Of Political Violence People 2

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

bs_bs_banner

Political Psychology, Vol. 43, No. 1, 2022


doi: 10.1111/pops.12743

Predictors of Political Violence Outcomes among Young People: A


Systematic Review and Meta-­Analysis
Sara Jahnke
Friedrich Schiller University Jena
Åbo Akademi University

Katharina Abad Borger


Technical University Dortmund

Andreas Beelmann
Friedrich Schiller University Jena

The present meta-­analysis summarizes the international basis of evidence regarding links between psychologically
meaningful risk factors and political violence outcomes among adolescents and young adults. We synthesized
422 cross-­sectional effect sizes from 95 samples (67 index publications, 23 countries), using robust variance
estimation. The results of seven longitudinal studies and one intervention study are discussed in a narrative
manner. We detected significant effects for depression (r = .07, k = 10); empathy (r = −.16, k = 7); aggression
(r = .24, k = 10); identification (r = .21, k = 30); relative group deprivation (r = .19, k = 11); realistic threat
(r = .30, k = 27); symbolic threat (r = .28, k = 10); negative intergroup emotions (r = .25, k = 9); experiences
of discrimination (r = .11, k = 12); dissatisfaction with the police, political actors, and institutions (r = .11,
k = 32); and negative attitudes toward democracy (r = .17, k = 10). No significant effect was found for self-­
esteem, intolerance of uncertainty, narcissism, or exposure to intergroup conflict. The reviewed longitudinal
and intervention studies mainly confirm these cross-­sectional results. The script and our data are available in
an open online repository.
KEY WORDS: extremism, radicalization, political violence, adolescence, early adulthood

Political violence is a threat to security with far-­reaching effects on public health, economic pros-
perity, and well-­being (Alne & Serdarevic, 2020; Grossman, Khalil, & Ray, 2019; Vorsina, Manning,
Fleming, Ambrey, & Smith, 2017). In 2019, the most extreme acts of political violence claimed
approximately 15,952 lives worldwide (Global Terrorism Index of the Institute for Economics and
Peace, 2019). Although cases of terrorism-­related deaths are less frequent in Europe or North America
than in the Middle East or Northern Africa, political violence remains an important concern (Institute
for Economics and Peace, 2019). For example, the German Verfassungsschutz registered more than
2000 acts of right-­wing or left-­wing violence in 2018 (Bundesministerium des Innern, für Bau und
Heimat, 2013), whereas the U.S.-­based Uniform Crime Report counted 7120 hate crime–­related

111
0162-895X © 2021 The Authors. Political Psychology published by Wiley Periodicals LLC on behalf of International Society of Political Psychology
Published by Wiley Periodicals, Inc., 350 Main Street, Malden, MA 02148, USA, 9600 Garsington Road, Oxford, OX4 2DQ,
and PO Box 378 Carlton South, 3053 Victoria, Australia
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any
medium, provided the original work is properly cited.
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
112 Jahnke et al.

incidents within the same time period (Federal Bureau of Investigation, 2019). Additionally, recent
surveys have documented that a non-­negligible proportion of the population express sympathy for
political violence or readiness to engage in politically violent acts (McCauley & Moskalenko, 2017).
For instance, in a sample of Swiss adolescents, about 1 in 10 agreed that kidnappings or other acts
of violence are sometimes necessary to fight for a better world (Nivette, Eisner, & Ribeaud, 2017).
In recent years, numerous studies have been conducted to identify risk and protective factors
of political violence outcomes. Meta-­analysis is an effective tool to summarize results in a field
and reach evidence-­based conclusions about the significance and strength of effects. Yet, current
meta-­analyses only provide preliminary evidence with regard to psychologically meaningful factors,
as opposed to fixed and atheoretical markers of political violence outcomes such as age or gender
(Wolfowicz, Litmanovitz, Weisburd, & Hasisi, 2020). Furthermore, prior reviews typically are not
focused on people in adolescence or early adulthood, even though young people are considerably
more likely to support or engage in political violence than older people (Desmarais, Simons-­Rudolph,
Brugh, Schilling, & Hoggan, 2017; Wolfowicz et al., 2020). Therefore, the present meta-­analysis sets
out to establish a sound evidence base regarding factors associated with political violence outcomes
among people under the age of 30.

Political Violence and Political Radicalization: Previous Theories and a New


Developmental Focus

Political violence describes “the deliberate collective attempt to use force against persons or ob-
jects for political reasons,” that is, with the intent to undermine another group or community “de-
fined by self-­categorization” such as a political party, nation, socioeconomic class, or religious group
(Sageman, 2017, p. 14). Political violence is typically discussed within the broader context of political
radicalization (Neumann, 2013; Sageman, 2017), where it has been conceptualized as one possible,
yet neither necessary nor sufficient, aspect of radicalization by some authors or as the main or defining
outcome of this process by others. At the center of the radicalization concept lies the idea that “no-
body is born radical” or ready to pursue political violence (Klandermans, 2014, p. 17). Instead, acts
of political violence are preceded by a “change in beliefs, feelings, and behaviors in directions that in-
creasingly justify intergroup violence” (McCauley & Moskalenko, 2008, p. 416) or “a process through
which people become increasingly motivated to use violent means against members of an out-­group or
symbolic targets to achieve behavioral change and political goals” (Doosje et al., 2016, p. 79).
A new understanding of radicalization, in which anyone or almost anyone might come to embark
on a path toward political violence, has come to replace prior radicalization theories that focused on
mental illness or trauma (Borum, 2011; King & Taylor, 2011). Many modern radicalization models
have taken up the theoretical and empirical challenge to explain how “otherwise normal mental states
and processes” may drive people to engage in political violence (Borum, 2014, p. 286). Some of these
theories assume a more or less linear progress from a nonradicalized state to terrorism over a distinct
set of steps, such as the perception of justice and the displacement of aggression (e.g., the staircase to
terrorism model; Moghaddam, 2005). Such stage models have been criticized for being overly sim-
plistic, as pathways toward political violence are rarely orderly, predictable, or linear (McCauley &
Moskalenko, 2017). In fact, real-­life trajectories of antisocial behavior are characterized by equi-­and
multifinality, meaning that various pathways may lead to the same outcome and the same factors may
lead to different pathways and forms of outcomes (in this case, political violence; Beelmann, 2020).
To account for this observation, other frameworks typically allow for multiple causal processes.
For instance, McCauley and Moskalenko (2008) identify 12 distinct pathways through which indi-
viduals can come to embrace intergroup violence. Although these pathways can operate at the level of
the individual, the group, or the masses, the authors stress that processes of group identification and
intergroup threat are oftentimes more important than individual variables. Kruglanski et al. (2014)
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Political Violence Outcomes among Young People 113

single out the need to “be someone” as the crucial motivational source of radicalization. In their
significance quest model, people who seek to increase or restore significance may come to commit
politically violent acts if they are exposed to ideologies and social networks promoting violence to
reach political goals. The need for significance can be activated by a range of experiences, including
personal discrimination, trauma, humiliation, or group deprivation. Other influential models seek to
explain specific mechanisms of radicalization. For instance, based on the uncertainty identity, Hogg
(2014) argues that identification with radical groups may be particularly attractive for people who are
prone to feel uncertain because they offer a clear and unambiguous worldview (Doosje et al., 2016).
Given that most individuals become radicalized or engage in political violence before the age of
30, the previous frameworks are limited in that they are not informed by a perspective of develop-
mental psychology (e.g., Beelmann, 2020; Borum & Patterson, 2019). To advance our understanding
of the developmental processes leading to such outcomes, the social-­developmental model of radi-
calization offers one of the first developmentally informed perspectives on radicalization (Beelmann,
2020). According to this model, adverse developmental processes are linked to various interconnected
risk factors at the level of the individual (e.g., problematic self-­concept, bias in social-­cognitive in-
formation processing), the social environment (e.g., experiences of discrimination), and society (e.g.,
real intergroup conflicts). These risk factors may, however, be counteracted by risk-­reducing protec-
tive factors like empathy or democratic values. Radicalization results only if the combined effect of
risk factors outweighs the combined effect of protective factors, leading to four interrelated proximal
and necessary radicalization processes (identity problems, prejudice, ideology, and antisocial behav-
ior) with each specific developmental pathway from childhood to early adulthood.

Psychological Variables and Processes Associated with Political Violence

Mirroring the strong need for structuring and synthesizing the literature on political violence
and radicalization, there are numerous review articles on potential associated variables or pathways.
Vergani, Iqbal, Ilbahar, and Barton’s (2020) scoping review on push, pull, and personal factors offers
one of the largest collections of qualitative and quantitative studies to date. The authors found that
the majority of studies assesses push factors, that is, factors that drive people to political violence
(e.g., loss of legitimacy, relative deprivation, state repression), or pull factors, which make political
violence appealing or legitimate to some (e.g., identification, exposure to propaganda). Personal
factors, that is, factors that increase individual vulnerability for political violence (e.g., narcissism
or depression) are studied less frequently. Vergani et al. (2020) concede that push, pull, and personal
factors are closely interrelated in real-­life settings and often co-­occur. For instance, low self-­esteem
(personal factor) could lead to identification with a radical group (pull factor), which could foster
dissatisfaction with the government (push factor).
Based on systematic database searches, Lösel, King, Bender, and Jugl (2018) identified 30 pro-
tective factors of political violence outcomes. The protective factors pertain to the level of the indi-
vidual (e.g., empathy, self-­control, employment, perceived discrimination, depression), family (e.g.,
appreciative parenting behavior), school (e.g., bonding to school, good school achievement), peer
group (e.g., nonviolent peers, contact with foreigners), and society (e.g., low social capital). With few
exceptions, most of these factors have only been assessed in a single study.
Based on 57 publications, Wolfowicz et al. (2020) meta-­analytically synthesized results for 62
risk and protective factors (e.g., experienced racism, discrimination, perceived injustice, personal
strain, and social disconnectedness) for each level of political violence outcomes (i.e., attitudes,
willingness, behaviors). Although their focus on very narrow categories is commendable, this results
in 95 individual meta-­analyses, which are oftentimes based on as few as two or three effect sizes. To
achieve a research synthesis that is easier to interpret and generalize, the present synthesis will focus
on broader categories based on conceptual core features.
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
114 Jahnke et al.

Previous research syntheses converge on several important observations. First, they note a large
conceptual and methodological heterogeneity (Lösel et al., 2018; Wolfowicz et al., 2020). Second,
there is near-­universal consensus that adolescents and young adults are at the highest risk of harbor-
ing beliefs and attitudes in favor of political violence (Desmarais et al., 2017; Lösel et al., 2018).
Prior reviews are also tainted by important methodological limitations. In particular, many reported
difficulties with collecting a sufficient number of individual studies (Lösel et al., 2018). Even in
Wolfowicz and colleagues’ (2020) comprehensive meta-­analysis, the evidence base for most psycho-
logically relevant factors is relatively small. Additionally, previous reviewers did not assess small-­
study effects or the quality of the screening or coding procedures (Wolfowicz et al., 2020), which
might compromise the validity of their findings.

The Present Meta-­Analysis

The aim of the current study is to inform our understanding of factors associated with political
violence outcomes and to quantify the size and significance of these effects. We will gather evidence re-
garding psychological variables for which there exists a plausible rationale for a causal link to political
violence outcomes. Furthermore, we will focus on studies that predominantly sampled young people
(i.e., below age 30), as this is the group that is most vulnerable to radicalization. Based on theoretical
assumptions and a developmental understanding of radicalization, we expect that political violence out-
comes are linked to a multitude of heterogeneous variables. Following Vergani et al. (2020), these will
be categorized on the level of pull factors, push factors, and personal factors. Another primary concern
for the present analyses was to maximize the turnout of eligible studies and effect sizes. In contrast
to most previous reviews, we will consider studies in languages besides English and supplement the
search of electronic databases with methods including backward reference list searches, contacting
experts, and forward citation searches (Atkinson, Koenka, Sanchez, Moshontz, & Cooper, 2015).

Method

Eligibility Criteria

Dependent and Independent Variables

Our choice and classification of independent variables were guided by Verkuyten’s (2018) three P
model, which differentiated between push, pull, and personal factors. References had to include links
between a political violence outcome and one of the factors defined in Table 1. Although state gov-
ernments might also employ political violence to further their own or their followers’ interests at the
expense of outgroups (Schmid, 2004), the current review will only consider political violence perpe-
trated by nonstate actors. This includes not only right-­wing, left-­wing, and Islamist violence but also
political violence connected to intractable conflicts, such as the Israel–­Palestine conflict, or violence
for other specified or unspecified political goals. SJ and KAB made decisions about the inclusion and
categorization of each coded effect size and resolved conflicting opinions in team discussions.

Age

All references needed to include samples of young people below 30 years.1 This includes studies
with older participants, as long as the mean sample age did not exceed 30.
1
If mean ages in school surveys were not reported, we estimated them based on information about average ages in specific
school grades (e.g., ninth grade corresponds to 14.5 years). If mean ages in student surveys were not reported, we estimated
them to be 21 years.
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Political Violence Outcomes among Young People 115

Table 1. Coding of Factors

Factor Definition
Self-­esteem Rosenberg Self-­Esteem Scale and related scales
Narcissism Personality trait characterized by grandiosity
Depression Clinical definition (ICD or DSM), assessed with PHQ-­9 or other screening
inventories
Empathy Competence to view something from the perspective of someone else
(cognitively and emotionally)
Aggression Behavior that is intentionally directed toward harming or injuring others
Intolerance of uncertainty Individual preference of situations that are clear and unambiguous; in-
cludes need for cognitive closure
Identification Subjective degree of affiliation with or belonging to a social group
Exposure to content legitimizing politi- Being exposed to propaganda legitimizing political violence (offline or
cal violence online)
Relative group deprivation Belief that someone’s group is worse off compared to other groups in
society (Smith, Pettigrew, Pippin, & Bialosiewicz, 2012)
Realistic threat Belief that another group threatens the ingroup’s welfare (e.g., through
competition over power or resources; Stephan & Renfro, 2002)
Symbolic threat Belief that another group poses a threat to the values or views of the in-
group; may include concerns about language and moral beliefs (Stephan
& Renfro, 2002)
Negative intergroup emotions Experience of negative emotions on the ingroup’s behalf
Personal discrimination Experience of unfair disadvantage because of membership in a social group
Exposure to intergroup conflict Witnessing ingroup members fight or get wounded in a confrontation with
the outgroup (directly or indirectly through media reports)
Dissatisfaction with the police, political Negative beliefs regarding the legitimacy, efficiency, and trustworthiness of
actors, and institutions the police, political actors (i.e., parties, politicians), and political institu-
tions like the government or the juridical system
Negative attitudes toward democracy A rejection of democracy or its ethical foundations (i.e., human rights and
civic ideals like freedom of speech or free and fair elections)
Note. Scales were sorted into the domain with the best fit. At least 50% of the items had to match one of the above definitions.

Adequate Statistical Information

Effect sizes based on complex multifactorial or multivariate models are unfit for meta-­analyses
(Lipsey & Wilson, 2001). Therefore, and deviating from procedures in Wolfowicz et al. (2020), only
samples with bivariate effect sizes were included. To limit the loss of information due to this proce-
dure, we attempted to retrieve bivariate effect sizes directly from the authors if they were not readily
accessible.

Language, Publication Date, and Publication Type

We considered studies published in English, German, French, Italian, Norwegian, Swedish,


Danish, Dutch, Spanish, and Hebrew. Studies were included if they were published through, and
including, December 2019.

Literature Search

Our first attempts to search the literature were based on search strings used in previous reviews
(Christmann, 2012) and contained keywords like terrorism, radicalization, and jihad. These first
attempts led to an unmanageable number of mostly irrelevant articles. To increase precision, we
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
116 Jahnke et al.

proceeded by using combinations of the terms radicalization and extremism, political ideologies
(e.g., left-­wing) and variations of the word violence, terrorism-­related terms, and support. As this
search string was still yielding too many irrelevant articles, we used filters to selectively retrieve
quantitative studies and studies with young participants. Our final search string and all employed
filters are documented in Supplemental Material A.
We conducted our search in several relevant databases (see flow chart in Figure 1). The da-
tabase search was first conducted in November 2017 and updated in January 2020. The process
of study selection based on the combined information of both database searches is detailed in the
flow chart in Figure 1. The study selection and data extraction were conducted by SJ (who scanned
all retrieved sources). Additionally, KAB and a trained research assistant each scanned about 50%
of the database. Disagreements between raters were resolved in team discussions. For all eligible
studies, we used forward and backward search methods and contacted experts to retrieve fur-
ther documents and unpublished materials. We also scanned reviews and meta-­analyses that were
published before or while we were working on our own meta-­analysis (e.g., Lösel et al., 2018;
Wolfowicz et al., 2020). However, when accounting for additional publications based on samples
that were already included in our analysis, this did not lead to the identification of further relevant
publications.

Retrieving Additional Information

For studies published after 1990, we sought to include information not provided in published
document(s) whenever necessary. Typically, this involved requests for unreported bivariate effect
sizes, information about mean ages, and information about items from unpublished scales in cases
where scale descriptions were ambiguous or unclear. In total, we contacted 32 authors and received
additional effect sizes for 19 index publications.

Independent Effect Sizes

In many cases, particularly for large and laborious studies, results based on the same sample
are used in several publications. To avoid double-­counting, we designated one study as an index
publication but used all records available to us to retrieve a maximum amount of information.
Supplemental Material B provides an overview of index publications and additional publications
or communications. If authors published analyses based on overlapping samples, we chose the
analysis strategy that allowed us to retain the largest amount of information, that is, the one based
on the largest n. We deviated from this strategy for one factor only. Gender might obfuscate a link
between depression and political violence outcomes (because boys and men typically score higher
on political violence outcomes, but are less likely to have depression than girls and women). About
half of the studies reported correlations for depression separately for male and female partici-
pants, prompting us to favor gender-­separated effect sizes in the analysis for this factor whenever
possible.

Study Coding

Classification of Variable Domains

All included measures were classified by two independent raters (SJ and KAB). On the predictor
side, we coded the factors listed in Table 1.
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Political Violence Outcomes among Young People 117

Databases: PsychINFO/PsycArticles/Behavioral Science Collection (n = 722), PubMed (n =


529), WebofScience Core Collection (n = 2,170), ProQuest Dissertations and Theses Global
(n = 913), ERIC (n = 356), Sociological Abstracts (n = 1303), Psyndex (n = 297)

3,938 after removal of duplicates

3,938 records screened (Title, Abstract)

Full-text articles excluded 443 full-text Additional records:


publications
Not available (n = 12) Reference lists (12
assessed for
No quantitative data (n = 42) samples from 8
eligibility
Mean age 30 yrs. (n = 80) index publications)
No political violence outcome (n =
article citations (15
173)
samples from 11
Experimental study on death
salience or identity threat (n = 12) index publications)
Only macro-level analysis or no 56 samples expert
predictors in study (n = 3) from 39 index recommendations
None of the 13 variable domains publications /unstructured
(n = 43) searches (12
No bivariate information a (n = 13) samples from 9
not in one of the 10 pre-specified index publications)
languages (n = 6)
additional publications regarding
samples already in the database (n
= 20)b

95 samples from 67 index publications

Figure 1. Flow Chart. aFor sources published 1990 and later, we attempted to retrieve missing bivariate correlations by
contacting study authors. bAll sources were used to gather additional information. A list of additional publications can be
found in Supplemental Material B.

Reliability of the Coding Process

Based on recommendations put forward by Lipsey and Wilson (2001), we randomly selected
17 publications (25 samples, 77 effect sizes, 26% of all publications, and 20% of all effect sizes)
to be coded independently by the two raters (SJ and KAB). For the continuous variables (effect
sizes, mean ages, sample sizes, rate of females in the sample), the average correlation was r = .97
(SD = .03) and ranged from r = .93 to r = 1.00. There was complete agreement regarding the peer
review status, publication year, and status of the sample (Cohen’s κ = 1.00) and near complete agree-
ment for the country of the sample (Cohen’s κ = .995). Cohen’s κ ranged from .89 to .99 for the
classification of effect sizes, the ideological context, and the level of political violence outcomes. In
sum, these observations indicate high interrater reliability.
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
118 Jahnke et al.

Table 2. Characteristics of Included Samples

Description Number of Samples


Country of sample
Central Europe
Austria 1
Czech Republic 1
Germany 32
Poland 4
Switzerland 3
Western Europe
Belgium 4
Great Britain 7
France 2
Netherlands 6
Northern Europe
Norway 1
Sweden 3
Eastern Europe
Ukraine 1
Southern Europe
Greece 1
Spain 8
North America
USA 6
Canada 3
South America
Colombia 1
Middle East
Israel 4
Palestine 3
South Asia
Pakistan 1
Southeast Asia
Indonesia 1
India 1
Australia/Oceania
Australia 1
Social status
No subordinate status 61
Subordinate status 34
Type of violence
General political violence 41
Right-­wing violence 19
Islamist violence 16
Other intergroup violence 29
Left-­wing violence 10
Mixed 1
Level of violence
Attitudes 49
Willingness 43
Actual behavior 21
Mixed 3

Preparation of Effect Sizes

We chose Pearson’s r as the basis of our meta-­analysis since this was the most commonly
reported effect size across all studies. Following recommendations in Card (2012), we estimated
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Political Violence Outcomes among Young People 119

r from other reported effect sizes like Spearman’s ρ, point-­biserial correlations, and contingency
coefficients (note that uncorrected contingency coefficients were transformed to corrected contin-
gency coefficients before entry). If the study provided the information in frequency tables only, we
used Wilson’s (2001) Practical Meta-­Analysis Effect Size Calculator to calculate Pearson’s r. In the
case that detailed statistical information was missing but an effect was described as nonsignificant,
we conservatively estimated that r = 0. We z-­transformed correlation coefficients before entering
them into the meta-­analytic model and then back-­transformed them to r to facilitate interpretation.
All effect sizes included in the meta-­analyses were cross-­sectional and should therefore be regarded
as associations, not causal predictors. The R script and the associated data file are available at Open
Science Framework (Jahnke, Abad Borger, & Beelmann, 2020, see link in the References section).

Meta-­Analytical Procedure

Choice of Meta-­Analytic Model

We favored a random effects model because we anticipated substantial between-­study variability


(Victoroff, 2005). Traditional random effects models are based on the assumption that effect sizes are in-
dependent. This assumption is not justified for the current analyses because many samples included two or
more effect sizes based on data from the same participants. Robust variance estimation (RVE) represents
a new and viable alternative, which permits using dependent effect sizes and, hence, the full information
from the data set (Tanner-­Smith, Tipton, & Polanin, 2016). To compute r-­to-­z transformed coefficients
and sampling variances based on standard large-­sample approximation, we used the escalc-­function of
the R package metafor (Viechtbauer, 2010). All analyses based on RVE were calculated with the R pack-
age robumeta (Fisher & Tipton, 2015). We reduced the probability of inflated Type 1 error rates in RVE
analyses via Tipton’s (2015) small-­sample adjustment and estimated the intercepts (representing weighted
mean effects) separately for each predictor. Tanner-­Smith et al. (2016) warn that if small-­sample corrected
degrees of freedom are lower than 4, associated p-­values are not reliable. To avoid underestimating the
true Type 1 error, we changed the significance level to p < .01 for all analyses with df < 4.

Publication Bias

Publication biases occur when studies with significant effects or stronger outcomes have a higher
chance of being published than studies with smaller or nonsignificant findings. Meta-­analysts often
rely on visual inspection of funnel plots and other, more objective methods based on funnel plots
(e.g., Egger’s test) to assess small study effects. For lack of a better alternative, some researchers
have also applied these methods when dealing with dependent effect sizes, even though they techni-
cally require independent effect sizes. Rodgers and Pustejovsky (2019) have recently introduced a
new version of the traditional Egger test that is based on RVE and can be used for dependent samples.
Therefore, we conducted the “Egger Sandwich” test to investigate potential small study effects.

Sensitivity Tests

Effect size outliers can bias the results of meta-­analyses. To investigate whether effect size outli-
ers had a disproportionate impact on the outcome, we winsorized the bottom and top 5% of the effect
size distribution before each meta-­analysis, setting them to the 5th and 95th percentile, respectively
(see Lipsey & Wilson, 2001; Tanner-­Smith et al., 2016). We then compared the results based on the
original and the winsorized data set. Additionally, we investigated whether changing the assumed
size of the within-­study correlations led to different results for the intercept-­only models.
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
120 Jahnke et al.

Moderator Analyses

We conducted a series of explorative meta-­regressions for all factors based on k ≥ 10, and one
moderator variable at a time.

Results

Sample Characteristics

Descriptive information on the overall characteristics of all included samples is presented in


Table 2. For a detailed list of the predictor and outcome variables included for each sample and
their classification, see Supplemental Material C. For an overview of individual characteristics of
the included samples, see Supplemental Material D. The data in our review were drawn together
from 46 peer-­reviewed journal articles, four reports, one research memo, three master’s theses, 10
monographs, two book chapters, and one unpublished data set. Studies were published from 1975 to
2019 (note that these and the following statistics are limited to only index publications to ensure that
several publications based on the same data set would not be counted multiple times). Most index
publications were written in English (69%) or German (30%), and one publication was written in
Dutch (1%). Most studies were conducted in European (particularly Germany) or North American
countries (see Table 2). Mean ages ranged from 13.58 to 29.97 years within the included samples,
with a median mean age of 16.64 years (not weighted).
Most studies assessed unspecific political violence outcomes, followed by right-­wing and
Islamist violence. It was more typical to assess attitudes or behavioral intentions toward violence
than actual political violence (see Table 2 and Supplementary Material C). For 84% of effect sizes,
we were able to retrieve Pearson’s r as a measure of bivariate association. Forty-­four studies were
conducted among predominantly adolescent samples, and 51 among predominantly adult samples.
It is important to note that the younger samples involved a total of 74,633 cases, compared to only
34,516 cases in the predominantly adult samples.

Overall Aggregated Correlations

We describe results for personal, pull, and push factors separately. Results across all domains
are presented in Table 3.

Personal Factors

We detected significant overall effects for depression, empathy, and aggression. We found no
significant links between self-­esteem and political violence outcomes, or between intolerance of un-
certainty and political violence. Note that a significant effect for narcissism could not be established,
despite the fact that the four effect sizes that went into this estimate were positive and ranged from
r = .20 to r = .44. This can be explained by the small number of effect sizes/studies as well as the use
of random effects models. With the exception of narcissism, heterogeneity was severe across the six
constructs (i.e., I2 ≥ 75%).

Push Factors

We did not analyze exposure to content legitimizing political violence because this factor was
only assessed in a single study (Pauwels & Schils, 2016). For identification, we yielded a significant
small effect size and severe heterogeneity.
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Political Violence Outcomes among Young People 121

Table 3. Overall Effect Sizes

r (95% CI) k n df p I2
Personal
Self-­esteem −.05 [−.15,.06] 14 26 11.93 .326 91.41
Narcissism .23 [−.38,.70] 2 4 1 .135a 25.38
Depression .07 [.02,.12] 10 29 8.32 .011 87.48
Empathy −.16 [−.19, −.12] 7 26 4.24 <.001 81.94
Aggression .24 [.14,.33] 10 17 8.93 .001 96.07
Intolerance of uncertainty .03 [−.11,.17] 6 14 4.68 .603 84.43
Push
Identification .21 [.12,.29] 30 50 27.81 <.001 93.10
Exposure to content legitimizing political –­ –­ –­ –­ –­ –­
violenceb
Pull
Group relative deprivation .19 [.14,.25] 11 18 7.07 <.001 63.54
Realistic threat .30 [.22,.37] 27 50 25.88 <.001 98.24
Symbolic threat .28 [.18,.37] 10 12 8.92 <.001 96.02
Negative intergroup emotions .25 [.15,.35] 9 16 7.87 .001 82.26
Exposure to intergroup conflict .24 [.01,.45] 5 7 3.88 .046a 94.86
Experiences of discrimination .11 [.05,.17] 12 35 10.9 .003 93.58
Dissatisfaction with the police, political .11 [.07,.15] 32 103 30.87 <.001 96.30
actors, and institutions
Dissatisfaction with democracy .17 [.12,.21] 10 15 8.51 <.001 87.43
Note. ρ = .80, k = number of independent samples, n = number of effect sizes.
a
Significance level set from p < .05 to p < .01 because df < 4.
b
Excluded because k < 2 (Pauwels & Schils, 2016).

Pull Factors

We detected the highest overall effect size for realistic threat (r = .30). With the exception of
exposure to conflict, all variable domains showed significant associations with political violence
outcomes. Heterogeneity in the effect sizes was severe.

Small Study Effects

On visual inspection, only the funnel plots for intolerance of uncertainty showed signs of asym-
metry (see Supplemental Material E; note that all effect sizes were treated as if they were indepen-
dent). The plots appeared cylindrical, which indicates that heterogeneity overwhelmed the standard
error (Sterne et al., 2011). Utilization of the “Egger Sandwich” test detected a significant asymmetry
for four meta-­analyses (intolerance of uncertainty: β = 4.66, p = .025; aggression: β = 7.16, p = .002;
dissatisfaction with political actors, institutions, and the police: β = 2.27, p = .027; and dissatisfaction
with democracy: β = 2.44, p = .040).

Sensitivity Analyses

Sensitivity analyses demonstrated that mean effects are insensitive to changing values of ρ, as
we found no differences between the overall correlations, standard errors, and τ2 up until, and includ-
ing, the third decimal digit. In general, results obtained after winsorization did not differ substantially
from results based on non-­winsorized data sets (i.e., differences between the two analyses were less
than r = .03, and none of the results changed from nonsignificant to significant, or vice versa).
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
122 Jahnke et al.

Exploratory Moderator Analyses

Findings from exploratory moderator analyses are presented in Supplemental Material F (age,
publication year, rate of female participants, peer review, and subordinate status),2 Supplemental
Material G (political ideology: unspecific, right-­wing, and Islamist violence), and Supplemental
Material H (level of political violence: attitudes, willingness, and behavior). Effect sizes for depres-
sion were higher in peer-­reviewed journals compared to other forms of publication. For ideologies
associated with political violence, we discovered significantly stronger links between (1) identifica-
tion and (2) realistic threat on the one side and political violence on the other side for right-­wing vi-
olence compared to unspecific violence. Studies on mixed ideologies and mixed types of level of
political violence were not included in the combined models because they contributed too few sam-
ples and effect sizes. Lastly, effect sizes for the link between political violence outcomes and dissat-
isfaction with democracy were stronger for other ethnic, national, or religious violence compared to
unspecific types of political violence, as well as for samples with a surbordinate group status.

Narrative Review of Longitudinal Studies

Seven longitudinal studies and one intervention study assessed the links between political vi-
olence outcomes and at least one of the push, pull, and/or personal factors (Table 4). Dahl (2014)
found that political dissatisfaction predicted Swedish adolescents’ use of illegal political activity one
year later. In a study including all six annual waves of the same cohorts, Swedish adolescents’ prior
approval emerged as a significant predictor of future approval of political violent means (Dahl, 2020).
This effect is mediated by prior delinquency, but not antidemocratic sentiments. Hence, youth who
approve of political violence and later engage in aggressive behavior (with no political context) are
more likely to approve of political violence at a later point in time. Frindte, Boehnke, Kreikenbom,
and Wagner (2012) found no link between personal discrimination and political violence support
within two Muslim subsamples when controlling for prior political violence support. In Nivette et al.
(2017), aggressive behavior at age 15 emerged as one of the strongest predictors of sympathy toward
political violence at age 17. Within a structural equation modeling study on Czech adolescents, a
lower trust in political institutions predicted an increase in their willingness to engage in illegal and/
or violent political actions 1.5 years later (Šerek, Machackova, & Macek, 2018). Among German
university students with a migration background, negative intergroup emotions and various measures
of ethnic or religious identification were not predictive of a higher readiness to engage in illegal or
violent protest at a later point in time. In fact, one measure of group identity (religious identification)
emerged as a significant protective factor of later readiness to engage in political violence (Simon,
Reichert, & Grabow, 2013). In a sample of Protestant and Catholic adolescents in Northern Ireland,
Taylor and McKeown (2019) established a significant link between exposure to sectarian violence
(over the past three months) and participation in sectarian violence (over the past three months) at a
later point in time. Likewise, Taylor, Merrilees, Goeke-­Morey, Shirlow, and Cummings (2016) found
that adolescents from either of the two communities who reported more general aggression and/or
have been exposed to sectarian violence were more likely to engage in acts of political violence.
Regarding the methodological quality of these longitudinal studies, all attempted to control for
relevant confounding variables such as sex or education. Half of the studies additionally controlled
for prior differences in political violence outcomes (Dahl, 2014; Frindte et al., 2012, Šerek et al.,
2018; Simon et al., 2013). About half of the studies employed follow-­up periods of at least one year
2
A subordinate or lower status was assumed for samples composed entirely or mostly of ethnic or religious minority groups,
with the following exceptions: we coded a subordinate status for Catholic residents of Northern Ireland (as opposed to
Protestants) and for Palestinians in the context of the Israel–­Palestine conflict. In all other cases, a nonsubordinate status was
assumed.
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Political Violence Outcomes among Young People 123

(Frindte et al., 2012, Nivette et al., 2017; Šerek et al., 2018) or multiple annual assessments (Dahl,
2020; Taylor et al., 2016). The same set of studies also involved relatively large samples with N > 500
participants at the first assessment. None of the longitudinal studies tested or reported the effects
from predictors on political violence outcomes with corresponding effects in the opposite direction
(i.e., from political violence outcome to predictor at a later point in time), missing an opportunity to
establish the causal direction of the effect with more methodological rigor.
The intervention evaluated in Feddes, Mann, and Doosje (2015) involved training of social and
professional competencies, perspective taking, and self-­esteem. The authors found that political vi-
olence approval and willingness to engage in political violence decreased from the start to the end
of the training. Regression analysis controlling for T1 measures indicated that changes in empathy,
but not self-­esteem or narcissism, significantly predict positive attitudes toward ideology-­based vio-
lence. The association between empathy and attitudes toward ideology-­based violence by others was
negative, meaning that an increase in empathy predicted lower approval of political violence. When
the regression analysis was repeated, replacing attitudes toward ideology-­based violence with one’s
own violent intentions as the outcome, neither empathy nor self-­esteem or narcissism yielded signif-
icant results. The study is limited in its ability to make causal claims based on its very small sample
size and lack of a control group.

Discussion

Within the present meta-­analyses, we identified 95 eligible samples and a small set of longi-
tudinal and intervention studies. Overcoming limitations in previous reviews, the current analyses
provide a rigorous assessment of interrater reliability. Moreover, we were able to retrieve additional
effect sizes and publications that previous reviewers have missed. This pertained particularly to find-
ings from before 1990, studies published in non-­English languages, and studies focusing on various
forms of youth political participation. Hence, we were able to establish a sound international evi-
dence base regarding factors that not only mark but also explain the cause of political violence.
The variety of relatively small but significant effects indicates that only a few young people with
vulnerable personalities, or who are exposed to specific push or pull factors, will actually endorse or
engage in political violence. This is in line with the developmentally informed theory, which predicts
that radicalization is caused by a chronic imbalance of a large number of risk and protective factors,
rather than one or a few concrete causal processes (Beelmann, 2020). Thereby, our results highlight
the need to rely on a broad developmental framework when theorizing radicalization. Concerning hy-
pothesized relationships between specific risk or protective factors and political violence outcomes,
our findings confirm some, but not all, predictions from previous radicalization models.

Push and Pull Variables

Regarding realistic and symbolic threat and group relative deprivation, we found consistent
evidence for a link to political violence outcomes, both in terms of cross-­sectional and longitudinal
research. Also, both longitudinal studies and our cross-­sectional meta-­analysis support the idea that
political violence outcomes are correlated with dissatisfaction with current political actors or democ-
racy. We also found a significant positive cross-­sectional effect for ingroup identification. However,
the only longitudinal design including identification as a predictor variable (Simon et al., 2013) could
not corroborate a significant risk-­enhancing effect over time. Findings regarding exposure to inter-
group conflict were not consistent. Although we detected a small summary effect, we kept the null
hypothesis due to the (conservative) practice to reduce the alpha level when degrees of freedom fall
below a specific threshold. Yet, the two longitudinal studies including exposure to intergroup conflict
provide support that this factor is indeed related to a higher risk for political violence outcomes.
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
124 Jahnke et al.

Table 4. Overview of Longitudinal Studies

Sample Characteristics at t1

Publication nb MAge % Female Country Follow-­Up Factor Type and Name


a
Dahl (2014) 2012 15.04 51.4 SE 1 year Dissatisfaction with the
police, political ac-
tors, and institutions:
Lack of confidence in
institutions
Dahl (2020)a 1987 15.03 52.1 SE 5–­6 annual waves Aggression: Delinquency
Dissatisfaction
with democracy:
Anti‑democratic
sentiments
Feddes et al. 46 16.93 22 NL ca. 3 monthsb Self-­Esteem: Self-­Esteem
(2015) Narcissism: Narcissism
Empathy: Empathy
Group Relative
Deprivation: Collective
relative deprivation
Frindte et al. 923 21.2 52 DE ca. 1 year Experiences of dis-
(2012) crimination: Personal
discrimination
Nivette et al. 1214 15 50 CH 1–­2 years Aggression: Aggression
(2017)
Šerek et al. 768 15.97 54 CZ 1.5 years Dissatisfaction with the
(2018) police, political ac-
tors, and institutions:
Institutional trust
Simon et al. 341 24 66 DE ca. 8–­10 months Identification:
(2013) Ethnocultural iden-
tification, separatist
identification
Negative intergroup
emotions: Group-­based
anger
Taylor and 466 14.5 50 GB ca. 6 months Exposure to intergroup
McKeown conflict: Exposure to
(2019) sectarian antisocial
behavior
Taylor et al. 820 13.61 51 GB 4 annual waves Aggression: Aggression
(2016) Scale
Exposure to intergroup
conflict: Exposure to
sectarian antisocial
behavior
Note. CH = Switzerland, CZ = Czech Republic, DE = Germany, GB = United Kingdom, NL = Netherlands, SE = Sweden.
a
Based on a near-­identical participant pool.
b
Between measurement within 3-­month intervention, post measurement after completion of 3-­month intervention, follow-
­up measurement conducted among 12 adolescents 3 months after the end of the intervention, no follow-­up results reported.

Albeit significant, the overall effect of personal experiences of discrimination is smaller than the
overall effect of the perception of threat at an intergroup level. The only study that assessed this
link over time could not confirm personal experiences of discrimination as a predictor of political
violence outcomes. In sum, these findings are in line with expectations that intergroup processes and
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Political Violence Outcomes among Young People 125

perceptions (more so than actual experiences of unfair treatment) play a major role in explaining
political violence outcomes (e.g., McCauley & Moskalenko, 2008; Verkuyten, 2018).

Personal Variables

The present study also finds longitudinal and cross-­sectional evidence for a link between select
personal variables (i.e., empathy, depression, and aggression) and political violence outcomes. This
strengthens the concept that political violence and general violence among youth may have similar
causes (see, e.g., Beelmann, 2020, for a model that draws heavily from general youth delinquency re-
search). However, it is interesting to note that the effects of personal variables were typically smaller
than what we found for (intergroup) push or pull variables (with the exception of aggression). For
some variables, like self-­esteem and intolerance of uncertainty, neither cross-­sectional nor longitudi-
nal studies were in support of the proposition that these variables are connected to political violence
outcomes. While not disproving the hypothesis that intolerance of uncertainty or low self-­esteem can
motivate identification or affiliation with a radical group (Doosje et al., 2016; Hogg, 2014), this puts
into question the assumption that these factors increase the risk of actual political violence outcomes
as well. Although the two samples reporting on narcissism found a significant positive link to polit-
ical violence outcomes, we could not identify a significant summary effect based on this very small
set of studies.
Contrary to Wolfowicz et al. (2020), who found a risk-­reducing effect, we obtained a positive
(i.e., risk-­enhancing) link between depression and political violence outcomes. There are several
potential explanations for this difference: First, it may indicate differential developmental effects,
as irritability is more typical for depression in adolescents than in later points in life. Second, it
may point toward a third variable effect: Since women are generally less likely to support political
violence and more likely to have depression, this could explain why Wolfowicz et al. (2020) found a
protective effect of depression. This is not the case in the present research, as we coded associations
separately for male and female samples.
Overall, effect sizes did not statistically differ across different participant and publication char-
acteristics for a range of potential moderators (i.e., age, year of publication, rate of female partic-
ipants, peer review status, low social status, and level of political violence). We did, however, find
that correlations between right-­wing violence and identification or realistic threat may be higher
than correlations detected for other types of political violence. This could mean that the derogation
of outgroups and strong identification with the ingroup is particularly central to right-­wing ideology.
Nevertheless, given the speculative nature of these arguments and the exploratory focus of the tests,
these findings should be interpreted with caution.

Limitations

To establish risk and protective factors, as well as causal factors for political violence outcomes,
longitudinal or experimental studies are needed. Therefore, we included evidence retrieved from
longitudinal or intervention studies in a narrative review to allow readers to draw preliminary con-
clusions about the causal nature of the overall cross-­sectional effects. However, for the most connec-
tions, longitudinal evidence supported conclusions based on the results of the cross-­sectional studies.
It is not clear whether the present results generalize to older sections of the population. Findings
were, for the most part, similar in direction to effects reported in another meta-­analysis with no age
restrictions (Wolfowicz et al., 2020). Yet, as the period from childhood to young adulthood is most
important for the development of political violence outcomes, our focus on young samples is also an
asset of our review.
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
126 Jahnke et al.

We detected small study effects in four out of 15 cases. For intolerance of uncertainty, aggres-
sion, dissatisfaction with democracy, and dissatisfaction with political actors, institutions, and the po-
lice, the effect sizes based on small studies were lower than the effect sizes based on larger and more
precise studies. While a significant “Egger’s sandwich” test result can be indicative of publication
bias, alternative causes are possible, especially in the context of high heterogeneity. Nevertheless, we
believe that the risk for publication bias in our meta-­analyses is reasonably low. This assumption is
based on (1) the observational and correlational nature of the data (which typically did not represent
the main result of the publication) and (2) the extensive search strategies that we employed to locate
gray literature, such as searching dissertation databases, reference list and citation searches, contact-
ing experts, and unstructured Internet searches.
Finally, political violence outcomes were based on self-­report data only. Also, the majority of
the tested samples surveyed attitudes about or willingness to engage in political violence (as opposed
to actual violent behavior), and only very few addressed left-­wing violence. Despite these shortcom-
ings, the current meta-­analysis is the most comprehensive of its kind, as despite restricting our search
to young samples and bivariate effect sizes, we collected considerably more studies than previous
reviews focusing on political violence outcomes.

Implications for Prevention

All factors studied in the present meta-­analyses are common targets in youth-­centered radical-
ization prevention programs (Madriaza & Ponsot, 2016; Stephens, Sieckelinck, & Boutellier, 2021).
Although we find tentative overall support for most factors, intolerance of uncertainty and self-­
esteem do not appear suitable intervention targets. Based on an extended analysis of the outcomes of
youth prevention programs (Beelmann, Jahnke, Lutterbach, & Hercher, 2021) and the results of the
present synthesis of the literature, the following measures appear most promising: (1) programs that
promote the development of a positive social identity and a sense of belonging to nondeviant groups;
(2) contact interventions, whereby members of different social groups come together to follow a
shared goal to counteract negative intergroup attitudes and emotions; (3) interventions reducing ag-
gression and prejudice among children and adolescents, such as training of social cognitive skills like
empathy or interpersonal problem solving; and (4) educational interventions to strengthen trust in
political institutions and the support of democratic values.

ACKNOWLEDGMENTS

Correspondence concerning this article should be addressed to Sara Jahnke, Åbo Akademi
Fakulteten för humaniora psykologi och teologi, Tehtaankatu 2, Turku 20500, Finland. E-­mail: sara.
jahnke@abo.fi

REFERENCES

All references included in the quantitative meta-­analysis are listed in Supplemental


Material B.
Alne, R., & Serdarevic, N. (2020). The cost of terrorism: Network effects and the economic impact of child loss (SSRN
Scholarly Paper ID 3517441). Social Science Research Network. Retrieved from https://papers.ssrn.com/abstr​
act=3517441
Atkinson, K. M., Koenka, A. C., Sanchez, C. E., Moshontz, H., & Cooper, H. (2015). Reporting standards for literature
searches and report inclusion criteria: Making research syntheses more transparent and easy to replicate. Research
Synthesis Methods, 6(1), 87–­95. https://doi.org/10.1002/jrsm.1127
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Political Violence Outcomes among Young People 127

Beelmann, A. (2020). A social-­developmental model of radicalization: A systematic integration of existing theories and em-
pirical research. International Journal of Conflict and Violence, 14(1), 1–­14. https://doi.org/10.4119/ijcv-­3778
Beelmann, A., Jahnke, S., Lutterbach, S., & Hercher, J. (2021). Developmental prevention of radicalization and violent ex-
tremism: A systematic review of approaches and evaluation research to inform science and public policy. Manuscript
Submitted for Publication.
Borum, R. (2011). Radicalization into violent extremism I: A review of social science theories. Journal of Strategic Security,
4(4), 7–­36.
Borum, R. (2014). Psychological vulnerabilities and propensities for involvement in violent extremism. Behavioral Sciences
& the Law, 32(3), 286–­305. https://doi.org/10.1002/bsl.2110
Borum, R., & Patterson, T. D. (2019). Juvenile radicalization into violent extremism: Investigative and research perspectives.
Journal of the American Academy of Child & Adolescent Psychiatry, 58(12), 1142–­1148. https://doi.org/10.1016/j.
jaac.2019.07.932
Bundesministerium des Innern, für Bau und Heimat. (2013). Verfassungsschutzbericht 2018. Retrieved from https://www.
verfa​ssung​sschu​tz.de/de/downl​oad-­manager
Card, N. A. (2012). Applied meta-­analysis for social science research. New York, NY: Guilford Press.
Christmann, K. (2012). Preventing religious radicalisation and violent extremism: A systematic review of the research evi-
dence. Youth Justice Board for England and Wales. Retrieved from https://assets.publi​shing.servi​ce.gov.uk/gover​nment/​
uploa​ds/syste​m/uploa​ds/attac​hment_data/file/39603​0/preve​nting​-­viole​nt-­extre​mism-­syste​matic​-­review.pdf
Dahl, V. (2014). Breaking the law: Adolescents’ involvement in illegal political activity. Retrieved from http://urn.kb.se/resol​
ve?urn=urn:nbn:se:oru:diva-­33225
Dahl, V. (2020). Adolescents’ development of approval of violent political action: Evidence from six waves of longitudinal
data. Acta Politica, 55(4), 538–­559. https://doi.org/10.1057/s4126​9-­019-­00130​-­x
Desmarais, S. L., Simons-­Rudolph, J., Brugh, C. S., Schilling, E., & Hoggan, C. (2017). The state of scientific knowledge
regarding factors associated with terrorism. Journal of Threat Assessment and Management, 4(4), 180–­209. https://doi.
org/10.1037/tam00​00090
Doosje, B., Moghaddam, F. M., Kruglanski, A. W., de Wolf, A., Mann, L., & Feddes, A. R. (2016). Terrorism, radicalization
and de-­radicalization. Current Opinion in Psychology, 11, 79–­84. https://doi.org/10.1016/j.copsyc.2016.06.008
Feddes, A. R., Mann, L., & Doosje, B. (2015). Increasing self-­esteem and empathy to prevent violent radicalization: A lon-
gitudinal quantitative evaluation of a resilience training focused on adolescents with a dual identity. Journal of Applied
Social Psychology, 45(7), 400–­411. https://doi.org/10.1111/jasp.12307
Federal Bureau of Investigation. (2019). Incidents and offenses. Retrieved from https://ucr.fbi.gov/hate-­crime/​2018/topic​
-­pages/​incid​ents-­and-­offen​ses.pdf
Fisher, Z., & Tipton, E. (2015). Robumeta: An R-­package for robust variance estimation in meta-­analysis. Retrieved from
https://arxiv.org/abs/1503.02220
Frindte, W., Boehnke, K., Kreikenbom, H., & Wagner, W. (2012). Lebenswelten junger Muslime in Deutschland. Berlin:
Bundesministerium des Inneren.
Grossman, D., Khalil, U., & Ray, A. (2019). Terrorism and early childhood health outcomes: Evidence from Pakistan. Social
Science & Medicine, 237, 112453. https://doi.org/10.1016/j.socsc​imed.2019.112453
Hogg, M. A. (2014). From uncertainty to extremism: Social categorization and identity processes. Current Directions in
Psychological Science, 23(5), 338–­342. https://doi.org/10.1177/09637​21414​540168
Institute for Economics and Peace. (2019). Global Terrorism Index 2019: Measuring the impact of terrorism. Retrieved from
https://www.visio​nofhu​manity.org/wp-­conte​nt/uploa​ds/2020/11/GTI-­2019-­web.pdf
Jahnke, S., Abad Borger, K., & Beelmann, A. (2020, September 30). Predictors of political violence outcomes among young
people: A systematic review and meta-­analysis. Open Science Framework. Retrieved from https://osf.io/w8mbd/
King, M., & Taylor, D. M. (2011). The radicalization of homegrown jihadists: A review of theoretical models and social psy-
chological evidence. Terrorism and Political Violence, 23(4), 602–­622. https://doi.org/10.1080/09546​553.2011.587064
Klandermans, P. G. (2014). Identity politics and politicized identities: Identity processes and the dynamics of protest. Political
Psychology, 35(1), 1–­22. https://doi.org/10.1111/pops.12167
Kruglanski, A. W., Gelfand, M. J., Bélanger, J. J., Sheveland, A., Hetiarachchi, M., & Gunaratna, R. (2014). The psychology
of radicalization and deradicalization: How significance quest impacts violent extremism. Political Psychology, 35(S1),
69–­93. https://doi.org/10.1111/pops.12163
Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-­analysis. Thousand Oaks, CA: Sage.
Lösel, F., King, S., Bender, D., & Jugl, I. (2018). Protective factors against extremism and violent radicalization: A sys-
tematic review of research. International Journal of Developmental Science, 12(1–­2), 89–­102. https://doi.org/10.3233/
DEV-­170241
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
128 Jahnke et al.

Madriaza, P., & Ponsot, A.-­S. (2016). Preventing radicalization: A systematic review. Technical Report from the International
Centre for the Prevention of Crime (ICPC). https://doi.org/10.13140/​RG.2.1.4862.1682
McCauley, C., & Moskalenko, S. (2008). Mechanisms of political radicalization: Pathways toward terrorism. Terrorism and
Political Violence, 20(3), 415–­433. https://doi.org/10.1080/09546​55080​2073367
McCauley, C., & Moskalenko, S. (2017). Understanding political radicalization: The two-­ pyramids model. American
Psychologist, 72(3), 205–­216. https://doi.org/10.1037/amp00​00062
Moghaddam, F. M. (2005). The staircase to terrorism: A psychological exploration. The American Psychologist, 60(2), 161–­
169. https://doi.org/10.1037/0003-­066X.60.2.161
Neumann, P. R. (2013). The trouble with radicalization. International Affairs, 89(4), 873–­ 893. https://doi.
org/10.1111/1468-­2346.12049
Nivette, A., Eisner, M., & Ribeaud, D. (2017). Developmental predictors of violent extremist attitudes: A test of general strain
theory. Journal of Research in Crime and Delinquency, 54(6), 755–­790. https://doi.org/10.1177/00224​27817​699035
Pauwels, L., & Schils, N. (2016). Differential online exposure to extremist content and political violence: Testing the rela-
tive strength of social learning and competing perspectives. Terrorism and Political Violence, 28(1), 1–­29. https://doi.
org/10.1080/09546​553.2013.876414
Rodgers, M. A., & Pustejovsky, J. E. (2019). Evaluating meta-­analytic methods to detect selective reporting in the presence
of dependent effect sizes [Preprint]. MetaArXiv. https://doi.org/10.31222/​osf.io/vqp8u
Sageman, M. (2017). Turning to political violence: The emergence of terrorism. Philadelphia, PA: University of Pennsylvania
Press.
Schmid, A. P. (2004). Frameworks for conceptualising terrorism. Terrorism and Political Violence, 16(2), 197–­221. https://
doi.org/10.1080/09546​55049​0483134
Šerek, J., Machackova, H., & Macek, P. (2018). Who crosses the norms? Predictors of the readiness for non-­normative political
participation among adolescents. Journal of Adolescence, 62, 18–­26. https://doi.org/10.1016/j.adole​scence.2017.11.001
Simon, B., Reichert, F., & Grabow, O. (2013). When dual identity becomes a liability: Identity and political radicalism among
migrants. Psychological Science, 24(3), 251–­257. https://doi.org/10.1177/09567​97612​450889
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. https://doi.org/10.1177/10888​68311​430825
Stephan, W. G., & Renfro, C. L. (2002). The role of threat in intergroup relations. In D. M. Mackie & E. R. Smith (Eds.), From
prejudice to intergroup emotions: Differentiated reactions to social groups (pp. 191–­207). New York, NY: Psychology
Press.
Stephens, W., Sieckelinck, S., & Boutellier, H. (2021). Preventing violent extremism: A review of the literature. Studies in
Conflict & Terrorism, 44(4), 346–­361. https://doi.org/10.1080/10576​10X.2018.1543144
Sterne, J. A. C., Sutton, A. J., Ioannidis, J. P. A., Terrin, N., Jones, D. R., Lau, J., … Higgins, J. P. T. (2011). Recommendations
for examining and interpreting funnel plot asymmetry in meta-­analyses of randomised controlled trials. BMJ, 343,
d4002. https://doi.org/10.1136/bmj.d4002
Tanner-­Smith, E. E., Tipton, E., & Polanin, J. R. (2016). Handling complex meta-­analytic data structures using robust vari-
ance estimates: A tutorial in R. Journal of Developmental and Life-­Course Criminology, 2(1), 85–­112. https://doi.
org/10.1007/s4086​5-­016-­0026-­5
Taylor, L. K., & McKeown, S. (2019). Does violence beget violence? The role of family ethnic socialization and intergroup
bias among youth in a setting of protracted intergroup conflict. International Journal of Behavioral Development, 43(5),
403–­408. https://doi.org/10.1177/01650​25419​844036
Taylor, L. K., Merrilees, C. E., Goeke-­Morey, M. C., Shirlow, P., & Cummings, E. M. (2016). Trajectories of adolescent
aggression and family cohesion: The potential to perpetuate or ameliorate political conflict. Journal of Clinical Child &
Adolescent Psychology, 45(2), 114–­128. https://doi.org/10.1080/15374​416.2014.945213
Tipton, E. (2015). Small sample adjustments for robust variance estimation with meta-­regression. Psychological Methods,
20(3), 375–­393. https://doi.org/10.1037/met00​00011
Vergani, M., Iqbal, M., Ilbahar, E., & Barton, G. (2020). The three Ps of radicalization: Push, pull and personal. A systematic
scoping review of the scientific evidence about radicalization into violent extremism. Studies in Conflict & Terrorism,
43(10), 854. https://doi.org/10.1080/10576​10X.2018.1505686
Verkuyten, M. (2018). Religious fundamentalism and radicalization among Muslim minority youth in Europe. European
Psychologist, 23(1), 21–­31. https://doi.org/10.1027/1016-­9040/a000314
Victoroff, J. (2005). The mind of the terrorist: A review and critique of psychological approaches. Journal of Conflict
Resolution, 49(1), 3–­42. https://doi.org/10.1177/00220​02704​272040
Viechtbauer, W. (2010). Conducting meta-­analyses in R with the metafor package. Journal of Statistical Software, 36(3),
1–­48.
Vorsina, M., Manning, M., Fleming, C. M., Ambrey, C. L., & Smith, C. (2017). The welfare cost of terrorism. Terrorism and
Political Violence, 29(6), 1066–­1086. https://doi.org/10.1080/09546​553.2015.1111207
14679221, 2022, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/pops.12743 by University Of Trnava, Wiley Online Library on [22/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Political Violence Outcomes among Young People 129

Wilson, D. B. (2001). Practical meta-­analysis effect size calculator. Retrieved from http://www.campb​ellco​llabo​ration.org/
escal​c/html/Effec​tSize​Calcu​lator​-­Home.php
Wolfowicz, M., Litmanovitz, Y., Weisburd, D., & Hasisi, B. (2020). A field-­wide systematic review and meta-­analysis of
putative risk and protective factors for radicalization outcomes. Journal of Quantitative Criminology, 36(3), 407–­447.
https://doi.org/10.1007/s1094​0-­019-­09439​-­4

Supporting Information
Additional Supporting Information may be found in the online version of this article at the publish-
er’s web site:
A. Search Terms

B. Index Publications and Additional Sources of Information

C. Overview of All Predictor and Outcome Names and Associated Classifications for Each Sample

D. Overview of Included Samples

E. Funnel Plots

F. Moderator Analyses (Age, Publication Year, Rate of Female Participants, Peer Review Status,
Subordinate Status)

G. Moderation Coefficients for Different Ideologies associated with Political Violence

H. Moderator Analyses of Level of Political Violence

You might also like