Socioeconomic hardships, religiosity, and radicalization: A non-linear exploration

Abstract

Socioeconomic hardships are often advocated as the drivers of radicalization, but the existing research shows mixed evidence for this relationship. This study argues that socioeconomic hardships should increase the likelihood of radicalization only for sufficiently religious people, thereby implying a non-linear relationship. The study suggests that as a result of the lower opportunity costs of extreme acts during socioeconomic hardships, the likelihood of committing or supporting such acts for obtaining religiously inspired mental rewards should be the highest for the economically disadvantaged, religious individuals. The study tests this hypothesis through the non-linear threshold regression method developed by Hansen (2000), using survey data from the Khyber Pakhtunkhwa province of Pakistan. Results indicate that radicalization has a statistically significant positive relationship with indicators of the individual level socioeconomic conditions such as the perceptions of poor economic prospects, individual relative deprivation, injustice, and inequality only above the religiosity threshold. These findings support the hypothesis that socioeconomic hardships (or the perception of these) drive radicalization only in sufficiently religious people. This implies that apart from security-centric approaches, radicalization can also be deterred by increasing the opportunity costs of political violence through socioeconomic improvements. Moreover, promoting balanced/diverse religious views may also help check radicalization.

Keywords:

• radicalization
• threshold
• non-linearity
• blasphemy
• Pakistan

1. Introduction

Radicalization is a threat to the security and stability of nations. Radicalization is defined as the support/endorsement that one renders to the use of violence for accomplishing political or religious goals (Rink & Sharma, 2018). This definition is in line with the various definitions of the phenomenon provided by prominent scholars in the field as well as with the definitions advanced by various governmental agencies that emphasize radicalization as a process that entails justifying violence for attaining certain goals (Crossett & Spitaletta, 2010; McCauley & Moskalenko, 2008; Wilner & Dubouloz, 2010). Given the global rise in radicalization, it is imperative to ask which factors drive this process. The existing literature attribute radicalization to a diverse mix of psychological, economic, social, and political factors. Among these, economic factors receive considerably large attention from academics, policymakers, and journalists. The presumed link between economic factors and terrorism also plays a major role in international security and counter-terrorism policies. However, the empirical analysis of this relationship reveals a complicated picture.

Studies investigating the relationship between socioeconomic factors and radicalization can be grouped into three categories. The first strand detects no relationship between radicalization and socioeconomic factors such as poverty, economic disparity, inequality, discrimination, and economic marginalization (e.g., Bhui et al., 2014; Blair Graeme et al., 2013; Groppi, 2018; Rink & Sharma, 2018). The second strand finds that socioeconomic hardships such as economic marginalization, inequality, exclusion, and deprivation are instrumental in catalyzing radicalization (e.g., Bäck Emma et al., 2018; Ceder et al., 2011; Holla, 2020; Lindekilde et al., 2016; Macdougall et al., 2018). In contrast, the third strand finds that instead of the poor and marginalized, individuals from affluent backgrounds are more likely to radicalize and commit extremist violence (e.g., Bhui et al., 2014; Blair Graeme et al., 2013; Delia Deckard & Jacobson, 2015).

The existing literature tests a wide range of economic variables as plausible determinants of radicalization. Yet, there is no real consensus on the relationship between socioeconomic factors and radicalization. Most empirical studies on the phenomenon assume a linear relationship between radicalization and socioeconomic factors (Franc & Pavlović, 2021b). However, the linearity assumption may not hold up in all cases (Arin et al., 2021). For instance, Andreas et al. (2011a) studied the relationship between real GDP per capita and terrorism in 110 countries. They found that up to a certain level, an increase in per capita GDP results in more terrorism since higher income enhances the state’s repressive capacity, allowing only for clandestine (terrorist) activity instead of open rebellion. Afterward, more per capita income means less terrorism due to a rise in the opportunity cost of terrorism. Such “switch points” could also be salient in the relationship between individual-level socioeconomic factors and the individual decision to support or commit terrorist acts. However, most studies test this relationship through linear models using data on terrorist activity and aggregate socioeconomic indicators. Ignoring the potential switch points restrict studies to only one coefficient for the explanatory variable when there should be separate coefficients delimited by the range of the switch point (Arin et al., 2021).

This study proposes an alternative non-linear mechanism for the relationship between radicalization and the individual-level perceptions of socioeconomic conditions/prospects based on opportunity costs and the mental rewards of extreme acts. Socioeconomic hardships (or the perception of these) reduce the opportunity costs of extreme acts, making it attractive to gain mental rewards by supporting or participating in such acts (Andreas et al., 2011b). However, subscription to mental rewards of extreme acts may depend upon the degree of personal religiosity since religion bestows ideology and purpose. This is to say that a certain level or threshold of religiosity must be reached to perceive extreme acts as avenues of mental rewards during socioeconomic hardships. In other words, this study hypothesizes that socioeconomic hardships should drive radicalization only in sufficiently religious people.

To test the above hypothesis, this study uses the non-linear threshold regression method developed by Hansen (2000) that, splits the data, and searches for the existence of multiple regimes. Hansen’s (2000) method endogenously detects the presence of the possible switch points from the data when their existence is not known a priori. It further allows the coefficients to vary across regimes that lay below or above the switch points or threshold values. Several studies have used this method to investigate the existence of switch points in the relationship between different economic variables. Some examples include investigating the relationship between trade and economic growth using trade openness as a threshold variable (Papageorgiou, 2002), the relationship between economic growth and financial development using institutional quality as a threshold variable (Law et al., 2013), and the relationship between foreign direct investment and economic growth using financial market development as a threshold variable (Azman-Saini et al., 2010). Drawing on these, this study splits the data into different regimes or classes using religiosity as a threshold variable and estimates the effect of the varying degrees of religiosity on the relationship between socioeconomic factors and radicalization.¹

This study contributes to the literature in two ways. First, it uses novel survey data from the Khyber Pakhtunkhwa (K.P) province of Pakistan. K.P is one of the most marginalized and economically disadvantaged regions of Pakistan. It has also been the key recruitment ground for several Islamist militant groups such as the Afghan Taliban, Al-Qaeda, and the Pakistani Taliban. Moreover, it is currently experiencing a surge in various types of religious violence, such as suicide attacks, targeted killings of religious minorities, sectarian clashes, and blasphemy vigilantism. Given these factors, this study provides empirical evidence from a region that exhibits a unique combination of socioeconomic hardships, militant groups motivated by supreme values, and a high degree of religious violence. Second, it demonstrates the existence of a non-linear relationship between socioeconomic factors and radicalization using Hansen’s (2000) methodology. For instance, the study finds a statistically significant positive relationship between radicalization and the individual perceptions of economic prospects, relative deprivation, injustice, and inequality only above the religiosity threshold. This supports the hypothesis that socioeconomic hardships drive radicalization only in sufficiently religious people. This suggests that apart from security-centric approaches, radicalization can also be deterred by increasing the opportunity costs of political violence through socioeconomic improvements.

2. Socioeconomic hardships & radicalization in Khyber Pakhtunkhwa: a brief overview

Pakistan faces serious challenges posed by radicalization and religiously inspired extremism. This is evident from the scores of terrorism incidents faced over the past two decades. For instance, from 9/11 to date, Pakistan suffered 29,721 terrorism incidents, including 597 suicide attacks, and received 65,271 fatalities. However, these losses are disproportionately spread across Pakistan, with K.P., which houses 17% of the country’s population, receiving the biggest brunt (“Pakistan Bureau of Statistics.”, 2017). Out of the total terrorism incidents 13,265 (44.5%) occurred in K.P, coupled with 362 (60.6%) of all suicide attacks and 45,975 (70.4%) of all fatalities (South Asia Terrorism Portal, 2022).

The uneven impact of terrorism goes hand in hand with the relatively dismal state of socioeconomic development in K.P. For instance, in terms of gross regional product, K.P. ranks third among Pakistan’s four provinces, indicating a lower level of economic activity. In terms of provincial per capita income, K.P. ranks third, 9.3 % below the national average. Corollary to these economic facts, K.P. also exhibits a relatively poor state of human development. This is indicated by the low Human Development Index (H.D.I) score, which ranks K.P. third among the four provinces (Pasha, 2021). Likewise, K.P. further manifests the second-highest incidence of multidimensional poverty among the four provinces of Pakistan (Oxford Poverty and Human Development Initiative, 2021).

K.P. also has a long history of religious mobilization. People from this region recruited into Jihadist movements during various epochs. For instance, in 1826, Syed Ahmed Barelvi, an Islamic revivalist from India, came to N.W.F.P² (now K.P) and urged the local populace to join him in establishing an Islamic state and waging Jihad against the then neighboring Sikh empire. After securing the support and participation of the local Pashtun tribes, he launched his Jihadist campaign against the Sikh army in 1831 (Z. Khan & Ullah, 2018). In 1948, Pakistan and India fought their first war over the disputed territory of Kashmir. Instead of regular troops, Pakistan entered the war using Pashtun tribesmen from N.W.F.P. who joined the conflict against the “Hindu India” under Jihadist motivations (Yousaf, 2019). In 1979, the then-U.S.S.R. invaded Afghanistan. Under the aegis of the United States, Pakistan undertook the recruitment, indoctrination, and training of the anti-Soviet Mujahideen³ fighters predominantly in K.P (Ahmad, 2013). After 9/11, K.P became the recruitment ground, sanctuary, and launching pad of Al-Qaeda and the Afghan Taliban in their fight against the N.A.T.O forces in Afghanistan (Gunaratna & Nielsen, 2008). Currently, K.P is facing a surge in various types of religious violence, such as suicide attacks, targeted killings of religious minorities, sectarian clashes, and blasphemy vigilantism.

The above discussion indicates that the socioeconomic disparities are seemingly in tune with the religious mobilization in K.P. This line of reasoning is advocated by several studies that posit a link between socioeconomic hardships and radicalization in Pakistan (e.g., Ahmed et al., 2018; Yusuf, 2008). However, most studies propose such linkages primarily on the basis of macro/aggregate level data on socioeconomic indicators. Nevertheless, macro-level variables indicate the state of the overall social context surrounding the individuals. Moreover, numerous studies argue that the social context can greatly moderate the relationship between individual-level perceptions/variables and political beliefs (e.g., Federico & Malka, 2018; Jasko et al., 2020). This could also be salient in the case of K.P, which exhibits a confluence of poor socioeconomic conditions and a high degree of religious violence. Despite receiving considerable attention, there is a dearth of studies that test the linkages between individual-level socioeconomic conditions/perceptions and the individual decision to commit or support extreme acts. This study empirically tests these linkages using survey data from the K.P province of Pakistan.

3. Theoretical background

This study adopts the theoretical framework of Freytag et al. (2011), who explained the emergence of terrorism from the rational choice perspective. Terrorism is seen as a consequence of the socioeconomic conditions of the terrorists or their Umfeld. Terrorists and their supporters are assumed to be rational actors whose behavior is directed by the costs, benefits, and opportunity costs of extreme acts. Likewise, terrorism or its support is viewed as one of the many choices influenced by economic constraints. Of particular importance are the opportunity costs, which indicate the alternatives that one needs to sacrifice while committing or supporting extreme acts.

Under the opportunity costs framework, radicalization results from a trade-off between material and mental rewards. Material rewards, such as income, are obtained by refraining from extreme acts and participating in activities that produce material well-being. Conversely, mental rewards like social recognition, feelings of significance, power, and martyrdom are collected from supporting or committing extreme acts. Individuals opt for extreme acts to seek mental rewards as long as the resulting benefits exceed the (opportunity) costs. The socioeconomic conditions of individuals may significantly influence these cost-benefit calculations. For instance, economic hardships indicate low material rewards from non-violence. This renders the mental incentives associated with extreme acts attractive prospects for individuals.

In the case of religious (Islamist) radicalization, the opportunity cost considerations are likely to be influenced by the degree of individual religiosity since religion bestows ideology, purpose, and mental rewards/incentives. It is therefore intuitive to assume that a higher degree of religiosity would presumably translate into a greater propensity to political violence when faced with economic hardships. In other words, a certain level or threshold of religiosity must have to be reached to perceive extreme acts as avenues of mental rewards during socioeconomic downturns. This study tests this prediction using survey data of undergraduate students from the Khyber Pakhtunkhwa province of Pakistan. Specifically, it tests whether socioeconomic hardships drive radicalization only in sufficiently religious people. This is accomplished using a diverse range of variables drawn from the existing literature, such as economic prospects, relative deprivation, perceived inequality, and perceived oppression, etc. A systematic review of the relevant scholarly literature was conducted to identify various socioeconomic factors considered to be the pivotal drivers of radicalization by scholars in the field. Next, through consultation with local academics aware of the radicalization landscape in Pakistan, the most plausible of these variables were selected for empirical testing. These questions/items used to construct these variables measure aspects related to rational choice and collective social identity. The study also uses religiosity as a variable to account for the ideological component underlying radicalization. The study thus draws insights from three different perspectives which are considered to be complementary (James & Dawson, 2023). These variables are discussed in detail below. All variables measure the respondents’ perceptions regarding the situations depicted in the questions/construct used to construct these.

3.1. Economic prospects

Literature on the phenomenon frequently mentions that poor economic prospects can considerably influence the radicalization process. For instance, Lehmann and Tyson (2022) developed a model of the strategic interaction between the state and radical groups. Specifically, they theorized the effects of the mutual anticipation of each another’s choices on the consequent responses. They argued that the government’s policy to enhance economic growth improves the economic prospects of the citizens, which significantly deters the radicalization process. Freytag et al. (2011) studied the socioeconomic determinants of terrorism in 110 countries and found that economic development and improvements in future prospects can significantly reduce terrorism. Holla (2020) studied the relationship between the perception of economic marginalization and radicalization through a correlation research design using survey data of Somali Muslims in Kenya. The study found that the perception of marginalization is related to a rise in radicalization. However, unlike the current study, Holla (2020) did not study the conditional effect of religiosity on this relationship. Yusuf (2008) studied the process of youth radicalization in Pakistan and considered the lack of socioeconomic opportunities to be an important catalyst of this process. Ahmed et al. (2018) assessed the determinants of terrorism in Pakistan based on the opinions of university students. A considerable number of students were of the view that the lack of socioeconomic development is associated with susceptibility to terrorism. Drawing on these accounts, this study tests the relationship between the perception of poor economic prospects and radicalization in the Khyber Pakhtunkhwa province of Pakistan.

3.2. Political marginalization

Opportunities for political participation increase the likelihood of economic success, which reduces the need to resort to extreme acts for voicing dissent (Andreas et al., 2011b; Wahl, 2019). Conversely, political marginalization deters the chances of economic prosperity and makes political violence more likely (H. E. Hansen et al., 2020). Choi and Piazza (2016) tested the relationship between political marginalization and terrorism using data from 130 countries. They found that countries where certain ethnic groups are politically excluded, are more likely to suffer from domestic terrorism. Meierrieks et al. (2021) studied terrorism in 99 countries and found political exclusion associated with terrorist activity. Dalacoura (2006) studied Islamist terrorism in the Middle East and argued that political exclusion contributes to the adoption of terrorist methods. Zeb and Shahab Ahmed (2019) also found political exclusion as an important factor behind terrorism in Pakistan. Drawing on these accounts, this study tests the relationship between radicalization and political marginalization in K.P.

3.3. Relative deprivation

Another factor that may drive individual radicalization through socioeconomic differences is the perception of relative deprivation. Relative deprivation is a situation when one considers himself subjected to unfair treatment or disadvantage. This generates anger, resentment, and a desire for revenge against the depriver (Bergen Diana et al., 2015; Obaidi et al., 2019). Macdougall et al. (2018) analyzed survey data from the U.S. and Netherlands and found that relative deprivation predicts willingness to join violent groups. Obaidi et al. (2019) assessed the relationship between relative deprivation and violent extremism among Muslims in Western countries. They found group-based relative deprivation associated with the endorsement of extremism. Pearson (2016) also found socioeconomic deprivation as an important catalyst for the actions of Roshonara Choudhry—a British lone-wolf terrorist who attacked a parliament member in 2010. Drawing on these accounts, this study tests the relationship between radicalization and measures of individual and collective relative deprivations in K.P.

3.4. Injustice

Socioeconomic differences may catalyze the perception of injustice among individuals, which can provoke anger and violence as compensatory reactions (Al-Saggaf, 2016; Van Brunt et al., 2017). Doosje et al. (2013) studied the process of radicalization in Dutch Muslims and found perceived injustice to be an important predictor of the radical belief system. Pauwels and Heylen (2017) studied Belgian adolescents and young adults and found that perceived injustice wields a significant impact on right-wing extremism. Based on semi-structured interviews with right-wing, left-wing, and religious extremists in Belgium, Schils and Verhage (2017) also found perceived injustice as a starting point of radicalization. Drawing on these accounts, this study includes perceived injustice in the empirical analysis.

3.5. Exclusion

The mental rewards of extreme acts may be particularly attractive for individuals experiencing social or economic exclusion. Exclusion induces significance loss which inspires a quest for significance seeking. In such a situation, subscription to extremist ideologies and groups contributes to personal significance and the incremental ingress into radical schemas. Bäck Emma et al. (2018) tested this in Swedish university students and found that social exclusion, and a subsequent inclusion by radical groups, results in the adaption of the group’s attitudes. Lindekilde et al. (2016) studied Danish Islamist foreign fighters and found exclusion as an important factor in their radicalization. Moreover, Hansen et al. (2020) examined sub-national terrorist violence and found that regions with excluded ethnic groups exhibit a higher risk of terrorism. Given the prevailing perception of exclusion in the study setting, this variable is included in the empirical analysis.

3.6. Inequality

An individual’s perception of socioeconomic inequality may also act as an important factor in the opportunity cost calculations underlying the radicalization process. Franc and Pavlović (2021)a reviewed the relevant literature and argued that the perception of social inequality is related to radical attitudes. Ceder et al. (2011) studied global horizontal inequalities and found that unequal societies face violent conflict more often than relatively equal ones. Ahmed et al. (2018) assessed the domestic triggers of terrorism in Pakistan and found inequality as an important factor in pushing individuals towards militant groups. By studying the 13 Daesh-affiliated Bulgarian Imams, Panayotov (2019) also found that individuals from unequal regions may be highly susceptible to radicalization. Drawing on these accounts, this study includes perceived inequality in the empirical analysis.

3.7. Oppression

The perception of oppression may also stem from the lack or access to the opportunities of economic prosperity. Such a perception is likely to make the mental rewards of political violence particularly attractive. Milla et al. (2013) studied Bali bombers and found the oppression of Muslims as a key motivation for their bombings. Manuel and Trujillo (2014) studied radicalism intentions in Spanish Muslim and Christian school students. They found perceived oppression and radicalism correlated in Muslim students. In K.P, people widely perceive that they are being oppressed by the dominant ethnic groups. Given this, the current study includes perceived oppression in the empirical analysis.

3.8. Religiosity

Since the mental rewards of religiously motivated extremist acts stem primarily from religion, religiosity can therefore play an important role in the radicalization process. For instance, Coid et al. (2016) investigated the population distribution of extremist views in a sample of young men (18–34 years) from England, Scotland, and Wales. They found a significant relationship between religiosity and anti-British extremist views. Dawson (2018) conducted interviews with Western foreign fighters in Syria and Iraq and their friends and families. They found that religiosity plays a substantial role in terrorism, particularly in the case of Jihadism. Rink and Sharma (2018) studied the determinants of religious radicalization using survey data from Kenya. They found religiosity to be an important predictor of radicalization. Beller and Kröger (2018) studied the predictors of support for extremist violence among Muslims from Albania, Bosnia and Herzegovina, Egypt, Indonesia, Kazakhstan, Kosovo, Malaysia, and Russia. They found support for extremist violence strongly associated with social religious activities. Drawing on these accounts, this study includes religioisity in the empirical analysis.

4. Material design and methods

4.1. Participants of the study

It is generally believed that radicalized individuals in Pakistan come from poor and highly religious backgrounds. They are also believed to be either illiterate or educated in religious seminaries, called Madrassahs. This stereotypical narrative is reiterated by journalists, security officials, and policymakers alike (Delavande & Zafar, 2015). While it was relatable to earlier generations of extremists, it is evident from recent incidents that the radicalization landscape in Pakistan has considerably evolved and transformed (Dawn, 2017). For instance, in the year 2015, several Al-Qaeda-affiliated militants attacked and killed 43 members of the Ismaeli Shia community traveling in a bus. It was later found out that these militants are highly educated individuals possessing university degrees (Zahid, 2015). Likewise, in 2017, a medical student from Pakistan, Naureen Leghari, left home and visited Syria for joining the militant Islamia State (IS) group. She received militant training there and came back to Pakistan to conduct a suicide attack on a Church on Easter (Firdous, 2017). In another incident in 2017, a journalism student, Mashal Khan, was alleged by fellow students of posting blasphemous content against Islam on Facebook. Over these allegations, he was later lynched inside the university campus by other students (Singay, 2020). In 2018, Faheem Ashraf, a college student killed his college principal who had reprimanded him for being absent from the college for several days to attend the rallies supporting the strict blasphemy laws of Pakistan. The student later admitted the murder without remorse claiming that his action was aimed at safeguarding the honor of the Prophet Muhammad (Yilmaz & Shakil, 2022). Similarly, in 2019, another student, Khateeb Hussain, killed his teacher on accusations of blaspheming against Islam and for supporting un-Islamic culture (Imran, 2019). From these incidents, it could be deduced that instead of the illiterate and impoverished, the highly educated individuals are increasingly becoming radicalized, and educational institutes are seemingly contributing to this process (Iqbal & Mehmood, 2021). Hence, university students appear as a relevant sample to study the drivers of radicalization. Therefore, primary data was collected from undergraduate students of 19 universities in the Khyber Pakhtunkhwa province of Pakistan through a survey conducted between December 2019 and March 2020. The study region can be seen in the map given in Figure 1.

Figure 1. Map of the study region.

It was initially planned that a pen-and-paper survey would be conducted. However, as a result of the COVID-19 pandemic and the worsening security situation amid the U.S-Taliban peace deal, it was not possible to travel to the study region. Therefore, Google Forms was used to administer the survey remotely. The questionnaire was pasted into Google Forms and its web link was generated. Afterward, through personal contacts, the university teachers in the target region were approached and requested for the implementation of the survey. Due to the pandemic, physical classes were on a halt. Therefore, the faculty members reached the participants via phone and obtained verbal consent after explaining that the survey is conducted solely for academic research and that their responses will remain anonymous since no personal information like name and address are collected. The questionnaire link was then shared with participants primarily through WhatsApp which they filled remotely. This procedure ensured the safety of the interlocuters and also helped mitigate the response bias since nobody was watching the participants filling out the online questionnaires. To ensure data security, the participants were not shown their or other partcipants’ responses after they submitted the survey form upon completion.

In total, 510 respondents took part in the survey. All participants identified themselves as Muslims. Nearly half of them are female (52%). Only a few of them are married (5.3%). In terms of age, most participants (92%) are between 18 and 23 years old. Complete descriptive statistics are given in Appendix 2.

4.2. Radicalization measure

Most studies on radicalization in Pakistan usually assess the phenomenon from a historical perspective and are mostly of descriptive nature. Hence, they do not contain a scale elaborate enough to empirically measure radicalization. Nevertheless, three studies are an exception to this general pattern.

The first of these studies is undertaken by Fair et al. (2012), who analyzed whether religiosity predicts support for political violence in Pakistan. The second study is undertaken by Blair Graeme et al. (2013), who assessed whether poverty is associated with militant support in Pakistan. These studies solicited the participants’ views regarding support for militant Islamist organizations to construct their dependent variables. The third study is undertaken by Bélanger et al. (2019), who assessed the predictors of ideologically-based violence in four countries, namely Pakistan, Canada, the U.S, and Spain. They constructed their dependent variable using a set of six questions/items for all the aforementioned countries. While these studies offer valuable information about the phenomenon, their radicalization scales/measures, nevertheless, miss certain important issues that are of crucial importance in understanding radicalization in contemporary Pakistan. Some of these issues are persecution/vigilantism over blasphemy allegations, the widespread resentment against members of the Ahmadiyya community, military operations inside Pakistan, and the U.S occupation of Afghanistan—all potentially serving as important catalysts of radicalization in Pakistan. Moreover, unlike the current paper, these three studies also lack a simultaneous focus on socioeconomic variables and religiosity. For instance, Fair et al. (2012) studied the relationship between religion and political violence but did not include economic variables in the empirical analysis. Likewise, Blair Graeme et al. (2013) assessed the relationship between poverty and militant politics but did not study the role of religiosity. On the other hand, Bélanger et al. (2019) focused entirely on psychological variables without including the economic and religious dimensions in their empirical analysis.

To build a comprehensive scale for measuring radicalization, nine items/questions are drawn from the questionnaire of Rink and Sharma (2018) who studied the drivers of religious radicalization in the Kenyan case. These nine items measure support for militant groups and violent behavioral intentions. Four main reasons inspired the choice of using Rink and Sharma’s (2018) survey instrument as a baseline. First, Rink and Sharma (2018) chose their survey participants from the Eastleigh district in Kenya. Eastleigh houses Somali immigrants and recruiters of the Somalia-based terrorist group, Al-Shabaab, in large numbers. Such parallels also exist in the context of K.P which too consists of a large number of Afghan refugees and recruiters of the Pakistani and Afghan Taliban. Second, poor socioeconomic conditions are viewed as one of the key drivers of radicalization in Eastleigh (Chepkong’a, 2020). Likewise, several scholars consider the dismal state of socioeconomic conditions in KP as one of the important catalysts of radicalization and violent extremism (e.g. Khan & Ahmed, 2017a; Naz et al., 2013; Wahab & Hussain, 2021). Third, Rink and Sharma (2018) drew on the inter-religious tensions existing between the Christians of Muslims of Kenya to assess the predictors of religious radicalization. Such tensions among the followers of different Islamic sects are also considered one of the main drivers of radicalization in Pakistan. Finally, Rink and Sharma (2018) operationalized and tested their radicalization scale in Eastleigh, Kenya, which exhibits religious violence as a recurring feature. Religious violence is an increasing phenomenon in K.P as well which currently faces a surge in suicide bombings, target killings, and persecution over allegations of blasphemy. Given these reasons, the questionnaire of Rink and Sharma (2018) appeared as an optimal choice for the current paper.

Considering the specific context of Pakistan, six additional items, that measure support for blasphemy vigilantism, are constructed and added to build the radicalization index. The radicalization scale of this study, which consists of 15 items, measures support for militant groups, violent behavioral intentions, and support for persecution/vigilantism under the allegations of blaspheming against Islam.

These 15 items/questions are combined and standardized⁴ to create the dependent variable of this study, i.e. the radicalization index (Cronbach’s alpha: .78). The survey questions are given in Appendix 1. A table containing full descriptive statistics regarding the responses to the survey questions is given in Appendix 2.

5. Method

This study hypothesizes that socioeconomic hardships increase the likelihood of radicalization for obtaining mental rewards only in sufficiently religious people. In other words, the relationship between socioeconomic factors and religious radicalization may be non-linear, and only after a certain point will religiosity make the mental rewards of political violence attractive while facing socioeconomic hardships. To test this hypothesis, this study uses the non-linear threshold regression approach developed by Hansen (2000). Using religiosity as a threshold variable, this technique splits the sample into two regimes or classes, which allows testing the relationship between socioeconomic factors and radicalization at different degrees of religious adherence. The empirical model of this study in the linear form takes the following shape:

(1) $Radicalizatio{n}_i={\propto }_0+{\propto }_{1}EcoVa{r}_i+{\propto }_2{X}_i+{\epsilon }_i$(1)

where $Radicalizatio{n}_i$ is the dependent variable for individual i, $EcoVa{r}_i$ represents the socioeconomic variables and $X_i$ is a vector of controls (gender, age, marital status, family size, and family income).

To test the non-linearity assumption discussed above, the threshold regression model is written as follows:

(2) $Radicalizatio{n}_i=\left\{\begin{array}{c}{\beta }_0^1+{\beta }_1^{1}EcoVa{r}_i+{\beta }_2^1{X}_i+e_i,Religiosity\le \gamma \\ {\beta }_0^2+{\beta }_1^{2}EcoVa{r}_i+{\beta }_2^2{X}_i+e_i,Religiosity>\gamma \end{array}\right.$(2)

where Religiosity is the threshold variable used to split the sample into two regimes, and $\gamma$ is the unknown threshold parameter. First, the null hypothesis of linearity (H0: β1 = β2) is tested against the threshold model (equation 2) using the fixed bootstrap procedure.

6. Results

Results of the threshold regression are given in Table 1. Since religiosity is co-linear with several explanatory variables, it is used only as a threshold variable. The results of the pairwise correlation test are reported in Appendix 3.

Table 1. Threshold regression

	Outcome: Radicalization
	Religiosity
Variables	< −.330	> −.330
Economic Prospects	−0.106	0.136***
	(0.104)	(0.036)
Political Marginalization	0.012	0.027
	(0.095)	(0.030)
Individual Relative Deprivation	0.092	0.165***
	(0.119)	(0.042)
Collective Relative Deprivation	−0.281*	0.057
	(0.115)	(0.038)
Injustice	0.107	0.098***
	(0.084)	(0.038)
Exclusion	0.114	0.021
	(0.095)	(0.028)
Inequality	0.033	0.096*
	(0.129)	(0.047)
Oppression	0.084	−0.084*
	(0.096)	(0.037)
Controls
Gender (0 = Female, 1 = Male)	0.190	0.216***
	(0.140)	(0.043)
Age	0.110	0.017
	(0.070)	(0.024)
Marital Status	−0.284	0.209*
	(0.245)	(0.094)
Family Size	0.204*	0.004
	(0.101)	(0.025
Family Income	−0.104	−0.050***
	(0.069)	(0.018)
Bootstrap (P-value)
R-sq	0.319	0.234
Het (P-value)
No. of Observations	64	446

Note: Standard errors are reported in parenthesis (White Corrected).

***indicates significance at 1%.

**indicates significance at 5%.

*indicates significance at 10%.

The statistically significant bootstrap p-value in Table 1 signifies the rejection of the linearity/no-threshold assumption. This p-value corresponds to 5000 replications and a 15% trimming percentage. The existence of the threshold effect is also confirmed in Figure 2, where the F Sequence exceeds the critical value, thereby indicating the rejection of the linearity assumption.

Figure 2. Threshold test.

For comparison, an ordinary least squares (OLS) regression model was also estimated, which showed that radicalization has a statistically significant positive relationship with the perception of poor economic prospects, individual relative deprivation, perceived injustice, perceived inequality, gender, and marital status. The results of the OLS model are given in Appendix 4. However, Hansen’s (2000) method provides more nuanced conclusions, as discussed in the following sections.

6.1. Economic prospects

The respondents’ perception of poor economic prospects is measured using three survey questions/items. The first of these items is drawn from Rink and Sharma (2018), while the rest are designed specifically for this study. Responses to these questions were obtained on a five-point Likert scale which ranged from 1, strongly disagree, to 5, strongly agree. All three items are combined and standardized to create the index of poor economic prospects (Cronbach’s alpha: .48). The survey questions are given in Appendix 1.

Table 1 indicates that the perception of poor economic prospects has a statistically significant positive relationship with radicalization only above the religiosity threshold. This supports the hypothesis of this study that poor economic prospects fosters radicalization only in highly religious individuals. This loosely relates to Freytag et al. (2011), who studied the determinants of terrorism in 110 countries and found that depressed socioeconomic conditions reduce the opportunity costs of terrorism and are thus instrumental in catalyzing terrorism.

6.2. Political marginalization

Political marginalization is measured using three survey questions adapted from Rink and Sharma (2018). Responses to both items were obtained on a five-point Likert scale which ranged from 1, strongly disagree, to 5, strongly agree. These two items are combined and standardized to create the political marginalization index (Cronbach’s alpha: .18). The survey questions are given in Appendix 1.

Table 1 indicates that political marginalization lacks a statistically significant relationship with radicalization below and above the religiosity threshold. This finding is in line with Rink and Sharma (2018), who found a null relationship between political marginalization and radicalization in Kenya. It contests Rathore and Basit (2010), Basit (2015), and Naseer et al. (2019), who argued that political marginalization is a crucial driver of radicalization in Pakistan.

6.3. Relative deprivation

Individual relative deprivation is measured using six survey questions adapted from Doosje et al. (2013). Responses to these questions were obtained on a five-point Likert scale which ranged from 1, strongly disagree, to 5, strongly agree. All six items are combined and standardized to create the individual relative deprivation index (Cronbach’s alpha: .85). Likewise, collective relative deprivation is also measured using six questions adapted from Doosje et al. (2013). Responses to these questions were obtained on a five-point Likert scale which ranged from 1, strongly disagree, to 5, strongly agree. All six items are combined and standardized to create the collective relative deprivation index (Cronbach’s alpha: .87). The survey questions are given in Appendix 1.

Table 1 shows that individual relative deprivation has a statistically significant positive relationship with radicalization only above the religiosity threshold, indicating that relative deprivation drive radicalization only in highly religious individuals. This is in line with the relative deprivation hypothesis of Gurr (2015), which considers relative deprivation to be an important impetus for political violence. It also supports the propositions of Khan and Kiran (2012) and Khan and Ahmed (2017b), who considered deprivation a major cause of radicalization in Pakistan.

On the other hand, a statistically significant negative relationship is detected between collective relative deprivation and radicalization below the religiosity threshold. From this, it could be deduced that for less religious individuals, collective relative deprivation does not act as a catalyst for radicalization.

6.4. Injustice

The perception of injustice is measured using seven survey questions adapted from Doosje et al. (2013). Responses to all these questions were obtained on a five-point Likert scale which ranged from 1, strongly disagree, to 5, strongly agree. All seven items are combined and standardized to create the perceived injustice index (Cronbach’s alpha: .80). The survey questions are given in Appendix 1.

Table 1 indicates that perceived injustice has a statistically significant positive relationship only above the religiosity threshold, indicating that injustice catalyzes radicalization only in sufficiently religious people. This relates to Doosje et al. (2013) and Pauwels and Heylen (2017), who found perceived injustice significantly related to radicalization in Dutch and Belgian samples, respectively. It also supports Hussain et al. (2014), Khan (2015), and Tanoli (2018), who considered injustice as an important cause of radicalization in Pakistan.

6.5. Exclusion

The respondents’ perception of exclusion is measured using six survey questions adapted from Rich et al. (2013). Responses to all these questions were obtained on a five-point Likert scale which ranged from 1, strongly disagree, to 5, strongly agree. All six items are combined and standardized to create the exclusion index (Cronbach’s alpha: .87). The survey questions are given in Appendix 1.

Table 1 indicates that exclusion lacks a statistically significant relationship with radicalization below and above the religiosity threshold. Therefore, this study departs from Renström et al. (2020) and Bäck Emma et al. (2018), who found exclusion and radicalization empirically related.

6.6. Inequality

The respondents’ perception of socioeconomic inequality is measured using five survey questions specifically designed for this study. Responses to all these questions were obtained on a five-point Likert scale which ranged from 1, strongly disagree, to 5, strongly agree. All five items are combined and standardized to create the inequality index (Cronbach’s alpha: .48). The survey questions are given in Appendix 1.

Table 1 indicates that perceived inequality has a statistically significant positive relationship with only above the religiosity threshold. This suggests that inequality drives radicalization only in highly religious individuals. This finding supports the propositions of Azam and Aftab (2009) and MALIK (2009), who in their theoretical studies, considered inequality to be an important driver of militancy in Pakistan.

6.7. Oppression

Perceived oppression is measured using 10 survey questions adapted from Lobato (2017). Responses to all these questions were obtained on a five-point Likert scale which ranged from 1, strongly disagree, to 5, strongly agree. All 10 items are combined and standardized to create the perceived oppression index (Cronbach’s alpha: .92). The survey questions are given in Appendix 1.

Table 1 indicates that perceived oppression has a statistically significant negative relationship with radicalization only above the religiosity threshold. This means that oppression reduces the likelihood of radicalization in highly religious individuals. This is in line with the widely held belief among Muslims that considers oppression a test of faith from Allah, which entails great rewards for steadfastness. The Quran mentions that Allah will reward the oppressed in the hereafter if they maintain righteousness while suffering and pardon the oppressor instead of retribution (Vasegh, 2009). Likewise, the Hadiths⁵ also promise significant rewards in the hereafter for facing oppression, such as loading the sins of the oppressed onto the oppressor (al-Bukhari, 846). This suggests that highly religious individuals may find greater meaning in oppression.

6.8. Sociodemographic controls

Among the sociodemographic controls, gender has a statistically significant positive coefficient only above the religiosity threshold in Table 1. This means that highly religious men are more prone to radical world views. Conversely, age lacks a statistically significant relationship with radicalization below and above the religiosity threshold. All the participants of the study are university-going undergrad students. Studying at the university allows individuals to form pro-social bonds as well as improve future prospects. These reasons may have rendered the effect of younger age less effective on the prospects of becoming radicalized. Moreover, marital status has a statistically significant positive relationship with radicalization only above the religiosity threshold. This implies that for highly religious couples, the marital relationship act as a mutually reinforcing echo chamber for transmitting concerns, ideology, and dissent. Marriage allows couples to push each other to think and act more radically than each would do on their own (Alexander, 2015). Family size has a statistically significant positive relationship with radicalization only below the religiosity threshold. This relates to Rodermond and Weerman (2021), who studied the impact of family characteristics on susceptibility to terrorism using data of 226 individuals suspected of terrorist intent in the Netherlands. They found that terrorist suspects, on average, come from larger families than individuals from the rest of the population. The current study finds that susceptibility to radicalization prevails only in less religious larger families. It is intuitive to assume that highly religious families are more tightly knitted since they are generally concerned about transmitting their ideology and groupthink, ensuring control and integration. Conversely, families with large size and lower religiosity indicate lower parental control and commitment, thereby increasing the risk of radicalization. Finally, income has a statistically significant negative relationship with radicalization only above the religiosity threshold. This suggests that economic improvements decrease the likelihood of radicalization even in highly religious individuals.

6.9. Robustness check

As a robustness check, Model 2 is re-estimated by adding religiosity as both threshold and explanatory variable. Religiosity is measured using the two-item scale adapted from Rink and Sharma (2018). These two items are combined and standardized to create the religiosity index (Cronbach’s alpha: .03). The survey questions are given in Appendix 1.

The existence of the threshold effect is also confirmed in this specification, as shown in Appendix 5. The results are given in Appendix 6. Compared to the previous results (i.e., Table 1), this specification brings no new insights. The only changes that occur are in the linear regression. For instance, in the linear regression, perceived inequality now has a statistically significant positive relationship with radicalization. Conversely, family income is negatively related to radicalization. Moreover, marital status now lacks a statistically significant relationship with radicalization. However, linear regression is not the main focus of this study. On the other hand, no major change is detected in the threshold regression, and results remain nearly the same in terms of signs and statistical significance. The only exception is religiosity, which as an explanatory variable, has a statistically significant negative relationship with radicalization only above the threshold value. This suggests that higher religiosity tends to instill greater tolerance among individuals. This is in line with Clingingsmith et al. (2009), who observed greater tolerant and peaceful attitudes in Pakistani Muslims after they performed Hajj,⁶ an indicator of serious devotion and adherence to religion.

6.9.1. Squared terms

As a robustness check for non-linearity, squared terms are added in the linear regression for all the variables with thresholds. Squared terms allow for checking the existence of switch points in quadratic relationships. The table in Appendix 7 confirms the presence of such switch points, which are called inflection points. These are also shown graphically in Appendix 8. However, since squared terms capture only one type of switch point (i.e., inflection points), therefore Hansen (2000) is preferred since it can detect more than one type of unknown switch points (Arin et al., 2021).

7. Discussion and conclusion

This study tests the relationship between socioeconomic factors and radicalization using survey data from the Khyber Pakhtunkhwa province of Pakistan. The study finds a statistically significant positive relationship between radicalization and the indicators of the individual level socioeconomic conditions, such as the perceptions of economic prospects, individual relative deprivation, injustice, and inequality above the religiosity threshold. These findings indicate that socioeconomic hardships (or the perception of these) drive radicalization only in sufficiently religious people. Moreover, the study detects a statistically significant negative relationship between family income and radicalization. This implies that apart from security-centric approaches, radicalization can also be deterred by increasing the opportunity costs of political violence through socioeconomic improvements. Moreover, promoting balanced and diverse religious views may also help check the radicalization process. Apart from these main findings, two takeaways from the results merit discussion.

First, as discussed above, the study detects a positive relationship between radicalization and the perceptions of poor economic prospects, relative deprivation, injustice, and inequality only above the religiosity threshold. The participants of this study come from a region that exhibits a dismal state of aggregate socioeconomic conditions and a high degree of religious violence, as outlined in Section 2. Hence, it can be argued that a mutually constitutive relationship may exist between regional contextual factors and individual perceptions. This relates to Federico and Malka (2018) and Jasko et al. (2020), who argued that the relationship between individual-level variables and political beliefs is moderated by the surrounding social context. This suggests that policies that foster regional development, diversity, and inclusion may also reduce the risk of radicalization. The significance of socioeconomic factors above the religiosity threshold also indicates that their relationship with radicalization is far from simple, and therefore needs further investigation through future research.

Second, the study finds that the explanatory variables are associated with the outcome variable in a threshold-dependent way. The threshold parameter (i.e., religiosity) acts as a change point, enabling the modeling of the non-linearity in the relationship between radicalization and independent variables. From this, it can be deduced that instead of religiosity per se, a certain degree of religious adherence may be required to subscribe to the mental rewards of religiously motivated political violence while facing economic hardships.

For future research, two possible directions are suggested. First, this study tests the relationship between radicalization and perceived socioeconomic conditions. Hence, it furnishes evidence on the link between socioeconomic factors and support for political violence from the micro perspective. However, the aggregate/regional socioeconomic conditions may greatly influence these individual-level perceptions. Since the current study is based on individual-level/micro data from a single region, it cannot capture the effect of aggregate socioeconomic factors on individual (radical) perceptions. To better understand the impact of the context on the individuals, future studies may combine the micro data of individual-level support for violent groups/acts with indicators of aggregate socioeconomic conditions. Second, to test the generalizability of these results, future studies may empirically test the framework of this paper in other regions with considerable degrees of religious radicalization.

Submission

This manuscript is submitted to Cogent Social Sciences.

Acknowledgments

The author would like to thank Professor Dr. Marcel Thum, Faculty of Business and Economics, Technische Universität Dresden (TUD), for reading several drafts of this paper and providing useful comments, suggestions, advice, and encouragement. The author is also grateful to Professor Dr. Kerim Peren Arin, Zayed University Abu Dhabi, for providing valuable advice and comments on the methodological aspects of this paper.

Disclosure statement

No potential conflict of interest was reported by the author.

Supplemental material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/23311886.2023.2286042

Additional information

Funding

The author is grateful for the financial support (doctoral scholarship) provided by the Dresden Leibniz Graduate School (DLGS) and the Leibniz Institute of Ecological Urban and Regional Development (IOER).

Notes on contributors

Fahim Nawaz

Fahim Nawaz was a doctoral candidate and scholarship holder at the Dresden Leibniz Graduate School (DLGS), which is a joint interdisciplinary facility of the Leibniz Institute of Ecological Urban and Regional Development (IOER), and the Technische Universität Dresden (TUD), Germany. He currently works as a Lecturer at the Department of Economics, University of Peshawar, Pakistan

Notes

1. Before applying the method developed by Hansen (2000), squared terms were added in the linear regression which confirmed the existence of inflection points. However, squared terms capture only one type of switch point (i.e., inflection points), therefore Hansen (2000) is preferred since it can detect more than one type of unknown switch points. The results of the regression with squared terms in given in Appendix 7.

2. North-West Frontir Province.

3. A term used by Muslims in religious context for individuals engaged in a holy struggle for serving Islam, such as Jihad.

4. The procedure employed here involves adding the scores of all items and then dividing them by the number of items inovled in creating the index.

5. Hadiths refer to the sayings and traditions of the Islamic Prophet Muhammad. Hadiths are considered an important source of Shariah (Islamic law), ranking second only to the Quran.

6. Islamic pilgrimage to Mecca.

Appendix 1:

Questionnaire/Survey Instrument

Dear Participant,

This survey will be used in a university research project purely for academic purposes. Your participation is completely voluntary, and you will not be asked for personal information like your name, address, etc. Please note that your responses will not be shared with anyone else. We will be thankful for your participation.

Sociodemographic Controls

Table

Variable	Statement
Age 1 2 3 4 5	What is your age in years? 18–19 20–21 22–23 24–25 26 and Above
Gender 0 1	What is your gender? Female Male
Marital Status 0 1	What is your marital status? Unmarried Married
Family Size 1 2 3 4 5	How many people live in your house as one family? 1–5 6–10 11–15 16–20 21 and above
Family Income 1 2 3 4 5	In terms of total income, at which level can your family be placed? Rs. 5,000–10,000 Rs. 11000–20,000 Rs. 21000–30,000 Rs. 31000–40,000 Above Rs. 40000

Radicalization

Instruction: For each of the statements below, please indicate how much you agree or disagree by choosing the appropriate option using (✓).

1=Strongly Disagree 2=Disagree 3=Undecided 4=Agree 5=Strongly Agree

Table

S. No	Statement	Strongly Disagree	Disagree	Undecided	Agree	Strongly Agree
1	Imagine that a Pakistani Imam says that all Pakistani Muslims should support the Afghan Taliban in their fight against the American army in Afghanistan. How much would you agree to the Imam’s demand?
2	Imagine that a Muslim person is killed by an American drone strike in Afghanistan or the tribal areas (FATA) of Pakistan. One of his relatives decides to join the Afghan Taliban to take revenge. How much would you agree that he made the right decision?
3	In 2007, Pakistan conducted a military operation in Laal Masjid (Red Mosque), during which the head Imam, Maulana Abdur Rashid, died. Afterward, Osama Bin Laden described him as a hero of Islam and declared war against the Pakistan army. To what extent do you agree that the declaration of Osama Bin Laden is acceptable?
4	The Quran has ordered that a Muslim should not kill himself. This is what the Quran says (Quran 2:29): “And do not kill yourselves. Indeed, Allah is to you ever merciful.” Have you heard about this?	Yes (11)	No (0)
5	Imagine that a Muslim man has problems with his family. He decides to kill himself. To what extent do you agree that this is acceptable?
6	Imagine that a Muslim man is beaten by the Pakistani police. He decides to wear a suicide jacket and attack the police. To what extent do you agree that this is acceptable?
7	Imagine that a Muslim man is beaten by the Pakistani police. He decides to place a bomb at the police station. To what extent do you agree that this is acceptable?
8	The Hadith forbids a Muslim from killing women and children. This is what the Hadith says (Kitab al-Jihad): “Do not kill women, children, the old or the infirm (weak).” Have you heard about this?	Yes (1)	No (0)
9	Imagine that a Muslim man has a severe problem with his daughter. He decides to kill her. To what extent do you agree this is acceptable?
10	Imagine that after an Imam was killed, a Muslim man attacks a bus filled with men, women, and children. To what extent do you agree that this situation is acceptable?
11	Imagine that after an Iman was killed, a Muslim man attacks a bus filled with adult men. To what extent do you agree that this is acceptable?
12	In 2011, the Governor of Punjab (Salman Taseer) was killed by his police bodyguard, Mumtaz Qadri. The killer claimed that it was his religious duty to kill the Governor because he spoke against the blasphemy law. Do you agree that the claims of Mumtaz Qadri are acceptable?
13	In 2017, Mashal Khan, a student at Abdul Wali Khan University Mardan, was blamed for posting blasphemous content against Islam on Facebook. Rather than handing him over to the police, he was killed by a crowd of people inside his university over these allegations. Do you agree that this situation is acceptable?
14	In the blasphemy case against Aasia Bibi, an Imam in Peshawar offered a reward of five lakh rupees to anyone who kills her. To what extent do you agree that this is acceptable?
15	Imagine that a Pakistani Imam says that one’s faith/belief in Islam allows him to use violence against the Ahmadis. How much would you agree with the Imam’s statement?

Economic Prospects

Instruction:

The statements given below relate to your perception regarding yourself in various situations. You are requested to tell us how much you agree or disagree by choosing the appropriate option using (✓).

1=Strongly Disagree2=Disagree3=Undecided4=Agree5=Strongly Agree

Table

S. No	Statement	Strongly Disagree	Disagree	Undecided	Agree	Strongly Agree
1	I believe I have little chance of becoming wealthy in Pakistan.
2	My family is not fairly treated in the existing economic situation.
3	My family members have to struggle frequently to find good employment due to our ethnicity/caste (e.g., Pashtun, Punjabi, Sindhi, Baluch, etc.).

Political Marginalization

1=Strongly Disagree 2=Disagree 3=Undecided 4=Agree 5=Strongly Agree

Table

S. No	Statement	Strongly Disagree	Disagree	Undecided	Agree	Strongly Agree
1	I plan to vote in the next election.
2	The Pakistani government does not represent my interests.

Individual Relative Deprivation

1=Strongly Disagree 2=Disagree 3=Undecided 4=Agree 5=Strongly Agree

Table

S. No	Statement	Strongly Disagree	Disagree	Undecided	Agree	Strongly Agree
1	I don´t think I get as many chances as others in Pakistan.
2	It makes me angry when I think of how I am treated compared to others in Pakistan.
3	I think I am less well off than others in Pakistan.
4	I have the feeling that I am being discriminated.
5	If I compare myself with others in Pakistan, I have the feeling that I am being treated unfairly.
6	I think I can buy less than others in Pakistan.

Collective Relative Deprivation

Table

S. No	Statement	Strongly Disagree	Disagree	Undecided	Agree	Strongly Agree
1	I think my ethnic group (e.g., Pashtuns) is less well off than other groups in Pakistan.
2	It makes me angry when I think of how my ethnic group is treated compared to other groups in Pakistan.
3	I believe people from my ethnic group are discriminated.
4	If I compare my ethnic group with other groups in Pakistan, I have the feeling that we are being treated unfairly.
5	I think that people from my ethnic group don’t get as many opportunities as others in Pakistan.
6	I think people from my ethnic group cannot buy more than others in Pakistan.

Injustice

Table

S. No	Statement	Strongly Disagree	Disagree	Undecided	Agree	Strongly Agree
1	I think I am usually treated fairly.
2	I think that in my life, I am treated the same as everyone else.
3	I think people generally listen to me well.
4	I think that when people make decisions about me, they are always well prepared.
5	I think I am usually treated with respect.
6	If I disagree with something, I get enough opportunity to have my say.
7	I think that when I have complaints about something, my opinion is generally listened to carefully.

Exclusion

Table

S. No	Statement	Strongly Disagree	Disagree	Undecided	Agree	Strongly Agree
1	In general, others spend time with me for relaxing and enjoying.
2	In general, others invite me to their club, organization, or association.
3	In general, others include me in their plans for the holidays/vacations.
4	In general, others make an effort to get my attention.
5	In general, others invite me to go out to eat with them.
6	In general, others invite me to join them for weekend activities, hobbies, or events.

Inequality

Table

S. No	Statement	Strongly Disagree	Disagree	Undecided	Agree	Strongly Agree
1	I am satisfied with the economic situation of my family.
2	In comparison with the community where I live, the economic situation of my family is worse.
3	In comparison with the community where I live, the access of my family to the government services is better.
4	I am concerned about the inequality that exists around me and my family.
5	There is a serious need for re-distribution of resources within my society.

Oppression

Table

S. No	Statement	Strongly Disagree	Disagree	Undecided	Agree	Strongly Agree
1	My ethnic group (e.g., Pashtuns) is considered inferior.
2	My ethnic group is often looked down upon.
3	Some people treat us unjustly.
4	The dominant ethnic groups want to humiliate us.
5	We are denied our equal rights.
6	We feel humiliated.
7	The dominant ethnic group keeps us from living the way we want.
8	My ethnic group gets controlled too much.
9	The dominant ethnic groups want to physically hurt us.
10	My ethnic group is often verbally abused.

Religiosity

1. If you had to choose one label to describe yourself from the following options, which would it be?

1. Male

2. Female

3. Muslim

4. Caste/Tribe

5. Pakistani

2. How often do you visit your place of worship to offer prayers?

1. Daily

2. Weekly

3. Monthly

4. Sometimes

5. Rarely

Appendix 2:

Descirptive Statistics

Table

Variables	Frequency	Percent	Cumulative	Mean	Std. Dev.	Min	Max
AGE (YEARS)				2.202	.898	1	5
18–19	117	22.94	22.94
20–21	216	42.35	65.29
22–23	138	27.06	92.35
24–25	35	6.86	99.22
26 and above	4	0.78	100.00
GENDER				1.522	.5	1	2
Male	244	47.48	47.48
Female	266	52.16	100.00
MARITAL STATUS				.053	.224	0	1
Unmarried	483	94.71	94.71
Married	27	5.29	100.00
FAMILY SIZE				1.992	.848	1	5
1–5	126	24.71	24.71
6–10	308	60.39	85.10
11–15	43	8.43	93.53
16–20	20	3.92	97.45
21 and above	13	2.55	100.00
FAMILY INCOME (MONTHLY)				3.924	1.214	1	5
Rs. 0–10,000	20	3.92	3.92
Rs. 11000–20,000	55	10.78	14.71
Rs. 21000–30,000	111	21.76	36.47
Rs. 31000–40,000	82	16.08	52.55
Above Rs. 40000	242	47.45	100.00
RADICALIZATION				0	.496	−1.027	2.234
Support Taliban
Strongly Disagree	61	11.96	11.96
Disagree	99	19.41	31.37
Undecided	152	29.80	61.18
Agree	123	24.12	85.29
Strongly Agree	75	14.71	100.00
Join Taliban
Strongly Disagree	65	12.75	12.75
Disagree	151	29.61	42.35
Undecided	123	24.12	66.47
Agree	112	21.96	88.43
Strongly Agree	59	11.57	100.00
Support Bin Laden’s Declaration
Strongly Disagree	76	14.90	14.90
Disagree	123	24.12	39.02
Undecided	193	37.84	76.86
Agree	83	16.27	93.14
Strongly Agree	35	6.86	100.00
Quranic verse
No	40	7.84	7.84
Yes	470	92.16	100.00
Kill Self
Strongly Disagree	364	71.37	71.37
Disagree	115	22.55	93.92
Undecided	16	3.14	97.06
Agree	11	2.16	99.22
Strongly Agree	4	0.78	100.00
Attack Police
Strongly Disagree	283	55.49	55.49
Disagree	157	30.78	86.27
Undecided	51	10.00	96.27
Agree	11	2.16	98.43
Strongly Agree	8	1.57	100.00
Bomb Police Station
Strongly Disagree	313	61.37	61.37
Disagree	150	29.41	90.78
Undecided	32	6.27	97.06
Agree	8	1.57	98.63
Strongly Agree	7	1.37	100.00
Hadith
No	56	10.98	10.98
Yes	454	89.02	100.00
Kill Daughter
Strongly Disagree	350	68.63	68.63
Disagree	109	21.37	90.00
Undecided	37	7.25	97.25
Agree	6	1.18	98.43
Strongly Agree	8	1.57	100.00
Attack Men, Women & Children
Strongly Disagree	332	65.10	65.10
Disagree	123	24.12	89.22
Undecided	48	9.41	98.63
Agree	4	0.78	99.41
Strongly Agree	3	0.59	100.00
Attack Adult Men
Strongly Disagree	286	56.08	56.08
Disagree	159	31.18	87.25
Undecided	52	10.20	97.45
Agree	5	0.98	98.43
Strongly Agree	8	1.57	100.00
Kill Governor
Strongly Disagree	50	9.80	9.80
Disagree	88	17.25	27.06
Undecided	153	30.00	57.06
Agree	109	21.37	78.43
Strongly Agree	110	21.57	100.00
Kill Mashal
Strongly Disagree	177	34.71	34.71
Disagree	135	26.47	61.18
Undecided	106	20.78	81.96
Agree	59	11.57	93.53
Strongly Agree	33	6.47	100.00
Kill Asia Bibi
Strongly Disagree	92	18.04	18.04
Disagree	101	19.80	37.84
Undecided	166	32.55	70.39
Agree	84	16.47	86.86
Strongly Agree	67	13.14	100.00
Violence Against Ahmadis
Strongly Disagree	66	12.94	12.94
Disagree	89	17.45	30.39
Undecided	153	30.00	60.39
Agree	118	23.14	83.53
Strongly Agree	84	16.47	100.00
POOR ECONOMIC PROSPECTS				0	.703	−1.768	1.634
Economic Prospects
Strongly Disagree	42	8.24	8.24
Disagree	98	19.22	27.45
Undecided	63	12.35	39.80
Agree	211	41.37	81.18
Strongly Agree	96	18.82	100.00
Unfair Treatment
Strongly Disagree	43	8.43	8.43
Disagree	156	30.59	39.02
Undecided	116	22.75	61.76
Agree	166	32.55	94.31
Strongly Agree	29	5.69	100.00
Struggle
Strongly Disagree	72	14.12	14.12
Disagree	160	31.37	45.49
Undecided	93	18.24	63.73
Agree	141	27.65	91.37
Strongly Agree	44	8.63	100.00
POLITICAL MARGINALIZATION				0	.743	−1.535	1.769
Vote
Strongly Disagree	53	10.39	10.39
Disagree	53	10.39	20.78
Undecided	67	13.14	33.92
Agree	163	31.96	65.88
Strongly Agree	174	34.12	100.00
Representation
Strongly Disagree	29	5.69	5.69
Disagree	96	18.82	24.51
Undecided	146	28.63	53.14
Agree	158	30.98	84.12
Strongly Agree	81	15.88	100.00
INDIVIDUAL RELATIVE DEPRIVATION
Chances
Strongly Disagree	37	7.25	7.25
Disagree	153	30.00	37.25
Undecided	117	22.94	60.20
Agree	148	29.02	89.22
Strongly Agree	55	10.78	100.00
Treatment
Strongly Disagree	44	8.63	8.63
Disagree	169	33.14	41.76
Undecided	93	18.24	60.00
Agree	143	28.04	88.04
Strongly Agree	61	11.96	100.00
Less Well Off
Strongly Disagree	54	10.59	10.59
Disagree	175	34.31	44.90
Undecided	131	25.69	70.59
Agree	121	23.73	94.31
Strongly Agree	29	5.69	100.00
Discrimination
Strongly Disagree	59	11.57	11.57
Disagree	194	38.04	49.61
Undecided	117	22.94	72.55
Agree	111	21.76	94.31
Strongly Agree	29	5.69	100.00
Comparison
Strongly Disagree	67	13.14	13.14
Disagree	211	41.37	54.51
Undecided	90	17.65	72.16
Agree	107	20.98	93.14
Strongly Agree	35	6.86	100.00
Buying
Strongly Disagree	46	9.02	9.02
Disagree	191	37.45	46.47
Undecided	108	21.18	67.65
Agree	127	24.90	92.55
Strongly Agree	38	7.45	100.00
COLLECTIVE RELATIVE DEPRIVATION
Group Less Well Off
Strongly Disagree	87	17.06	17.06
Disagree	170	33.33	50.39
Undecided	118	23.14	73.53
Agree	95	18.63	92.16
Strongly Agree	40	7.84	100.00
Group’s Treatment
Strongly Disagree	39	7.65	7.65
Disagree	119	23.33	30.98
Undecided	139	27.25	58.24
Agree	152	29.80	88.04
Strongly Agree	61	11.96	100.00
Group Discrimination
Strongly Disagree	43	8.43	8.43
Disagree	135	26.47	34.90
Undecided	131	25.69	60.59
Agree	151	29.61	90.20
Strongly Agree	50	9.80	100.00
Group Unfairly Treated
Strongly Disagree	51	10.00	10.00
Disagree	156	30.59	40.59
Undecided	114	22.35	62.94
Agree	131	25.69	88.63
Strongly Agree	58	11.37	100.00
Group Opportunities
Strongly Disagree	57	11.18	11.18
Disagree	162	31.76	42.94
Undecided	103	20.20	63.14
Agree	134	26.27	89.41
Strongly Agree	54	10.59	100.00
Buying
Strongly Disagree	69	13.53	13.53
Disagree	175	34.31	47.84
Undecided	130	25.49	73.33
Agree	107	20.98	94.31
Strongly Agree	29	5.69	100.00
INJUSTICE
Fair Treatment
Strongly Disagree	11	2.16	2.16
Disagree	60	11.76	13.92
Undecided	76	14.90	28.82
Agree	283	55.49	84.31
Strongly Agree	80	15.69	100.00
Equal Treatment
Strongly Disagree	13	2.55	2.55
Disagree	81	15.88	18.43
Undecided	82	16.08	34.51
Agree	261	51.18	85.69
Strongly Agree	73	14.31	100.00
Listening
Strongly Disagree	14	2.75	2.75
Disagree	57	11.18	13.92
Undecided	54	10.59	24.51
Agree	292	57.25	81.76
Strongly Agree	93	18.24	100.00
Decision
Strongly Disagree	12	2.35	2.35
Disagree	80	15.69	18.04
Undecided	138	27.06	45.10
Agree	227	44.51	89.61
Strongly Agree	53	10.39	100.00
Respect
Strongly Disagree	6	1.18	1.18
Disagree	34	6.67	7.84
Undecided	60	11.76	19.61
Agree	306	60.00	79.61
Strongly Agree	104	20.39	100.00
Disagreeing
Strongly Disagree	11	2.16	2.16
Disagree	70	13.73	15.88
Undecided	114	22.35	38.24
Agree	255	50.00	88.24
Strongly Agree	60	11.76	100.00
Complaints
Strongly Disagree	21	4.12	4.12
Disagree	59	11.57	15.69
Undecided	106	20.78	36.47
Agree	269	52.75	89.22
Strongly Agree	55	10.78	100.00
EXCLUSION
Time Spending
Strongly Disagree	16	3.14	3.14
Disagree	55	10.78	13.92
Undecided	93	18.24	32.16
Agree	238	46.67	78.82
Strongly Agree	108	21.18	100.00
Invitation
Strongly Disagree	34	6.67	6.67
Disagree	114	22.35	29.02
Undecided	111	21.76	50.78
Agree	198	38.82	89.61
Strongly Agree	53	10.39	100.00
Holiday Plans
Strongly Disagree	36	7.06	7.06
Disagree	78	15.29	22.35
Undecided	93	18.24	40.59
Agree	231	45.29	85.88
Strongly Agree	72	14.12	100.00
Attention
Strongly Disagree	27	5.29	5.29
Disagree	90	17.65	22.94
Undecided	109	21.37	44.31
Agree	219	42.94	87.25
Strongly Agree	65	12.75	100.00
Eating Invitation
Strongly Disagree	24	4.71	4.71
Disagree	70	13.73	18.43
Undecided	67	13.14	31.57
Agree	270	52.94	84.51
Strongly Agree	79	15.49	100.00
Weekend Invitations
Strongly Disagree	26	5.10	5.10
Disagree	87	17.06	22.16
Undecided	88	17.25	39.41
Agree	238	46.67	86.08
Strongly Agree	71	13.92	100.00
INEQUALITY
Satisfaction
Strongly Disagree	39	7.65	7.65
Disagree	93	18.24	25.88
Undecided	72	14.12	40.00
Agree	221	43.33	83.33
Strongly Agree	85	16.67	100.00
Worse Economic Situation
Strongly Disagree	58	11.37	11.37
Disagree	182	35.69	47.06
Undecided	129	25.29	72.35
Agree	107	20.98	93.33
Strongly Agree	34	6.67	100.00
Government Services
Strongly Disagree	33	6.47	6.47
Disagree	111	21.76	28.24
Undecided	130	25.49	53.73
Agree	198	38.82	92.55
Strongly Agree	38	7.45	100.00
Inequality Concern
Strongly Disagree	28	5.49	5.49
Disagree	133	26.08	31.57
Undecided	168	32.94	64.51
Agree	147	28.82	93.33
Strongly Agree	34	6.67	100.00
Redistribution
Strongly Disagree	18	3.53	3.53
Disagree	47	9.22	12.75
Undecided	93	18.24	30.98
Agree	215	42.16	73.14
Strongly Agree	137	26.86	100.00
OPPRESSION
Inferior
Strongly Disagree	44	8.63	8.63
Disagree	156	30.59	39.22
Undecided	140	27.45	66.67
Agree	128	25.10	91.76
Strongly Agree	42	8.24	100.00
Looked Down Upon
Strongly Disagree	41	8.04	8.04
Disagree	147	28.82	36.86
Undecided	156	30.59	67.45
Agree	126	24.71	92.16
Strongly Agree	40	7.84	100.00
Unjust Treatment
Strongly Disagree	44	8.63	8.63
Disagree	153	30.00	38.63
Undecided	99	19.41	58.04
Agree	172	33.73	91.76
Strongly Agree	42	8.24	100.00
Humiliated by Dominant Groups
Strongly Disagree	47	9.22	9.22
Disagree	193	37.84	47.06
Undecided	122	23.92	70.98
Agree	115	22.55	93.53
Strongly Agree	33	6.47	100.00
Rights
Strongly Disagree	38	7.45	7.45
Disagree	154	30.20	37.65
Undecided	96	18.82	56.47
Agree	150	29.41	85.88
Strongly Agree	72	14.12	100.00
Felt Humiliated
Strongly Disagree	51	10.00	10.00
Disagree	212	41.57	51.57
Undecided	106	20.78	72.35
Agree	105	20.59	92.94
Strongly Agree	36	7.06	100.00
Living
Strongly Disagree	44	8.63	8.63
Disagree	165	32.35	40.98
Undecided	130	25.49	66.47
Agree	131	25.69	92.16
Strongly Agree	40	7.84	100.00
Control
Strongly Disagree	43	8.43	8.43
Disagree	157	30.78	39.22
Undecided	121	23.73	62.94
Agree	147	28.82	91.76
Strongly Agree	42	8.24	100.00
Hurt
Strongly Disagree	63	12.35	12.35
Disagree	221	43.33	55.69
Undecided	107	20.98	76.67
Agree	86	16.86	93.53
Strongly Agree	33	6.47	100.00
Abuse
Strongly Disagree	55	10.78	10.78
Disagree	180	35.29	46.08
Undecided	119	23.33	69.41
Agree	120	23.53	92.94
Strongly Agree	36	7.06	100.00
PRAYER FREQUENCY
Prayer Frequency
Daily	395	77.45	77.45
Weekly	37	7.25	84.71
Monthly	12	2.35	87.06
Sometimes	21	4.12	91.18
Rarely	45	8.82	100.00
Identity
Male	17	3.33	3.33
Female	9	1.76	5.10
Muslim	404	79.22	84.31
Caste/Tribe	13	2.55	86.86
Pakistani	67	13.14	100.00

Appendix 3:

Pairwise Correlation Results

Table

Variables	(1)
(1) Religiosity	1.000
(2) Radicalization	−0.010
	(0.820)
(3) Economic Prospects	0.073
	(0.097)
(4) Political Marginalization	0.063
	(0.153)
(5) Individual Relative Deprivation	0.101*
	(0.023)
(6) Collective Relative Deprivation	0.040
	(0.366)
(7) Injustice	−0.126*
	(0.004)
(8) Exclusion	−0.063
	(0.156)
(9) Inequality	0.002
	(0.966)
(10) Oppression	0.022
	(0.615)
(11) Age	−0.040
	(0.368)
(12) Gender	−0.023
	(0.598)
(13) Marital Status	−0.093*
	(0.035)
(14) Family Size	−0.047
	(0.292)
(15) Family Income	0.023
	(0.609)

Appendix 4:

Linear Regression

Table

Outcome: Radicalization
Variables	Linear
Economic Prospects	0.103***
	(0.035)
Political Marginalization	0.031
	(0.030)
Individual Relative Deprivation	0.149***
	(0.041)
Collective Relative Deprivation	−0.005
	(0.041)
Injustice	0.110***
	(0.036)
Exclusion	0.001
	(0.028)
Inequality	0.102**
	(0.044)
Oppression	−0.024
	(0.040)
Controls
Gender (0 = Female, 1 = Male)	0.211***
	(0.041)
Age	0.016
	(0.023)
Marital Status	0.149*
	(0.089)
Family Size	0.035
	(0.030)
Family Income	−0.057
	(0.017)
Bootstrap (P-value)	0.017
R-sq	0.183
Het (P-value)	0.009
No. of Observations	510

***indicates significance at 1%.

**indicates significance at 5%.

*indicates significance at 10%.

Appendix 5:

Threshold Test (Robustness Check)

Appendix 6:

Threshold Regression (Robustness Check)

Table

	Outcome: Radicalization
	Religiosity
Variables	Linear	< −.330	> −.330
Economic Prospects	0.104***	−0.113	0.139***
	(0.035)	(0.108)	(0.035)
Political Marginalization	0.031	0.011	0.026
	(0.030)	(0.096)	(0.030)
Individual Relative Deprivation	0.150***	0.111	0.169***
	(0.041)	(0.133)	(0.041)
Collective Relative Deprivation	−0.005	−0.296**	0.061
	(0.042)	(0.123)	(0.038)
Injustice	0.110***	0.108	0.087**
	(0.037)	(0.084)	(0.039)
Exclusion	0.001	0.100	0.023
	(0.028)	(0.108)	(0.028)
Inequality	0.103	0.033	0.099**
	(0.044)	(0.129)	(0.046)
Oppression	−0.024	0.076	−0.089**
	(0.040)	(0.100)	(0.037)
Religiosity	−0.006	0.106	−0.074**
	(0.031)	(0.270)	(0.037)
Controls
Gender (0 = Female, 1 = Male)	0.211***	0.194	0.218**
	(0.041)	(0.143)	(0.043)
Age	0.016	0.105	0.014
	(0.023)	(0.070)	(0.024)
Marital Status	0.147	−0.271	0.191**
	(0.090)	(0.244)	(0.095)
Family Size	0.035	0.203**	0.001
	(0.030)	(0.101)	(0.025)
Family Income	−0.057***	−0.108*	−0.045**
	(0.017)	(0.066)	(0.018)
Bootstrap (P-value)	0.000
R-sq	0.183	0.321	0.241
Het (P-value)	0.011
No. of Observations	510	64	446

Note: Standard errors are reported in parenthesis (White Corrected).

***indicates significance at 1%.

**indicates significance at 5%.

*indicates significance at 10%.

Appendix 7:

Linear Regression with Squared Terms and Inflection Points

Table

Radicalization	Coef.	Inflection Point
Economic Prospects	.097**
	(.039)	1.383
Economic Prospects^2	−.035
	(.041)
Political Marginalization	.037
	(.03)
Individual Relative Deprivation	.137***
	(.039)	2.333
Individual Relative Deprivation^2	−.029
	(.034)
Collective Relative Deprivation	−.002
	(.04)	−.094
Collective Relative Deprivation^2	−.01
	(.035)
Injustice	.136***
	(.037)	−1.164
Injustice^2	.058**
	(025)
Exclusion	.004
	(.028)
Inequality	.108**
	(.043)	−.883
Inequality^2	.061
	(.048)
Oppression	−.015
	(.037)	−.051
Oppression^2	−.15***
	(.03)
Controls
Gender (0 = Female, 1 = Male)	.229***
	(.041)
Age	.025
	(.023)
Marital Status (0 = Unmarried, 1 = Married)	.179**
	(.089)
Family Size	−.099
	(.105)	424.060
Family Size^2	.03
	(.022)
Family Income	−.049
	(.113)	1.621
Family Income^2	0
	(.016)

Note: Standard errors are reported in parenthesis (Robust).

***indicates significance at 1%.

**indicates significance at 5%.

*indicates significance at 10%.

Appendix 8:

Inflection Points (Graphical Illustration)