The Impact of Ideological Ambiguity on Terrorist Organizations

Abstract

Organizations that have a clear and unambiguous focus acquire greater legitimacy, which raises their capacity for mobilization. Using data on terrorist organizations, this paper explores two empirical implications of this claim: A terrorist organization’s survival and lethality will be threatened to the extent that it has an ambiguous ideological identity. Analyses using panel data from the Extended Data on Terrorist Groups (EDTG) test these arguments for 474 global terrorist organizations observed over 1970–2016. The key empirical predictions are that ambiguity inhibits lethality and curtails survival. This paper finds support for these claims, controlling for competition from rivals and allies, ethno-nationalist or Islamic ideological orientation, and a variety of other measures of organizational capacity.

This paper proposes that ideological ambiguity diminishes the performance and shortens the lifetime of terrorist organizations (TOs). Two parallel lines of theoretical work inspire this claim. The first comes from a theoretical perspective on conceptual ambiguity that treats concepts as expectations about the properties of entities, such as organizations (Hannan et al. 2019: 5–7). Less clearly recognized entities are conceptually ambiguous and harder to interpret (Negro and Leung 2013). This implies that adverse consequences can result if high levels of conceptual ambiguity arise when classifying an object (Kovács and Hannan 2015; Hannan et al. 2019). Empirical work in this tradition finds that having a clearly focused identity improves the performance of many types of organizations, including restaurants, film productions, and environmental social movement organizations (Hannan et al. 2019; Hsu 2006; Kovács and Hannan 2010, 2015; Negro, Koçak, and Hsu 2010; Olzak and Johnson 2019). Conversely, ambiguous organizations tend to confront more obstacles (Hsu, Negro, and Perretti 2012).

A second line of work that draws similar inferences about the consequences of an organization’s lack of focus comes from the study of social movement organizations (hereafter SMOs). In social movement theory, a SMO that lacks focus risks being perceived as having a fuzzy or indeterminant identity (Fassioto and Soule 2017). Having a strong identity assists SMOs in demarcating group boundaries, recruiting, and reinforcing solidarity (Levy 2021; Luna 2017; Polletta and Jasper 2001). SMOs perceived as having a distinctive and well-focused identity thrive (Beck and Schoon 2019; Gamson 1975; McCarthy and Zald 1977; Wang, Rao, and Soule 2019). Theories of SMOs also specify that issue specialization assists organizations in performing routine activities and attaining goals because it streamlines coordination processes (McCarthy and Zald 1977: 1234). When such organizations are well understood, they are also more likely to be perceived as authentic members of a category (Luna 2017; Piazza and Wang 2020; Walker and Stepick 2020). Such perceptions further increase an organization’s access to valuable resources.

TOs can be usefully considered a subset of SMOs: both tend to be organized around a distinctive collective identity and express goals of political or social change (Beck 2008; Beck and Schoon 2019; Crenshaw 1987). Recognition that a particular movement is representative of a distinct type of SMO improves the likelihood that it will resonate with an audience (Benford and Snow 2000). Such resonance communicates a SMO’s purpose to current and potential supporters.

Social movement scholars have also argued that SMOs with contradictory or ambiguous ideological affiliations have a diminished capacity to mobilize (Fassioto and Soule 2017). For example, environmental social movement organizations focusing on two issues related to environmental justice and endangered species would be expected to have shorter life spans than organizations that focus on just one of these categories (Olzak and Johnson 2019). It follows that having an ambiguous ideological identity ought to hamper TOs in the same manner. An example of a TO with high ambiguity is one that was assigned to three ideologies is the Jordanian Islamic Resistance, which received labels of Islamic, anti-globalization, and racist/genocidal. A TO that received an assignment of a singular, leftist ideology is the Baader-Meinhof Group (see Table 1).

Table 1.

Ideological categories and levels of ambiguity assigned to 756 organizations.

Ideological labels assigned by archivists	Frequency	Percent	Ambiguity
Ethno-nationalist only	213	28.2	0
Leftist/Marxist only	119	15.7	0
Islamic only	122	16.1	0
Religious-not Islamic only	6	.66	0
Right-wing only	15	2	0
Environmental only	5	.66	0
Anarchist only	5	.66	0
Racist/Genocide only	3	.39	0
Other/none only	22	3	0
Tribal only	7	.9	0
Other and tribal	1	.13	.671
Other and Islamic	1	.13	.572
Other and religious-not Islamic	1	.13	.658
Leftist and ethno-nationalist	28	3.7	.635
Leftist and anarchist	18	2.4	.374
Leftist and anti-globalization	2	.26	.217
Leftist and racist	3	.39	.452
Leftist and other	3	.39	.485
Leftist and Islamic	2	.26	.727
Leftist and religious-not Islamic	1	.13	.366
Leftist and tribal	1	.13	.390
Ethno-nationalist and tribal	13	1.7	.254
Ethno-nationalist and Islamic	71	9.4	.713
Ethno-nationalist and religious-not Islamic	6	.66	.235
Ethno-nationalist and racist/Genocide	8	1.1	.304
Ethno-nationalist and anarchist	1	.13	.241
Ethno-nationalist and right	6	.66	.392
Ethno-nationalist and other	11	1.5	.333
Ethno-nationalist and environment	1	.13	.102
Religious-not Islamic and racist	1	.13	.674
Islamic and tribal	5	.66	.522
Right and Islamic	11	1.5	.560
Right and racist	11	1.5	.670
Right and other	3	.39	.683
Right and anti-globalization	1	.13	.442
Right and religious-not Islamic	3	.39	.622
Islamic and anti-globalization	3	.39	.439
Islamic and racist	2	.26	.554
Anti-globalization and racist	1	.13	.538
Environment and other	1	.13	.434
Environment and leftist	1	.13	.173
Ethno-nationalist and Islamic and racist	1	.13	.845
Ethno-nationalist and Islamic and right	1	.13	.861
Ethno-nationalist and not Islamic and racist	2	.26	.510
Ethno-nationalist and tribal and Islamic	1	.13	.823
Ethno-nationalist and left and Islamic	1	.13	1.01
Ethno-nationalist and leftist and racist	2	.26	.826
Ethno-nationalist and right and racist	1	.13	.653
Ethno-nationalist and leftist and anti-globalization	1	.13	.719
Ethno-nationalist and not Islamic and right	4	.53	.587
Left and anti-globalization and anarchist	2	.26	.554
Right and religious-not Islamic and racist	1	.13	1.04
Racist and Islamic and anti-globalization	1	.13	.640

The paper tests these ideas using the newly available dataset on TOs from the Extended Data on Terrorist Groups (hereafter EDTG). EDTG is an organizational-level dataset based on the Global Terrorism Database (GTD). The GTD provides a chronological listing of researcher verified terrorist events and, if known, the names of perpetrator organizations, and the numbers killed or injured. The original GTD dataset does not contain pertinent information on organizational characteristics, such as ideology,¹ rivals, allies, tactical diversity, or the timing of an organization’s demise. The EDTG dataset offers an unusually comprehensive picture of many of these key organizational features of a set of TOs operating worldwide over an extended period, 1970 through 2016 (see Hou, Gaibulloev, and Sandler 2020).

This paper investigates two outcomes relevant to a TO’s performance: the counts of the civilian casualties it causes and its survival over time. The paper proceeds by first describing relevant findings from prior research that motivate this study. It then offers an argument for why conceptual ambiguity might hamper the performance and survival of TOs. The next section discusses the research design, data, and measures used to analyze the effect of conceptual ambiguity. The results support the core theoretical claim that ambiguity lowers fatalities and increases the rate of organizational endings of terrorist organizations.

Prior Research on Terrorist Organizations

An underlying assumption here is that TOs engage in lethal (as well as other) tactics to spread fear (Valentino 2014). A second underlying assumption is that members of terrorist organizations want them to survive (Gamson 1975). Accordingly, studies of TOs have focused mainly on the number of attacks and/or fatalities generated (e.g., Asal and Rethemeyer 2008; Hou et al. 2020) or on the longevity of the organizations (e.g., Blomberg, Engel, and Sawyer 2010; Blomberg, Gaibulloev, and Sandler 2011; Cronin 2009).

What do we know about the tactical performance and survival of terrorist groups? Scholarship on terrorist organizations finds that more lethal TOs have an international scope, or more intra- and inter-group rivalries, or dominate and out-perform rivals, or hold a central network position (Asal, Rethemeyer, and Young. 2016; LaFree, Dugan, and Miller 2015; Phillips 2019; Sandler 2014; Valentino 2014; Young and Dugan 2014). Others find that the demise of a TO’s leader and the scope of its transnational activities significantly decrease its chances of survival (Hou et al. 2020; Olzak 2016).²

TOs that espouse ideologies related to religious and/or ethnic identities kill more victims (Asal and Rethemeyer 2008; Asal and Rethemeyer 2009; Asal and Phillips 2015; Berman 2009; Blomberg et al. 2011; Cronin 2009) and survive longer than those associated with other ideologies (Hou et al. 2020; Jones and Libicki 2008; Phillips 2014; Piazza 2009).

Only a few theories have been offered to explain the performance and survival of ethno-nationalist and religious TOs (for examples, see Berman 2009 and Olzak 2016). The theoretical argument presented here suggests that the lethality and survival of ethnic and religious TOs do not rest on their ethnic or religious nature. Instead, they thrive because they have a distinct and recognizable ideological identity. Such an argument implies that TOs that are neither religious nor ethnic but that also have this characteristic ought to benefit in the same way.

The present work builds on my previous research that found terrorist organizations that focus on a single ideological category benefit more than ones that span ideological categories (Olzak 2016). Category spanning theory holds that entities, such as organizations, that span categories will have diminished appeal and receive lower evaluations when compared to organizations that have a singular identity (Kovács and Hannan 2010). The present research refines that work theoretically by reframing the issue in terms of conceptual ambiguity. An object perceived as having high conceptual ambiguity creates disfluency (Rosch 1975; Oppenheimer 2008) and diminishes evaluations (Alter and Oppenheimer 2008; Oppenheimer and Frank 2008 ; Winkelman et al. 2003; Winkelman et al. 2006). The following sections of this paper offer and test the argument that TOs with high conceptual ambiguity ought to experience negative consequences because disfluency hampers interpretability. To my knowledge, this is the first empirical test of this hypothesis.

The present paper also differs from Olzak (2016) by operationalizing conceptual ambiguity precisely in formal terms as the probability of a given categorical assignment (Hannan et al. 2019). Compared to Olzak (2016), the present research analyzes a larger and more updated dataset that includes time-varying control variables observed over a longer period. This improves over my earlier analysis that used cross-sectional covariates.

The Argument: Ambiguity and its Consequences

Ambiguity refers to the difficulty of interpretation. In technical terms, the ambiguity of an object is the evenness of the probabilities being classified in each of the set of relevant concepts (Hannan et al. 2019: 168). In the present study, concepts refer to the ideological categories published in historical accounts by scholars, journalists, and archives that have characterized TOs by their ideologies.

Ambiguity is a characteristic of a position in a semantic space (defined as the set of values of the relevant features of an entity). Hannan et al. (2019: 164) propose that “objects represented as positions with a high categorization probability for only one concept have low conceptual ambiguity; positions that have an even distribution of categorization probabilities have maximal conceptual ambiguity.”

The key substantive claim is that ambiguity lowers fluency, the ease of processing information about an object or an organization. Research in organizational sociology also finds that disfluency gives rise to negative affect and lowers appeal (Negro and Leung 2013). Based on these findings, ambiguity ought to have a similar effect on the valuations of terrorist organizations.

Ambiguous and Focused Identities

TOs are routinely referred to by one or more ideological identities, such as leftist, ethno-nationalist, Islamic, and so on (e.g., Asal and Rethemeyer 2008; della Porta 2013; Jones and Libicki 2008). References to ideological categories locate an organization in semantic space, described above in terms of concept categorization. A TO’s affiliation with recognizable ideological categories helps others make sense of a TO’s goals and activities (Beck and Schoon 2019, della Porta 2013; Luna 2017).

Why would the tactical performance of TOs be affected by having a focused ideological identity? The recognition of an organization as a typical instance of a category influences how people think about it, especially how they process information about it. Having a highly focused identity allows an organization’s culture to be easily interpreted (Sørensen 2002). When organizations are well understood, they are also more likely to be perceived as reliable (Hannan and Freeman 1989; Hannan, Pólos, and Carroll 2007; Hsu et al. 2009; Selznick 1960). Perceptions of reliability further increase a TO’s members’ confidence in performing difficult tasks, which is relevant when considering the array of dangerous activities that are part of a TO’s repertoire.

A focused ideology offers a way to place the motivation for violence within an understandable context (Chou 2016). Affiliation with a well-known ideological identity—for instance, leftist—provides a recognizable rationale for committing terrorist acts (della Porta 2013). Conversely, a TO that has equal probabilities of being classified Islamic, anti-globalization, and racist/genocidal would be harder to interpret.

Having a well-bounded identity enhances perceptions of authenticity, which increases its cognitive legitimacy.³ As Luna (2017) argues, “authenticity requires coherence.” Having an ambiguous identity raises questions about authenticity (Hsu, Hannan, and Pólos 2011; Walker and Stepick 2020). Less authentic TOs seen as weak become more vulnerable to attack by authorities and rivals (Davenport 2015). Such vulnerability diminishes a TO’s chances of survival (della Porta 2013).

Ambiguity and Longevity

TOs frequently dissolve due to acts of betrayal, internal conflict, attacks by rivals or authorities, and leadership struggles (Asal, Brown, and Dalton 2012; Chou 2016; Crenshaw 1991, 2011; Cronin 2009; Davenport 2015; della Porta 2013; LaFree and Dugan 2009; Young and Dugan 2014). Compared to TOs with an ill-defined or contradictory ideological focus, a TO with a highly focused identity ought to have a greater capacity for suppressing internal conflicts. Such a focus also would help resolve disputes over leadership succession (Cronin 2009). Like all SMOs, TOs rely on mechanisms that increase internal solidarity and loyalty, which can be undermined if some of its members express doubt about an organization’s focus.

Organizations that straddle different ideologies are more likely than focused ones to attract a heterogeneous pool of followers. Such heterogeneity risks a loss of cohesion (Wang et al. 2019). Membership heterogeneity in turn increases an organization’s vulnerability to internal conflicts (Asal et al. 2012; Cronin 2009; della Porta 2013). Indeed, Davenport (2015) claims that along with repression, internal disputes are a key cause of a SMO’s demise.

Organizations with ambiguous identities likely develop fissures that undermine leadership authority (La Free and Dugan 2009). A TO leader’s diminished authority hampers efforts to accomplish hazardous missions, such as suicide bombings. In contrast, a leader’s authority that is reinforced by a cohesive ideology allows TO members to act decisively when facing a threat (Berman 2009). If this argument holds, then ideological clarity would also improve a TO’s chance of survival (Conrad and Greene 2015; Cronin 2009).

Ambiguity and Recruitment

Finally, as noted above, ambiguity generally lowers evaluations of an organization (Hannan et al. 2019; Kovács and Hannan 2010). Diminished evaluations could hamper an organization’s ability to attract newcomers. Having a focused ideological identity ought to streamline the recruitment process by framing difficult actions (such as suicide bombing) as a duty required by a movement’s core ideology. According to Webber, et al. (2020: 108), “Ideologies, then, can promote violence if they associate political goals with the need to feel significant.” TOs with coherent and well-understood ideologies are better positioned to make these claims.

Recruits regularly enter TOs and existing members leave them, which creates difficulties for maintaining continuity and commitment (Cronin 2009; della Porta 2013). When an audience assigns membership in a single recognizable ideology to a TO, it can attract those most committed to that particular cause.

TOs try to shape these claims of authenticity and coherence by regularly posting mission statements and manifestos on social media. Kinney, Davis, and Zhang (2018) analyzed some of these postings and found that a TO’s recruitment process is streamlined when these postings signal its ideological commitment and cognitive legitimacy. Coordination costs will be lower to the extent that a TO’s ideological identity can be taken for granted (Post, Sprinzak, and Denny 2003). Overall, having a sharp ideological focus engages members who can be induced to commit terrorism. The specific hypotheses are:

H1: Terrorist organizations that have high conceptual ambiguity will cause fewer casualties per year than organizations that have relatively low concept ambiguity.

H2: Terrorist organizations that have high conceptual ambiguity will be less likely to survive and will be more likely to cease operating as an organization.

Data and Measures

This analysis uses a publicly available and comprehensive data set (EDTG), created by Hou et al. (2020), which includes information on 760 terrorist organizations listed by names and dates of attack included in the Global Terrorism Database (GTD). The GTD Codebook (2019: 10) “defines a terrorist attack as the threatened or actual use of illegal force and violence by a non-state actor to attain a political, economic, religious, or social goal through fear, coercion, or intimidation.”

Using these criteria, Blomberg et al. (2011) began with a listing of terrorist attacks that occurred from 1970 through 2007. They then identified names of TOs if they were associated with these events and verified by GTD. Their inclusion criteria closely adhere to GTD’s inclusion criteria listed above. In addition, Blomberg et al. (2011) eliminated all incidents designated as “Doubt Terrorism Proper” by GTD (Hou et al. 2020: 203; GTD Codebook 2019: 11). This yielded a dataset of 367 TOs.

EDTG researchers expanded these data to include a total of 760 TOs that operated between 1970 through 2016.⁴ The unit of analysis is an organizational year, where observations begin in 1970. Each organization’s event-history record begins with its first attack (observed after January 1, 1970) and ends at the end of December 2016. The EDTG dataset collected information on organizational features, taking care to identify numerous spelling differences, name changes, and other relevant characteristics.⁵

Using LEXIS/NEXIS, Mapping Militants, Rand Corporation, and Wikipedia, I then cross-validated all information on each organization listed in EDTG (including searches for alternative spellings in native languages). Many scholars, the media, and archives identify TOs by adding initials after their name (e.g., South West African People’s Organization or SWAPO). These initials greatly assist in distinguishing among similarly named organizations. Location of operations, leadership names, and other corroborating factors were also used here to cross-check the data. Due to missing data on size, fatalities, the number of attacks, and other measures, Huo et al.’s (2020: 217) full model analyzed 339 organizations from the EDTG dataset. I tracked down some of this missing data and my final dataset includes 474 organizations with full information on ideological assignments, survival, lethality, and control variables. By the end of 2016, 47 percent (225) of these organizations were still active.

Lethality

One measure of tactical performance of TOs used by a variety of researchers counts the number of fatalities generated.⁶ This research tradition treats civilian fatalities as part of a TO’s tactical repertoire, but it does not assume fatalities are always a key or sole objective (Asal et al. 2015: 402). The analysis builds on this tradition by exploring the effects of ambiguity on lethality: the annual number of civilian deaths from attacks in which a named organization was listed as responsible by GTD, cross-validated by researchers Hou et al. (2020) and re-checked by this author.

Note that some researchers (e.g., Horowitz and Potter 2014; Olzak 2016) have used the ratio of deaths per attack, while others use a straightforward count of fatalities. To distinguish the influence of lethality from the volume of attacks, I specify a model that also includes the number of attacks generated. The reasoning here is that number of civilian deaths in a year depends upon an organization generating one or more attacks in that same year (Clauset and Gleditsch 2012). Hou et al. (2020: 5) define this measure as “the total number of deaths caused by a terrorist group in a given year.” Following GTD, the EDTG data also allows me to distinguish deaths of terrorists from non-terrorists.⁷ My analysis focuses only on fatalities of non-terrorists.

Terrorist Organizations Endings

The second dependent variable is the hazard of ending by various means, with an organization’s observed lifetime measured in years that elapse from an organization’s first observed attack (beginning in 1970) through the last recorded attack of violence or the end of the observation period. An organization’s ending event was first coded by EDTG researchers and validated by this author, using all sources listed in the next section.

Given my theoretical focus on organizations, I treat a TO as having ended if a TO no longer engages in terrorist activity as an organizational entity (e.g., that it no longer operates as the same, single TO).⁸ This ending category combines subtypes of endings coded by EDTG that meet these criteria: defeat by military or police, splintered from within, merge with other groups, or inactive (EDTG defines inactive if a TO had no recorded attacks for five consecutive years).⁹

The second major type of ending occurs when TOs ceases terrorist activity by joining the political system and/or meeting their objectives. These TOs do not necessarily cease organizational activities in the same way as ones that are defeated militarily or ended by internal splits. For instance, leaders and their followers may join an existing political party or establish a new one, sometimes using the same name.¹⁰ Thus, I analyze two qualitatively ways a TO might end: (1) by ending terrorist activity as an organization (2) by entering the political system (or meeting its goals).

Sources of Information on Ideological Assignments

Coding of all ideological categories assigned was conducted by this researcher by systematically searching all archives by each organization’s name.¹¹ I used aliases recommended by prior researchers, including different spellings uncovered during these searches. Most datasets on terrorist groups (e.g., Asal, Rethemeyer, and Anderson 2008; Huo et al. 2020; Jones and Libicki 2008), assign just one ideological label for each organization. In contrast, other sources such as Mapping Militants (Crenshaw 2020), Tracking Terrorism (2020), Wikipedia, the South Asian Terrorism Portal, the Guardian, and the BBC often assign more than one ideological affiliation.

Evaluating my theoretical argument requires that I take all ideological affiliations into account that were attributed to a TO. This procedure yielded a finite set of ideological labels that were assigned by these sources either separately or in combination: ethno-nationalist, leftist/Marxist, right-wing, religious-Islamic, religious-not-Islamic, anti-globalization, environmental, anarchist, racist/genocidal, tribal, and other (e.g., not elsewhere classified). Given the heightened interest surrounding violence involving Islamic TOs (Asal et al. 2015), I distinguish Islamic-Religious from Non-Islamic Religious (many scholars use only a single label of “religious”). An organization was listed as being assigned to a given ideological category only if it appeared listed by that label by a source, either singly or in combination with other categories. For example, the South Asian Terrorist Portal describes the group “All Adivasi National Liberation Army” as having a tribal identity:¹²

“The All Adivasi National Liberation Army (AANLA) was formed in the second half of 2006. The AANLA claims to be fighting to safeguard the tribal culture of the plantation workers whose ancestors were brought from northern India by British colonialists. The outfit demands Scheduled Tribe (ST) status for the Adivasi community and rehabilitation of the displaced members of its community.”

The full dataset of 756 organizations shows that the number of ideological affiliations ranges from one to three categories (see Table 1 for the distribution of all combinations listed).

Measuring Ideological Focus and Ambiguity

The measure of ambiguity requires estimates of the probability of categorization in each ideological category. The categorical probability is the likelihood that an object will be perceived to be a more or less typical illustration of a particular concept, c. The objective is to measure the ambiguity of an object with respect to a cohort of relevant concepts. The ambiguity A, of an object $o$ at position x in the feature space, is defined as the entropy of the set of categorization probability over the full cohort of 756 TOs

A (o) = - \sum_{c \in C} P (c, x) \ln P (c | x)

(1)

where denotes the relevant cohort of concepts (here ideologies). Calculation of categorization probabilities can be done by using Bayes’rule¹³

P (c | x) = \frac{π (x | c) p (c)}{\sum_{c \in C} π (x | c) p (c)}

(2)

where

π (x | c)

is the so-called concept likelihood and p(c) is the prior that on any object chosen at random will turn out to be an instance of c. This prior is set equal to the observed proportion.

However, in this analysis, as well as most other prior studies on ambiguity, only categorizations (label assignments) are available. This paper follows the procedure introduced by Hannan et al. (2019) to estimate categorization probabilities, P(c|x), from vectors of label assignments using estimates of typicalities. The key idea is to use the following approximation for the categorization. probability

P^{*} (c, o) = \frac{τ (c, o) p (c)}{\sum_{c^{'} \in C} τ (c', o) p (c')}

(3)

where

τ (c, o)

denotes the typicality of o as an instance of the concept c. The approximation is justified by the observation that the concept likelihood and typicality represent the similarity of the object and the concept.

The typicality of an object with respect to a concept is defined as goodness of representation (Rosch 1975). Typical objects have feature values that are common for instances of the concept. Lacking feature values, prior research has approximated typicality from vectors of label assignments. Kovács and Hannan (2015: 257) argue that the typicality of an object in any relevant concept falls with: “(1) the number of labels used to describe it; and (2) the distances among the schemas associated with those labels.” These authors propose that the typicality of the object o as a c be measured as follows

T (c, o) = \frac{l (c, o)}{1 + \sum_{c, c'} l (c, o) l (c^{'}, o) d (c, c^{'})}

(4)

where l(c,o) is an indicator that equals one if the object o is labeled as a c and equals zero otherwise; and d (c,c´) is the distance between the concepts c and c′. According to this measure, an organization with a single categorization has typicality of one in the assigned label and zero for all of the others. An organization with two categorical assignments, say c and c′, has typicality equal to 1/(1+ d (c,c´)) in each of the concepts. Note that this construction sets typicality to zero in all unassigned concepts.

For label-only data, a natural choice of distance metric is Jaccard distance. We use the matrix of Jaccard similarity. Let i denote the set of objects labeled as $c_{i}$ and let j denote the set labeled as $c_{j}$ . $| i \cap j |$ denotes the number of organizations categorized as belonging to category i and label j, and $| i \cup j |$ denotes the number of organizations that are categorized as belonging to category i and/or category j, then this Jaccard distance is defined as $d (c_{i}, c_{j}) = 1 - | i \cap j | / | i \cup j |$ .

In a final step, we set the prior on membership in c in equation (2) to the empirical frequency of such memberships. We then calculate ambiguity following equation (1) above.

Consider the example of Harakat Ansar Iran (HAI). This organization is listed by three sources: TRAC, The United States Institute of Peace, and The Combatting Terror Center at West Point. These sources assign to HAI the labels: tribal (T), Islamic (I), and ethno-nationalist (E). We first calculate the Jaccard matrix for the entire cohort of organizations to obtain the distance (d) for these three pairs: For (T, I), this distance is 0.98, for (T, E) it is 0.96, and for (E, I) it is 0.86. The priors on category memberships (p(c) are set to the observed frequencies: .04 for Tribal ideology, .29 for Islamic, .49 for ethno-nationalist ideology. We then estimate the typicality of HAI’s membership with respect to a given ideology. Using equation (4), we get the typicality of HAI as a tribal ideology that that equals 0.26. Therefore

P^{*} (T, H A I) = \frac{τ (T) p (T)}{τ (T) p (T) + τ (I) p (I) + τ (E) p (E)} = \frac{0.26 * 0.04}{0.26 * 0.04 + 0.26 * 0.49 + 0.26 * 0.29} = 0.049

(5)

Using equation (4), we conduct similar calculations for the other two categories that have been assigned: Islamic and Ethno-nationalists. These calculations yield $P^{*} (I, H A I) = 0.35$ and $P^{*} (E, H A I) = 0.60$ . These big differences are due to the large differences in the priors on the concepts. Finally, I use these probabilities of membership in all ideologies assigned to HAI to calculate the ambiguity, using equation (1). The ambiguity of this organization turns out to be

A (H A I) = - (0.049 *ln (0.049) + . 35 *ln (0.35) + 0.60 *ln (0.60)) = 0.82

(6)

To maximize information on ideological assignments, the measure of ambiguity for any organization is calculated across all 756 organizations.¹⁴ Table 1 displays all the labels attributed to these TOs. First, note that the vast majority (68.3%) of organizations have a singular focus, while 29.2% have been assigned only two labels, with a small (2.5) percent spanning three categories. For TOs assigned to more than one ideological label, ambiguity depends on the number of labels, the relative frequency of that combination, and most importantly, the distribution of ideologies in all other pairs of labels. Note that ambiguity is not merely a function of a large number of ideological categories assigned to a TO. For example, the assigned combination of Leftist and Religious-Not Islamic ideologies is less ambiguous than the combination of Leftist and Religious-Islamic ideologies (the ambiguity measures are .366 versus .727, respectively).

Control Variables

To avoid conflating ambiguity with other factors that have been shown to be relevant in analyses of TOs, it is important to include some of the key controls used by researchers in this tradition. Consistent with the organizational perspective used here, I focus on specific organizational characteristics of allies and rivals, tactical diversity, and organizational size.

Though rivals might seem to be a serious threat to TOs, research by Phillips (2015) finds that TOs operating in the same geographical location with other terrorist groups and within a similar ideological sphere are significantly less likely to end than TOs without rivals. Nemeth (2014) finds that nationalist and religious groups respond with more violence when they compete with rivals. Asal et al. (2016) find that TOs with a similar ideology (especially religious ones) and that operate in the same region are likely to establish alliances.

The EDTG dataset provides an updated annual count of rivals if the focal terrorist group competes with another terrorist group. Over 85% of all organization-years had missing data on both rivals and allies. Some of this missing information was found using Phillips’ (2015) data.¹⁵ Organizations with available data on rivals and allies are likely to be different from those that lack this information. Because this measure is theoretically relevant (e.g., Phillips 2015; 2019), I include two measures: (1) the absolute count of rivals (or allies), which equals 0 if information is missing or if the count is known but is zero, and (2) a dummy variable that equals 1 if information on rivals is available, 0 otherwise. (The effect of the counts of rivals and allies is conditional on the observed non-missing count).

The EDTG project provided two estimates of a TO’s size: a time-varying measure of size, and a single measure of peak size, which is defined as the largest recorded number of members at some point during an organization’s existence. The time-varying measure for size is missing for over 95% of the organization-years, which precludes any serious analysis. As many have noted, a single measure of peak size ignores temporal variation, but since it is available for a larger number of observations, many researchers decide to include it (Phillips 2015; Hou et al. 2019; Blomberg et al. 2011). I identified cases with missing “peak size” in EDTG and supplemented information on this measure using information found on the Mapping Militants website (Crenshaw 2020), the South Asian Terrorism Portal (SATP), and Jones and Libicki (2008). Virtually all studies of lethality and survival of TOs find that an organization’s peak size improves its performance and extends its longevity (Asa et al. 2015; Asal and Rethemeyer 2008; Blomberg et al. 2010; Blomberg et al. 2011; Clauset and Gleditsch 2012).

Social movement theories suggest that SMOs that have a broader tactical repertoire flourish (Wang, Piazza, and Soule 2018; Heaney and Rojas 2014). For TOs, tactical diversity might improve their flexibility in the face of counterterrorism and rival attacks. The EDTG dataset provides a time-varying measure of tactical diversity as an index calculated by one minus the Herfindahl-Hirschman index

D i v e r s i t y = 1 - \sum_{i} S_{i j t}^{2}

(7)

where

S_{i j t}^{2}

is the share of the ith type of terrorist attack in total identified attacks for group j in year t. ETGD lists nine tactics: assassination, hijacking, kidnapping, barricade incident, bombing/explosion, armed assault, facility/infrastructure attack (unknown is omitted from the calculation).

Many scholars argue that organizations operating internationally differ substantially from organizations specializing in domestic targets. The EDTG provides a measure that improves over previous work by creating a time-varying indicator of the proportion of a TO’s attacks that were transnational in a year. It is calculated as the number of transnational terrorist attacks divided by the sum of domestic and transnational events attributed to that group (see also Enders, Sandler, and Gaibulloev 2011).¹⁶ In order to control for a TO’s base country, all models are clustered on ETDG measure “base,” in which each base country listed for a TO is taken into account.

The survival analysis depends upon observing each group’s terrorist activity followed to its ending (if a TO survives to the end of 2016, the last spell is treated as censored). The EDTG dataset uses a clock that begins observing attacks in 1970. A TO’s duration is calculated in years that begin in the year of its first attack and followed through to its demise (or through 2016 if it is still active). This variable ranges from 1 (if an organization ends in its first year of verified terrorist activity) to 47. This variable is highly skewed and so its natural log was calculated to take this into account in the analysis of fatalities.

As noted above, TOs are observed by GTD from 1970-on. Some TOs listed in GTD as active in 1970 were active before that date. To distinguish this cohort from other TOs, I consulted the Rand Database of Worldwide Terrorism Incidents, Mapping Militants, BAAD1, and Wikipedia to assess whether a given organization committed terrorist acts before 1970. This variable was coded 1 if an organization was active prior to 1970, 0 otherwise.

Leaders of TOs confront a variety of threats, including that of arrest, or outright assassination. The EDTG dataset includes a dummy variable that indicates that a leader was either killed or arrested in a given year. To ensure that such an event is not coterminous with an organization’s demise, I calculated the bivariate correlation between the year an organization ended and the year it experienced a loss of leadership. This correlation is .06. The descriptive statistics for all measures can be found in table A1 in the Online Appendix A. The arrest or death of a leader had identical effects on an organization’s likelihood of ending, and so I coded a single measure as 1 if an organization experienced either one of these events in a specific year.¹⁷

Estimation

The analysis of lethality uses annual counts of non-terrorist fatalities attributed to a given organization by the GTD researchers. EDTG researchers further cross-validated this information, using a variety of news media sources (see Hou et al. 2020). These counts (like most event count data) are overdispersed, which means that standard deviations exceed the mean values. For longitudinal count data with overdispersion, negative binomial panel models are appropriate and so were chosen for this first analysis (Cameron and Trivedi 2013; Hilbe 2011; Rabe-Hesketh and Skrondal 2012).

I specify mixed-level negative binomial models (menbreg in Stata 14.2) because TOs are nested within one or more base countries of operations. TOs can and do operate in multiple countries and so EDTG provides complete information on all countries that each TO has a base of operation. For example, the TO Al-Zulfikar is listed by EDTG as having bases in Afghanistan, Libya, Pakistan, India, and Syria.¹⁸ In the analysis of civilian deaths I use mixed-level negative binomial estimates with TO-level one measures and country/base as the second level.¹⁹ In the competing risks analysis, I cluster on this measure (following ETDG’s usage, which lists any and all of the base countries of a given TO).

The second dependent variable is the hazard of ending, with the lifetime measured in years that elapse from an organization’s first observation through its ending date (validated by Huo et al. 2020 and again by this author). Given the importance of aging found in prior analyses (Young and Dugan 2014; Blomberg et al. 2011), it seems important to specify a model that allows the effects to vary by age. I begin the analysis with a flexible piecewise exponential model, which allows the hazard to vary in an unconstrained manner at preselected ages (which I determined by the cumulative hazard curves and other diagnostics). I divide the age range into three intervals: 0–2 years, 2–25, and 25+ years. Survival times vary broadly, from one to 47 years, with a mean survival of 13.7 years. Use of the piecewise exponential parameterization, in which each age segment has its own baseline rate of failure, allows us to examine all the possibilities of age dependence.

Results

Civilian Fatalities

Table 2 presents the mixed-level negative binomial regression results. In Table 2, the first column reports results including the measures of allies, and the second column reports results including the measures for rivals. The two dummy variables indicating that rivals or allies have missing data are correlated .99. This necessitates separating the analysis of the effects of allies and rivals.

Table 2.

Mixed-level negative binomial estimates of the effects of ambiguity on civilian deaths attributed to terrorist organizations, 1970 Through 2016.

Covariates	Ambiguity and allies	Ambiguity and rivals
Ambiguity	−.529**	−.566**
Ambiguity	(.169)	(.170)
N of allies \| N Rivals (conditional)	.036	−.652**
N of allies \| N Rivals (conditional)	(.034)	(.188)
Ally or rival counts observed (0,1)	.386***	.498**
Ally or rival counts observed (0,1)	(.100)	(.090)
Peak size in 1000s	.032**	.034**
Peak size in 1000s	(.012)	(.012)
Tactical diversity	.355*	.360*
Tactical diversity	(.156)	(.156)
Proportion transnational acts	−.190	−.189
Proportion transnational acts	(.105)	(.105)

Ethnic/Nationalist (0,1)	.098	.107
Ethnic/Nationalist (0,1)	(.114)	(.114)
Islamic (0,1)	.600***	.607***
Islamic (0,1)	(.145)	(.145)
Ln duration	−.038	−.037
Ln duration	(.044)	(.044)
Post 9/11 (0,1)	−.596***	−.578***
Post 9/11 (0,1)	(.094)	(.094)
Exists before 1970 (0,1)	.146	.134
Exists before 1970 (0,1)	(.132)	(.132)
Log N of attacks	1.036***	1.041***
Log N of attacks	(.027)	(.026)
Constant	.116	.094
Constant	(.190)	(.190)
Log likelihood	−6758.6	−6754.6
Observations	6501	6501
Number of organizations	474	474

Notes: *** p < 0.001, ** p < 0.01, * *p < 0.05. Level two: country base (or bases) of operations.

How does the ambiguity argument fare? Compared to organizations with a narrow focus, organizations with high ambiguity have significantly diminished lethality, as expected by hypothesis 1. Concretely, the coefficient of −.529 in column 1 shows that when compared to an organization assigned a single ideology of ethno-nationalist (where ambiguity is 0), a TO assigned to both ethno-nationalist and leftist categories (where its ambiguity is .635) would have its lethality reduced by 29% (net of the effects of the other covariates and controlling for the number of attacks). Column 2 (which includes measures of rivals) shows that a one-unit increase in ambiguity would decrease fatalities by 43%, net of the effect of a host of other characteristics (see Table 2 column 2). The control variables show similar effects found in prior research: Islamic ideology, tactical diversity, and the overall number of attacks generated by an organization increase its lethality (e.g., see Huo et al. 2020).

Based on past research we might expect that allies improve and rivals constrain an organization’s performance (Asal et al. 2016; Bapat and Bond 2012; Horowitz and Potter 2014). In Table 2 (column 1), we see that the effect of allies on fatalities is not statistically significant. In Figure 1(a), the point estimate for the number of allies is close to zero. But in column 2 of Table 2 the number of rivals (conditional on the observed non-missing count) significantly decreases lethality. Specifically, the coefficient of -.652 indicates that the presence of just one rival reduces a group’s fatalities by about 48%.

Figure 1.

(a) Effect of ambiguity and allies on civilian deaths. (b) Effect of ambiguity and rivals on civilian deaths.

Why would the presence of rivals reduce lethality, but the presence of allies does not improve it? One possible answer is that less lethal terrorist organizations that have more rivals actively seek out more powerful allies to bolster their position (Asal et al. 2016; Bapat and Bond 2012; Phillips 2019). For example, the New York Times reporters Goldbaum and Schmitt (2021) wrote that in 2021, because the ISIS organization in Africa was declining in strength, it has been “forging alliances with local militant groups …that have pumped up their profiles, fund-raising and recruitment.”²⁰

Figures 1(a) and (b) present the point estimates and 95% confidence intervals from the mixed negative binomial regression analysis in Table 2. These figures highlight the negative effect of ambiguity on fatalities while controlling for several other relevant factors. In Figure 1(b), the negative effect of rivals is especially striking.

Figure 2.

(a) Effects of ambiguity and allies on hazard of ending. (b) Effects of ambiguity and rivals on hazard of ending.

The effects of other control variables are unsurprising: larger organizations have a significant and positive effect on fatalities, as others have found (Gaibulloev and Sandler 2013; Hou et al. 2020; but see Clauset and Gleditsch 2012).²¹ Figures 1(a) and (b) show that tactical diversity significantly increases the lethality of organizations.²² The decline in fatalities post 9/11 hints at the possibility that counterterrorism since then has been effective at reducing fatalities.

I performed one more cross-check of the data on attacks and fatalities. The GTD Codebook (2019: 4) cautions the data collection periods might introduce differences that could affect the counts of attacks and fatalities. I created four dummy variables to estimate the effect of different coding intervals. Table A2 in the Online Appendix A reports the results, which show that all of the coding periods before 2012 had a significantly higher count of non-terrorist fatalities when compared to the post-2012 data collection period. This may be due to coding protocols, or it may reflect actual declines in the number of attacks and fatalities. It is useful to see that the effect of ambiguity on non-terrorist fatalities remains negative and statistically significant (and these measures did not affect any other of the results).

In sum, the influence of ambiguity on fatalities is robust across different specifications, net of the non-trivial effects of other organizational features. The results remain the same when the outlier case of al-Qaeda is removed. Together, these findings lend support for the argument that ideological clarity aids the performance of one key tactic used by terrorists.

Competing Risk Specification of the Hazard of Endings

The data on each organization, i, is updated annually, beginning with the year of the first observed event through its specific ending year (or remains censored). Given that different durations and ending types can be identified, it seems reasonable to consider a flexible model that compares the competing risks of ending (λi(t| x i) in one of the two substantively important ways.²³ I use a piecewise exponential hazard model that estimates the effects of the covariates ( x) on the rate of endings, which takes into account that the rate can change over the course of an organization’s lifespan. The standard form is

λ i (t | x i) = λ 0 (t) exp (x b)

(8)

I partition duration into J intervals with 0=τ0<τ1<…<τJ=∞0=τ0<τ1<…<τJ=∞. I define the jth interval as [τj−1,τj)[τj−1,τj), extending from the (j−1) (j−1)st boundary to the jj-th. The piecewise baseline hazard is constant within each interval (or pieces), so that

λ 0 (t) = λ j for t interval [Τ j - 1, Τ j)

(9)

Thus, I model the baseline hazard λ0(t) using J parameters λ₁…,λ_J, each representing the risk for the reference group in one particular interval compared to ending by some other means. Since the risk is assumed to be piecewise constant, the corresponding survival function is called a piecewise exponential. The first 2-year interval captures the theoretical argument that an organization’s first years of operation are most hazardous. The second cut-off point was chosen based upon inspection of the cumulative hazard plot (Nelson-Aalen), in which the hazard of ending levels off after 25 years.

As discussed above, I rely on organizational perspectives that treat organizational endings as such if an organization no longer operates with the same identity (Carroll and Hannan 1994; 2004). I analyze two qualitatively different types of endings. The first category considers a TO as ended if it no longer exists as an independent, named entity on its own for five consecutive years. The second ending type involves a TO joining the political system (e.g., becoming a political party or joining a coalition with a party) or attaining their goals.

Table 3 presents the results of piecewise exponential estimation of the hazard of ending in two ways: by military force, splintering, merger, or fading away or by entering politics. The first two columns (and the last two columns) differ by whether information on allies or rivals is included. We see that ambiguity increases an organization’s chances of disbanding by military, splintering, merger, or fading away, but it has no significant effect on entering politics. For example, in Table 3, the coefficient of .474 in column (1) shows that a TO with an ambiguity level of one standard deviation above the mean (.48) would increase an organization’s demise by about 26% when compared to single-issue TO.

Table 3.

Piecewise exponential estimates of the hazard of two types of endings.

Covariates		Military force, splinters, merges, or fades away (2)	Enters Politics (3)	Enters Politics (4)
0< 2 years active	−1.846***	−1.830***	−3.146***	−3.164***
0< 2 years active	(.215)	(.214)	(.387)	(.394)
2 > 25 years active	−2.481***	−2.480***	−4.268***	−4.292***
2 > 25 years active	(.189)	(.189)	(.351)	(.358)
25> years active	−2.630***	−2.607***	−3.611***	−3.634***
25> years active	(.258)	(.258)	(.476)	(.486)
Ambiguity	.474*	.472*	−.112	−.106
Ambiguity	(.238)	(.239)	(.469)	(.479)
N Allies (condit)	−1.257	—	−13.482***	—
N Allies (condit)	(.727)		(.400)
Ally counts avail. (0,1)	−.968*	—	.418	—
Ally counts avail. (0,1)	(.442)		(.355)
N Rivals (condit) (* 10)	—	−1.238***	—	.048
N Rivals (condit) (* 10)		(.069)		(.072)
Rival counts avail. (0,1)	—	−1.591**	—	-.231
Rival counts avail. (0,1)		(.487)		(.435)
Leader arrest/kill	1.314***	1.303***	1.135**	1.104*
Leader arrest/kill	(.281)	(.280)	(.506)	(.504)
Peak size 1000s	−.141*	−.140*	.032	.034
Peak size 1000s	(.056)	(.056)	(.038)	(.039)
Tactical diversity	−2.389***	−2.383***	−2.101*	−2.119*
Tactical diversity	(.565)	(.567)	(.821)	(.825)
Proport. Transnat.	.240	.232	1.566***	1.610***
Proport. Transnat.	(.209)	(.209)	(.392)	(.400)
Ethnic/Nationalist (0,1)	−.443**	−.440**	−.420	−.431
Ethnic/Nationalist (0,1)	(.155)	(.156)	(.247)	(.248)
Islamic (0,1)	−.295	−.323	−1.623**	−1.759**
Islamic (0,1)	(.206)	(.205)	(.612)	(.617)
N Civilian deaths	−.010*	−.010*	−.002	−.002
N Civilian deaths	(.005)	(.005)	(.004)	(.003)
Exists before 1970	−1.161**	−1.164**	−.270*	−.298
Exists before 1970	(.419)	(.419)	.320	(.329)
Ln attacks	.173***	.172***	.068	.061
Ln attacks	(.026)	(.026)	(.456)	(.046)
Observations	6503	6503	6503	6503
N of TOs ended	170	170	79	79

Notes: *** p < 0.001, ** p < 0.01, *p < 0.05 (Two-tailed tests). Observations clustered on countries of operation. Robust standard errors are in parentheses.

Figure 2(a) and (b) present the point estimates using information from Table 3. These figures illustrate that the effects of ambiguity increase the more conflict-ridden endings compared to the risk of ending by entering politics or reaching their goals. Figure 2(a) incorporates the effects of allies, while Figure 2(b) incorporates the effects of rivals.

In both figures, the effect of ambiguity increases the hazard of a TO ending as an organizational enterprise that engages in terrorist acts. Figure 2(a) and (b) show that conceptual ambiguity has a positive and statistically significant effect on the likelihood of a TO experiencing an organizational death. These results support hypothesis 2. Indeed, among all factors that increase the death rate of a TOs, only the removal of a leader has a more powerful effect than ambiguity (Cronin 2009; Hou et al. 2020).²⁴ Ambiguity has no significant effect on the hazard of ending by entering the political system—indeed the pattern of results for political endings is quite different from the patterns seen in the analysis of organizational endings.

The impact of rivals on a TO’s demise is straightforward. The presence of rivals significantly decreases the chances of ending as an organizational entity. This is consistent with Brian Phillips’ findings (2015). Phillips argues that rivals can benefit a TO by encouraging recruitment, facilitating competition, and encouraging learning that sharpens a group’s skills, or by providing an oppositional identity that raises in-group solidarity.

Other factors also matter. Columns 1 and 2 of Table 3 (and Figure 2(a) and (b)) show that organizations with a larger share of attacks devoted to international targets are not more likely to end by military defeat, split, merger, or by fading away. In contrast, columns 3 and 4 of Table 3 indicate that TOs with a larger proportion of international attacks are significantly more likely to end by entering politics. Ethno-nationalist TOs resist ending as an organizational entity. Islamic TOs resist entering politics. Tactical diversity significant reduces the chances a TO will end by any means (see also Olzak and Ryo 2007; Wang et al. 2018).

Robustness Checks

The Online Appendix B reports results from other plausible methods of estimation that are also suitable for use with count data (for the fatalities analysis) and other competing risk models using proportional hazard models (see Supplemental Figures B1 and B2 for the competing risk results). No differences were found when compared to those reported here.

Discussion and Conclusions

TOs that have an ambiguous combination of ideological labels attributed to them accumulate fewer fatalities and cease engaging in terrorist activity as an organizational entity. Ambiguity does not affect the rate at which a TO enters the political process.

Many of the findings regarding control variables are consistent with previous work on terrorist organizations. In particular, the beneficial results of organizational size, duration, ethno-nationalist, and religious orientation for lethality and survival of TOs reported here have been documented in prior research by leading scholars in this field (Asal and Rethemeyer 2008; Gaibulloev and Sandler 2019; Hou et al. 2020; Jones and Libicki 2008). Findings reported by Hou et al. (2020) are replicated here, especially those concerning the effects of size, and religious organizations on lethality. Supporting many of the pioneering efforts of Crenshaw, Cronin, Sandler, Asal, Rethemeyer, and others, I find that Islamic organizations are significantly more lethal than organizations affiliated with other ideologies.

These results expand our understanding of the role of ideology in TOs. The ambiguity argument offers a theoretical explanation for why some religious and ethno-nationalist terrorist organizations have endured. They persist (at least in part) due to their ideological clarity. Such clarity improves their ability to attract those most committed to a cause. Scholars interested in how organizations attract, socialize, and retain individual members of TOs might build on these ideas.

As with all studies, this one has limitations. Only two consequences have been studied here. Other possible outcomes that ought to be investigated in future research might analyze whether efforts by counterterrorist forces are more successful when they are directed against TOs with ambiguous ideological identities. The results regarding the effects of age dependence suggest another avenue to explore. Ambiguity may also influence TOs at other crucial turning points in their lifecycle, such as those surrounding leadership succession, leadership loss, and internal betrayals.

Another limitation is that unknown perpetrators are, by definition, omitted from this dataset. It seems likely that organizations or individuals that do not claim responsibility for terrorist attacks are different from others. For example, such events might indicate the activity of a lone wolf or deranged person, with little or no attachment to ideology.²⁵ Future research might usefully explore these types of acts of terrorism.

Because terrorist organizations are clandestine organizations, there is considerable missing data on organizational size, rivals, and membership growth. Such limitations reduce the sample size available for analysis. At the same time, these limitations offer exciting opportunities for researchers to fill in these gaps.

Keeping these limitations in mind, it is tempting to draw some policy implications from the findings regarding ambiguity. Some social movement scholars find that SMOs benefit by mobilizing support from a diverse set of constituents (Heaney and Rojas 2014). The results shown here suggest that such diversity comes with a cost (Wang et al. 2018). TOs that want to survive should develop identities that are sharply focused and well understood.

From the vantagepoint of those combatting terrorism, the knowledge that organizations with a crisp focus are harder to conquer could inform strategies of counterterrorism. The perspective offered here suggests that these TOs survive not just because they are religious or ethnic in character, but because their membership claims are unambiguous. By this argument, counterterrorist campaigns could gain leverage over single-ideology TOs if they provide creditable evidence that raises doubt among their supporters about a TO’s claim of ideological purity (Innes, Dobreva, and Innes 2021). Conversely, counterterrorism efforts could publicize information about those TOs with multiple ideologies that could suggest that their identities are in conflict, vague, or even contradictory.

Another policy insight comes from studies of the population dynamics that shape the nature and diversity of other types of organizational populations. This perspective analyzes factors that affect the survival and mortality rates of organizations (Carroll and Hannan 2004). For example, the more recent cohort of TOs analyzed here have fewer ideological combinations assigned to them compared to ideologies assigned to an earlier cohort of TOs (Olzak 2016: 566). One reason for this might be due to the winnowing effects of ambiguity, which would be consistent with the fact that the surviving TOs have a narrow ideological focus. The results offered here suggest that ambiguous TOs are more likely to unravel on their own, but that single-focus organizations will be much harder to defeat.

Supplemental Material

sj-pdf-1-jcr-10.1177_00220027211073921 Supplemental Material for The Impact of Ideological Ambiguity on Terrorist Organizations

Supplemental Material, sj-pdf-1-jcr-10.1177_00220027211073921 for The Impact of Ideological Ambiguity on Terrorist Organizations by Susan Olzak in Journal of Conflict Resolution

Footnotes

Acknowledgments

The author would like to thank Todd Sandler, Donfang Hou, and Krushav Gaibulloev for their ETDG dataset and for helpful advice during the course of this research. The author thanks Daria Tabea Lenz, Carolina Sculti, Sarah A. Soule, and Michael Hannan for providing useful advice and suggestions on earlier drafts.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

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

ORCID iD

Susan Olzak

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Alter

Adam L.

Oppenheimer

Daniel M.

. 2008. “Easy on the Mind, Easy on the Wallet: The Roles and Familiarity and Processing Fluence on Valuation Judgements.” Psychonomic Bulletin Review 15: 985-990.

Asal

Victor

Brown

Michell

Dalton

Angela

. 2012. “Why Split? Organizational Splits Among Ethnopolitical Organizations in the Middle East.” Journal of Conflict Resolution 56: 94-117.

Asal

Victor

Phillips

Brian J.

. 2015. “What Explains Ethnic Organizational Violence? Evidence from Eastern Europe and Russia.” Conflict Management and Peace Science (published online November 27, 2015). doi: 10.1177/0738894215614504. (accessed January 10, 2016).

Asal

Victor H.

Gill

Paul

Rethemeyer

R. Karl

Horgan

John

. 2015. “Killing Range: Explaining Lethality Variance within a Terrorist Organization.” Journal of Conflict Resolution 59: 401-427. doi: 10.1177/0022002713508927. (accessed January 1, 2015).

Asal

Victor H.

Park

Hyun Hee

Rethemeyer

R. Karl

Ackerman

Gary

. 2016. “With Friends Like These...Why Terrorist Organizations Ally.” International Public Management Journal 19 (1): 1-30.

Asal

Victor

Rethemeyer

R. Karl

. 2008. “The Nature of the Beast: Organizational Structures and the Lethality of Terrorist Attacks.” The Journal of Politics 70: 437-449.

Asal

Victor

Rethemeyer

R. Karl

. 2009. “Islamic Use and Pursuit of CBRN Terrorism.” Pp. 335-358 in Ackerman

Gary

Tamsett

Jeremy

(eds.) Jihadists and the Weapons of Mass Destruction. Boca Raton FL: CRC Press.

Asal

Victor

Rethemeyer

R. Karl

Anderson

Ian

. 2008. “Big Allied and Dangerous”(BAAD1) Version 1. http://www.albany.edu/pvc/current_projects.shtml (accessed June 1, 2011).

Asal

Victor

Rethemeyer

R. Karl

Young

Joseph

. 2016. “Insurgency BAAD: Dynamics of Terrorism and Counterterrorism Campaigns.” National Consortium for the Study of Terrorism and Responses to Terrorism. Research Brief (accessed January 11, 2016). https://www.start.umd.edu/pubs/START_BAAD_CT_October2015.pdf?utm_source=START+Announce&utm_campaign=19dc6a919f-START_Newsletter_Jan2016&utm_medium=email&utm_term=0_a60ca8c769-19dc6a919f-14084733 (accessed October 30, 2015).

10.

Bapat

Navin

Bond

Kanisha D.

. 2012. “Alliances Amongst Militant Groups.” British Journal of Political Science 42 (4): 793-824.

11.

Beck

Colin J

. 2008. ‘‘The Contribution of Social Movement Theory to Understanding Terrorism,’’ Sociology Compass 2 (5): 1565–1581. doi: 10.1111=j.1751-9020.2008.00148.x, 1576

12.

Beck

Colin J.

Schoon

Eric

. 2019. “Terrorism and Social Movements.” In Wiley-Blackwell Companion to Social Movements, edited by Snow

David A.

Soule

Sarah A.

Kriesi

Hanspeter

, McCammon

Holly J.

, 698–713. 2nd ed. Hoboken NJ: John Wiley & Sons.

13.

Benford

Robert D.

Snow

David A.

. 2000. “Framing Processes and Social Movements: An Overview and Assessment.” Annual Review of Sociology 26: 611-38.

14.

Berman

Eli

. 2009. Radical, Religious, and Violent. New Brunswick NJ: Rutgers University Press.

15.

Blomberg

Brock S.

Engel

Rozlyn C.

Sawyer

Reid

. 2010. “On the Duration and Sustainability of Terrorist Organizations.” Journal of Conflict Resolution 54: 303-330.

16.

Blomberg

Brock S.

Gaibulloev

Khusrav

Todd

Sandler.

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

17.

Cameron

A. Colin

Trivedi

Praven K.

. 2013. Regression Analysis of Count Data, 2nd ed. Cambridge MA: Cambridge University Press.

18.

Capellan

Joel A

. 2015. “Lone Wolf Terrorist or Deranged Shooter? A Study of Ideological Active Shooter Events in the United States, 1970–2014.” Studies in Conflict & Terrorism 38 (6): 395-413. DOI: 10.1080/1057610X.2015.1008341

19.

Carroll

Glenn R.

Hannan

Michael T.

. 2004. The Demography of Corporations and Industries. Princeton NJ: Princeton University Press.

20.

Chou

Winston

. 2016. “Seen Like a State: How Illegitimacy Shapes Terrorism Designation.” Social Forces 94 (5): 1129-1152.

21.

Clauset

Aaron

Gleditsch

Kristian Skrede

. 2012. “The Developmental Dynamics of Terrorist Organizations.” PLoS One 7 (11): e48633. doi:10.1371/journal.pone.0048633. (accessed January 15, 2016).

22.

Conrad

Justin

Green

Kevin

. 2015. “Competition, Differentiation, and the Severity of Terrorist Attacks.” The Journal of Politics 77: 546-561.

23.

Cox

David R.

Snell

Eberhard Joachim

. 1989. Analysis of Binary Data. Boca Raton FL: Chapman and Hall.

24.

Cox

David R.

Oakes

David

. 1984. Analysis of Survival Data. Boca Raton FL: Chapman and Hall.

25.

Crenshaw

Martha

. 1987. “Theories of Terrorism: Instrumental and Organizational Approaches.” Journal of Strategic Studies 10 (4): 13-31.

26.

Crenshaw

Martha

. 1991. “How Terrorism Declines.” Terrorism and Political Violence 3: 69-87.

27.

Crenshaw

Martha

. 2011. Explaining Terrorism: Causes, Processes, and Consequences. New York: Routledge.

28.

Crenshaw

Martha

. 2020. “Mapping Militant Organizations.” Stanford CA: Stanford University. http://web.stanford.edu/group/mappingmilitants/cgi-bin/groups (accessed June 20, 2015).

29.

Cronin

Audrey Kurth

. 2009. How Terrorist Groups End. Princeton NJ: Princeton University Press.

30.

Davenport

Christian

. 2015. How Social Movements Die. New York: Cambridge University Press.

31.

della Porta

Donatella

. 2013. Clandestine Political Violence. New York: Cambridge University Press.

32.

Enders

Walter

Todd

Sandler

Gaibulloev

Khusrav

. 2011. “Domestic Versus Transnational Terrorism: Data, Decomposition, and Dynamics.” Journal of Peace Research 48: 319-337.

33.

Fassioto

Magali

Soule

Sarah A.

. 2017. “Loud and Clear: The Effect of Protest Signals on Congressional Attention.” Mobilization: An International Journal 22 (1): 17-38.

34.

Gaibulloev

Khusrav

Sandler

Todd

. 2013. “Determinants of the Demise of Terrorist Organizations.” Southern Economic Journal 79: 774-92. DOI: 10.4284/0038-4038-2012.269

. 2019. “What We Have Learned about Terrorism since 9/11.” Journal of Economic Literature 57(2): 275-328.

36.

Gamson

William

. 1975. The Strategy of Social Protest. Homewood Ill: Dorsey Press.

37.

Global Terrorism Database (GTD) . 2019. Global Terrorism Database. Codebook: Inclusion Criteria and Variables. College Park MD: National Consortium for the Study of Terrorism and Responses to Terrorism. http://www.start.umd.edu/gtd/using-gtd/downloads/Codebook.pdf (accessed March 10, 2019).

38.

Goldbaum

Christina

Schmitt

Eric

. 2021. “In Bid to Boost Its Profile, ISIS Turns to Africa’s Militants.” The New York Times (published April 8, 2021, page 1). https://www.nytimes.com/2021/04/07/world/africa/ISIS-Africa-Mozambique.html?smid=url-share

39.

Hannan

Michael T.

Carroll

Glenn R.

. 1992. Dynamics of Organizational Populations: Density, Legitimation, and Competition. Oxford UK: Oxford University Press.

40.

Hannan

Michael T.

Freeman

John

. 1989. Organizational Ecology. Cambridge MA: Harvard University Press.

41.

Hannan

Michael T.

Hsu

Greta

Kovács

Bálasz

Le Mens

Gaël

Negro

Giacomo

Pólos

Lászlo

Pontikes

Elizabeth G.

Sharkey

Amanda J.

. 2019. Concepts and Categories: Foundations for Sociological and Cultural Analysis. NY: Columbia University Press.

42.

Hannan

Michael T.

Pólos

Lászlo

Carroll

Glenn R.

. 2007. Logics of Organization Theory: Audiences, Codes, and Ecologies. Princeton NJ: Princeton University Press.

43.

Heaney

Michael T.

Rojas

Fabio

2014. “Hybrid Activism: Social Movement Mobilization in a Multi-Movement Environment.” American Journal of Sociology 119: 1047-1103.

44.

Hilbe

Joseph M

. 2011. Negative Binomial Regression. 2nd ed. Cambridge MA: MIT Press.

45.

Horowitz

Michael C.

Potter

Philip B. K.

. 2014. “Allying to Kill: Terrorist Intergroup Cooperation and the Consequences for Lethality.” Journal of Conflict Resolution 58: 199-225.

46.

Hou

Donfang

Gaibulloev

Khusrav

Todd

Sandler

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

47.

Hsu

Greta

. 2006. “Jacks of All Trades and Masters of None: Audiences’ Reactions to Spanning Genres in Feature Film Production.” Administrative Science Quarterly 512: 420-450. http://dx/doi.org/10.2189/asqu.51.3.420.

48.

Hsu

Greta

Hannan

Michael T.

Koçak

Özgeçan

. 2009. “Multiple Category Memberships in Markets: An Integrative Theory and Two Empirical Tests.” American Sociological Review 74 (1): 150-169.

49.

Hsu

Greta

Hannan

Michael T.

Pólos

Lászlo

. 2011. “Typecasting, Legitimation, and Form Emergence: A Formal Theory.” Sociological Theory 29 (2): 97-123. DOI: 10.1111/j.1467-9558.2011.01389.x

50.

Hsu

Greta

Negro

Giacomo

Perretti

Fabrizio

. 2012. “Hybrids in Hollywood: A Study of the Production Performance of Genre Spanning Films.” Industrial and Corporate Change 21 (2): 459-487.

51.

Innes

Martin

Dobreva

Diyana

Innes

Helen

. 2021. “Disinformation and Digital Influencing After Terrorism: Spoofing, Truthing and Social Proofing.” Contemporary Social Science 16 (2): 241-255.

52.

Jones

Seth G.

Libicki

Martin C.

. 2008. “How Terrorist Groups End.” Santa Monica CA: Rand Corporation.

53.

Kalbfleisch

John D.

Prentice

Ross L.

. 2002. The Statistical Analysis of Failure Time Data. 2nd ed. New York NY: John Wiley & Sons.

54.

Kinney

Alexander B.

Davis

Andrew P.

Zhang

Yongjun

. 2018. “Theming for Terror: Organizational Adornment in Terrorist Propaganda.” Poetics 69: 27-40.

55.

Kovács

Balász

Hannan

Michael T.

. 2010. “The Consequences of Category Spanning Depend on Contrast.” In Research in the Sociology of Organizations, edited by Hsu

Greta

Negro

Giacomo

Koçak

Özgecan

, Vol. 31, 175-201. Bingley, UK: Emerald Book.

56.

Kovács

Balász

Hannan

Michael T.

. 2015. “Conceptual Spaces and the Consequences of Category Spanning.” Sociological Science 2: 258-286. DOI: 10.15195/v2.a13

57.

LaFree

Gary

Dugan

Laura

. 2009. “Research on Terrorism and Countering Terrorism.” Crime and Justice 38: 413-477.

58.

LaFree

Gary

Dugan

Laura

Miller

Erin

. 2015. Putting Terrorism in Context. NY: Routledge.

59.

Levy

Ido

. 2021. “Lethal Beliefs: Ideology and the Lethality of Terrorist Organizations.” Terrorism and Political Violence (published online October 2021). DOI: 10.1080/09546553.2021.1977282

60.

Luna

Zakiya

. 2017. “Who Speaks for Whom? (Mis)Representation and Authenticity in Social Movements.” Mobilization: An International Quarterly 22 (3): 435-450.

61.

McCarthy

John D.

Zald

Mayer

. 1977. “Resource Mobilization and Social Movements: A Partial Theory.” American Journal of Sociology 82: 1212-1241.

62.

Miller

Erin

. 2017. "Global Terrorism Database Ideological Motivations of Terrorism in the United States". DOI: 10.7910/DVN/SACQNK, Harvard Dataverse, V2, UNF:6:V5eEse3JC2uXnbHeqx6w5A== [fileUNF]

. 2010. “Research on Categories in the Sociology of Organizations.” Research in the Sociology of Organizations 31: 1-35.

. 2013. “‘Actual’ and Perceptual Effects of Category Spanning.” Organization Science 24: 684-696.

65.

Nemeth

Stephen

. 2014. “The Effect of Competition on Terrorist Group Operations.” Journal of Conflict Resolution 58: 336-62.

66.

Olzak

Susan

. 2016. “The Effect of Category Spanning on the Lethality and Longevity of Terrorist Organizations.” Social Forces 95 (2): 559-584. doi: 10.1093/sf/sow081

. 2019. The Risk of Occupying a Broad Niche for Environmental Social Movement Organizations.” Mobilization: An International Quarterly 24 (2): 177-198.

. 2007. “Organizational Diversity, Vitality and Outcomes in the African American Civil Rights Movement.” Social Forces 85: 1561-1592.

69.

Oppenheimer

Daniel M

. 2008. “The Secret Life of Fluency.” Trends in Cognitive Science 2: 237-241.

2008. “A Rose in Any Other Font Would Not Smell as Sweet: Effects of Perceptual Fluency on Categorization.” Cognition 106 (3): 1178-1194.

71.

Phillips

Brian J

. 2014. “Terrorist Group Cooperation and Longevity.” International Studies Quarterly 58 (June) 336-347.

72.

Phillips

Brian J

. 2015. “Enemies with Benefits? Violent Rivalry and Terrorist Group Longevity.” Journal of Peace Research 52: 62-75.

73.

Phillips

Brian J

. 2019. “Terrorist Group Rivalries and Alliances: Testing Competing Explanations.” Studies in Conflict & Terrorism 42 (11): 997-1019.

. 2020. “Claim Specialization, Tactical Diversity and the Protest Environment in the Success of U.S. Antinuclear Activism.” Mobilization 25 (1): 93-114.

75.

Piazza

James A

. 2009. “Is Islamic Terrorism More Dangerous?: An Empirical Study of Group Ideology, Organization, and Goal Structure.” Terrorism and Political Violence 21: 62-88.

. 2019. “Islamic Terrorism, Diaspora Links and Casualty Rates.” Perspectives on Terrorism 13 (5): 2-21.

. 2001. “Collective Identity and Social Movements.” Annual Review of Sociology 27: 283-305.

. 2003. “The Terrorist in Their Own Words: Interviews with 35 Incarcerated Middle Eastern Terrorists.” Terrorism and Political Violence 15: 171-184.

. 2012. Multilevel and Longitudinal Modeling Using Stata: Volume II: Categorical Responses, Counts, and Survival . 3rd ed. College Station TX: Stata Press.

80.

Rosch

Eleanor H.

1975. “Cognitive Representations of Semantic Categories.” Journal of Experimental Psychology: General 104: 192-233.

81.

Sandler

Todd

. 2014. “The Analytical Study of Terrorism: Taking Stock.” Journal of Peace Research 51: 257-71.

82.

Selznick

Phillip

. 1960 The Organization Weapon. Glencoe IL: Free Press.

83.

Sørensen

Jesper B

. 2002. “The Strength of Corporate Culture and the Reliability of Firm Performance.” Administrative Quarterly 47(1): 70-91.

84.

Valentino

Benjamin A

. 2014. “Why We Kill: The Political Science of the Political Violence Against Civilians.” Annual Review of Political Science 17: 89-103.

. 2020. “Valuing the Cause: A Theory of Authenticity in Social Movements.” Mobilization: An International Quarterly 25(1): 1-25.

. 2018. “Boundary-Spanning in Social Movements: Antecedents and Outcomes.” Annual Review of Sociology 44: 167-187.

. 2019. “Crossing Categorical Boundaries: A Study of Diversification by Social Movement Organizations.” American Sociological Review 84 (3): 420-458.

. 2020. “Ideologies That Justify Political Violence.” Current Opinion in Behavioral Sciences 34: 107-111.

. 2003. “The Hedonic Marking of Processing Fluency: Implications for Evaluative Judgement.” In The Psychology of Evaluation: Affective Processes in Cognition and Emotion, edited by Musch

Kaluer

K.C.

, 195-223. Mahwah NJ: Erlbaum.

. 2006. “Prototypes Are Attractive Because They Are Easy on the Mind.” Psychological Science 17: 799-806.

2014. “Survival of the Fittest: Why Terrorist Groups Endure.” Perspectives on Terrorism 8 (2): 1-22.

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.

0.00 MB

0.52 MB