Abstract
The present research note studies the interaction between the ability to exercise self-control and extremist moral beliefs with regard to the explanation of violent extremism. Although some evidence exists for the interaction between moral beliefs and self-control in the explanation of adolescent offending, no previous study has studied this interaction effect in a survey of young adults and with regard to politically or religiously motivated violence. This study therefore extends the existing literature by testing a key proposition of Situational Action Theory. We use a large-scale web survey of young adults in Belgium. The results support the hypothesis that the effect of the ability to exercise self-control is conditional upon one’s extremist beliefs. The results are stable across extremism-specific measures of extremist beliefs.
Introduction
One of the most influential criminological theories of the 20th century is without doubt Self-Control Theory, originally developed by Gottfredson and Hirschi (1990) in their seminal book A General Theory of Crime. The central assumption of this theory is that low self-control increases the risk of offending together with other deviant and imprudent behaviors. According to Gottfredson and Hirschi, low self-control is the primary cause of all crimes and analogue acts. Self-control is a major research line, also beyond the realm of criminology (Baumeister, Vohs, & Tice, 2007).
A large number of studies show that low self-control is associated with offending in different samples and in different types of studies as, for example, cross-sectional and longitudinal (e.g., Antonaccio & Tittle, 2008; Burton, Cullen, Evans, Alarid, & Dunaway, 1998; Chapple, 2005; Finkel & Campell, 2001; Longshore, 1998; Ribeaud & Eisner, 2006; Shoepfer & Piquero, 2006; Vazsonyi, Pickering, Junger, & Hessing, 2001; Wikström & Butterworth, 2006; for an overview, see Pratt & Cullen, 2000).
However, though self-control is highly correlated with offending, it is possible that self-control has a different impact on offending for individuals with different characteristics and backgrounds. This issue that self-control, or broader, mechanisms of control may have conditional effects has been recognized before (see Agnew, 2003; Hay & Meldrum, -2015). In their original seminal work, Gottfredson & Hirschi themselves argue that the effect of self-control is only dependent on the opportunities for crime and probably therefore neglected the possibility that there are interaction effects with other risk factors for offending (which are not important as direct mechanisms in their reasoning).
A recently developed cross-level integrative theory of crime causation that explicitly states that the effect of self-control is strongly depending on one’s level of morality is the Situational Action Theory (SAT; for example, Wikström, Oberwittler, Treiber, & Hardie, 2012). SAT argues that all forms of crime (or in the theory’s words rule-breaking) are the outcome of how individuals perceive their alternatives for action (influenced by their morality) and make their choices (influenced by their ability to exert self-control) when confronted with different types of settings. Morality (an overall term for moral values and moral emotions such as shame and guilt) is seen as the main individual level cause of why some people see crime as an action alternative and others do not. Thus, seeing crime as an alternative is assumed to be primarily a question of morality and not of low self-control (Wikström & Treiber, 2007).
According to SAT, crimes are moral actions and therefore should be analyzed and explained as such. Acts of crime are equated with the breaking of moral rules of conduct (defined in law). Moral rules are rules that stipulate what it is right or wrong to do in a particular circumstance. What a theory of crime causation ultimately should explain is therefore why people follow and break moral rules stated in law. In SAT, the ability to exercise self-control is only relevant in individuals who see crime as an action alternative (i.e., have low morality). When individuals who see crime as an action alternative make use of their ability to exercise self-control, their level of offending should decrease depending on their ability to exercise self-control. SAT proposes to make a theoretical distinction between abilities on one hand and the actual use of self-control on the other hand. However, although it definitely deserves merits to distinguish between abilities and the actual making use of self-control, it is a daunting task to translate this difference into empirical research. In the present study, we deal with the traditional concept of self-control (ability) as measured by impulsiveness and thrill-seeking behavior. When we use the term self-control (or low self-control), we thus refer to the ability.
The question whether morality and self-control interact in the explanation of offending has been examined in a number of empirical studies that explicitly tested this hypothesis and were guided by SAT. The results from these studies are mixed and can be categorized into three groups. The first group is based on one study that did not find support of the interaction between morality and self-control (Gallupe & Baron, 2014). This study is based on a rather small street youth sample (N = 300) with a majority of males. The dependent variable used in this study is based on a single item of drug use, which may make it extremely difficult to detect an interaction effect.
The second group includes studies that all found full support for an interaction effect between morality and ability to self-control (Hirtenlehner & Hardie, 2016; Hirtenlehner & Kunz, 2016; Svensson, Pauwels, & Weerman, 2010; Wikström & Svensson, 2010; Pauwels, 2015). One study found a stable interaction effect across subgroups (gender–immigrant background). All five studies that found support for this interaction effect used last-year self-reported offending scales, were based on school-based samples, and targeted youths in early to mid-adolescence. One of the major limitations of these five studies is that they all are based on cross-sectional data, hampering a causal interpretation.
The third group is based on some studies that found partial support for the hypothesis. In one of these studies, support was found for projections of offending (total frequency scale), but this study did not find support for the projections of property and violent offending scales (Antonaccio & Tittle, 2008; Bruinsma, Pauwels, Weerman, & Bernasco, 2015; Eifler, 2015). In another one, support for the hypothesis was found for cross-sectional analyses but not for longitudinal, and in the third study, support was found for the overall sample but not for subgroups. However, these studies have their limitations. First, two studies that have focused on the interaction used intentions to commit crimes as the dependent variable but not actual offending. To say that you have the intentions to commit a crime does not necessarily mean that you actually will do it. Second, all studies are limited in their sampling strategies. Antonaccio and Tittle (2008) pointed out themselves that there are some limitations of their data and that they “cannot be sure of the accuracy of the data” (p. 503), and Shoepfer and Piquero (2006) used a rather small sample of university students.
Apparently, all of these aforementioned partial tests of SAT are based on measures of traditional delinquency (street crimes, that is, things that adolescents usually get involved in). That is a valid criticism of many tests of criminological theories. On one hand, it is understandable that theories are tested on these kinds of crime, as they are quite common and thus deserve both theoretically and empirically attention. On the other hand, a negative consequence is that other types of rule-breaking are not scrutinized. One of the kinds of rule-breaking that is less studied is politically and religiously motivated violence. In the present, post-9/11 context, it remains remarkable to see how few empirical studies of violent extremism have been conducted. This is a valid criticism of studies of violent extremism that is recognized in studies on political and religious violence. It has been stated before that studies of political violence can learn from criminological studies (Decker & Pyrooz, 2015; De Waele & Pauwels, 2016).
This Study
Against this background, this research article contributes to the literature in three ways: (a) It provides additional insights into the interaction between morality and the ability to exercise self-control, (b) it applies this key idea of a recent criminological theory on politically or religiously motivated violence, and (c) it provides preliminary insight into the robustness of this interaction effect by simultaneously studying the interaction between morality and self-control for three different scales that tap into the moral support for political violence: moral support for right-wing extremist violence, left-wing extremist violence, and religiously motivated violence. We submit that only a few studies apply criminological theories to violent extremism, while it has been argued previously that SAT is a fruitful approach to explain also acts of violent extremism (Bouhana & Wikström, 2008, 2010, 2011). Against this background, we hypothesize that there is an interaction effect between morality and self-control with regard to the explanation of politically and/or religiously motivated violence. More precisely, we expect that self-control has a strong effect on offending for individuals with low levels of morality and has no effect in individuals who do not see politically or religiously motivated violence as an action alternative.
This study is as far as we know the first study that tests the interaction between the ability to exercise self-control and moral beliefs with regard to three subscales that measure moral beliefs toward the use of violence by extremist groups. The subscales refer to the use of violence by (a) nationalist/separatist, (b) left-wing, and (c) religiously motivated extremist groups to gain political goals. The effects are studied on self-reported violent extremism (violence toward persons and property). Data are used from a large Belgian survey of young adults (Pauwels & Schils, 2016; Schils & Pauwels, 2014, 2016).
Data
Data were collected (a) through a classic paper-and-pencil survey of pupils in the third cycle of secondary education in Antwerp and Liège (aged 16 to 18) and (b) through a web survey of young adults—both students and young adults who have left school (aged 16 to 24). The paper-and-pencil survey was restricted to the cities of Liège and Antwerp for practical reasons: Liège and Antwerp are, except for Brussels, the two largest cities of Belgium (+100,000 inhabitants). All schools in the third cycle of the secondary education in Antwerp and Liège were contacted and invited to participate in the study. A total of 34 schools in Antwerp and 32 schools in Liège were contacted. 1 The web survey consisted of a self-administered questionnaire that is conducted online. Access could be gained through a link to the survey’s web page on Facebook. This survey mode requires almost no organization, does not cause disruption of working time, and leaves the decision to participate entirely to the students. As the web survey was meant to reach both students and nonstudying young adults, posters were placed visibly in different strategic places that attract a high number of the target population, such as popular pubs and bars. In addition, flyers were distributed in buildings of virtually all faculties of the university and university colleges in Antwerp, Ghent, Louvain-la-Neuve, and Liège, and pamphlets were distributed among the students. The central faculties and administrational services for students of all universities and university colleges of Flanders, Liège, and Louvain-la-Neuve were sent an email invitation with a request to circulate the web link to the questionnaire’s Facebook page. This method proved to be most effective. Many additional organizations, associations, and local youth clubs were contacted with a request to distribute the survey to their members, to reach youth who are no longer in school. This last tactic was particularly effective in Wallonia, where 32 youth associations were contacted for this purpose.
Although the possibility of distributing the questionnaire via online platforms or mailing lists has significantly contributed to the survey response, one must still have some reservations with regard to the response. Although web surveys seem to be increasingly popular in social science research, there are some questions left with regard to the systematic bias that might result from exclusively using the world wide web as a sample frame. We acknowledge that the researcher cannot completely monitor the processes of response selection, and we must admit that we cannot verify the conditions under which the questionnaire is completed (the presence of others, anonymity, etc.). In addition, the initiative to participate in the survey is entirely left to the respondent. The impossibility of monitoring, response selection, self-selection, and undercoverage (Internet availability) are important drawbacks. It should, however, be mentioned that these issues (preparedness to answer survey questions, willingness to report) are central to the more traditional survey modes as well. It is probably fair to state that the web survey may contribute more to explanatory research (studies of the causes and correlates) than to prevalence studies (studies that try to gain insight into the prevalence of attitudes and behavior). Still, web surveys are increasingly accepted as a valid and reliable tool of measuring self-reported delinquency with their data quality measuring up that of paper-and-pencil surveys. The fact that the questionnaire web page was visible on Facebook meant that a high number of respondents could be reached in a very short time. The web survey was online between September 2012 and December 2012, and the response was huge, with 3,653 respondents in Flanders and 2,367 respondents in Wallonia, making a total of 6,020 respondents.
Measurement of Constructs
Dependent variable
Political violence was measured using two subscales: one that measures acts of violence committed toward persons and another that measures acts of violence committed toward property (damaging or destroying things). Self-reported political violence toward property was measured by asking the respondents whether they had ever “written on a wall a political message or politically oriented graffiti,” “participated in a political action that was not allowed,” “thrown stones at the police during a political protest or demonstration,” “destroyed something on the streets because of your political or religious belief,” “damaged someone’s property,” or “set something on fire.” The scale is derived from a Belgian study of violent extremism. Alpha for this scale is .87. Self-reported political violence toward persons was measured by asking the respondents whether they had ever “fought with someone because of political or religious belief,” “threatened someone on the Internet because of political or religious belief,” “threatened someone on the streets because of political or religious belief,” or “hit a foreigner.” Alpha for the political violence toward persons is .87. Alpha for the combined scale is .77. Both subscales are highly correlated (r = .69). We used a 4-point scale (never, once, 2-3 times, and 4 times or more).
Independent variables
Three different measures that tap into extremist moral beliefs have been used in this study. Religious extremist beliefs (5-point scale): “I endorse that some religious fundamentalists use violence against the people who have the power in Belgium,” “I endorse religious fundamentalists who disrupt the order,” and “I endorse religious fundamentalists who use violence against others.” Cronbach’s alpha is .91. Left-wing extremist beliefs (5-point scale): “I endorse that some anti-globalists use violence against the people who have the power in Belgium,” “I endorse anti-globalists who disrupt the order,” and “I endorse religious fundamentalists who use violence against others.” Cronbach’s alpha is .83. Nationalist–separatist extremist beliefs (5-point scale): “I endorse that some nationalist/separatist extremists use violence against the people who have the power in Belgium,” “I endorse nationalist/separatist extremists who disrupt the order,” and “I endorse nationalist/separatist extremists who use violence against others.” Cronbach’s alpha is .89. Inspiration for these scales is found in the work of Doosje, Loseman, and Bos (2013). Impulsiveness (5-point scale): “I always say what I think, even if it is not nice or smart”; “If I want something, I do it immediately”; “I lose my temper easily”; and “When I am really angry, other people better stay away from me.” Cronbach’s alpha is .63. Thrill-seeking behavior (5-point scale): “I sometimes find it exciting to do things that could be dangerous,” “I often do things without thinking of the consequences,” “Sometimes I will take a risk just for the fun of it.” Cronbach’s alpha is 0.73. The impulsivity and thrill-seeking behavior scales were correlated (r = .36), and both scales were merged into an overall low self-control scale. The items were derived from the well-known Grasmick, Tittle, Bursik, and Arneklev (1993) scale. The scale was reversed to demonstrate the effect of the ability to exercise self-control.
Gender is coded 1 for males and 0 for females. Immigrant background is coded 1 when the respondent and both parents are of Belgian descent and 0 if at least one of the parents is not of Belgian descent.
Analytic Strategy
Violent extremism is measured in terms of a total frequency scale in this study. A negative binomial regression (or other count-data model) is usually considered to be the appropriate analytical procedure for this kind of data. But although a negative binomial regression best matches the skewed and discrete nature of the employed response variable, our focus on interactive relationships suggests that we should not use this analytical technique. Recent advances in the statistical analysis of interaction effects in highly skewed data have demonstrated that the established practice of testing interaction effects by adding product terms to the model equations, which works well in the context of ordinary least squares (OLS) regression, cannot be applied to nonlinear models. Sometimes, interaction effects are misinterpreted. To control for spurious moderator effects, we rely on a modeling strategy proposed by Lubinski and Humphreys (1990) which has been used in several criminological inquiries (Hirtenlehner, Pauwels, & Mesko, 2015). Lubinski and Humphreys recommend introducing the quadratic terms of the predictor variables involved in the interaction into the model equations—in addition to the product term between extremist moral beliefs and self-control and the main effects of extremist moral beliefs and self-control. The quadratic terms of both predictors are a means of controlling for nonlinearity of the relationship of the predictors to the outcome variable. This allows for an estimation of the interaction effect that is not obscured by artifacts related to the nonnormality of the outcome variable. Additional analyses reveal that the interaction coefficient survived this critical test, but the magnitude of the interaction term decreased. To avoid complexity, only linear regression models were fitted using SPSS 23 (generalized linear modeling). The scale variables were standardized before entering the equation and computing the multiplicative interaction terms (Jaccard, Turrisi, & Wan, 1990). All metric scale scores have been standardized before entering the equation. The regression models in the table show the main effects and the multiplicative term between the two standardized metric scales (Extremist Beliefs × Self-Control). The dummy variables were also standardized. In addition, the results are visualized. We present the effect of self-control on self-reported political violence for three groups based on their score on extremist beliefs. The three scales of extremist beliefs are categorized based on the mean and standard deviation (−1 = low moral support for violent extremism = <1 SD, 0 = medium moral support for violent extremism = between −1 SD and +1 SD, 1 = high levels of moral support for violent extremism = +1 SD). This has been a common practice to demonstrate the differential effect of one metric dependent variable by subgroups of another variable (see, for example, Svensson & Pauwels, 2010; Wikström & Butterworth, 2006).
Results
Table 1 shows the net effect of the right-wing extremist beliefs scale (β = .20), the capacity to exercise self-control (β = −.34), and the multiplicative interaction term (β = −.32). Both control variables (male and immigrant background) have a significant effect. The positive effect of immigrant background on nationalist–separatist extremism may be odd at first sight, but nationalist/separatist extremism was defined as groups that fight for independence that have an extreme nationalistic character. Extreme nationalism is not exclusively related to native Belgians but also exists in immigrant groups (e.g., Grey Wolves—Turkish nationalists, Nationalist Youths from Eastern European countries, Separatist Kurds, and so on). 2
Generalized Linear Regression of Violent Extremism on the Ability to Exercise Self-Control, Nationalist/Separatist Extremist Beliefs, and Their Interaction.
Note. AIC = Akaike information criterion.
p < .05. **p < .01. ***p < .001.
In Figure 1, the relationship between self-control and self-reported political violence is demonstrated for the three categories of right-wing extremist beliefs. The figure clearly shows that the regression line becomes steeper when extremist beliefs increase. This is partially in line with SAT. The regression line is, however, not flat for the group that has the lowest score on right-wing extremist beliefs. The explained variance is derived from an OLS model and is modest: 9.60%.

Interaction between the ability to exercise self-control and nationalist extremist beliefs.
Table 2 shows the relationship between the overall left-wing extremist belief extremism scale (β = .31) and self-control (β = −.34) and the interaction term (β = −.38). The model shows controlling for gender and immigrant background, the main effects and interaction term have significant effects on self-reported violent extremism. The explained variance is 11.5%.
Generalized Linear Regression of Violent Extremism on the Ability to Exercise Self-Control, Left-Wing Extremist Beliefs, and Their Interaction.
Note. AIC = Akaike information criterion.
p < .05. **p < .01. ***p < .001.
In Figure 2, the relationship between self-control and self-reported political violence is demonstrated for the three categories of left-wing extremist beliefs. The figure clearly shows that the regression line becomes steeper when extremist beliefs increase. Again, this is partially in line with SAT. The effect of low self-control is much more pronounced in the group that has the highest scores on left-wing extremist beliefs.

Interaction between the ability to exercise self-control and left-wing extremist beliefs.
Table 3 shows the relationship between religious extremist belief (β = .25) and the capacity to exercise self-control (β = −.32) and the interaction term (β = −.36). The model shows controlling for gender and immigrant background, the main effects and interaction term have significant effects on self-reported violent extremism. The explained variance is also rather low (10.9%).
Generalized Linear Regression of Violent Extremism on the Ability to Exercise Self-Control, Religious Extremist Beliefs, and Their Interaction.
Note. AIC = Akaike information criterion.
p < .05. **p < .01. ***p < .001.
In Figure 3, the relationship between self-control and self-reported political violence is demonstrated for the three categories of religious extremist beliefs. The figure clearly shows that the regression line becomes steeper when extremist beliefs increase. Also with regard to religious extremist beliefs, this finding is partially in line with SAT. The effect of self-control is much more pronounced in the group that has the highest scores on left-wing extremist beliefs.

Interaction between the ability to exercise self-control and religious extremist beliefs.
Conclusion and Discussion
The main objective of this study was to examine whether self-control and extremist moral beliefs interact in the explanation of violent extremism. The study revealed main and interactive effects. The main effects are not surprising and could be expected from Self-Control Theory (impulse control makes it easier to respond to provocations, in casu with violent extremism). The effect of thrill-seeking in the explanation of violent extremism has been underscored in literature reviews and empirical studies, but mainly with regard to right-wing extremism (Bjørgo, 1997; Borum, 2007; Horgan, 2004).
The question of whether extremist moral beliefs and the ability to exercise self-control may interact in the explanation of individual differences in violent extremism has not been examined in empirical research on violent extremism to date. We assumed that the ability to exercise self-control would have a stronger effect on violent extremism for individuals with high levels of extremist beliefs, whereas for individuals who lack extremist beliefs, the effect of self-control would be weaker. To test these hypotheses, data were drawn from Belgium’s largest survey of political violence. The results of this study show that both extremist moral beliefs and self-control are related with violent extremism. That both of these factors are related with offending is something that it is compatible with some previous research that tested this key assumption of SAT. Furthermore, the main finding of this study is that there is a statistical interaction between extremist moral beliefs and self-control in the explanation of violent extremism with regard to three different measures of extremist moral beliefs. Although this study replicates an interaction effect that is central to SAT, there still remain some questions of interpretation. If one is willing to accept the thesis that individuals that have low moral beliefs (i.e., those that share extremist beliefs) are people who see violent extremism as an alternative, then the findings are in line with SAT. However, SAT is much more complex with regard to the conditional effect of controls. According to SAT, and in line with dual process theory, and the model of frame selection (Kroneberg, Heintze, & Mehlkop, 2010), individuals either deliberately or habitually choose an action alternative. According to SAT, self-control is irrelevant to those individuals that habitually respond to a provocation or temptation. This hypothesis is extremely hard to test.
The finding we presented in this study has, however, been done in studies of adolescent (violent) offending. Thus, we found an interesting similarity between explaining crime and explaining violent extremism. This study raises a number of questions regarding the applicability of a general model of offending on violent extremism. It is interesting to see the similarities, but that does not automatically mean that one model can be applied to the explanation of violent extremism. We think there are important differences with regard to the causes of the causes, that is, the mechanisms that explain how an individual adopts extremist moral beliefs (Schils & Pauwels, 2016). However, the logic of general models like SAT seems to be applicable to political violence. We hope that this empirical study has demonstrated that there is a lot to gain by studying a serious threat to our democratic society from an integrative theoretical perspective. Future studies should focus on empirical tests of the mechanisms that explain why some people tend to morally support the use of violence by terrorist/extremist groups (e.g., see Pauwels & De Waele, 2014). In doing so, several lines of inquiry can be combined to increase our understanding of violent extremism.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
