economic deprivation, ideology and terrorism a system

37
1 Economic deprivation, ideology and terrorism A system justification approach of domestic radicalization in the USA (1948-2016) Simon Varaine Phd candidate in Political science Univ. Grenoble Alpes, CNRS, Science Po Grenoble*, PACTE, 38000 Grenoble, France * School of Political Studies Univ. Grenoble Alpes Sciences Po Grenoble 1030 Avenue Centrale 38400 Saint-Martin-d'Hères - France [email protected] +33 6 07 90 02 23 Acknowledgement I would like to thank Raul Magni-Berton, Laurent Bègue and Céline Belot for their insightful comments and advises. This research is part of a Phd funded by the Communauté Université Grenoble Alpes.

Upload: others

Post on 25-Dec-2021

1 views

Category:

Documents


0 download

TRANSCRIPT

1

Economic deprivation, ideology and terrorism

A system justification approach of domestic radicalization in the USA

(1948-2016)

Simon Varaine

Phd candidate in Political science

Univ. Grenoble Alpes, CNRS, Science Po Grenoble*, PACTE, 38000 Grenoble, France

* School of Political Studies Univ. Grenoble Alpes

Sciences Po Grenoble

1030 Avenue Centrale

38400 Saint-Martin-d'Hères - France

[email protected]

+33 6 07 90 02 23

Acknowledgement

I would like to thank Raul Magni-Berton, Laurent Bègue and Céline Belot for their insightful

comments and advises. This research is part of a Phd funded by the Communauté Université

Grenoble Alpes.

1

Economic deprivation, ideology and terrorism

A system justification approach of domestic radicalization in the USA

(1948-2016)

Abstract

Objectives

Studies generally find no relationship between economic deprivation and terrorist activities,

leading to the conclusion that economic conditions have no role in the emergence of terrorist

movements. The present paper challenges this conclusion based on system justification

theory. It argues that collective deprivation affects participation into terrorism, but in different

directions depending on the ideology of terrorist movements: conservative terrorism should

mobilize more under times of collective deprivation while progressive terrorism should

mobilize more under times of collective improvement.

Methods

I test this hypothesis on the PIRUS database about domestic terrorists (N=1,527) in the United

States from 1948 to 2016. I analyse whether the proportion of right-wing (versus left-wing)

terrorists at a given year depends on collective deprivation in the US, operationalized through

long-term recession of the national income and long-term growth of inequalities.

Results

Multi-level logistic regression analyses confirm that right-wing terrorism mobilizes more

under periods of long-term economic deprivation, while left-wing terrorism mobilizes more

under improving economic conditions. Besides, the effect of collective deprivation appears to

be of socio-tropic nature since it is not mediated through the terrorist’s individual level of

deprivation. In contrast, results do not support the view that Islamist terrorism is affected by

collective deprivation.

Conclusions

The study (1) challenges the view that economic conditions have no role in triggering terrorist

activity and (2) supports recent development of system-justification theory implying that

ideology matters in the explanation of collective action.

Key words deprivation, terrorism, system justification, threat, radicalization

2

1. INTRODUCTION

For the past decades, the common idea that terrorism emerges from situations of economic

deprivation has been disproven by several empirical studies. At the collective level, indicators

of economic contractions and inequality are not good predictors of the occurrence of terrorist

events (Piazza, 2006, Kurrild-Klitgaard et al., 2006, Abadie, 2006). At the individual level,

terrorists are not mostly drawn from impoverished backgrounds (Krueger & Malečková,

2003, Sageman, 2004, Krueger, 2008, Russell & Miller, 1977, Clark, 1983, Speckhard &

Akhmedova, 2006). As a consequence, various scholars have concluded that economic

deprivation has no role on the emergence of terrorist movements (Krueger, 2007, Piazza,

2006, Krieger & Meierrieks, 2011).

Recently, this conclusion has been challenged by the claim that collective deprivation has

opposite effects on terrorism depending on the ideology of terrorist groups (Varaine, 2018).

According to this hypothesis, collective deprivation - i.e. long-term economic recession and

long-term growth of inequalities - fosters conservative oriented terrorism while it decreases

progressive oriented terrorism. The underlying causal mechanism is based on system

justification theory (Jost & Banaji, 1994, Jost et al., 2011): collective deprivation, by

generating threat and uncertainty, motivates the adherence to system-justifying beliefs that are

the core of conservative ideologies. This in turn affects individual motives for collective

action - including violent collective action lead by terrorist groups: threatening situations

decrease the individual propensity to engage in progressive collective action and increase the

propensity to engage in conservative collective action (Jost et al., 2017a). Empirically,

Varaine (2018) found evidence of such an opposite effect of collective economic deprivation

on the mobilization level of conservative and progressive radical movements in the case of

France from 1882 to 1980.

This hypothesis meets a new body of research focusing on the determinants of distinct

terrorist ideologies. Indeed, recent studies have demonstrated that terrorism has heterogeneous

determinants depending on its ideology (Kis-Katos et al., 2014, Brockhoff et al., 2016), which

may explain the persisting lack of consensus on the core causes of terrorism. The system-

justification hypothesis follows this argument by predicting that collective deprivation

differently affects conservative and progressive terrorist ideologies. Moreover, this new

hypothesis offers a powerful explanation of the historical and geographical variations in the

ideological motives of terrorism. For instance, the 1929 financial crisis is likely to be a key

cause of the level of mobilization and violence of terrorist fascist leagues in the 1930s in

various European countries. In the opposite direction, the left-wing wave of terrorism in the

3

1970s - incarnated for example by the Red Brigades in Italy, the Red Army Fraction in

Germany, the Weather Underground in the US and the Red Army in Japan - occurred after an

unprecedented period of economic prosperity. Nowadays, the re-emergence of right-wing and

xenophobic violent groups in Western countries coincides with economic stagnation and rise

of inequalities.

Yet, the empirical examination of this hypothesis has so far been limited. Firstly, the first

study testing the hypothesis (Varaine, 2018) was based on an ad hoc database suffering from

reliability and representativeness problems, thus requiring replication on other historical

cases. Secondly, we lack empirical clues to ascertain the individual causal mechanism.

Indeed, a key question is whether collective deprivation affects participation into terrorism

through the experience of individual deprivation or through the perception of a collective

threat. While the classic economic theory of terrorism predicts that individual deprivation

foster terrorism because it lowers the opportunity cost of participation into terror groups

(Freytag et al., 2011), the system justification hypothesis implies a socio-tropic mechanism.

Indeed, studies show that collective threats are far more determinant than individual threats in

affecting individuals’ ideological orientation (Feldman & Stenner, 1997, Stenner, 2005,

Onraet et al., 2013b), hence collective economic deprivation should affect the individual’s

propensity to participate into conservative or progressive terrorism independently from his/her

personal background. Thirdly, we have no evidence on whether the system justification

hypothesis account for the emergence of Islamist terrorism, whilst it is nowadays one of the

most prevalent form of political violence.

The present paper addresses these three issues together. Firstly, it provides a new empirical

test of the system justification hypothesis on the PIRUS database (START, 2017) about

domestic terrorists (N=1,417) in the United States from 1948 to 2016. I analyse whether the

proportion of right-wing (versus left-wing) terrorists at a given year depends on collective

deprivation in the US, operationalized through long-term recession of the national income and

long-term growth of inequalities. Multi-level logistic regression analyses confirm the

prediction that right-wing terrorism mobilizes more under periods of long-term economic

deprivation, while left-wing terrorism mobilizes more under improving economic conditions.

Secondly, the analyses support the view that the effect of collective deprivation is of socio-

tropic nature since it is not mediated through the terrorist’s individual level of deprivation.

Thirdly, the paper tests whether collective deprivation is a driving force behind the rise of

Islamist domestic terrorism in the US. Indeed, Islamist ideologies share definitional features

of conservative ideologies, such as the defence of tradition through the mythical theme of the

4

golden age of Islam (Moghadam, 2009, Khosrokhavar, 2015), hence Islamist terrorism should

be positively related to contexts of collective deprivation. However, results do not support this

view, possibly because the reference group of potential Islamist recruits does not maps onto

the national group.

In the next section, I develop the literature on the economic determinants of terrorism and

detail the rationale and empirical background of the theory of a differential effect of collective

deprivation on conservative and progressive terrorism. I develop in section three the

hypotheses of the paper, including the socio-tropic mechanism and the implications of the

theory on the case of Islamist terrorism. I then detail my empirical strategy based on the

PIRUS database. In section five, I present the results and discuss them in section six.

2. LITERATURE

2.1. The inconsistent effect of economic deprivation on terrorism Terrorism is commonly defined as the use of violence by sub-national actors to achieve

political ends (Krueger, 2007). Radicalization refers to the individual process leading to the

justification and, eventually, the use of political violence (McCauley & Moskalenko, 2008).

The idea that economic deprivation is a root cause of political violence has a long-lasting

history, going from Aristotle to Tocqueville and Marx, and persists in the recent literature on

terrorism. The alleged deprivation-terrorism nexus is generally based on one of two

mechanisms. According to the opportunity cost argument, derived from the economics of

crime (Becker, 1968), economic deprivation lowers the cost for individuals to withdraw from

economically productive occupations hence increasing the number of potential recruits for

terror groups (see for instance Freytag et al., 2011). According to the grievance argument,

derived from frustration aggression theory (Dollard et al., 1939), economic deprivation and

especially deprivation relative to other individuals and groups generates a sense of discontent,

which fuels the propensity for political violence (see for instance Gurr, 1970).

2.1.1. Evidence at the collective level

Empirically, the two mechanisms both yield the prediction that societies facing recessions and

high inequality generate more terrorism. Indeed, low growth and unequal societies have by

definition a higher share of poor individuals, which are easy recruits for terror groups

according to the opportunity cost argument. Alternatively, economic recessions and high

inequality are supposed to induce discontent against the political power or upper classes and

hence motivate political violence, according to the grievance argument.

5

However, results on this matter are disputed. While various cross-national studies do find that

societies facing economic recessions generate more terrorism (Blomberg et al., 2004, Caruso

& Schneider, 2011, Freytag et al. 2011, Ghatak & Gold, 2017, Kis-Katos, Liebert & Schulze,

2011�), others fail to find a significant relationship between economic growth and terrorism

(Drakos & Gofas, 2006, Kurrild-Klitgaard et al., 2006, Piazza, 2006, Krueger & Laitin,

2008). Using Granger-models to analyse domestic terrorism in seven European states from

1951 to 2004, Gries, Krieger and Meierrieks (2011) found that economic growth is causally

related to terrorism only in three out of seven countries. Recently, Choi (2014) found

evidence that industrial (and not agricultural) economic growth is negatively related to

domestic and transnational terrorism and positively related to suicide terrorism in a sample of

127 countries for 1970-2007.

Regarding the effect of inequality, results are also debated: whilst past research did not find

any relationship between inequality and terrorism (Abadie, 2006, Piazza, 2006, Kurrild-

Klitgaard et al., 2006) recent studies have contested this conclusion and supported the view

that inequality is a predictor of domestic terrorism (Krieger & Meierrieks, 2016, Piazza,

2017).

In sum, at a collective level, indicators of economic depressions and inequality do not appear

to be consistent and robust predictors of terrorism. This conclusion does not give much

support to both the opportunity-cost and grievance arguments.

2.1.2. Evidence at the individual level Individual data on terrorists’ backgrounds also cast doubts on the alleged deprivation-

terrorism nexus. Indeed, compared to their home countries’ population, terrorists are not

mostly drawn from the unprivileged. For instance, Krueger & Malečková (2003) found

members of lower class to be under-represented and educated individuals over-represented in

samples of Hezbollah members in Lebanon, Israeli Jew terrorists and Palestinian suicide

bombers. Similar conclusions have been drawn from samples of transnational jihadists

(Sageman, 2004), US homegrown jihadists (Krueger, 2008), Basque separatist terrorists

(Clark, 1983), Chechen suicide bombers (Speckhard & Akhmedova, 2006) and diverse

terrorist groups for the 1966-1976 period (Russel & Miller, 1977). Economic background and

level of education actually seem to vary across terrorists’ ideologies - leftist terrorists

generally coming from more educated and affluent backgrounds than rightists (Chermak &

Gruenewald, 2015, Smith & Morgan, 1994).

6

The overall observation clearly refutes the opportunity cost argument that terrorists engage in

result of a lack of economic opportunities1. All in all, ideological goals seem to matter more

than material considerations to explain terrorism: to understand terrorism, as Krueger

concludes, “voting is a better analogy than crime” (2007: 146).

2.2. A system justification approach of the economic deprivation-terrorism nexus Based on the preceding results, various studies have concluded that economic deprivation

does not matter in the explanation of terrorism (Krueger, 2007, Piazza, 2006, Krieger &

Meierrieks, 2011). Yet, this conclusion has recently been challenged by the claim that

collective economic deprivation affects terrorism in different directions depending on the

ideology of terrorists: according to this new hypothesis, collective deprivation - i.e. long-term

economic recession and growth of inequalities - is positively related to the mobilization of

conservative terrorism and negatively related to the mobilization of progressive terrorism.

2.2.1. Theoretical foundation: system justification theory The theoretical foundation of this hypothesis lies in system justification theory. According to

system justification theory, people share inherent motivations to see the social system in

which they live as legitimate (Jost & Banaji, 1994, Jost et al., 2012), which they satisfy by

endorsing politically conservative cognitions that justify the societal status quo. In particular,

conservative cognitions are better able than progressive cognitions (i.e. which challenge the

societal status quo) to address universal existential motivation to reduce threat and epistemic

motivation to reduce uncertainty (Jost et al., 2003, Hennes et al., 2012, Jost et al., 2017b, Jost

et al., 2007). These psychological needs vary across chronic and situational conditions,

affecting the attractiveness of conservative versus progressive cognitions. For instance,

individual variation in existential needs - measured through the perception of a dangerous

world - and epistemic needs - measured through the need for cognitive closure - are strongly

related to the endorsement of conservative beliefs (see Jost, 2017 for a recent meta-analysis).

Similarly, experimentally induced threat and uncertainty have been shown to increase the

adherence to conservatism (see for instance Duckitt & Fischer, 2003, Nail et al., 2009,

Thórisdóttir & Jost, 2011).

2.2.2. Collective deprivation and system justification

1 Regarding the grievance argument, it remains possible that people resort to terrorism as the result, not of personal deprivation but on the behalf of economic deprivation experienced at a societal level. Yet, as presented earlier, the inconsistent relationship between collective deprivation and terrorism does not give much support to this alternative grievance argument.

7

System justification theory offers a key explanation of the observed relationship between

collective economic deprivation and conservatism. Indeed, studies have for long noticed that

contexts of economic threat, such as recession periods, are related with enhanced right-wing

attitudes (Onraet et al., 2013a), far-right votes (Funke et al., 2016, De Bromhead et al., 2012)

and authoritarian behaviors (Sales, 1972, Sales, 1973, Doty et al., 1991). Moreover, individual

perceptions of collective economic threat are positively related to conservative attitudes

(Rickert, 1998, Feldman & Stenner, 1997, Onraet et al., 2013b). From a system justification

perspective, this so called authoritarianism-economic threat nexus (Rickert, 1998) is due to

the fact that system-justifying ideologies are inherently better at addressing the increased

feelings of threat induced by economic contractions. Experimental evidence supports this

view: Duckitt & Fischer (2003) found that the relationship between an induced socio-

economic declining scenario and authoritarianism was entirely mediated by the perception of

a societal threat.

2.2.3. The system justification model of collective action

Recently, Jost, Becker, Osborne and Badaan (2017) proposed integrating system justification

theory into the socio-psychological model of collective action (see van Zomeren et al., 2008).

They distinguished two ideological forms of collective action - system-challenging versus

system-supporting protest - and argued that these forms have different socio-psychological

antecedents. An important prediction of their model is that the occurrence and ideological

orientation of collective action should vary with the very factors that have been proven to

affect system justification. For instance, Hennes, Nam, Stern and Jost (2012) have found that

individual needs to reduce threat and uncertainty, by increasing system justification, were

positively related to support for the Tea Party - a conservative political movement - and

negatively related to support for Occupy Wall Street - a progressive political movement.

Similarly, Jost et al. (2012) found that a manipulations inducing uncertainty reduced

collective actions intentions among progressive political activists. Looking at more extreme

collective actions, Pauwels & Heylen (2017) recently found that perceived in-group threat, by

increasing authoritarianism, is eventually related to self-reported participation into right-wing

violence in a sample of young Belgian.

2.2.4. The system justification approach of the deprivation-terrorism nexus Based on the system justification model of collective action, it has been proposed that

collective deprivation, by generating threat and uncertainty, should increase the mobilization

level of conservative collective action and decrease the mobilization level of progressive

collective action (Varaine, 2018). Albeit formulated for collective action in general, this

8

hypothesis applies for both normative and non-normative forms of collective action including

terrorism (Tausch et al., 2011). It is worth noting that this hypothesis is about long-term

variation of the economic situation of the majority of a given society, which may be captured

through the long-term variation of wealth and inequalities. The reason is that conjuncture

variations are not likely to shape in-depth psychological motivations for system justification.

Indeed, studies demonstrating a link between collective deprivation and conservatism

typically compare long-term threatening periods (Sales, 1972, Sales, 1973, Doty et al., 1991,

Funke et al., 2016) or long-term experimental scenario (Duckitt & Fischer, 2003).

This system justification hypothesis has important empirical implications for the study of

terrorism. Indeed, if corroborated, it would shed a new light on the historical and geographical

variations of terrorists’ ideologies. For instance, at a global level, various scholars have

noticed the huge expansion of left-wing oriented terrorism in the 1970s, often called the “third

wave” of modern terrorism (Rapoport, 2004, Shughart, 2006), and its decline after the 1980s.

This phenomenon was insofar mostly related to the Cold War and the Soviet financing of

terror groups (O’brien, 1996, Robinson et al., 2006). Yet, this explanation did not explain why

anti-communist and right-wing terrorism, albeit also financed by the United States (George,

1991), did not rise such a strong level of mobilization and activism. The present hypothesis

provides an alternative explanation, linking the popular success of left-wing radical groups to

unprecedented global rates of development among this period.

Similarly, the hypothesis fits geographical variations at a same historical period of the level of

terror experienced by countries. Taking again the 1970s-1980s period, one may observe that

Western democracies experienced variable mobilization and intensity of violence arousing

from New Left terror groups: huge in Italy with the Red Brigades, high in West Germany

with the Red Army Fraction, more limited in France with Action Direct and the United States

with the Weather Underground, and very contained in United Kindgom (see della Porta, 2006,

START, 2016). This ranking strikingly maps with the rate of economic development

experienced during the preceding decades by those countries: between 1950 and 1970, the

GDP per capita nearly quadrupled in Italy, tripled in West Germany, more than doubled in

France, less than doubled in the USA, and increased by a half in United Kingdom (The

Maddison-Project, 2013).

Empirically, the system justification hypothesis meets a new body of research investigating

the determinant of distinct terrorist ideologies. This research generally demonstrates the

heterogeneity of causes of different ideologies of terror groups (Kis-Katos et al., 2014,

Brockhoff et al., 2016, Robinson et al., 2006). Yet, these studies have so far focused on the

9

effect of short-term economic indicators. Thus, to my knowledge, there has been no empirical

study directly testing the system justification hypothesis with the exception of Varaine (2018).

This study corroborated the hypothesis in the case of French radical movements - including

terrorist groups - from 1882 to 1980: long-term recession and long-term elevation of

inequality was positively related to the number of members of progressive radical

conservative movements and negatively related to the number of members of progressive

radical movements.

3. AIMS OF THE PRESENT RESEARCH

The present research aims at expanding the empirical examination of the system justification

hypothesis in three ways. Firstly, it offers an alternative test of the hypothesis on new data.

Secondly, the present research investigates the individual mechanism, at the level of terrorists,

by which collective deprivation affects conservative and progressive terrorism. Thirdly, the

paper analyses the relevance of the system justification hypothesis to explain Islamist

terrorism. I detail below the rationale for this three contributions.

3.1. A new empirical test of the system justification hypothesis

The previous empirical study testing the system justification hypothesis was based on an ad

hoc database suffering from reliability and representativeness problems (Varaine, 2018), thus

requiring replication on other historical cases. Hence, the first aim of the present study is to

test the hypothesis on a new historical case, based on more reliable data.

Moreover, the present research tests the system justification hypothesis in an alternative way.

The previous study focused on the effect of collective deprivation on the absolute number of

adherent of conservative and progressive radical movements. Another empirical formulation

of the hypothesis is that collective deprivation affects the main ideological orientation of

terrorists, that is the share of conservative versus progressive radical activists at a given

historical period. Indeed, the system justification hypothesis predicts that collective

deprivation generates a perception of social threat and uncertainty, which motivates the

individual adherence to system-justifying beliefs. This eventually increases the propensity for

conservative radicalization while it decreases the propensity for progressive radicalization.

Hence, under conditions of collective deprivation, terrorists should be mainly motivated by

conservative ideologies, while it should be mainly motivated by progressive ideologies under

times of collective improvement.

H1: collective deprivation increases the share of conservative terrorists (and decreases the

share of progressive terrorists).

10

As in the previous study, collective deprivation is defined as the long-term variation of the

wealth of the majority of the population - which decomposes into both (a) the evolution of the

absolute level of wealth and (b) the evolution of economic inequality. These two factors -

collective wealth and inequality - should have opposite relationship with the orientation of

terrorism. Indeed, a positive growth of inequality at a constant level of absolute wealth means

that the majority of the population has became poorer.

H1a: long-term growth of wealth decreases the share of conservative terrorists (and

increases the progressive terrorists).

H1b: long-term growth of inequality increases the share of conservative terrorists (and

decrases the progressive terrorists).

3.2. The socio-tropic effect of economic deprivation on terrorism

There is currently no evidence on the individual mechanism by which collective deprivation

affects conservative versus progressive terrorism. A crucial question is whether collective

deprivation affects radicalization through the activist’s own experience of economic

deprivation or through the activist’s perception of a collective threat. In other word, it is worth

ascertaining whether the effect of collective deprivation is of ego-tropic or socio-tropic

nature.

The classic opportunity-cost argument implies an ego-tropic mechanism: collective

deprivation induce more terrorism because the higher share of deprived people enjoys a lower

opportunity cost of quitting economically productive occupation (Freytag et al., 2011). In

contrary, the system justification hypothesis implies a socio-tropic mechanism. Indeed, one

premise of the hypothesis is that terrorism, contrary to crime, is by essence an ideological

phenomenon (see Krueger, 2007) and hence should be reactive to the very factors that

activate ideological motivations. To this regard, various studies demonstrate that collective

threats - including collective deprivation - are far more determinants that individual threats in

shaping ideological orientations (Feldman & Stenner, 1997, Stenner, 2005). For instance, the

meta-analysis of Onraet et al. shows that situations and perceptions of collective threats,

including economic threat, relate strongly to conservative attitudes, while the relationship is

weak for personal threats: they conclude that “threat on the level of society, rather than

anxiety originating from one’s private life, is the key factor in explaining the relationship

between threat and right-wing attitudes” (2013b, 245).

This conclusion perfectly fits findings from the literature on collective action. Indeed,

research has for long noticed that individual perceptions about the in-group situation matters

11

more than perception about one’s situation in explaining collective action intention (Koomen

& Fränkel, 1992, Walker & Mann, 1987, Smith et al., 2012).

Regarding the system justification hypothesis, this means that collective deprivation should

affect the ideological orientation of terrorists independently from their own experience of

economic deprivation:

H2: the effect of collective deprivation on the share of conservative terrorists (and the share

progressive terrorists) is not mediatized by the terrorists’ experience of economic deprivation.

Decomposing collective deprivation into wealth and inequality, this yields:

H2a: the effect of long-term growth of wealth on the share of conservative terrorists (and the

share progressive terrorists) is not mediatized by the terrorists’ experience of economic

deprivation.

H2b: the effect of long-term growth of inequality on the share of conservative terrorists (and

the share progressive terrorists) is not mediatized by the terrorists’ experience of economic

deprivation.

3.3. Economic deprivation and Islamist terrorism The system justification hypothesis directly aims at accounting for the mobilization of right-

wing versus left-wing oriented terrorism. However, the question remains open whether it may

account for the mobilization of terrorist ideologies that do not directly enter the left-right

dimension. The issue is especially sensible for Islamist terrorism, which is nowadays one of

the most prevalent form of political violence. I argue that the hypothesis may contribute

explaining the mobilization of Islamist terrorist to the extent that most Islamist ideologies

share definitional traits that categorize them as conservative. Indeed, salafi jihadism,

nowadays the most common Islamist terrorist ideology (Moghadam, 2009), has a clear

reactionary agenda in the sense that it advocates to a return to a mythicized golden age of

Islam (Moghadam, 2009, Torres et al., 2006, Khosrokhavar, 2015). More generally, salafi

jihadist beliefs fits the general outlook of authoritarian ideologies (Altemeyer, 1996, Stenner,

2005), through the recurring valorisation of obedience, social hierarchy, gender-based

divisions, moral purity and punitiveness against deviance (Khosrokhavar, 2015, Heghammer,

2017). Hence, as a conservative ideology, Islamist beliefs should be more appealing under

threatening social contexts, such as collective deprivation:

H3: collective deprivation increases the share of Islamist terrorists.

Decomposing collective deprivation into wealth and inequality, this yields:

H3a: long-term growth of wealth decreases the share of Islamist terrorists.

H3b: long-term growth of inequality increases the share of Islamist terrorists.

12

4. METHOD To test the hypotheses, the present study focuses on the ideological orientation of domestic

terrorists in the United States from 1948 to 2016. I analyse whether the probability for

terrorists to be left-wing, right-wing and Islamist depends on the level of collective

deprivation experienced in the US during their radicalization process.

4.1. Data Analyses are based on the Profiles of Individual Radicalization in the United States (PIRUS)

dataset, collected by START (National Consortium for the Study of Terrorism and Response

to Terrorism). This dataset gathers systematic information on 1,867 radical activists in the

United States from 1948 to 2016. To my knowledge, this is currently the only attempt to

capture data on the whole universe of individual radical activists in a country for such a long

period, thus providing sufficient historical variation to test the effect of year-level variables.

Moreover, the historical and geographical focus of the dataset fortunately allows keeping

relatively constant variables that has proven to affect radical movements, such the

centralization level (Dreher & Fischer, 2011), the state capacity (Li & Schaub, 2004), long-

term cultural features (Wiedenhaefer et al., 2007), and the level of democracy (Abadie, 2006).

Individuals are included in the dataset if they enter one of the following criteria: 1) they were

arrested / indicted / killed by public forces as a result of an ideological crime, 2) they were

members of an organization that was designated as terrorist by the United States, or 3) they

were members of an organization whose leader has been indicted as a result of an ideological

crime. These criteria map onto the definition of terrorism as the use of violence, including

violence against property, by sub-national actors to achieve political ends.

The dataset focuses on domestic terrorism, which refers to individuals who were in the United

States during most of the duration of their radicalization process. This focus on domestic

terrorism is convenient for testing the present hypotheses. Indeed, collective deprivation in a

country should affect the ideological orientation of people belonging to the national group,

nor of people coming from abroad.

The PIRUS database is a collective effort based on publicly available sources. An inter-coder

reliability test performed on 10% of the cases yielded an alpha score of .76 (Jasko et al.,

2017), indicating that the dataset respects common standards of reliability (>.70). The use of

publicly available sources of information surely implies multiple biases, which are hard to

evaluate as far as the dimension of the population of interest is unknown. Nonetheless, I do

not see major reasons to expect that the under- or over-reporting of some activists would

13

deeply alter the conclusions of my analyses. There are indeed good reasons to expect that the

year and the ideology of the activists respectively imply reporting biases: for example, past

events are likely to be under-reported because archival press records are more limited and less

easy of access than current press contents; similarly, right-wing activists may be differently

reported in public sources because their actions often target minority groups while left-

wingers generally attack governmental and private company targets. But, such biases should

not affect my analyses since I compare the temporal evolution of the relative frequency of

ideological orientations. In other words, there is no serious reason to expect that the reporting

biases of certain ideologies systematically vary across time, in parallel to the level of

collective deprivation.

4.2. Individual-level variables

Ideology. The dependent variable is the ideological orientation of radical activists. The

PIRUS dataset divides activists into four kinds of ideology: left-wing, right-wing, Islamist

and single-issue. I excluded single-issue activists from the sample because it gathers very

distinct political goals (e.g. anti-abortion, Puerto Rican independence, etc.) so it was difficult

to assert whether those activists were conservative or progressive.

To test H1 and H2, I analyse the probability for radical activists to be right-wing compared to

left-wing, which narrows the sample to N=1,070. The dependent variable is thus coded

0=left-wing (30.28% of the sample) and 1=right-wing (69,72% of the sample). This binary

opposition constitutes a fair measure of the distinction between progressive and conservative

terrorism. Indeed, most of right-wing terror groups in the PIRUS database can be defined as

conservative, albeit in an extreme sense, to the extent that they advocate for changing the

system to go back to an idealized past (Lipset & Raab, 1970, Parker & Barreto, 2014). This

reactionary orientation can take many forms: through the valorisation of the ethnical origins

of the nation (e.g. white supremacist groups such as the Klu Klux Klan, National Alliance, the

World Church of the Creator), the defence of traditional values and way of life (e.g. religious

fundamentalist groups such as the Army of God), the demise of the current institutions to

return to a past state of freedom (e.g. anti-federalist groups such as the Sovereign Citizen

Movement). In contrary, most of left-wing terror groups can be considered as progressive to

the extent that they advocate for ending-up with traditional societal arrangements to create a

new system. Similarly, this revolutionary orientation may take many forms: through

advocating of end of the capitalist economic system (e.g. anti-capitalist groups such as the

Weather Underground), the end of ethnical and identity-based discriminations (e.g. Black

14

power groups such as the Black Panthers, the Black Liberation Group), the re-definition of

moral and legal rights (e.g. animal rights groups such as the Animal Liberation Front).

To test H3, instead of right-wing orientation, I analyse the probability for radical activists to

be Islamists compared to left-wing, which narrows the sample to N=781. The dependent

variable is thus coded, 0=left-wing (41.49% of the sample) and 1=Islamist (58.51% of the

sample).

Individual deprivation. To test whether the effect of collective deprivation affects the

ideological orientation of terrorists through individual deprivation (H2), I used measures of

the activists’ level of deprivation. The first is the social class of the activist when he/she was

adult which is coded 0=low (e.g. receives welfare, lives close to the poverty line, regularly

unemployed or at best works a blue collar job, lives in subsidized housing), 1=middle (e.g.

does not receive welfare, lives in lower-middle or middle class neighbourhood, has steady

professional employment, owns or holds a mortgage on a house, has college degree), 2=high

e.g. works a high-income, white-collar job, lives and owns a house in a middle or upper class

neighborhood, can afford luxury items, has college degree or is self-employed as a successful

entrepreneur). The second variable, using the same coding, is about the activists’ childhood.

The first variable measures individual absolute deprivation, while the second, when included

as a covariate of the first in the regression model, allow measuring the effect of the individual

deprivation relative to his/her past status.

Control variables. Previous studies found that several socio-economic characteristics are

distinctive of right-wing versus left-wing terrorism (Chermak & Grunewald, 2015, Russel &

Miller, 1977, Smith & Morgan, 1994): compared to left-wing terrorists, right-wing terrorists

are generally less educated, older; they are also more likely to be male and to have a criminal

record. I controlled for these variables and for other variables included in the PIRUS dataset

which could also affect the dependent variable such as military experience and belonging to

an ethnic minority.

5.3. Year-level variables

Long-term collective deprivation. I used two indicators of collective deprivation: the long-

term variation of the national wealth, measured by the growth of the mean American fiscal

income, and the long-term variation of inequalities, measured by the growth of the share of

the national pre-tax income owned by the 10% richest (WID, 2017). By long-term, I mean

around 10 years, as in the experimental scenario used by Duckitt & Fischer (2003). However,

I am agnostic on what long-term precisely mean - e.g., whether it is 6 or 14 years. In

consequence, similarly to the previous study (Varaine, 2018), I computed different temporal

15

specifications for the independent variables: the growth of the wealth and inequalities over 5,

10, 15 and 20 years. The main analyses presented in the paper use the 10 years specification.

Analyses with the other temporal specifications are presented in the online appendix and I

will produce comments about them in order to check the consistency of the results at different

temporal specifications. Regarding the directions of the relationships, H1 predicts that the

lower the increase of wealth (H1a) and the higher the increase of inequalities (H1b) the higher

the probability for the radical activists at a given period to be right-wing oriented (compared

to left-wing).

Control variables. I included other year-level variables to control for alternative hypotheses

that could account for the share of right-wing versus left-wing ideology among the radical

activists at different historical periods. First, I included a measure of the ideological position

of the governors coded 0=Republican (right-wing) presidency, 1=Democrat (left-wing)

presidency. Indeed, one may expect that right-wing movements are more likely to mobilize

activists under a Democrat presidency (and conversely under a Republican presidency).

Besides, whilst my focus is on economic collective threat, various study found that other

kinds of threat are related to surges of conservatism, and may thus increase the probability for

the activists of a given period to be right-wing oriented. First, various studies indicate that

existential threat increases conservatism (Jost et al., 2003). Consistent with this view, Onraet

et al. (2013a) found that a composite indicator including the national homicide rate was

positively related with conservative views on cross-national surveys. Thus I included the one-

year lagged homicide rate as a control variable. Second, some studies indicate that

authoritarian attitudes crystallize under conditions of threat affecting the perceived

homogeneity of the in-group (Stenner, 2005), in particular the share of ethnic minorities

(Velez & Lavine, 2017, McCann, 2008). To capture this, I used the one-year lagged

immigration rate as a proxy. Albeit it is important to control for these variables, the literature

is divided on whether these threats increase conservatism or affirm attitudes of both

conservative and liberals (Stenner, 2005, Velez & Lavine, 2017): notably, McCann (2008)

found that a composite state-level indicators combining crime and non-White percentage of

the population increase authoritarianism in conservative states and decrease it in liberal states.

4.4. Temporal connection between individual radicalization and year-level variables To test the effect of year-level collective deprivation on the probability for a given terrorist to

be conservative versus progressive, I had to determine for each individual the year in which

he/she became radical. Hopefully, the PIRUS dataset includes information on each radical

activist’s date of exposure, which refers to his/her earliest mention by the sources as a radical

16

(see the inclusion criteria): for example, it may be the day the activist carried out an illegal

action or the earliest day the sources reveal that he/she was part of a designated terrorist

organization. I used this information to merge year-level indicators with the PIRUS dataset:

that is, for each activist, year-level variables refer to the year the activist became radical. Fig

1. plots the number of radical activists by year of exposure and ideology.

[Insert figure 1 around here]

One may oppose that using the year of exposure of the activist is not the best choice on

theoretical grounds. Indeed, collective deprivation is likely to affect radicalization at an earlier

phase: it is likely to trigger cognitive and emotional disposability before the radicalization

actually happens, and to have limited effects on the individuals’ motivations and actions as

soon he/she starts radicalizing. During this second phase, factors such as network (Sageman,

2004), in-group interactions (McCauley & Moskalenko, 2008) and political opportunity

structure (della Porta, 2006) should be of greater influence. Hopefully again, the PIRUS

dataset includes a variable estimating the radicalization duration. The variable includes three

categories: less than a year, between one and five years, more than five years. I produced

analyses using this information to estimate the level of collective deprivation at the start of the

individual radicalization2. The information on the radicalization duration was only available

for a limited number of activists (N=782), thus narrowing the sample of right-wing and left-

wing activists to N=400. In consequence, the analyses based on this procedure, which are

presented in the online appendix, simultaneously provide an alternative test and a robustness

check on a sub-sample of activists. I will produce comments about the outcomes in the section

relative to the robustness checks.

4.5. Handling missing values The main issue with the PIRUS dataset is that various variables include a non-trivial number

of missing values, due to the lack of public information for some characteristics of radical

activists. Table 1 summarizes the share of missing values among the variables that I

previously described. Hopefully, the share of missing values is not of concern for both the

dependent variable (ideology) and the main explanatory variables (the growth of wealth and

inequalities). In contrary, the amount of missing values is of concern in the case of some

2 To obtain an estimation of the year in which the activists started to radicalize, I subtracted to the year of exposure 1 years for people whose radicalization lasted less than a year, 3 for people whose radicalization lasted between one and five years and 5 years for people whose radicalization lasted more than five years.

17

control variables (education, previous criminal activity and military experience) and the

measures of individual deprivation (especially childhood deprivation).

[Insert table 1 around here]

Jasko, Lafree and Kruglanski (2017) faced the same issue when using the PIRUS dataset, and

handled it by using a multiple imputation procedure3. This method is particularly suited for

the present case because listwise deletion would dramatically reduce the sample, despite the

fact that the main variables have almost no missing values. In consequence, I used the same

procedure as Jasko et al. (2017) to impute multiple values only for my secondary variables:

that is individual control variables and individual indicators of deprivation. To do this, I

generated 25 datasets with imputations based on chained equation using all the individual

level predictors used by Jasko et al. (2017).

5. RESULTS In this section, I present the results of multi-level logistic regression analyses of the ideology

of radical activists. I first provide general comments about my modelling strategy and the

outcomes related to the control variables. I then detail the results about the relationship

between radical ideology and indicators collective deprivation for 10 years, the effect of

individual deprivation and the analyses related to Islamist terrorism. I then comment

robustness checks based on different temporal specifications, using an alternative starting year

of radicalization and without multiple-imputed values.

5.1. General comments

To test H1 and H2, I performed multi-level logistic regression models to predict the

probability for a given activist to be right-wing versus left-wing: that is, I nested individual

activists into year-groups with random effects. Indeed, I had enough groups (Nb of years=61)

and observations per group (mean=17.54) to respect common standards for multi-level

modelling. Table 4 and the regression tables in the online appendix B present the

unstandardized coefficients from regression models in ascendant steps: step I estimates an

empty model; step II and III respectively include the collective deprivation indicators (growth

of wealth and growth of inequalities), and step IV includes these variables together; step V

3 This method generates multiple datasets in which the missing values are filled with various predicted values based on multiple regression models including other variables from the dataset. That is, the method maximises all the available information in the dataset to estimate predictions of the missing cases. Contrarily to other methods imputation (such as interpolation or simple imputation), it does not fill the missing cases with one specific value, but with a chosen number of estimations which reflect the degree of uncertainty about the true value.

18

and VI respectively include year-level and individual level control variables, and step VII

includes both; finally step VIII includes the indicators of individual deprivation.

[Insert table 2 around here]

Estimates from the empty model in table 2 confirm that there is considerable year-level

variance in the dataset, and thus strongly justify the multi-level modelling strategy. Indeed,

the log of the group-level variance is significantly different from zero (lnsig2u=.86, p=.002)

and the Intra-class correlation (ICC) indicates that 41.76% of the total variance in the sample

is accounted by the year-level variance. This proportion is considerable: this means that

nearly half of the variation in the probability for the activists of the sample to be right-wing

depends on the year in which they became radicals. In other words, historical factors are of

great matter to explain the ideological orientation of radical activists.

Before focusing on the test of my hypotheses, I shall comment the outcomes relative to the

control variables. First, regarding the year-level variables, results from table 2 indicate that

the political orientation of the presidency is the only year-level control variable yielding a

significant effect on the probability for a radical exposure to be right-wing oriented. As

expected, radical exposures are more likely to be right-wing under a Democrat presidency (in

model VIII: B=1.02, p=.004). In contrast neither the lagged immigration rate (B=-.05,

p=.730) nor the lagged homicide rate (B=-.03, p=.803) display significant relationships with

the ideological orientation or radical activists, suggesting that existential and normative

threats do not specifically increase right-wing radicalization. In fact, this could either mean

that such threats do not affect radicalization or that existential and normative threats radicalize

both conservatives and progressives in opposite directions (Stenner, 2005, McCann, 2008,

Velez & Lavine, 2017).

Besides, most coefficients of individual-level variables are significant and in the expected

direction: having completed a higher degree than high school (In model VIII, B=-1.57,

p=.012) and belonging to an ethnic minority (B=-2.40, p<.001) both decrease the activists’

probability of being right-wing compared to left-wing, while being a man (B=1.65, p<.001),

relatively old (B=.08, p<.001) and having been previously involved in violent non-political

crimes (B=.93, p=.025) increase one’s likelihood of being right-wing. The relationship

between military experience and with right-wing orientation does not attain statistical

significance (B=.53, p=.162).

5.2. Main results

5.2.1. Collective deprivation and right-wing versus left-wing radicalization

19

In line with H1, results from table 2 consistently show that indicators of long-term collective

deprivation affect the ideological orientation of radical activists. Both the growth of wealth

and the growth of inequalities since 10 years display the expected significant relationship with

the probability for a radical activist to be right-wing oriented, and those relationships remain

significant when including both variables in the same model (step IV) and when controlling

for other relevant variables (steps V to VII). In line with H1a, the higher the growth of the

average fiscal income over 10 years, the lower the probability for radical activists of being

right-wing (in model VIII: B=-6.42, p<.001); in line with H1b, the higher the growth of the

share of the pre-tax national income owned by the 10% richest over 10 years the higher the

probability for radical activists of being right-wing (in model VII: B=7.82, p=.010). Plus,

these variables have a considerable explanatory power: while the share of variance accounted

by the year-level was 41.76% in the empty model, the inclusion of the collective deprivation

variables reduces this share to 24.38%. In particular, the long-term growth of the national

wealth has a substantive explanatory power: the inclusion of this variable withdraws the

significance of the log of the year-level variance (in model II: lnsig2u=.42, p>.10). This

means that, when taking into account the long-term growth of national wealth, the remaining

year-level variance in the share of right-wing activists is likely to be due to chance.

5.2.2 The socio-tropic effect of collective deprivation Model VIII includes indicators of individual-level deprivation. If the relationship between

collective deprivation and the ideological orientation of radical activists was mediated by

individual deprivations (the ego-tropic mechanism), the coefficients of these indicators would

be significant and their inclusion would reduce the absolute value of the coefficients of the

collective deprivation indicators. This is clearly not the case. Both the activist’s social class as

an adult and as a child do not significantly alter his/her probability of being right-wing, and

the inclusion of those variable has no effect on the coefficients of the collective deprivation

indicators - that is, the growth of wealth (H2a) and the growth of inequality (H2b). Thus the

evidence strongly supports the hypothesis (H2) that the effect of collective deprivation on

radicalization is of socio-tropic nature: that is, it does not affect radicalization through the

activist’s own experience of deprivation.

[Insert figures 2 and 3 around here] Based on model VIII, I predicted the probability for a radical activist of being right-wing

depending on the level of long-term deprivation, at average value for other covariates. Fig. 2

and 3 respectively plot those predictions depending on the growth of wealth and the growth of

inequalities for 10 years. These predictions clearly support H1. In line with H1a, the predicted

20

probability for a radical activist of being right-wing (compared to left-wing) goes from 95%

when the average American fiscal income has decreased of 10% in 10 years, such as in 2010,

to 46% when the average American fiscal income has increased of 40%, such as in 1968.

Conversely, the predicted probability for a radical activist of being right-wing (compared to

left-wing) goes from 61% when the share of the American pre-tax income owned by the 10%

richest has decreased of 5%, such as in 1960, to 92% when the average American fiscal

income has increased of 20%, such as in 1988.

5.2.3. The effect of collective deprivation on Islamist domestic terrorism

To test H3, I performed similar multilevel logistic regression analyses than for H1 and H2,

except that the dependent variable is now the probability for a radical activist to be Islamist

(compared to left-wing). To gain space, table 3 directly presents models including all

covariates - that is, except for the dependent variable, model I of table 3 is the same than

model VIII of table 2.

[Insert table 3 around here]

At a first look, results seem to support H3. Indeed, in model I of table 3, the coefficients

related to the collective deprivation indicators are significant and in the expected direction: in

line with H3a, the higher the growth of national wealth for the past 10 years the lower the

probability for radical activists to be Islamist (versus left-wing) (B=-18.94, p<.001); in line

with H3b, the higher the growth of inequalities the higher the probability for radical activists

to be Islamist (versus left-wing) (B=24.73, p=.004).

Nonetheless, model II of table 3, which controls for a simple linear year effect, contradicts

this view. Indeed, estimates from model 2 indicate that the probability for radical activists is

growing over time (the coefficient of the year linear variable is significantly positive, B=.31,

p<.001), and that controlling for this simple time trend entirely capture the previous effects of

the collective deprivation variables, which are no longer significant (for the variation of the

national wealth, B=2.10, p=.598 ; for the variation of inequalities, B=-8.20, p=.268). This

means that the statistical relationships in model I were simply reflecting the fact that both the

share of Islamist activists, and the level of collective deprivation have grown over time in the

case under study. In contrary, model III of table 2 indicates that controlling for a linear year

effect does not alter the previous results regarding the share of right-wing (versus left-wing)

domestic terrorists. Indeed, the estimates are similar to the coefficient of model VIII in table

2. In conclusion, the analyses do not support the view that collective deprivation increases the

share of Islamists among domestic terrorists (H3). I will discuss this result in the conclusion

of the paper.

21

Interestingly, results from the individual control variables indicate that Islamist domestic

terrorist display mostly similar socio-demographic features than left-wingers, whilst we

noticed that there was great differences in the profiles of right-wing versus left-wing domestic

terrorists. Indeed, the only stable and significant differences are the share of people from

ethnic minorities which is unsurprisingly higher among Islamists compared to left-wingers (in

model II, B=4.02, p<.001) and military experience: Islamists are more likely to have had a

military experience than left-wingers, even though this effect is small both in size and

significance (in model II, B=1.50, p=.050). The other socio-demographic variables have no

significant effect on the probability to be Islamist compared to left-wing.

5.3. Robustness checks Finally, I performed a series of tests to check the robustness of the results regarding H1 and

H2.

5.3.1. Tests on different temporal specifications As the previous study (Varaine, 2018), I performed similar regression models using different

temporal specifications for the indicators of collective deprivation: that is, the growth of the

wealth and the growth of inequalities for 5, 15 and 20 years. The results are presented in the

online appendix. Firstly, in line with H1, no analyses report significant opposite relationships

at any temporal specification: that is, the coefficients of the indicators of collective

deprivation are either significant in the expected direction or non-significant. I consider that

this result corroborates H1.

Moreover, the growth of the national wealth displays a significant relationship with the

activists’ ideology in every temporal specification, which indicates a very robust relationship.

The growth of inequalities has a consistent significant effect on the activists’ ideology only in

short-term specifications (the variation for 5 and 10 years) while it has mostly non-significant

effects in long-term specifications (the variation for 15 and 20 years).

5.3.2. Alternative measure of the period of radicalization

To further check the robustness of the results, I first performed similar regression models on

the sub-sample of radical activists for which I had information on their radicalization duration

(N=400) and using the year they started to radicalize as the basis for merging the data with the

year-level variables. Results are presented in the online appendix. They are mostly similar to

the results of the main analyses.

5.3.2. Tests without multiple imputed values

Then, I also tested my models on the original dataset (without multiple imputed values)

excluding the control variables for which the amount of missing cases was of concern

22

(education, previous criminal activity and military experience). The results are presented in

the online appendix. Again, they are mostly similar to the results of the main analyses.

Notably, the effect of the growth of inequalities for 5 years and the effect of the growth of the

national wealth at every temporal specification remains significant on a narrow sample of 159

activists (that is, when including the individual deprivation indicators without multiple

imputed values). This indicates that the results, especially regarding the growth of the national

wealth, are strongly robust.

6. CONCLUSION

Contrary to a widespread opinion, research has not given much support to the theory that

economic deprivation is a core cause of terrorism, leading to the conclusion that terrorism has

little to do with economic conditions (Krueger, 2007, Piazza, 2006, Krieger & Meierrieks,

2011). The present paper challenges this conclusion based on system justification theory.

Contrary to classic approaches of the terrorism-deprivation nexus based on the economic of

crime (Freytag et al., 2011) and frustration aggression theory (Gurr, 1970), the present theory

argues that collective deprivation has differential effects on terrorism depending on its

ideology: conservative terrorism should mobilize more in periods of economic deprivation

whilst progressive terrorism should mobilize more in times of economic improvement.

Empirical results corroborate this view, by showing that the ideological orientation of

domestic radical activists in the USA since World War II is closely related to the level of

economic deprivation experienced at the national level: long-term recession of the national

wealth and growth of inequality are related to a conservative orientation of domestic

terrorism. These results match with recent findings on French radical movements (Varaine,

2018).

The present work supports recent developments of system justification theory implying that

ideology matters in the explanation of collective action (Jost et al., 2017). Indeed, collective

action, including violent action by terrorist groups (Tausch et al., 2011), is by essence

ideological insofar as it reveals individual preferences about the prevailing social system (Jost

et al., 2017). For that reason, individual propensity to join collective action should be related

to the very factors that affect ideological motivations. This conjecture was so far mostly

supported by studies on normative or moderately non-normative collective action intentions

(Jost et al., 2012, Hennes et al., 2012, Osborne & Sibley, 2013). The present study indicates

that system-justification theory is also relevant to explain radical collective action.

23

The present paper highlights that research on terrorism should treat it as an ideological

phenomenon, or as stated by Krueger, that to understand terrorism “voting is a better analogy

than crime” (2007: 146). The present study has focused on collective economic deprivation,

because it has for long been shown that collective deprivation is a key determinant of

ideological orientation (Sales, 1972, Sales, 1973, Doty et al., 1991, Onraet et al., 2013a), and

voting choices in particular (Funke et al., 2016, De Bromhead et al., 2012). Yet, there are

good reasons to expect that other factors have different relationship with different ideological

forms of terrorism. Indeed, recent studies have demonstrated that many variables that had

been thought to uniformly cause terrorism in fact have distinct effects on left-wing, right-

wing, separatist or religious terrorism (Kis-Katos et al., 2014, see also Brockhoff et al., 2016).

This heterogeneity of terror may explain the current lack of consensus on the core causes of

terrorism (Kis-Katos et al., 2011, de Mesquita, 2008). Hence, future research would benefit

from disaggregate terrorism into different ideological categories.

This paper has two main limitations. The first is about the psychological causal mechanism

linking collective deprivation and radicalization. The present evidence indicates that the effect

of collective deprivation is not ego-tropic: that is, it does not affect radicalization through the

activist’s experience of deprivation, but more probably through the perception of a social

threat. This finding is consistent with the literature on ideology, which shows that collective

threats are more determinant than individual threats in influencing individual conservative

attitudes (Onraet et al., 2013b, Feldman & Stenner, 1997, Stenner, 2005). Yet, the present

data on radical activists does not allow testing directly the psychological mechanism that I

have put forward, which would require measuring mediating psychometric factors such as

individual needs to reduce threat and uncertainty (Jost et al., 2003, Hennes et al., 2012).

Disparate evidence supports the alleged causal mechanism: for instance, Pauwels and Heylen

(2017) found that perceptions of group-level threat, through increasing right-wing

authoritarianism, predicted youth involvement in right-wing violence, and Rieger et al. (2017)

found that individuals primed with a scenario of collective economic threat felt more

uncertain, which, in interaction with authoritarianism, increased their interest in right-wing

extremist propaganda. However, more experimental and survey data is needed to conclude

that collective deprivation does increase conservative radicalization through the experience of

threat and uncertainty.

Secondly, cross-sectional research would be needed both in general to assert the robustness of

the findings and in particular to test whether the theory accounts for Islamist terrorism.

Indeed, whilst the empirical analyses do not support it, it would be premature to entirely reject

24

the hypothesis that collective deprivation drives Islamist domestic terrorism. Actually, one

may conclude from the present analyses that the data is insufficient to conclude, because

Islamist terrorism is a too recent phenomenon in the case under study to relate it to historical

explanatory factors. This means that the hypothesis should be tested in a cross-sectional rather

than in a longitudinal way. Nonetheless, one could also argue from the data that collective

deprivation is not likely to be a major cause behind Islamist domestic terrorism because, if it

was so, indicators of collective deprivation would kept explaining some of the proportion of

Islamist terrorism even when controlling for a simple linear time trend. Indeed, the level of

collective deprivation has not followed a perfectly linear pattern during the last years in the

case under study. Thus, if the hypothesis was true, small changes from the linear evolution of

the level of collective deprivation would have affected the share of Islamist terrorists, which

is not the case.

From a theoretical point of view, another possibility is that the absence of effect of collective

deprivation on the share of Islamist domestic terrorists is due to the fact that the reference

group of potential Islamist recruits does not map onto the American national group. Indeed, a

premise of the theory is that collective threat affects the individuals’ ideological orientation

only if they minimally identify with the group under threat. In the case of radical Islamists, it

is possible that this mechanism occurs, but not on the basis of the American in-group. In

consequence, it should not be so surprising for the mobilization level of American Islamists to

be unrelated with the American-level deprivation, which I analyse in this paper. In contrast, it

would be relevant for future research to test whether Islamist domestic terrorism is affected by

deprivation experienced by American Muslims in particular, or by Muslim populations from

abroad.

7. REFERENCES Abadie, A. (2006). Poverty, political freedom, and the roots of terrorism. American Economic

Review, 96(2), 50-56.

Altemeyer, B. (1996). The authoritarian specter. Harvard University Press.

Becker, G. S. (1968). Crime and punishment: An economic approach. In The economic

dimensions of crime (pp. 13--68). Palgrave Macmillan, London.

Blomberg, S. B., Hess, G. D., & Weerapana, A. (2004). Economic conditions and terrorism.

European Journal of Political Economy, 20(2), 463-478.

25

Brockhoff, S., Krieger, T., & Meierrieks, D. (2016). Heterogeneous terrorism: determinants of

left-wing and nationalist-separatist terrorism in Western Europe. Peace Economics, Peace

Science and Public Policy, 22(4), 393-401.

Caruso, R., & Schneider, F. (2011). The socio-economic determinants of terrorism and

political violence in Western Europe (1994–2007). European Journal of Political Economy,

27, S37-S49.

Chermak, S., & Gruenewald, J. A. (2015). Laying a foundation for the criminological

examination of right-wing, left-wing, and Al Qaeda-inspired extremism in the United States.

Terrorism and Political Violence, 27(1), 133-159.

Choi, S. W. (2014). Economic growth and terrorism: domestic, international, and suicide.

Oxford Economic Papers, 67(1), 157-181.

Clark, R. P. (1983). Patterns in the Lives of ETA Members. Studies in Conflict & Terrorism,

6(3), 423-454.

De Bromhead, A., Eichengreen, B., & O'Rourke, K. H. (2012). Right-wing political

extremism in the Great Depression (No. w17871). National Bureau of Economic Research.

della Porta, D. (2006). Social movements, political violence, and the state: A comparative

analysis of Italy and Germany. Cambridge University Press.

Dollard, J., Miller, N. E., Doob, L. W., Mowrer, O. H., & Sears, R. R. (1939). Frustration

and aggression. New Haven: Yale University Press.

Doty, R. M., Peterson, B. E., & Winter, D. G. (1991). Threat and authoritarianism in the

United States, 1978–1987. Journal of personality and social psychology, 61(4), 629.

Drakos, K., & Gofas, A. (2006). In search of the average transnational terrorist attack venue.

Defence and Peace Economics, 17(02), 73-93.

Dreher, A., & Fischer, J. A. (2011). Does government decentralization reduce domestic

terror? An empirical test. Economics Letters, 111(3), 223-225.

Duckitt, J., & Fisher, K. (2003). The impact of social threat on worldview and ideological

attitudes. Political Psychology, 24(1), 199-222.

Feldman, S., & Stenner, K. (1997). Perceived threat and authoritarianism. Political

Psychology, 18(4), 741-770.

Freytag, A., Krüger, J. J., Meierrieks, D., & Schneider, F. (2011). The origins of terrorism:

Cross-country estimates of socio-economic determinants of terrorism. European Journal of

Political Economy, 27, S5-S16.

Funke, M., Schularick, M., & Trebesch, C. (2016). Going to extremes: Politics after financial

crises, 1870–2014. European Economic Review, 88, 227-260.

26

George, A. (Ed.). (1991). Western state terrorism. Cambridge: Polity Press.

Ghatak, S., & Gold, A. (2017). Development, discrimination, and domestic terrorism:

Looking beyond a linear relationship. Conflict Management and Peace Science, 34(6), 618-

639.

Gries, T., Krieger, T., & Meierrieks, D. (2011). Causal linkages between domestic terrorism

and economic growth. Defence and Peace Economics, 22(5), 493-508.

Gurr, Ted. (1970). Why men rebel. Princeton University Press.

Hegghammer, T. (Ed.). (2017). Jihadi Culture. Cambridge University Press.

Hennes, E. P., Nam, H. H., Stern, C., & Jost, J. T. (2012). Not all ideologies are created equal:

Epistemic, existential, and relational needs predict system-justifying attitudes. Social

Cognition, 30(6), 669-688.

Jasko, K., LaFree, G., & Kruglanski, A. (2017). Quest for significance and violent extremism:

The case of domestic radicalization. Political Psychology, 38(5), 815-831.

Jost, J. T. (2017). Ideological asymmetries and the essence of political psychology. Political

psychology, 38(2), 167-208.

Jost, J. T., & Banaji, M. R. (1994). The role of stereotyping in system‐justification and the

production of false consciousness. British journal of social psychology, 33(1), 1-27.

Jost, J. T., Becker, J., Osborne, D., & Badaan, V. (2017a). Missing in (collective) action:

Ideology, system justification, and the motivational antecedents of two types of protest

behavior. Current Directions in Psychological Science, 26(2), 99-108.

Jost, J. T., Chaikalis-Petritsis, V., Abrams, D., Sidanius, J., Van Der Toorn, J., & Bratt, C.

(2012). Why men (and women) do and don’t rebel: Effects of system justification on

willingness to protest. Personality and Social Psychology Bulletin, 38(2), 197-208.

Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. J. (2003). Political conservatism as

motivated social cognition. Psychological bulletin, 129(3), 339.

Jost, J. T., Liviatan, I., Van Der Toorn, J., Ledgerwood, A., Mandisodza, A., & Nosek, B. A.

(2011). System justification: A motivational process with implications for social conflict. In

Justice and conflicts (pp. 315-327). Springer, Berlin, Heidelberg.

Jost, J. T., Napier, J. L., Thorisdottir, H., Gosling, S. D., Palfai, T. P., & Ostafin, B. (2007).

Are needs to manage uncertainty and threat associated with political conservatism or

ideological extremity?. Personality and social psychology bulletin, 33(7), 989-1007.

Jost, J. T., Stern, C., Rule, N. O., & Sterling, J. (2017b). The politics of fear: is there an

ideological asymmetry in existential motivation?. Social cognition, 35(4), 324-353.

27

Khosrokhavar, F. (2015). Inside jihadism: Understanding jihadi movements worldwide.

Routledge.

Kis-Katos, K., Liebert, H., & Schulze, G. G. (2011). On the origin of domestic and

international terrorism. European Journal of Political Economy, 27, S17-S36.

Kis-Katos, K., Liebert, H., & Schulze, G. G. (2014). On the heterogeneity of terror. European

Economic Review, 68, 116-136.

Koomen, W., & Fränkel, E. G. (1992). Effects of experienced discrimination and different

forms of relative deprivation among Surinamese, a Dutch ethnic minority group. Journal of

community & applied social psychology, 2(1), 63-71.

Krieger, T., & Meierrieks, D. (2011). What causes terrorism?. Public Choice, 147(1-2), 3-27.

Krieger, T., & Meierrieks, D., Does Income Inequality Lead to Terrorism? (March 23, 2016).

CESifo Working Paper Series No. 5821. Available at SSRN:

https://ssrn.com/abstract=2766910

Krueger, A. B. (2007). What Makes a Terrorist. Princeton University Press.

Krueger, A. B. (2008). What makes a homegrown terrorist? Human capital and participation

in domestic Islamic terrorist groups in the USA. Economics Letters, 101(3), 293-296.

Krueger, A. B., & Laitin, D. D. (2008). Kto kogo?: A cross-country study of the origins and

targets of terrorism. Terrorism, economic development, and political openness, 148-173.

Krueger, A. B., & Malečková, J. (2003). Education, poverty and terrorism: Is there a causal

connection?. Journal of Economic perspectives, 17(4), 119-144.

Kurrild-Klitgaard, P., Justesen, M. K., & Klemmensen, R. (2006). The political economy of

freedom, democracy and transnational terrorism. Public Choice, 128(1-2), 289-315.

Li, Q., & Schaub, D. (2004). Economic globalization and transnational terrorism: A pooled

time-series analysis. Journal of Conflict Resolution, 48(2), 230-258.

Lipset, S. M., & Raab, E. (1970). The politics of unreason: right wing extremism in America,

1790-1970 (Vol. 5). New York: Harper & Row.

McCann, S. J. (2008). Societal threat, authoritarianism, conservatism, and US state death

penalty sentencing (1977-2004). Journal of Personality and Social Psychology, 94(5), 913.

McCauley, C., & Moskalenko, S. (2008). Mechanisms of political radicalization: Pathways

toward terrorism. Terrorism and political violence, 20(3), 415-433.

de Mesquita, E. B. (2008). The political economy of terrorism: A selective overview of recent

work. The Political Economist, 10(1), 1-12.

Moghadam, A. (2009). Motives for martyrdom: Al-Qaida, Salafi Jihad, and the spread of

suicide attacks. International Security, 33(3), 46-78.

28

Nail, P. R., McGregor, I., Drinkwater, A. E., Steele, G. M., & Thompson, A. W. (2009).

Threat causes liberals to think like conservatives. Journal of Experimental Social Psychology,

45(4), 901-907.

O'brien, S. P. (1996). Foreign policy crises and the resort to terrorism: A time-series analysis

of conflict linkages. Journal of Conflict Resolution, 40(2), 320-335.

Onraet, E., Van Hiel, A., & Cornelis, I. (2013a). Threat and Right‐Wing Attitudes: A Cross‐

National Approach. Political Psychology, 34(5), 791-803.

Onraet, E., Van Hiel, A., Dhont, K., & Pattyn, S. (2013b). Internal and external threat in

relationship with right‐wing attitudes. Journal of Personality, 81(3), 233-248.

Parker, C. S., & Barreto, M. A. (2014). Change They Can't Believe In: The Tea Party and

Reactionary Politics in America-Updated Edition. Princeton University Press.

Pauwels, L. J., & Heylen, B. (2017). Perceived Group Threat, Perceived Injustice, and Self-

Reported Right-Wing Violence: An Integrative Approach to the Explanation Right-Wing

Violence. Journal of interpersonal violence, 0886260517713711.

Piazza, J. A. (2006). Rooted in poverty?: Terrorism, poor economic development, and social

cleavages. Terrorism and political Violence, 18(1), 159-177.

Rapoport, D. C. (2004), The Four Waves of Modern Terrorism, in Cronin, A. K., & Ludes, J.

M. Ludes, Attacking Terrorism: Elements of a Grand Strategy, Georgetown University Press,

46--73.

Rickert, E. J. (1998). Authoritarianism and economic threat: Implications for political

behavior. Political Psychology, 19(4), 707-720.

Rieger, D., Frischlich, L., & Bente, G. (2017). Propaganda in an insecure, unstructured world:

How psychological uncertainty and authoritarian attitudes shape the evaluation of right-wing

extremist internet propaganda. Journal for Deradicalization, (10), 203-229.

Robison, K. K., Crenshaw, E. M., & Jenkins, J. C. (2006). Ideologies of violence: The social

origins of Islamist and leftist transnational terrorism. Social Forces, 84(4), 2009-2026.

Russell, C. A., & Miller, B. H. (1977). Profile of a Terrorist. Studies in conflict & terrorism,

1(1), 17-34.

Sageman, M. (2004). Understanding terror networks. University of Pennsylvania Press.

Sales, S. M. (1972). Economic threat as a determinant of conversion rates in authoritarian and

nonauthoritarian churches. Journal of Personality and Social Psychology, 23(3), 420.

Sales, S. M. (1973). Threat as a factor in authoritarianism: An analysis of archival data.

Journal of Personality and Social Psychology, 28(1), 44.

29

Shughart, W. F. (2006). An analytical history of terrorism, 1945–2000. Public Choice, 128(1-

2), 7-39.

Smith, B. L., & Morgan, K. D. (1994). Terrorists right and left: Empirical issues in profiling

American terrorists. Studies in Conflict & Terrorism, 17(1), 39-57.

Smith, H. J., Pettigrew, T. F., Pippin, G. M., & Bialosiewicz, S. (2012). Relative deprivation:

A theoretical and meta-analytic review. Personality and Social Psychology Review, 16(3),

203-232.

Speckhard, A., & Akhmedova, K. (2006). The making of a martyr: Chechen suicide

terrorism. Studies in conflict & Terrorism, 29(5), 429-492.

START (2017), Profiles of Individual Radicalization in the United States (PIRUS), [Data

file]. Retrieved from (http://www.start.umd.edu/data-tools/profiles-individual-radicalization-

united-states-pirus).

START (2016), Global Terrorism Database, [Data file]. Retrieved from

(https://www.start.umd.edu/gtd).

Stenner, K. (2005). The authoritarian dynamic. Cambridge University Press.

Tausch, N., Becker, J. C., Spears, R., Christ, O., Saab, R., Singh, P., & Siddiqui, R. N. (2011).

Explaining radical group behavior: Developing emotion and efficacy routes to normative and

nonnormative collective action. Journal of personality and social psychology, 101(1), 129.

The Maddison Project (2013), [Data file]. Retrieved from

(http://www.ggdc.net/maddison/maddison‐project/home.htm).

Thórisdóttir, H., & Jost, J. T. (2011). Motivated closed‐mindedness mediates the effect of

threat on political conservatism. Political Psychology, 32(5), 785-811.

Torres, M. R., Jordán, J., & Horsburgh, N. (2006). Analysis and evolution of the global

jihadist movement propaganda. Terrorism and Political Violence, 18(3), 399-421.

Van Zomeren, M., Postmes, T., & Spears, R. (2008). Toward an integrative social identity

model of collective action: A quantitative research synthesis of three socio-psychological

perspectives. Psychological bulletin, 134(4), 504.

Varaine, S. (2018). Bad times are not good times for revolutions: Collective deprivation and

the mobilization level of French radical movements (1882–1980). Journal of Community &

Applied Social Psychology.

Velez, Y. R., & Lavine, H. (2017). Racial diversity and the dynamics of authoritarianism. The

Journal of Politics, 79(2), 519-533.

Walker, L., & Mann, L. (1987). Unemployment, relative deprivation, and social protest.

Personality and Social Psychology Bulletin, 13(2), 275-283.

30

Wiedenhaefer, R. M., Dastoor, B. R., Balloun, J., & Sosa-Fey, J. (2007). Ethno-psychological

characteristics and terror-producing countries: Linking uncertainty avoidance to terrorist acts

in the 1970s. Studies in Conflict & Terrorism, 30(9), 801-823.

WID (2017), World Wealth & Income Database, [Data file]. Retrieved from

(http://wid.world/data).

Table 1: share of missing values among the variables for the sample of

right-wing and left-wing radical activists Valid N % missing Year-level variables Δ average fiscal income since 10 years 1026 4,1% Δ share of the national income owned by the richest 10% since 10 years 1026 4,1% Left-wing government 1057 1,2% Net migration rate (lagged) 1067 0,3% Homicide rate (lagged) 1023 4,4% Individual-level variables Ideology 1070 0,0% Education 328 69,3% Age 1022 4,5% Gender 1070 0,0% Military experience 611 42,9% Previous criminal activity 534 50,1% Ethnicity 971 9,3% Social stratum (adulthood) 451 57,9% Social stratum (childhood) 214 80,0%

Table 2: unstandardized coefficients from multi-level logistic regression analyses of right-wing radical exposure (left-wing as reference)

Predictors I II III IV V VI VII VIII Year-level variables Δ wealth over the past 10 years -6.67 (1.47)*** -6.21 (1.34)*** -6.81 (1.23)*** -6.14*** (1.45) -6.42*** (1.40) -6.42*** (1.40) Δ inequality over the past 10 years 8.56 (2.79)*** 7.44 (2.46)*** 9.40 (2.62)*** 5.66** (2.63) 7.75** (3.02) 7.82*** (3.02) Left-wing government 0.81 (0.29)*** 1.02*** (0.35) 1.02*** (0.36) Net migration rate (lagged) -0.04 (0.14) -0.05 (0.17) -0.06 (0.17) Homicide rate (lagged) -0.10 (0.09) -0.03 (0.10) -0.03 (0.10) Individual-level variables Education (Less than high school as reference) High school -0.24 (0.41) -0.30 (0.41) -0.28 (0.42) More than high school -1.63*** (0.60) -1.63*** (0.60) -1.58*** (0.61) Age 0.08*** (0.01) 0.08*** (0.01) 0.08*** (0.01) Male (Female as reference) 1.62*** (0.34) 1.62*** (0.34) 1.65*** (0.35) Military experience (None as reference) 0.53 (0.37) 0.59 (0.38) 0.53 (0.38) Previous criminal activity (None as reference) Previous (nonviolent) minor activity 0.44 (0.33) 0.44 (0.33) 0.41 (0.35) Previous (nonviolent) serious activity 0.47 (0.44) 0.53 (0.45) 0.44 (0.48) Previous violent crime 1.01** (0.42) 1.02** (0.41) 0.93** (0.41) Minority status (No as reference) -2.40*** (0.37) -2.38*** (0.37) -2.40*** (0.40) Social stratum, adulthood (Low as reference) Middle 0.46 (0.58) High 0.81 (0.80) Social stratum, childhood (Low as reference) Middle 0.16 (0.60) High -0.01 (0.98) Constant 0.99 (0.23)*** 2.05 (0.31)*** 0.37 (0.29) 1.41*** (0.34) 0.17 (1.14) -1.37** (0.63) -3.71*** (1.43) -4.42*** (1.50) lnsig2u 0.86 (0.26)*** 0.42 (0.29) 0.66 (0.28)** 0.06 (0.34) -0.35 (0.37) 0.12 (0.39) -0.19 (0.42) -0.22 (0.44) ICC 42% 32% 37% 24% 17% 26% 20% 20% Nomber of observations 1,070 1,026 1,026 1,026 1,023 1,026 1,023 1,023 Number of groups (Years) 61 60 60 60 58 60 58 58

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

1

Table 3: unstandardized coefficients from multi-level logistic regression analyses of Islamist and right-wing radical exposure (left-wing as reference)

Outcome Islamist Right-wing Predictors I II III

Year-level variables Δ wealth over the past 10 years -18.94*** (3.95) 2.10 (3.99) -8.17*** (2.01) Δ inequality over the past 10 years 24.73*** (8.69) -8.20 (7.40) 9.93*** (3.51) Left-wing government -0.44 (0.68) 0.11 (0.56) 0.90** (0.37) Net migration rate (lagged) 0.25 (0.39) -0.34 (0.29) 0.06 (0.19) Homicide rate (lagged) -1.29*** (0.24) 0.33 (0.33) -0.14 (0.14) Year (linear) 0.31*** (0.07) -0.03 (0.02) Individual-level variables Education (Less than high school as reference)

High school -0.07 (0.83) -0.03 (0.80) -0.28 (0.42) More than high school -1.17* (0.69) -1.08 (0.68) -1.58*** (0.61) Age 0.01 (0.02) 0.01 (0.02) 0.08*** (0.01) Male (Female as reference) 0.92 (0.58) 0.93* (0.55) 1.66*** (0.35) Military experience (None as reference) 1.49* (0.77) 1.50** (0.76) 0.53 (0.38) Previous criminal activity (None as reference)

Previous (nonviolent) minor activity -0.59 (0.55) -0.60 (0.53) 0.40 (0.35) Previous (nonviolent) serious activity -0.82 (0.88) -0.70 (0.84) 0.43 (0.48) Previous violent crime -0.27 (0.82) -0.37 (0.77) 0.93** (0.41) Minority status (No as reference) 4.22*** (0.53) 4.02*** (0.50) -2.38*** (0.40) Social stratum, adulthood (Low as reference) Middle -0.24 (0.83) -0.28 (0.81) 0.46 (0.58) High 0.17 (1.19) 0.19 (1.18) 0.81 (0.80) Social stratum, childhood (Low as reference) Middle 0.60 (0.75) 0.56 (0.72) 0.17 (0.60) High 0.20 (1.03) 0.11 (1.05) -0.01 (0.99)

Constant 5.41* (2.87) -629.14*** (137.98) 48.23 (43.24)

lnsig2u 0.98* (0.58) -3.04 (11.86) -0.22 (0.44) ICC 45% 1% 20% Nomber of observations 709 709 1,023 Number of groups (Years) 51 51 58

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1