Do Failed or Weak States Favor Resident Terrorist Groups’ Survival?

Abstract

Employing two alternative measures of state failure, we investigate how state weakness influences resident terrorist groups’ survival. Theoretically, state failure favors resident terrorist groups’ survival, while state territorial control fosters resident groups’ termination until some control threshold. Empirically, we uncover a robust negative relationship between a country’s weakness and its control of terrorism through the lens of the resident terrorist groups’ survival prospects. The discovered relationship withstands a host of robustness tests – e.g., alternative estimates and samples. We apply an instrument designed to address endogeneity concerns. In particular, our novel instrument for failed states consists of the interaction between natural disasters and ethnic fractionalization. As a state’s percentage of territorial control increases, resident terrorist groups are more prone to ending until some threshold control percent. Our analysis can guide counterterrorism policy by exploiting the nuanced theoretical determinants identified here that foster resident groups’ termination in failing states.

Keywords

weak and failing states resident terrorist groups’ termination territorial control and state failure domestic and transnational terrorism instrument

The role of failed or weak states in facilitating domestic and transnational terrorism remains hotly debated over 20 years since the four skyjackings on 11 September 2001 (henceforth, 9/11), launched by al-Qaeda based in Afghanistan. Following 9/11, the US administration viewed such states as providing safe havens to terrorist groups, bent on attacking US interests at home and abroad (LaFree, Dugan, and Fahey 2007; National Intelligence Strategy 2005; Patrick 2011; Rotberg 2002; Takeyh and Gvosdev 2002). Despite careful empirical and theoretical analyses, the relationship between failing states and the promotion of terrorism remains contradictory and controversial (see especially Coggins 2015; George 2018; Hehir 2007; Pašagić 2020; Piazza 2008; Plummer 2012).¹ An understanding of this relationship and its drivers remains an essential ingredient to developing effective counterterrorism policy to confront the continuing terrorist threat. If resident terrorist groups in such weak states are the main drivers of terrorism there and abroad, then identifying the determinants of the survival of these groups can inform some effective policy making. Theoretically, a state’s viability and its control of home territory affect the survival prospects of resident terrorist groups and, thus, the indigenous terrorist threat.

There are essential aspects of the terrorism-weak state nexus that remain unexplored. Even though terrorist groups based in weak or failing states are the perpetrators of terrorist attacks (e.g., Hezbollah in Lebanon, al-Shabaab in Somalia; al-Qaeda in the Arabian Peninsula in Yemen, Jemaah Islamiya in Indonesia, and al-Qaeda in Iraq), the survival prospects of such resident groups are not yet known. Those terrorist groups are subnational collectives whose members employ terrorism to pursue a political objective (Phillips 2014). Knowledge of resident terrorist groups’ prognosis for survival in weak states is key to ascertaining how to limit the associated terrorism threat. Any understanding of those terrorist groups’ prospects must address many empirical challenges that include an alleged nonlinear relationship between state weakness in maintaining territorial control and resident groups’ termination (see, e.g., Patrick 2011; Piazza 2008). In addition, the relationship between state weakness and resident terrorist group survival involves a two-way causality that must be addressed with an appropriate instrument – merely lagging independent variables is not sufficient. Moreover, identifying the sought-after relationship requires a dataset that links resident terrorist groups and their actions to failed or weak states over time.

The primary purpose of our study is to meet those challenges to discern theoretically and empirically the survival prognosis of terrorist groups based in failed or weak states during 1970–2016. To do so, we utilize a relatively new dataset on terrorist groups that identifies their birth and demise over time, various endings (if relevant), ideology, and terrorist campaign variables. For the task at hand, a dataset based on groups’ characteristics and actions, rather than a dataset based on attacks, is most appropriate. To address the endogeneity of the studied relationship, we construct an appropriate novel instrument for state weakness or failure. Moreover, we permit the anticipated nonlinear relationship between state weakness and resident terrorist group termination.

For both discrete and continuous state fragility variables, we find that state failure has a negative influence on resident terrorist groups’ termination, while the government’s extent of territorial control has an inverted U-shaped relationship on resident terrorist groups’ termination, respectively. Those relationships hold for a rich set of controls, whose estimated coefficients generally agree with the literature on the determinants of terrorist groups’ ending. Moreover, the estimated relationships are robust to a host of alternative estimation techniques. Most importantly, the estimated relationship between state failure and resident terrorist groups’ conclusion remains when we instrument the state fragility indicators. Unlike the extant literature which tests state failure’s influence on terrorist violence, the current study examines how state failure or territorial control affects the prospects of the central actor, the resident terrorist group.

Literature Review

The whole notion of weak, failing, fragile, or failed state is controversial and subject to interpretation (see, e.g., Coggins 2015; Hehir 2007; Menkhaus 2003; Newman 2007; Pašagić 2020, Plummer 2012). Generally, a consensus of the literature associates state weakness or failure with two key governing challenges: (1) an impaired capacity to maintain public order and well-being (Patrick 2011), and (2) an inability to exercise power and sovereignty throughout the country (Newman 2007). Failing states’ incapacity to govern adversely affects the country’s security, political actions, economy, and social welfare (Patrick 2011). Security failures can result in conflict, political violence, terrorism, coups, and human rights abuses, while political challenges weaken government effectiveness, rule of law, civil and political rights, and corruption safeguards. Failed or weak states’ economic losses may involve low gross domestic product (GDP) per capita, reduced GDP growth, high inflation, and poor regulatory practices (Patrick 2011). Additionally, weak states may display high infant mortality, public good underprovision, poor education attainment, and less-developed institutions (Patrick 2011; Piazza 2008). Weak institutions when combined with economic discrimination of minorities can fuel terrorism (Campos and Gassebner 2013; Ghatak 2016). In addition to terrorism, failing states are often plagued by criminal activity (Patrick 2011; Rotberg 2002). When a failing state loses its monopoly over the deployment of military force throughout the countryside, the country relinquishes some control of its territorial integrity. That loss may provide safe havens to terrorist groups, insurgents, and criminal gangs from which to plan and base their attacks (Takeyh and Gvosdev 2002).

Given the numerous dimensions of state failure, researchers must employ alternative measures of this concept in empirical analyses on the consequences of such failures (Plummer 2012). Because many alternative failure metrics are highly correlated, they must often be employed separately in multiple empirical models. That procedure is followed in earlier articles that investigate state weakness or failure as a determinant of domestic and/or transnational terrorism (Coggins 2015; George 2018; Lai 2007; Newman 2007; Piazza 2008). For instance, Coggins (2015) employs three measures of human insecurity concerns [i.e., infant mortality, low GDP per capita, and small human development index (HDI)], three measures of state capacity (i.e., government effectiveness, rule of law, and corruption control), and two representations of political collapse. Similarly, George (2018) uses GDP per capita, an aggregate state fragility index, a government effectiveness measure, rule of law indicator, and corruption control metric in a series of models to capture alternative representations of state failure.

Hehir (2007) and Newman (2007) question the post-9/11 conventional wisdom (National Intelligence Strategy 2005) that failing states provide sanctuaries for terrorist groups to engage in campaigns at home and abroad. As a response, Newman (2007) investigates 84 terrorist groups in terms of various measures of home governance. Although some formidable terrorist groups resided in failing or failed states, he also highlights that many such states did not contain terrorist groups, while numerous viable states were home to terrorist groups. His thought-provoking study does not engage in statistical inference, which is also true of Hehir (2007). By contrast, Lai (2007) offers a panel analysis for 1968-1998 to examine whether weak states, which generally imposed low operating cost on resident terrorists, generated a large number of transnational terrorist incidents on other states.² Lai (2007) presents support for the weak states’ association with the dissemination of transnational terrorist incidents abroad (also, see Campos and Gassebner 2013). In Lai’s analysis, civil wars, interstate wars, GDP per capita, and the number of phone lines per thousand people serve as measures of state capacity. The use of war measures as independent variables raises endogeneity concerns in Lai (2007) and some subsequent studies owing to two-way causality between terrorism and civil wars. Moreover, clearer indicators of state failure would provide more support for this pioneering article’s conclusion.

In a subsequent article, Piazza (2008) puts forward a time-series, cross-sectional negative binomial analysis of 197 countries for 1973-2003 to investigate further the alleged relationship between state failure and terrorist attacks. His study is concerned with two hypotheses: (1) failed and failing states serve as the location for transnational terrorist attacks, and (2) failed and failing states are the source of transnational terrorist incidents abroad. Piazza’s (2008) dependent variables are the counts of transnational terrorist attacks, based on either the venue or perpetrator country. Independent variables include an aggregate state failure index, ethnic wars, revolutionary wars, genocides, politicides, adverse regime changes, regime measures,³ and other country-specific variables. Piazza’s (2008) findings indicate that failing states are more apt to host terrorist groups that conduct transnational terrorism at home and abroad; however, the prospects or survival of such groups are not addressed. Thus, Piazza (2008) supports the conventional linkage between governing fragility and transnational terrorism. Like Lai (2007), Piazza’s use of war variables raises endogeneity worries. His study avoids selection bias by including both weak and non-weak states.

Coggins (2015) further studies the relationship between failing states and terrorism with panel data for 153 countries during 1999-2008. As indicated earlier, she employs multiple indicators of human insecurity, state incapacity, and political collapse to represent state weakness or failure. Some of her interesting results differ markedly from earlier findings. In particular, she indicates that most failing states were not necessarily predisposed to either domestic or transnational terrorism based on venue (location), perpetrators’ origin country, or target state for most measures of state weakness. Her panel analysis finds that politically collapsed⁴ and highly corrupt states not only experienced more terrorism but also dispatched more terrorists abroad. Like earlier studies on state failure and terrorism, her study is silent about survival prospects of terrorist groups residing in failing states.

Finally, George (2018) presents a follow-up study on the relationship between state failure and transnational terrorism in which the emphasis is either on the perpetrators’ proximity to the target or on the complexity of the attack. George’s (2018) measures of transnational terrorism include within-country attacks, contiguous-country attacks, and noncontiguous attacks. His indicators of state weakness involve high levels of infant mortality, low GDP per capita, government ineffectiveness, absence of rule of law, high corruption, and large aggregate state failure index. Among many results, he shows that weak states experienced more logistically complex transnational terrorist events. Moreover, fragile states were the venue for homegrown attacks against foreigners.

To date, the extant literature casts some light on the nuances of the alleged linkage between weak states and terrorist attacks. However, this literature does not address the survival prospects of terrorist groups residing in failed or failing states, which is our focus. The emerging literature on terrorist groups’ survival was made possible with the expanding data on terrorist events, compiled by the Global Terrorism Database (GTD) (National Consortium for the Study of Terrorism and Responses to Terrorism 2020), International Terrorism: Attributes of Terrorist Events (ITERATE) (Mickolus et al. 2021), and RAND (2012) Database of Worldwide Terrorism Incidents. Global Terrorism Database contains both domestic and transnational terrorist attacks for 1970-2019; ITERATE surveys just transnational terrorist incidents for 1968-2020; and RAND includes only transnational attacks before 1997 and both types of attacks for 1998-2006. These attack-based event datasets contain various observations on the perpetrating terrorist groups, when known, and their attack characteristics. To foster terrorist group analyses still further, Cronin (2009) and Jones and Libicki (2008) utilize RAND terrorist event data and other sources to assemble terrorist group datasets that record groups’ ideology, size, goals, start and end dates, base country or countries, allied groups, and other pertinent variables. Hou, Gaibulloev, and Sandler (2020) update Jones and Libicki (2008) through 2016, while linking incidents to those in GTD instead of RAND, and, in so doing, provide a dataset on terrorist group characteristics and longevity, needed for the analysis here.

In a pioneering article based on ITERATE transnational terrorist events, Blomberg, Engel, and Sawyer (2010) identify key considerations determining terrorist groups’ survival, using time-to-failure methods. If a terrorist group is defeated by the authorities, splinters, joins the political process, merges, or ends operations, then the original group no longer survives (Jones and Libicki 2008). In a subsequent article, Blomberg, Gaibulloev, and Sandler (2011) show that terrorist groups’ longevity is enhanced with group size, a religious fundamentalist ideology, and a diversified attack campaign. Terrorist groups based in the Middle East and North Africa (MENA) survived longer relative to those based in other regions. Moreover, group survival was bolstered by democratic institutions.

Many articles on terrorist groups’ survival followed. Phillips (2015), Gaibulloev and Sandler (2014), and Gaibulloev, Hou, and Sandler (2020) find empirically that intergroup competition supported terrorist groups’ survival. By contrast, Phillips (2014) establishes that allied terrorist groups bolstered one another’s survival. In exploring the determinants of terrorist groups’ locational choice, Gaibulloev (2015) shows the desirability for terrorist groups to locate in weak countries as a mean to foster their longevity.

Missing from the literature on terrorist group survival is how these groups fare in failed states and those with limited control over their territory, which is analyzed here.

Theoretical Considerations

To build a theoretical rationale for our hypotheses, we draw, in part, on the literature that relates host countries’ governance prowess to the prospects for resident terrorist attacks. We contend that factors that bolster resident groups’ terror campaigns can also bode favorably on the longevity of these groups. Failed or failing states can reduce resident groups’ operating costs, thereby fostering larger terror campaigns (Lai 2007; Pašagić, 2020), which, in turn, lift the groups’ viability prospects and their visibility. Weak and, particularly, failed states offer few impediments to stymie resident terrorists’ operations (Pašagić 2020; Takeyh and Gvosdev 2002). As Patrick (2011) points out, weak states do not possess a monopoly over the deployment of paramilitary and military forces. The absence of such a host-government monopoly greatly favors resident terrorist groups’ survival, especially since military force is a key source of terrorist group termination (Gaibulloev and Sandler 2014; Jones and Libicki 2008). Absence of government control over national territory is an essential aspect of state failure favoring resident terrorist groups.⁵ Given ungoverned areas, resident terrorist groups can carve out territory for training camps, staging areas, field headquarters, hiding hostages, safe havens, and terrorism infrastructure from which to support and launch their attacks at home and abroad (Takeyh and Gvosdev 2002). Such terrorist assets foster their ability to maintain operations, plan attacks, attract members, and ultimately survive. As ungoverned areas grow in size, the resident terrorists pose a more diffused target to offensive government operations, which offers greater invulnerability to the terrorists. States with constrained territorial control are less adept at neutralizing resident terrorists, thus bolstering resident terrorist groups’ longevity (Piazza 2021).

Given weak states’ lack of governance effectiveness (Coggins 2015; George 2018; Rotberg 2002), resident terrorist groups can act with impunity in such states. Failing states’ inherent governance and territorial control challenges result in weak public institutions (Newman 2007), which include law enforcement and the judicial system. Consequently, resident terrorist groups’ operations are augmented as the government confronts roadblocks in capturing and bringing resident terrorists to justice. Failed and weak states are associated with insufficient bureaucracy, lack of rule of law, and rampant corruption (Patrick 2011; Rotberg 2002), all of which hamper the state from seizing the initiative against resident terrorists. One characteristic of state weakness – e.g., corruption – can give rise to grievances that can encourage the recruitment of new terrorists, which fosters the terrorist groups’ viability. When resident terrorist groups unleash large successful attack campaigns in a governance-challenged state, groups’ recruitment and retention of members increase, which, then, augments the groups’ survival. In fact, Takeyh and Gvosdev (2002) view failed states as bolstering resident terrorist recruitment.

Lawless places in weak states offer business opportunities through illicit activities to resident terrorist groups. Such endeavors include drug dealing, kidnapping, counterfeiting, and smuggling (Pašagić 2020, 20). Obviously, illicit activities can fund the resident groups and promote their longevity. With safe havens, resident terrorists have sanctuaries from which to hold kidnapped victim, whose ransoms can finance the groups’ operations. In addition, lawless areas can offer terrorists improved access to sophisticated weapons (George 2018), which makes the resident groups more formidable opponents and bolsters their survival prospects. Nongoverned areas also provide sanctuaries for the terrorist groups from which to elude government counterterrorism offensives. Such safe places make the terrorist groups more invulnerable to home and foreign military forces. If resident terrorists mount attacks abroad, there is always the risk of foreign counterterrorism actions in the failed states. The above rationales for the survival of resident terrorist groups in failed or weak states lead to the first hypothesis (H1):

H1:

State failure and weakness reduce the likelihood of resident terrorist group termination.

Although there has been some speculation along the lines of H1 in the qualitative literature on state failure (e.g., Menkhaus 2003; Pašagić 2020), there are no statement or tests of H1, thus making it a novel hypothesis.

We next offset some arguments that might be raised against H1 because not everyone views a failed state as an ideal environment for resident terrorist groups’ operations. For example, Pašagić (2020) points out that the population of a failed state may be hostile to foreigners, but this concern is more germane to foreign terrorists trying to engage in attacks in failed states. Resident terrorists in failed states are generally domestic operatives who know the lay of the land. In fact, domestic terrorist attacks far outnumber transnational ones in most countries (Gaibulloev and Sandler 2023). In Afghanistan during 1996-2001, al-Qaeda formed a bond with the ruling Taliban who provided a safe haven to al-Qaeda from which to operate even though al-Qaeda’s leadership contained foreigners.

Another consideration that allegedly works against H1 involves failed states being relatively target-poor countries because of the lack of Westerners, diplomats, and aid workers (Menkhaus 2003). Nevertheless, failed states still attract some of these kinds of targets as the many kidnappings of Westerners in Iraq during 2004 and in other failing states (e.g., Mali since 2012 by al-Qaeda in the Islamic Maghreb) demonstrate. Western journalists are drawn to failed states for newsworthy reporting and, in so doing, they present soft and high-value targets for resident terrorist groups – e.g., the kidnapping in 2012 (and subsequent beheading in 2014) of US journalist James Foley in Syria by the Islamic State in Iraq and Syria (ISIS). Aid workers and their organizations are also drawn to failed states, making for a target-rich environment (Murdie and Stapley 2014). Failed state may be susceptible to foreign counterterrorism missions (George 2018; Menkhaus 2003) but this risk can be minimized by resident terrorists by not targeting the assets of a country with formidable power-projection capabilities and the willingness to use them. When such third-party missions occur, foreign troops stimulate local grievances and backlash that further support the popularity and survival of resident terrorist groups (Piazza and Choi 2018).

H1 is now extended to allow for a continuum of failing states. As a state’s percentage of territorial control increases from zero, resident terrorist groups confront less conducive conditions for survival so that their termination rate rises. Previously identified conducive factors for resident terrorist groups’ survival become less supportive as ungoverned areas shrink.

The relationship between resident terrorist groups’ termination on the vertical axis and the extent of territorial control (in percentage) on the horizontal axis may be roughly linear, with group termination rising as the state’s territorial control increases from 0 to some high control percentage. If, however, the influence of rising territorial control to challenge resident terrorist groups experiences diminishing returns, then the relationship between group termination and territorial control will assume an inverted U-shaped curve. As a consequence, terrorist groups’ ending rises at a decreasing rate with the government’s enhanced territorial control in which groups’ termination rate may attain a peak at some high percentage of territorial control to be identified empirically. Beyond the peak, the resident terrorist groups’ ending rate falls. The implied terrorist groups’ resurgence may arise from a tactical advantage in blending in and setting their targeting agenda when territorial control is high countrywide. This follows because the terrorists will not necessarily be attracted to a few areas when territorial control is high almost everywhere. Alternatively, the enhanced survivability of resident terrorist groups may be traced to the government letting down its guard when it perceives to possess nearly full territorial control.⁶

There are other grounds for the hypothesized inverted U-shaped relationship between governmental territorial control and resident terrorist groups’ termination.⁷ For example, as the government’s control attains some threshold, the government may see less need to challenge the terrorists, thereby bolstering somewhat resident groups’ survival prospects. The downturn in more-developed states’ (i.e., those with greater territorial control) ability to challenge their limiting group demise. Similarly, more-in-control states face greater executive constraints that limit their ability to go after the resident terrorist group. Thus, we have H2:

H2:

State territorial control exerts an inverted U-shaped relationship on resident terrorist groups’ termination. That is, resident groups are less inclined to end operations in countries with little or no control over their territory. As governmental territorial control grows, resident groups’ likelihood of termination increases until some threshold state control, after which greater control may reduce groups’ ending operations.

Methodology and Data

Data

We test the hypotheses using cross-sectional time series data on terrorist groups derived from the Extended Data on Terrorist Groups (EDTG) database (Hou, Gaibulloev and Sandler 2020), averaged over 5-year intervals for the period 1970-2016. Since state failure changes very slowly over time, grouping years into 5-year intervals allows for increased within-sample variation. The test sample includes 627 terrorist groups based in 99 countries.⁸ The dependent variable measures terrorist groups’ termination using a dichotomous indicator coded one for the year in which a group terminates from being defeated, splintering, merging, joining the political process, or ceasing operations. For our sample, approximately 13.3 percent of terrorist groups ended at some point during the observation interval.

To measure state failure, we employ two independent variables, selected to capture the two crucial defining characteristics of state failure in the literature. The first is lack of government capacity to maintain public order and foster public well-being (Patrick 2011). To encapsulate that state-failure characteristic, we employ a dichotomous measure derived from the Polity 2 dataset (Marshall, Jaggers, and Gurr 2021), coded 1 for terrorist groups based in countries experiencing state collapse (i.e., the −77 score from Polity 2).⁹ When state failure applies, there is no de facto national government authority to oversee economic, security, or political activities, leaving a stateless territory. A complete state collapse is a rare event. In our sample, only 11.9 percent of terrorist groups are based in states experiencing such collapse. The second defining characteristic of state failure is the government’s inability to exercise power, policing, and sovereignty throughout its national territory (Newman 2007; Takeyh and Gvosdev 2002). With the second operationalization of state failure, proportions of the country are outside of the state’s effective governance and control, which is true of Iraq, Lebanon, Mali, Somalia, Yemen, and other sample countries during some intervals of 1970-2016. We operationalize that state-failure measure with the percentage of territory controlled (Territory control) by the central government in the terrorist group’s base country. The variable’s measurement is derived from the Varieties of Democracy (V-Dem) database and is calculated on a continuous scale, which ranges from 31.6 to 100 in our sample (Coppedge et al. 2020).¹⁰ Terrorist groups’ base countries, where the central government exerts authority over a smaller percent of national territory, are more likely to exhibit state failure.¹¹ To permit a nonlinear influence, we allow for linear and quadratic measures of the percent of territory controlled.

Other terrorist-group and base-state features may also affect resident group survival or longevity (e.g., Gaibulloev and Sandler 2013, 2014; Daxecker and Hess 2013); we include several controls when estimating the determinants of resident terrorist group termination. In particular, we control for terrorist groups’ ideological type, based on categorical variables for groups espousing left-wing, right-wing or nationalist-separatist (henceforth, nationalist) ideologies for which religious ideology serves as the reference category. Past studies indicate that religious terrorist groups are more prone to survival. As an indicator of terrorist groups’ capacity, the share of transnational attacks (Share trans. Attacks) conducted by resident groups serves as a control and corresponds to the ratio of their transnational terrorist attacks to the sum of their domestic and transnational terrorist attacks. We also control for attack mode diversity (Attack diversity), another indicator of a groups’ capacity, measured by one minus the Hirschman-Herfindahl index of all possible attack modes employed by the terrorist groups in a given year.¹² Greater attack diversity keeps the authorities guessing, thus adding to the terrorist groups’ longevity. Data for all of these group-level attributes are derived from the EDTG database (Hou, Gaibulloev, and Sandler 2020).

In the estimations, we also hold constant several base-country features, which include the natural log-transformed indicators of base country gross domestic product per capita, log(GDP/POP) and national population, log(POP), both taken from the World Bank (2021). Based on Polity 2 (Marshall, Jaggers, and Gurr 2021), the base-country political regime is controlled. The Polity2 value varies from -10 (full autocracy) to +10 (full democracy). Also, we control for the degree of ethnic group fractionalization (Ethnic fractionalization) in the base country, derived from Alesina et al. (2003). Base-country geographic features that may affect terrorist group survival are also controlled and include the natural log-transformed average elevation of the base country, log(Elevation), and dichotomous variables coded 1 for base countries located in the tropics or are landlocked (Central Intelligence Agency 2018; Gallup, Mellinger and Sachs 2001; Gallup, Sachs and Mellinger 1999). Finally, we also include dichotomous categorical variables for the location of the base country in the seven major world regions as designated by the World Bank (2021): East Asia and Pacific (EAP); Europe and Central Asia (ECA); Latin America and Caribbean (LAC); South Asia (SAS), sub-Saharan Africa (SSA), and North America (NA) where MENA is the reference category. Descriptive statistics are presented in Table 1.

Table 1.

Descriptive Statistics.

Variables	N	Mean	SD	Min	Max
SF	1955	0.07	0.25	0	1
Territory control	1947	85.30	14.09	31.67	100
Left wing	1944	0.27	0.44	0	1
Nationalist	1944	0.43	0.50	0	1
Religious	1944	0.27	0.44	0	1
Right wing	1944	0.04	0.19	0	1
Share trans. Attacks	1942	0.10	0.22	0	1
Attack diversity	1966	0.08	0.17	0	0.80
log(GDP/POP)	1874	7.97	1.45	5.14	11.22
log(POP)	1993	17.92	1.69	11.19	28.23
Polity2	1939	4.19	5.88	−10	10
Ethnic fractionalization	1975	0.45	0.23	0.012	0.88
EAP	1999	0.10	0.29	0	1
ECA	1999	0.18	0.38	0	1
LAC	1999	0.10	0.30	0	1
MENA	1999	0.17	0.38	0	1
NA	1999	0.03	0.17	0	1
SAS	1999	0.31	0.46	0	1
SSA	1999	0.12	0.32	0	1
log(Elevation)	1986	6.39	0.67	3.41	7.85
Tropics	1986	0.37	0.42	0	1
Landlock	1999	0.10	0.29	0	1
Natural disaster	1999	5.17	5.92	0	34.5

We construct an event study plot to provide a sense of how the number of terrorist groups varies around state-failure timings. Figure 1 displays the raw averages of the number of terrorist groups operating in countries that experienced (solid line) or did not experience (dashed line) a state collapse during the 5 years before and after the state collapse if relevant. In particular, for weak states, the plot traces the unweighted average of the number of terrorist groups in the 5 years before the state failure $(t = - 5, - 4, \dots, - 1)$ , in the year the state failure was first recorded $(t = 0)$ , and in the 5 years after the failure was initially recorded $(t = 1,2, \dots, 5)$ . Non-failed states, however, do not have pre- and post-failure periods, which complicates the construction of an event study plot for them. Therefore, we follow Baskaran, Min, and Uppal (2015) to construct synthetic averages for non-failed states as follows. For each failed state, we compute the corresponding average number of terrorist groups in all non-failed states in period $t = (- 5, - 4, \dots, 4,5)$ . For each period, we then collapse the average number of groups in non-failed countries across all failed states to obtain a synthetic value of the number of terrorist groups in non-failed states.¹³

Figure 1.

Event study plot.

The figure suggests that the number of terrorist groups in failed states is steadily lower than in non-failed states in the pre-failure period. A year prior to failure, weak states begin experiencing a steep increase in the number of terrorist groups.¹⁴ This trend continues until the third year after the recorded state collapse and then levels off. By then, the number of terrorist groups operating in weak states exceeds the number of groups in the non-failed states subsample. Hence, weak states appear to be a favorable environment for terrorist groups to emerge. Whether this translates into better terrorist group survivability remains an empirical question that we pursue.

Methods

We implement discrete-time hazard models to evaluate the impact of state failure on resident terrorist group survival (Allison 1982; Jenkins 1995). We prefer discrete-time hazard method to conventional continuous-time survival analysis because our terrorist group data, though intrinsically continuous, is grouped into discrete intervals (years).¹⁵ Let $T_{i}$ be a random variable indicating the length of time that a terrorist group is active. Consider a terrorist group that survives until $t_{p}$ . Conditional on explanatory variables, the probability of the terrorist organization concluding in a time interval, $[t_{p}, t_{p + 1})$ , $p = 1,2, \dots,$ and $t_{1} = 0$ , is:

P_{i p} = \Pr (T_{i} < t_{p + 1} | T_{i} \geq t_{p}, x_{i p}) = Φ (x_{i p} β + μ (p)),

(1)

where

x_{i p}

is a vector of explanatory variables, including state failure, for a terrorist group

i

at time

p

β

is a vector of coefficients,

μ (p)

is an unknown function of duration, and

Φ (\cdot)

is a cumulative distribution function. We use two alternative specifications of

μ (p)

: a quadratic specification,

μ (p) = μ_{1} t_{p} + μ_{2} t_{p}^{2}

, and a piecewise constant specification with a set of period dummies, assuming that the periods share the same hazard rate. We specify

Φ (\cdot)

as a logistic cumulative distribution and perform the logit estimator. In the robustness analysis, we perform the random-effects logit estimator to control for unobserved heterogeneity.¹⁶

Instrumental Variables Approach

Endogeneity, stemming from reverse causality and omitted variables, presents a threat to identifying the causal relationship. First, resident terrorist groups’ survivability may cause a country to fail. Although we are not aware of such evidence, it cannot be ruled out. Second, some omitted factors may jointly determine state weakness and a group’s likelihood to fail. We control for a host of confounders in our main and robustness regressions but cannot account for all possible joint correlates; hence, omitted variables bias remains a potential concern.

To address the endogeneity concern, we apply an instrumental variables approach. Our instrument is an interaction between natural disasters and ethnic fractionalization. We obtain data on the annual number of natural disasters from the Emergency Events Database (EM-DAT) (Centre for Research on the Epidemiology of Disasters 2022), which includes all geophysical, meteorological, hydrological, climatological, biological, and extraterrestrial disasters that meet one of the following criteria: at least 10 deaths, at least 100 affected people, a state of emergency declaration, or call for international assistance. The summary statistics for natural disaster data, converted into 5-year averages, is reported in Table 1.

Natural disasters can test the strength of the government, thereby influencing its control over its own territory. First, political instability may arise because of the perceived failure of government to provide equitable and adequate assistance to the disaster-hit areas that erodes government’s legitimacy in the eyes of the public (Olson and Drury 1997; Pelling and Dill 2006).¹⁷ Second, disasters can worsen pre-existing societal divisions. Politically weak or hostile areas tend to receive less government investment on disaster prevention. Moreover, marginalized groups are disproportionally affected by disasters (Cohen and Werker 2008; Pelling and Dill 2006). The literature shows that natural disasters affect terrorism (Berrebi and Ostwald 2011; Kingdon and Gray 2022), raising the possibility that disasters may influence the likelihood of terrorist groups ending through variables other than state weakness. This is why we interact the number of disasters in a terrorist group’s base country with the latter’s level of ethnic fractionalization.¹⁸

For an instrument to be valid, it must be correlated with the instrumented variable and the exclusion restriction hold. The degree to which natural disasters promote state failure may be bolstered by the country’s ethnic division, which adds stresses to a country already besieged by a natural disaster. Additionally, natural disasters in ethnically divided societies are more difficult to mitigate and may exacerbate group grievances (Nel and Righarts 2008). Conversely, natural disasters have been shown to disrupt and undermine violent nonstate actors (Salehyan and Hendrix 2014), pacify intrastate conflicts by unifying divided countries (Slettebak 2012), and prompt governments to end civil conflicts through concessions to armed actors (Kreutz 2012). Later, we will show statistically that our instrument is relevant.

Concerning the excludability restriction, we are using the number of natural disasters, rather than the resulting damages as the latter may hinge on factors that can threaten the exogeneity condition. Terrorist groups’ survivability does not determine the frequency of natural disasters. Unobserved factors that affect terrorism are also not expected to be correlated with the frequency of natural disasters. Still, the criteria used for inclusion of a natural disaster event into EM-DAT could be questioned. For example, the average number of natural disaster events are greater in low-income sample countries because the number of people affected is generally higher in poor countries, making natural disaster events there more likely to be included in the dataset. However, state of emergency is declared frequently in wealthy countries so based on this criterion more events in developed countries are expected to be selected so that the inclusion criteria do not necessarily explain the difference. Furthermore, we control for income per capita, which is a measure of a country’s level of development. Next, some countries are more vulnerable to natural disasters than others owing to their geographical locations; hence, we control for regional fixed effects and a country’s geographical characteristics. To account for any time-specific global shocks that may correlate with both the disasters and terrorist group’s survivability, we add period dummy variables as a robustness check. Conditional on these, the occurrence of natural disasters is plausibly exogenous to terrorist groups’ longevity. Further, we hold the average levels of natural disasters and ethnic diversity constant by adding the number of natural disasters and a measure of ethnic diversity to our regressors; only their interaction is used as an instrument. This allows us to control for potential independent effects of natural disaster and ethnic diversity on a terrorist groups’ termination. Hence, the identification strategy relies on the assumption that there are no omitted relevant variables that are simultaneously correlated with both natural disasters and ethnic diversity. This cannot be completely ruled out, although we control for likely covariates that might threaten the excludability condition. Our results should be interpreted with those caveats in mind.

Results and Robustness

We begin by examining the impact of state failure on resident terrorist groups’ demise based upon a simple logistic regression with duration variables and no other control variables – see Models 1 and 2 in Table 2. The duration dependence is specified as quadratic. To measure state weakness (strength), we use a dummy variable for state collapse in Model 1 and the percentage of territory controlled in Model 2. The coefficient in Model 1 is negative, indicating that a state collapse reduces a resident terrorist group’s likelihood of ending. Similarly, the positive coefficient in Model 2 suggests that as a country strengthens (loosens) control over its own territory, the probability of a terrorist group terminating increases (decreases). These findings support our first hypothesis that failed or weak states are favorable to resident terrorist groups’ survivability. Since low income per capita is used as a measure of state failure (Coggins 2015; George 2018; Lai 2007), we note that our results hold without income per capita (Models 1-2) and with income per capita (Models 3-6). In the latter case, income per capita is not a significant determinant of resident terrorist groups’ demise when our state failure measure is included. We also find that the correlation between our state failure measures and income per capita is low.

Table 2.

Binary Response Regressions of Terrorist Group Termination.

Variable	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
SF	−0.601* (0.308)		−1.410** (0.648)
Territory control		0.020*** (0.006)		0.025** (0.010)	0.526*** (0.149)	0.026*** (0.007)
Territory control squared					−0.003*** (0.001)	−0.000*** (0.000)
Left wing			1.313*** (0.345)	1.306*** (0.348)	1.305*** (0.349)	0.102*** (0.025)
Nationalist			0.983*** (0.313)	0.941*** (0.310)	0.970*** (0.310)	0.060*** (0.016)
Right Wing			1.532*** (0.478)	1.494*** (0.471)	1.473*** (0.468)	0.135** (0.058)
Share trans. Attacks			2.638*** (0.327)	2.598*** (0.338)	2.667*** (0.351)	0.460*** (0.052)
Attack diversity			−1.382** (0.623)	−1.296** (0.640)	−1.249* (0.643)	−0.161*** (0.048)
log(GDP/POP)			0.145 (0.144)	0.096 (0.148)	0.172 (0.153)	0.014 (0.013)
log(POP)			−0.111 (0.090)	−0.089 (0.090)	−0.147 (0.091)	−0.011 (0.009)
Polity2			−0.012 (0.021)	−0.007 (0.022)	−0.005 (0.022)	−0.000 (0.002)
Ethnic fractionalization			0.450 (0.641)	0.727 (0.636)	0.780 (0.657)	0.091 (0.059)
EAP			−0.657 (0.473)	−0.700 (0.481)	−0.721 (0.476)	−0.058 (0.036)
ECA			−0.279 (0.347)	−0.325 (0.373)	−0.377 (0.364)	−0.027 (0.038)
LAC			−0.067 (0.450)	−0.107 (0.459)	−0.186 (0.466)	−0.002 (0.052)
NA			−0.018 (0.480)	−0.222 (0.511)	0.057 (0.506)	−0.004 (0.067)
SAS			−0.732 (0.491)	−0.865 (0.566)	−0.925* (0.556)	−0.063 (0.042)
SSA			−1.074* (0.559)	−1.159** (0.579)	−1.124* (0.592)	−0.115** (0.052)
log(Elevation)			−0.069 (0.162)	−0.090 (0.160)	−0.138 (0.167)	−0.017 (0.016)
Tropics			−0.187 (0.423)	−0.120 (0.416)	−0.254 (0.426)	−0.015 (0.038)
Landlock			0.358 (0.364)	0.386 (0.362)	0.210 (0.364)	0.028 (0.034)
NT	1955	1947	1678	1670	1670	1670

Notes: Significance levels: *** is < 0.01, ** is < 0.05, and * is < 0.10. Robust standard errors are in parentheses. Constant and duration variables are suppressed. The duration dependence pattern is specified as quadratic. SF is state failure variable and NT is the sample size. Models 1-5 are estimated using the logistic regression, whereas Model 6 is estimated using the linear probability model.

We re-estimate Models 1 and 2 by adding a full set of control variables and our baseline results are confirmed in Models 3 and 4. The effects of covariates are consistent with findings in the literature. Religious fundamentalist terrorist groups (the excluded category) are more likely to survive than terrorist groups with alternative ideologies. Terrorist groups that perpetrate a larger share of transnational terrorist attacks are more likely to end, whereas terrorist groups that diversify their terrorist attacks are less inclined to end.

Consistent with the second hypothesis, we investigate the nonlinear relationship by adding the squared term of a measure of territorial control to the full specification (Model 5). The impact of Territory control is positive, while that of Territory control Squared is negative. Given that the squared term is challenging to interpret in nonlinear models, we plot the predicted probability of a terrorist group’s terminating at different values of Territory control, which ranges between 32 and 100 percent in our sample (Table 1). Figure 2 displays the plot with 95% confidence intervals.

Figure 2.

Predicted probability of terrorist group termination (95% CIs).

At a lower value of Territory control, the impact of territorial control is statistically not significant. As a governmental territorial control grows beyond 53 percent, the predicted probability of a resident terrorist group’s termination rises, first at an increasing rate and then at a decreasing rate, and reaches its peak at about 16 percent, which corresponds to the government controlling around 87 percent of its territory. Thereafter, the predicted probability of a resident terrorist group’s termination, in response to government’s expansion of its territorial authority, falls but remains positive and above 10 percent. The nonlinear finding is consistent with H2. In Model 6, we apply the Linear Probability Model and support the result.

We perform a number of robustness checks. First, we re-estimate Models 3 and 4 in Table 2 by replacing the quadratic duration variable with a piecewise constant specification using decade dummies. Second, we apply the random-effects logit estimator to control unobserved heterogeneities. Third, around 13 percent of terrorist groups in our sample concluded during the study period. Given that the outcome is rare, we implement complementary log-log and the random-effects complementary log-log regressions. We also test the nonlinear relationship using the random-effects complementary log-log estimator. Our main conclusion, which is reported in Table 3, holds with full results displayed in online Appendix Table A1.

Table 3.

Binary Response Regressions of Terrorist Group Termination.

Variable	Logit	Logit	RE	RE	CLL	CLL	RE CLL	RE CLL	RE CLL
SF	−1.067 (0.658)		−1.177* (0.709)		−0.985* (0.572)		−1.025* (0.598)
Territory control		0.022** (0.010)		0.026** (0.013)		0.017* (0.009)		0.019* (0.011)	0.521*** (0.146)
Territory control squared									−0.003*** (0.001)
NT	1678	1670	1678	1670	1678	1670	1678	1670	1670

Notes: Significance levels: *** is < 0.01, ** is < 0.05, and * is < 0.10. Robust standard errors are in parentheses. Constant, control, and duration variables are suppressed. The duration dependence pattern is specified using decade dummies. See Models 3-5 in Table 2 for the list of included control variables. SF is state failure variable and NT is the sample size. RE, CLL, and RE CLL denote random-effects logit, complementary log-log, and random-effects complementary log-log, respectively.

Further, we restrict our analysis to a small subsample of terrorist groups that ended at some point during our sample period. We also exclude terrorist groups that were born before the beginning of our sample period, 1970. Generally, our main conclusion is maintained; these regression outputs are reported in online Appendix Tables A2 and A3. Finally, we expand Models 3 and 4 by controlling for civil wars, corruption, and infant mortality. Our civil war variable is the annual count of armed conflicts taking place within a country involving the state and one or more internal opposition groups (Gleditsch et al. 2002). The corruption indicator denotes the perceived overall corruption level from 0 (no corruption) to 100 (pervasive corruption) (Standaert 2015). Data on infant mortality, measured by the number of infants dying before reaching age one per 1000 live births, is obtained from the World Bank (2021). Our results still hold (available upon request).

Instrumental Variables Approach

So far, we find a robust and statistically significant correlation between state failure and terrorist groups’ likelihood of ending. However, the identification strategy does not allow us to account for possible endogeneity of state failure; thus, we turn to the instrumental variables approach. Table 4 presents the first- and second-stage outputs for the main variables of interest, where online Appendix Table A4 reports the full results for the second stage. We apply the interaction between natural disasters and ethnic fractionalization (Models 1-4) and the logarithm of natural disasters and ethnic fractionalization (Models 5-8) as alternative instruments for state failure. We focus on the linear relationship; lack of an additional instrument does not allow us to test the quadratic relationship. The first-stage results show that our instrument is statistically significantly correlated with state failure (large first-stage F-statistics), signifying our instrument’s relevance.

Table 4.

Instrumental variables regressions of terrorist group termination.

Variable	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
SF	−0.865** (0.355)		−0.488* (0.284)		−0.607** (0.258)		−0.221** (0.110)
Territory control		0.011*** (0.004)		0.017* (0.010)		0.012** (0.005)		0.008* (0.004)
Natural disaster	−0.009*** (0.002)	−0.010*** (0.002)	0.011*** (0.004)	0.006** (0.003)
log(Natural Disaster)					−0.042*** (0.016)	−0.057*** (0.022)	0.033** (0.013)	0.031** (0.015)
Ethnic fractionalization	−0.046 (0.056)	0.115 (0.101)	0.174* (0.094)	0.447* (0.239)	−0.056 (0.053)	0.182 (0.140)	0.079 (0.060)	0.217* (0.121)
First-stage regressions
ND*EF	−0.031*** (0.008)	2.311*** (0.280)	−0.037*** (0.009)	1.059*** (0.260)
log(ND)*EF					−0.121*** (0.031)	5.426*** (0.945)	−0.237*** (0.027)	5.952*** (0.840)
Natural disaster	0.006* (0.003)	−0.327*** (0.113)	0.024*** (0.004)	−0.409*** (0.113)
log(Natural Disaster)					−0.002 (0.013)	1.608*** (0.448)	0.064*** (0.011)	−0.843** (0.389)
Ethnic fractionalization	0.249*** (0.054)	−33.262*** (1.978)	0.409*** (0.078)	−27.984*** (2.145)	0.261*** (0.046)	−31.406*** (1.525)	0.365*** (0.060)	−27.135*** (1.693)
NT	1933	1925	1678	1670	1933	1925	1678	1670
First-stage F statistic	14.15	68.20	17.93	16.63	14.84	32.95	74.65	50.24

Notes: Significance levels: *** is < 0.01, ** is < 0.05, and * is < 0.10. Robust standard errors are in parentheses. Constant, control, and duration variables are suppressed. The duration dependence pattern is specified as quadratic. Models 1-2 and 5-6 are baseline models that include constant and duration variables. Models 3-4 and 7-8 are fully specified – see Models 3-4 in Table 2 for the list of included control variables. SF is state failure variable, NT is the sample size and ND*EF (log(ND)*EF) denotes interaction between natural disaster and ethnic fractionalization (log of natural disaster and ethnic fractionalization).

The instrumental variables findings support our main conclusion: state failure (greater governmental territorial control) reduces (increases) the probability of a resident terrorist group’s ending. A state collapse is associated with about 48 percentage-points reduction in the probability of a resident terrorist group’s termination (Model 3). A one-percent increase in territorial control raises the likelihood of a resident terrorist group’s termination by 1.7 percentage points.

As a robustness check, we replace the duration variable with period dummies to control for the time effects, and the results hold (available upon request). We then create two placebo instruments to rule out the possibility that our findings are artifacts of spurious correlations. First, we randomly shuffle the values of the interaction between natural disasters and ethnic diversity across terrorist groups’ base countries for a fixed year. For example, in a given year, the value of the instrument corresponding to a country housing terrorist group A may be assigned to a country housing terrorist group B. Natural disasters and ethnic diversity in a base country should not systematically determine state failure in another base country for a specific year. Therefore, the absence of statistical relationship lends support to the soundness of our research design. Similarly, we randomly rearrange the values of the instrument across the sample so that the values of the instrument are reordered across both base countries and years. The results are reported in online Appendix Table A5. Across models, small values of the first-stage F-statistics and statistically insignificant relationship between placebo instruments and state failure signify that the placebo instruments are not relevant. As expected, the effect of state failure is not statistically significant; the randomly constructed instruments do not identify the effect of state failure on terrorist group’s survivability.

Concluding Remarks

At the outset, we develop and present a theory in which failed states favor the survivability of resident terrorist groups owing to a host of identified factors that stymie government’s counterterrorism actions while facilitating resident groups’ campaigns. Moreover, as a state’s control of its territory expands as a percent, we argue that resident terrorist groups’ survivability decreases until some threshold control percentage is attained. Our novel theory can facilitate counterterrorism by identifying the Achille’s heel of the agent – resident terrorist groups –causing terrorism in failed states.

Unlike the extant literature on failed or weak states and terrorism, we investigate this alleged theoretical association by focusing on the prospects of weak-state-resident terrorist groups based on two measures of state failure. For the main empirical models, a nonlinear, inverted U-shaped relationship characterizes a state’s extent of territorial control and the demise likelihood of its resident terrorist groups. Namely, greater state control of its territory increases, at a diminishing rate, the likely termination of resident terrorist groups up to a threshold level of territorial control. Our explicit analysis of a nonlinear relationship between state failure and terrorism, although mentioned, has not been previously tested or theorized in the literature (see, e.g., George 2018; Piazza 2008). In answer to our question, do failed or weak states favor resident terrorist groups’ survival, our analysis indicates that these weak states do favor such groups’ longevity, and quantifies that influence in terms of both a discrete and continuous measure of state failure. To address inherent endogeneity concerns, previously recognized but never addressed, we develop a novel instrument that identifies a robust relationship between state weakness and resident terrorist groups’ survivability. The causal relationship here is supported by our identification strategy and myriad robustness tests. Additionally, our instrumental approach is bolstered by placebo tests. The findings from this paper suggest that countries confronted by terrorists from abroad must take an interest in eradicating resident terrorist groups in weak states. Such efforts will require cooperation among targeted countries and those that house the terrorist groups. Of course, an alternative strategy would be to foster the host states’ ability to control its territory.

Supplemental Material

Supplemental Material - Do Failed or Weak States Favor Resident Terrorist Groups’ Survival?

Supplemental Material for Do Failed or Weak States Favor Resident Terrorist Groups’ Survival? by Khusrav Gaibulloev, James A. Piazza, and Todd Sandler in Journal of Conflict Resolution

Supplemental Material

Supplemental Material - Do Failed or Weak States Favor Resident Terrorist Groups’ Survival?

Supplemental Material for Do Failed or Weak States Favor Resident Terrorist Groups’ Survival? by Khusrav Gaibulloev, James A. Piazza, and Todd Sandler in Journal of Conflict Resolution

Footnotes

Acknowledgements

While assuming full responsibility for the content, the authors gratefully acknowledge helpful comments from two anonymous reviewers and Paul Huth on an earlier draft.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

James A. Piazza

Todd Sandler

Supplemental Material

Supplemental material for this article is available online

Notes

References

Alesina

Alberto

Devleeschauwer

Arnaud

Easterly

William

Kurlat

Sergio

Wacziarg

Romain

. 2003. “Fractionalization.” Journal of Economic Growth 8 (2): 155-94.

Allison

Paul D.

1982. “Discrete-Time Methods for the Analysis of Event Histories.” Sociological Methodology 13: 61-98.

Baskaran

Thushyanthan

Min

Brian

Uppal

Yogesh

. 2015. “Election Cycles and Electricity Provision: Evidence from a Quasi-Experiment with Indian Special Elections.” Journal of Public Economics 126: 64-73.

Berrebi

Claude

Jordan

Ostwald

. 2011. “Earthquakes, Hurricanes, and Terrorism: Do Natural Disasters Incite Terror?” Public Choice 149 (3): 383-403.

Blomberg

S. Brock

Engel

Rozlyn C.

Sawyer

Reid

. 2010. “On the Duration and Sustainability of Transnational Terrorist Organizations.” Journal of Conflict Resolution 54 (2): 303-30.

Blomberg

S. Brock

Gaibulloev

Khusrav

Sandler

Todd

. 2011. “Terrorist Group Survival: Ideology, Tactics, and Base of Operations.” Public Choice 149 (3–4): 441-63.

Campos

Nauro F.

Gassebner

Martin

. 2013. “International Terrorism, Domestic Political Instability, and the Escalation Effect.” Economics & Politics 25 (1): 27-47.

Central Intelligence Agency . 2018. The World Factbook. Washington, DC: Central Intelligence Agency. https://www.cia.gov/the-world-factbook/. Accessed April 2, 2022.

Centre for Research on the Epidemiology of Disasters . 2022. The Emergency Events Database (EM-DAT). UCLouvain. https://www.emdat.be/. Accessed on April 3, 2022.

10.

Coggins

Bridget L.

2015. “Does State Failure Cause Terrorism? An Empirical Analysis (1999–2008).” Journal of Conflict Resolution 59 (3): 455-83.

11.

Cohen

Charles

Werker

Eric D.

. 2008. “The Political Economy of “Natural” Disasters.” Journal of Conflict Resolution 52 (6): 795-819.

12.

Coppedge

Michael

Gerring

John

Knutsen

Carl Hendrik

Lindberg

Staffan I.

Jan

Teorell

David

Altman

Michael

Bernhard

Agnes

Cornell M.

Steven

Fish

Lisa

Gastaldi

Haakon

Gjerløw

Adam

Glynn

Allen

Hicken

Anna

Lührmann

Seraphine

F. Maerz

Kyle

L. Marquardt

Kelly

McMann

Valeriya

Mechkova

Pamela

Paxton

Daniel

Pemstein

Johannes

von Römer

Brigitte

Seim

Rachel

Sigman

Svend-Erik

Skaaning

Jeffrey

Staton

Aksel

Sundtröm

Eitan

Tzelgov

Luca

Uberti

Yi-ting

Wang

Tore

Wig

Daniel

Ziblatt

. 2020. V-Dem Codebook v11.l. Varieties of Democracy (V-Dem) Project. https://www.v-dem.net/vdemds.html. Accessed on November 9, 2021.

13.

Cronin

Audrey K.

2009. How Terrorism Ends: Understanding the Decline and Demise of Terrorist Campaigns. Princeton, NJ: Princeton University Press.

14.

Daxecker

Ursula E.

Hess

Michael L.

. 2013. “Repression Hurts: Coercive Government Responses and the Demise of Terrorist Campaigns.” British Journal of Political Science 43 (3): 559-77.

15.

Dreher

Axel

Gassebner

Martin

Paul

Schaudt

. 2020. “The Effect of Migration on Terror: Made at Home or Imported from Abroad?” Canadian Journal of Economics 53 (4): 1703-44.

16.

Enders

Walter

Sandler

Todd

. 2012. The Political Economy of Terrorism. 2nd ed. New York: Cambridge University Press.

17.

Enders

Walter

Sandler

Todd

Gaibulloev

Khusrav

. 2011. “Domestic Versus Transnational Terrorism: Data, Decomposition, and Dynamics.” Journal of Peace Research 48 (3): 319-37.

18.

Gaibulloev

Khusrav

. 2015. “Terrorist Group Location Decision: An Empirical Investigation.” Oxford Economic Papers 67 (1): 21-41.

19.

Gaibulloev

Khusrav

Hou

Dongfang

Todd

Sandler

. 2020. “How Do the Factors Determining Terrorist Groups’ Longevity Differ from Those Affecting Their Success?” European Journal of Political Economy 65: 1-15.

20.

Gaibulloev

Khusrav

Piazza

James A.

Sandler

Todd

. 2017. “Regime Type and Terrorism.” International Organization 71 (3): 491-22.

21.

Gaibulloev

Khusrav

Sandler

Todd

. 2013. “Determinants of the Demise of Terrorist Organizations.” Southern Economic Journal 79 (4): 774-92.

22.

Gaibulloev

Khusrav

Sandler

Todd

. 2014. “An Empirical Analysis of Alternative Ways That Terrorist Groups End.” Public Choice 160 (1–2): 25-44.

23.

Gaibulloev

Khusrav

Sandler

Todd

. 2023. “Common Myths of Terrorism.” Journal of Economic Surveys 37 (2): 271-301. doi: 10.1111/joes.12494.

24.

Gallup

John L.

Mellinger

Andrew D.

Sachs

Jeffrey D.

. 2001. Geography Datasets. Center for International Development, Harvard University.

25.

Gallup

John L.

Sachs

Jeffrey D.

Mellinger

Andrew D.

. 1999. “Geography and Economic Development.” International Regional Science Review 22 (2): 179-32.

26.

George

Justin

. 2018. “State Failure and Transnational Terrorism: An Empirical Analysis.” Journal of Conflict Resolution 63 (3): 471-95.

27.

Ghatak

Sambuddha

. 2016. “Challenging the State: Effect of Minority Discrimination, Economic Globalization, and Political Openness on Domestic Terrorism.” International Interactions 42 (1): 56-80.

28.

Ghatak

Sambuddha

Gold

Aaron

. 2017. “Development, Discrimination, and Domestic Terrorism: Looking Beyond a Linear Relationship.” Conflict Management and Peace Science 34 (6): 618-39.

29.

Gleditsch

Nils Petter

Wallensteen

Peter

Eriksson

Mikael

Sollenberg

Margareta

Strand

Håvard

. 2002. “Armed Conflict 1946-2001: A New Dataset.” Journal of Peace Research 39 (5): 615-37.

30.

Hehir

Aidan

. 2007. “The Myth of the Failed State and the War on Terror: A Challenge to the Conventional Wisdom.” Journal of Intervention and Statebuilding 1 (3): 307-32.

31.

Hou

Dongfang

Gaibulloev

Khusrav

Sandler

Todd

. 2020. “Introducing Extended Data on Terrorist Groups (EDTG), 1970–2016.” Journal of Conflict Resolution 64 (1): 199-225.

32.

Jenkins

Stephen P.

1995. “Easy Estimation Methods for Discrete-Time Duration Models.” Oxford Bulletin of Economics and Statistics 57 (1): 129-38.

33.

Jones

Seth G.

Libicki

Martin C.

. 2008. How Terrorist Groups End: Lessons for Countering al Qa’ida. Santa Monica, CA: Rand Corporation.

34.

Lai

Brian

. 2007. “Draining the Swamp: An Empirical Examination of the Production of International Terrorism, 1968–1998.” Conflict Management and Peace Science 24 (4): 297-310.

35.

Kalyvas

Stathis

. 2006. The Logic of Violence in Civil War. New York: Cambridge University Press.

36.

Kingdon

Ashton

Gray

Briony

. 2022. “The Class Conflict Rises When You Turn Up the Heat: An Interdisciplinary Examination of the Relationship between Climate Change and Left-Wing Terrorist Recruitment.” Terrorism and Political Violence 34 (5): 1041-56.

37.

Kreutz

Joakim

. 2012. “From Tremors to Talks: Do Natural Disasters Produce Ripe Moments for Resolving Separatist Conflicts?” International Interactions 38 (4): 482-502.

38.

LaFree

Gary

Dugan

Laura

Fahey

Susan

. 2007. “Global Terrorism and Failed States.” In Peace and Conflict 2008, edited by Hewitt

J. Joseph

Wilkenfeld

Jonathan

Gurr

Ted Robert

, 39-54. London: Routledge.

39.

Marshall

Monty G.

Keith

Jaggers

Gurr

Ted R.

. 2021. Polity IV Project: Dataset and Users’ Manual. Vienna, VA: Center for Systemic Peace: Polity IV Project. https://www.systemicpeace.org/inscrdata.html. Accessed on November 9, 2021.

40.

Menkhaus

Ken

. 2003. “Quasi-States, Nation-Building, and Terrorist Safe Havens.” Journal of Conflict Studies 23 (2): 7-23.

41.

Mickolus

Edward F.

Sandler

Todd

Murdock

Jean M.

Flemming

Peter

. 2021. International Terrorism: Attributes of Terrorist Events, 1968–2020 (ITERATE). Pointe Vedra, FL: Vinyard Software.

42.

Murdie

Amanda

Stapley

Craig S.

. 2014. “Why Target the ‘Good Guys’? The Determinants of Terrorism against NGOs.” International Interactions 40 (1): 79-102.

43.

National Consortium for the Study of Terrorism and Responses to Terrorism (START) . 2020. “Global Terrorism Database (GTD).” https://www.start.umd.edu/gtd. Accessed on October 3, 2020.

44.

National Intelligence Strategy . 2005. Transformation through Integration and Innovation. US Office of the Director of National Intelligence. https://stacks.stanford.edu/file/druid:xr966fk6098/NISOctober2005.pdf. Accessed on November 9, 2021.

45.

Nel

Phillip

Righarts

Marjolein

. 2008. “Natural Disasters and the Risk of Violent Civil Conflict.” International Studies Quarterly 52 (1): 159-85.

46.

Newman

Edward

. 2007. “Weak States, State Failure, and Terrorism.” Terrorism and Political Violence 19 (4): 463-88.

47.

Olson

Richard Stuart

Drury

A. Cooper

. 1997. “Un-Therapeutic Communities: A Cross-National Analysis of Post-Disaster Political Unrest.” International Journal of Mass Emergencies and Disasters 15 (2): 221-38.

48.

Pašagić

Aleksandar

. 2020. “Failed States and Terrorism.” Perspectives on Terrorism 14 (3): 19-28.

49.

Patrick

Stewart

. 2011. Weak Links: Fragile States, Global Threats, and International Security. New York: Oxford University Press.

50.

Pelling

Mark

Dill

Kathleen

. 2006. “Natural Disasters as Catalysts of Political Action.” Media Development 53 (4): 7-10.

51.

Phillips

Brian J.

2014. “Terrorist Group Cooperation and Longevity.” International Studies Quarterly 58 (2): 336-47.

52.

Phillips

Brian J.

2015. “Enemies With Benefits? Violent Rivalry and Terrorist Longevity.” Journal of Peace Research 52 (1): 62-75.

53.

Piazza

James A.

2008. “Incubators of Terror: Do Failed and Failing States Promote Transnational Terrorism?” International Studies Quarterly 52 (3): 469-88.

54.

Piazza

James A.

2021. “The Impact of Rural-Urban Economic Disparities on Terrorist Organizations’ Survival and Attacks.” Defence and Peace Economics 32 (6): 726-41.

55.

Piazza

James A.

Choi

Seung-Whan

. 2018. “International Military Interventions and Transnational Terrorist Backlash.” International Studies Quarterly 62 (3): 686-95.

56.

Plummer

Chelli

. 2012. “Failed States and Connections to Terrorist Activity.” International Criminal Justice Review 22 (4): 416-49.

57.

RAND . 2012. Database of Worldwide Terrorism Incidents. Accessed on August 12, 2012. https://www.rand.org/nsrd/projects/terrorism-incidents.html

58.

Rotberg

Robert I.

2002. “Failed States in a World of Terror.” Foreign Affairs 81 (4): 127-40.

59.

Salehyan

Idean

Hendrix

Cullen S.

. 2014. “Climate Shocks and Political Violence.” Global Environmental Change 28 (September): 239-50.

60.

Standaert

Samuel

. 2015. “Diving the Level of Corruption: A Bayesian State-Space Approach.” Journal of Comparative Economics 43 (3): 782-803.

61.

Slettebak

Rune T.

2012. “Don’t Blame the Weather! Climate-Related Natural Disasters and Civil Conflict.” Journal of Peace Research 49 (1): 163-76.

62.

Takeyh

Ray

Gvosdev

Nikolas

. 2002. “Do Terrorist Networks Need a Home?” Washington Quarterly 25 (3): 97-108.

63.

World Bank . 2021. World Development Indicators (WDI). Accessed on November 11, 2021. https://databank.worldbank.org/ddp/home.do

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.25 MB

0.00 MB

0.13 MB