2008), 83–115; Walter Enders, Gary A. Hoover, and Todd Sand- ler, ‘‘The Changing Nonlinear Relationship Between Poverty and Terrorism’’ Journal of Conflict Resolution (Forthcoming, 2015). doi:10.1177=0022002714535252.

49. Eyerman, ‘‘Terrorism and Democratic States’’ (see note 4 above). 50. Li, ‘‘Does Democracy Produce or Reduce Transnational Terrorist Incidents?’’ (see

note 5 above). 51. Wade and Reiter, ‘‘Does Democracy Matter?’’ (see note 1 above); Eyerman, ‘‘Terror-

ism and Democratic States’’ (see note 4 above). 52. Li, ‘‘Does Democracy Produce or Reduce Transnational Terrorist Incidents?’’ (see

note 5 above); Michael G. Findley and Joseph K. Young, ‘‘Terrorism and Civil War: A Spatial

Repression and Terrorism 117

and Temporal Approach to a Conceptual Problem,’’ Perspectives on Politics 10, no. 2 (2012): 285–305.

53. Eyerman, ‘‘Terrorism and Democratic States’’ (see note 4 above). 54. This is accomplished by inserting a lagged measure of the dependent variable in on

the right-hand side in all estimations. As a check of robustness, duplicate models are also run with this lagged dependent variable removed. These produce the same results as those por- trayed in the main analysis.

55. Walter Enders, Todd Sandler, and Khusrav Gaibulloev, ‘‘Domestic Versus Transna- tional Terrorism: Data, Decomposition and Dynamics,’’ Journal of Peace Research 48, no. 3 (2011): 319–337.

56. All are significantly correlated with one another at the p� .000 level. Variance inflation factor diagnostics are not necessarily suggestive of strong multicollinearity (mean VIF¼2.23). However, the condition number, calculated on eigenvalues extracted from the variables in the model, registers at 55.5, suggesting some concern that collinearity problems are affecting significance levels and signs of coefficients. D. A. Belsley, E. Kuh, and R. E. Welsch, Regression Diagnostics: Identifying Influential Data and Sources of Collinearity (New York, NY: John Wiley, 1980).

57. e.g., Eubank and Weinberg, ‘‘Does Democracy Encourage Terrorism?’’ (see note 1 above).

58. Countries characterized by labor repression are more likely to experience either high levels of terrorism or they are likely to experience none at all. This peculiar finding underscores the value of using the zero-inflated technique. A standard negative binomial regression analy- sis reveals a non-significant negative relationship between labor repression and domestic ter- rorism. It is likely that two different theoretical processes predict the likelihood that repression of labor rights will prompt a country to experience no terrorism versus a high level of terror- ism, and that the effect of these processes is washed in an estimation where zero values are pooled with counts of attacks, thereby masking a significant relationship between labor repression and terrorism.

59. When a control for the Muslim population of a country is included in the estimations, the core results remain the same. When a dichotomous indicator for Afghanistan, Iraq, and Pakistan is included—three countries that experience high levels of terrorism in the later per- iod of the data—the core results remain the same with two exceptions: religious repression becomes borderline significant and labor repression is not found to be significant in the count model only.

60. Walsh and Piazza, ‘‘Why Respecting Physical Integrity Rights Reduces Terrorism’’ (see note 11 above).

61. Dershowitz, ‘‘Is an Outright Ban the Best Way to Eliminate or Constrain Torture?’’ (see note 3 above); Yoo, ‘‘Counterterrorism and the Consitution’’ (see note 3 above).

118 J. A. Piazza