Motives for Punishing Powerful Vs. Powerless Offenders: The Mediating Role of Demonization

ABSTRACT

In the present research, we examine how power and group membership of an offender influence observers’ motives for punishment. As compared to powerless offenders, powerful offenders should elicit a stronger motivation of an observer to incapacitate them and protect society (i.e., utilitarian punishment motivation). Moreover, demonization of the offender (e.g., perceiving the offender as evil) should mediate the effect of power on punishing motives. Finally, we investigated whether group membership of an offender would moderate the effects of power on punishing motives. In three studies, we manipulated an offender’s power (high, low) and group membership (ingroup, outgroup, and – in Study 1 – ambiguous). Supporting our hypotheses, all three studies revealed that powerful offenders triggered stronger utilitarian punishment motivation as opposed to powerless offenders, while demonization of the offender mediated this effect. Moreover, Studies 1 and 2 showed that powerless offenders triggered stronger restorative punishment motivation as opposed to powerful offenders while low demonization of the offender mediated this effect. Contrary to our expectations, however, group membership did not moderate the effect of power on observer’s punishing motives. Theoretical and practical implications of these findings are discussed.

KEYWORDS:

• Power
• group membership
• utilitarian and restorative motives for punishment
• demonization

Power is traditionally defined as the ability to control resources and determine other people’s outcomes (Galinsky et al., 2015). Despite some positive effects of power on social judgment and moral behavior in asymmetric relationships (Overbeck & Park, 2001, 2006, see also, Fleischmann et al., 2019; Fousiani et al., 2021), power appears to have a direct causal impact on many forms of unethical behavior (Boles et al., 2000; Gruenfeld et al., 2008; Kipnis, 1972; Lammers et al., 2010; Pitesa & Thau, 2013) and utilitarian decision-making (Côté et al., 2013; Fleischmann et al., 2019): Powerholders are seen as more capable of causing harm given their extensive resource control (Fiske, 1993; Magee & Galinsky, 2008) and their ability to exert influence and produce intended effects (see, Weber, 1946). Although prior research has already shown that powerholders’ unethical behavior can be appalling and trigger strong punitive reactions (e.g., more severe punishment; Bowles & Gelfand, 2010; Fragale et al., 2009; Karelaia & Keck, 2013), the literature on how power of the offender shapes the motives for which observers seek punishment is considerably limited (Fousiani & van Prooijen, 2022a). Although people have a general need to punish with the aim to restore a sense of justice and fairness (Wenzel & Thielmann, 2006) they might also punish offenders with the ultimate goal to stop crime and protect society or to help offenders learn from their mistakes and reintegrate into society. Whether people apply these motives differently depending on the offender’s power position is yet unknown, however. In the present paper, we investigated an observer’s underlying motives for punishing powerful vs. powerless offenders and the explanatory mechanisms that drive these effects. This contribution is both conceptually and practically interesting as it points out biases in the observers’ motives to assign punishments based on offenders’ characteristics (e.g., power) even if those characteristics are not directly relevant to the inflicted harm. Importantly, this study is not only informative to justice decision-makers who need to be aware of this kind of biases but also to HR practitioners who need to take into account the influence of power and hierarchy-related characteristics of employees on the assessment of their performance and overall behavior.

Power-holders – due to their high standing, influence, and access to resources (Fiske, 1993; Magee & Galinsky, 2008; Weber, 1946) – are assigned with high responsibility to behave with integrity and to take care of the members of their group (Sassenberg et al., 2012, 2014; Scholl et al., 2018). Subsequently, when power-holders abuse their power, they are attributed stronger intentionality for the harm they have inflicted, are seen as highly self-concerned (Fragale et al., 2009) and are perceived as less moral than their powerless counterparts (Kakkar et al., 2020). Apparently offender’s power leads to a more biased justice reaction on the part of an observer, as observers recommend stronger punishments against powerful as opposed to powerless offenders (Fragale et al., 2009). Extending the prior literature, in this study we claim that as compared to powerless offenders, powerful offenders will elicit stronger motivation of an observer to incapacitate them and protect society. In contrast, as compared with powerful offenders, powerless offenders will trigger stronger motivation of an observer to help them improve and reintegrate into society. Furthermore, we argue that observer’s motives to incapacitate power-holders will be explained through a perception of power-holders as inherently evil beings who are intrinsically motivated to cause harm. Finally, we also examine the role of identity concerns of punishment in the context of powerful offenders, and investigate if people’s motives to punish powerful vs. powerless offenders will vary as a function of their group membership (ingroup vs. outgroup).

Motives for punishment

The literature identifies three distinct types of motives for punishment, namely utilitarian (Bentham, 1789), retributive (Kant, 1797), and restorative motives (De Beaumont & de Tocqueville, 1833; Saleilles, 1898). The main goal of utilitarian motives is to minimize the likelihood that an offense will transpire in the future and to reduce suffering in society via zero-tolerance punishments (see, Carlsmith & Darley, 2008; J.W. Van Prooijen, 2018; Nagin, 1998). Accordingly, utilitarian punishments aim at controlling the behavior of an offender by incapacitating them and by deterring future crimes (Carlsmith & Darley, 2008). Instead, the objective of retributive motives is not to prevent future offenses per se, but to pay back offenders’ past behavior proportionally for what they did (Goldberg et al., 1999; see also, J.W. Van Prooijen, 2018). Retributive justice, in other words, aims at restoring a moral balance by giving an offender their just deserts (Carlsmith & Darley, 2008; Gromet & Darley, 2009). Finally, people may punish harm-doers with restorative motives (De Beaumont & de Tocqueville, 1833; Saleilles, 1898). Restorative motives aim at helping an offender recognize the harm they have inflicted, stimulating an apology to the victim and repairing the relationship between the victim and the offender (Zehr, 1997).

Contrary to retributive motives, which focus on punishing past harm-doing, utilitarian and restorative motives both look forward and aim at stopping future crime. According to the literature people display a stronger tendency to punish offenders based on retributive rather than utilitarian or restorative motives. The preference for retributive punishments is commonly referred to as the “intuitive retributivism hypothesis” and underscores people’s need to punish past immoralities than to prevent future immoralities (Wenzel & Thielmann, 2006; see also, Carlsmith & Darley, 2008; Carlsmith et al., 2002; Darley & Pittman, 2003). Interestingly, in a recent study (Fousiani & Van Prooijen, 2022b) researchers found that power of a suspect of crime, whose guilt is uncertain, did not influence observers’ retributive motives for punishment. Accordingly, in this contribution, although we measure retributive motives and report the relevant statistics, we do not focus on how power influences people’s intuitive reaction to punish past harm-doing (i.e., retributive motives) as punishing past-harm-doing does not necessarily stop future crime. We claim that in the context of power, where valued resources come into play, future-oriented motives, which are best able to stop harm-doing are more relevant and more likely to be influenced by offender’s power. We therefore state hypotheses only for future-oriented punishing motives, namely utilitarian and restorative motives.

Power and motives for punishment

Given that power-holders have control over resources and other individuals’ outcomes (Fiske, 1993), power possession appears to be tightly linked to corruption, cheating, dishonesty and many other forms of unethical behavior (Blader & Yap, 2016; Boles et al., 2000; Gruenfeld et al., 2008; Kipnis, 1972; Piff et al., 2012). In other words, power-holders can abuse their control over resources to harm the community. The concern that power holders are both more capable and more likely to cause harm is associated with a utilitarian motive for punishment to protect society (Taylor et al., 1979; Thomas & Foster, 1975). For instance, prior research has shown that powerful offenders are punished more (as compared to powerless offenders; Bowles & Gelfand, 2010; Fragale et al., 2009; Kakkar et al., 2020; Karelaia & Keck, 2013) because they are seen as having a stronger intention to harm others (Dubois et al., 2015; Fragale et al., 2009; Galinsky et al., 2015; Lammers et al., 2015). Interestingly, Fousiani and van Prooijen (2022a) manipulated offender status as power (i.e., dominance) vs. prestige (i.e., recognition and admiration that an individual has in the eyes of an observer) and found that utilitarian motives drive punishment of an offender, but only if observers perceive the status of an offender as power rather than prestige (cf., Blader & Chen, 2012). Moreover, prior research has studied the effects of power of a suspect whose guilt is uncertain (Fousiani & Van Prooijen, 2022b) on motives for punishing offenders. Yet, the psychological processes underlying punitive responses differ between guilty offenders versus suspects whose guilt is uncertain, however (J. W. Van Prooijen, 2006), and no study has yet investigated the motives that drive punishment of a powerful as opposed to powerless offender whose guilt is beyond any doubt.

Based on the above theorizing, we expect that observers’ punishment motives toward powerful offenders involve incapacitating them to better protect society. We thus hypothesize that observers assign stronger utilitarian punishments to high power as opposed to low power offenders. Alternatively put, offenders low in power should trigger a decreased need for utilitarian punishment as their limited access to resources makes them less capable of harm, and loom less threatening and immoral, as compared to their powerful counterparts (Hypothesis 1). Although not a core hypothesis of this study, we also tested the idea that observers may be more willing to reintegrate low as opposed to high power offenders into society via restorative punishments as such offenders are less capable of causing severe harm due to their decreased access to resources. Differently put, although powerless offenders would be seen as a relatively smaller threat to society, observers would still want them to become better members of society; This might happen through restorative practices.

Although people prefer to punish past immoralities (see intuitive retributivism hypothesis; Wenzel & Thielmann, 2006; see also, Carlsmith & Darley, 2008; Carlsmith et al., 2002; Darley & Pittman, 2003), and thus display stronger retributive than utilitarian or restorative motives, there is no strong empirical evidence showing that power of an offender would influence retributive motives (see Fousiani & Van Prooijen, 2022b). After all, retributive motives are primarily based on perceiver’s intuitive desire to restore a sense of justice (“just deserts”; Carlsmith & Darley, 2008; Carlsmith et al., 2002), and also toward low-power offenders people are likely to seek justice and desire a punishment that is proportionate to the harm done (even when the absolute level of harm may be lower). Accordingly, we did not state hypotheses about the effects of power on retributive motives. Nevertheless, in order to achieve a complete design, we included this retributive motives in our measures and we report the relevant statistics in the analyses.

The mediating role of demonization of an offender

The above theorizing also suggests that offender’s power elicits a relatively strong belief among observers that a powerful offender is intrinsically evil. Perceived evilness (i.e., demonizing) is strongly associated with behaving immorally, showing decreased concern for others and taking pleasure into hurting others (Berkowitz, 1999; see also, Reicher et al., 2008), being short of uniquely human emotions (Leyens et al., 2000; Li et al., 2014), and being socially isolated (Baumeister, 1997) and having “evil” or “diabolical” character traits (Darley, 1992; Ellard et al., 2002; Van Prooijen & Van de Veer, 2010). Given that powerful offenders are seen as selfish and intentional offenders (Fragale et al., 2009) who lack moral credentials (Kakkar et al., 2020) they would also be seen as evil beings who enjoy breaking the rules. Such intention to commit harm and take pleasure into hurting others is more likely attributed to high-power offenders, who are held against relatively high moral standards to use resources prosocially, behave with integrity, and meet moral standards (Sassenberg et al., 2012, 2014). Importantly, what kind of characteristics are attributed to offenders shapes people’s punitive reactions to them (Cullen et al., 1985). Accordingly, a powerful offender should elicit stronger demonizing beliefs among observers than a powerless offender (Hypothesis 2). Relatedly, we hypothesized that high demonization of the offender should mediate the effect of (high) power on utilitarian punitive motives (Hypothesis 3). For exploratory purposes, we further tested the idea that low demonization should mediate the effect of (low) power on restorative motives.

The moderating role of an offender’s group membership

While the above line of reasoning highlights people’s motives to incapacitate offenders for instrumental concerns (i.e., punishing offenders that are more capable of causing harm), people also often punish offenders for identity concerns (i.e., protect group values and group image when threatened). People are more likely to forgive offenders who come from their own group than offenders who come from an outgroup (Otten, 2009), as they often treat ingroup members more favorably than outgroup ones (see also, ingroup favoritism for the differential treatment of ingroup vs. outgroup members; Taylor & Doria, 1981). Considering that utilitarian motives for punishment are exclusion-oriented strategies (see, Carlsmith & Darley, 2008) and restorative motives (Lacey & Pickard, 2015; Zehr, 1997) are inclusion-orientated (see also, Fousiani et al., 2019), one would expect that people display stronger utilitarian motives for punishing outgroup offenders and stronger restorative motives for punishing ingroup offenders. Nevertheless, when ingroup offenders behave immorally, they are regarded as undermining social cohesion in a community (see, Durkheim, 1964; Tyler, 1997; Vidmar, 2000) and as threatening the image of the ingroup (Brambilla et al., 2013; Pagliaro et al., 2013) since morality is a highly valued trait (Haslam, 2015) and people want their group to be seen as moral (Ellemers & Van den Bos, 2012; Leach et al., 2007).

According to this line of reasoning, research on the “black sheep effect” (BSE; Marques & Yzerbyt, 1988; Marques et al., 1988) shows that people treat ingroup transgressors more negatively than outgroup transgressors. Therefore, in their effort to restore the image of the group, people socially exclude deviant ingroup members (Marques & Yzerbyt, 1988; Marques et al., 1988). Consistent with this theorizing, Fousiani et al. (2019) found that observers assign stronger utilitarian than restorative or retributive punishments toward ingroup as opposed to outgroup offenders. For this reason, restorative motives are not as effective at enabling members of a group to manage the discomfort that was caused by the harm to the moral image of the group (Fousiani et al., 2019).

Based on the above line of reasoning, we expect punishment of powerful vs. powerless offenders to vary as a function of their group membership. Immorality enacted by powerful members of the ingroup is viewed as a particularly strong threat to one’s group identity, given that ingroup power-holders are often seen as group representatives (Tyler & Lind, 1992). We therefore predicted that high power as opposed to low power offenders should trigger more utilitarian motives for punishment, especially when those offenders come from the ingroup than an outgroup (Hypothesis 4). Yet, although immorality enacted by powerless outgroup members is less threatening to one’s group values as those members are not core members of the group, observers might want to punish these offenders with the ultimate goal to make them correct their behavior and become better citizens in society.¹ Accordingly, we tested the idea that observers display more restorative punishing motives toward powerless offenders especially when those offenders come from an outgroup rather than the ingroup.

Study 1

Study 1 investigated the effect of power (high vs. low) of an offender on observers’ punishment motives and demonizing beliefs about that offender, moderated by the group identity of the offender (ingroup vs. outgroup). Besides ingroup vs. outgroup conditions we included one more level in a more exploratory fashion: the ambiguous group membership.² Although not a core hypothesis in the present research, we expected observers to treat an offender with an ambiguous group membership similarly as an outgroup offender (see ingroup over-exclusion effect; Yzerbyt et al., 1995). The offense that we focused on in Study 1 was tax-evasion.

Methods

Participants

Three hundred and fifty British participants (241 females and 109 males; M_age = 37.89, SD = 9.73) living in the United Kingdom took part in this study. According to an a priori power analysis with G*power, based on a Multivariate Analysis of Variance (MANOVA), for our design a number of 270 participants was required in order to achieve a medium effect size (f = .25) and power .80%.³ The study took place online via Prolific and lasted 10 minutes approximately. Non-British participants could not take part in the survey. No participants were excluded from the study. all participants were paid £1.00 approximately.

Experimental design and procedure

We manipulated offender’s power in vignettes. Similar with Fousiani & Van Prooijen, 2022b, participants read a scenario which presented the main character (i.e., offender) as a powerful manager in the finance department of a large company and making decisions that his employees have to follow (high power) vs. a powerless bookkeeper in the finance department who has to follow his manager’s decisions (low power) (see online supplemental material for the complete manipulations). As a manipulation check, we asked participants to indicate the job role of the offender as either a “powerful manager,” a “powerless employee” or “other.” We also manipulated offender’s group membership such that the offender was born and raised in the UK by his British parents (in group – named Harry Smith), born and raised in India by his Indian parents (outgroup – named Rahul Acharya), or born and raised in the UK by his British mother and Indian father (ambiguous group membership – named Harry Acharya). As a manipulation check, we asked participants to indicate the ethnic background of the offender as either “British,” “Indian,” “half British/half Indian,” or “other.” The offender was presented as having defrauded the British authorities of a large amount of money in unpaid taxes (see online supplemental material for the full vignettes on Open Science Framework). Participants were randomly assigned to one of the six experimental conditions. After reading the vignettes, participants filled out the measures.

Measures

Demonization

To assess demonization of the offender we used the 5-item demonization scale by Fousiani and Van Prooijen (2019) and Van Prooijen and Van de Veer (2010). The scale was adjusted to the specifics of this study (e.g., “this act was caused entirely by the offender’s evilness”; α =.84).

Motives for punishment

We measured the various motives for punishment with the 16-item motives for punishment scale (Fousiani & Demoulin, 2019; Fousiani & van Prooijen, 2022; Fousiani & Van Prooijen, 2019; Fousiani et al., 2019; Fousiani & Van Prooijen, 2022a, 2022b). The scale assessed (a) retributive motives, b) restorative motives, and c) utilitarian motives for punishment and its sub-dimensions.⁴ Cronbach’s alpha was .92 for utilitarian, .92 for retributive, and .79 for restorative motives for punishment (see online supplemental material for the complete scales).⁵

All scales were measured in a 7-point Likert scale (1 = absolutely disagree, 7 = absolutely agree).

Results

Correlations between the study variables, means and standard deviations are presented in Table 1.

Table 1. Pearson correlations coefficients between study variables, means, and standard deviations (Study 1).

	1	2	3	4	M (SD)
1. Demonization	1	.60***	.14**	−.19**	2.99 (1.20)
2. Utilitarian motives		1	.30***	−.27***	3.90 (1.39)
3. Retributive motives			1	.21***	5.00 (1.31)
4. Restorative motives				1	5.90 (1.24)

Note. *p < .05; **p < .01; ***p < .001

Punishment motives

We conducted a 2 (offender’s power: high/low) x 3 (offender’s group membership: ingroup/outgroup/ambiguous) MANOVA with motives for punishment (utilitarian, retributive, restorative) as the main dependent variables. The multivariate effect of offender’s power on motives proved significant F(3,342) = 21.25, p < .001, η² = .157. In line with Hypothesis 1, observers assigned stronger utilitarian punishments against powerful than powerless offenders F(1,344) = 47.03, p < .001, η² = .12 (see, Figure 1). The univariate effect of power on retributive motives was also significant F(1,344) = 16.78, p < .001, η² = .047 and showed that offenders with high power are assigned stronger retributive punishments than offenders low in power. Finally, the univariate main effect of power on restorative motives also proved significant F(1,344) = 14.60, p < .001, η² = .041. People assigned stronger restorative punishments toward powerless than powerful offenders (Means and standard deviations in Table 2, Figure 2). The multivariate effect of group membership was not significant F(6,686) = 1.66, p = .129, η² = .014. Finally, contrary to Hypothesis 4, the interaction effect between power and group membership on punishment motives was not significant, F(6,686) = 1.73, p = .112, η² = .015.

Figure 1. Effects of power on utilitarian motives for punishment (Studies 1–3).

Figure 2. Effects of power on restorative motives for punishment (Studies 1–3).

Table 2. Means and standard deviations for the study variables across experimental conditions (Study 1).

	High power		Low power		Ingroup		Outgroup		Ambiguous
	M	SD	M	SD	M	SD	M	SD	M	SD
Demonization	3.27	1.16	2.67	1.17	3.14	1.26	2.92	1.22	2.89	1.11
Utilitarian motives	4.38	1.28	3.37	1.32	4.16	1.41	3.73	1.41	3.78	1.34
Retributive motives	6.15	1.05	5.62	1.37	5.81	1.33	5.89	1.26	5.99	1.13
Restorative motives	4.75	1.33	5.27	1.25	4.95	1.34	5.05	1.35	5.01	1.27

Note. All ratings were on 7-point scales ranging from 1 = absolutely disagree to 7 = absolutely agree.

The mediating role of demonization

In an ANOVA with demonization as dependent variable, the main effect of power was significant, F(1,344) = 20.00, p < .001, η² = .055. Supporting Hypothesis 2, observers demonized offenders more when they had high rather than low power. However, the main effect of group membership, F(2,344) = 1.26, p = .286, η² = .007, and the interaction effect, F(2,344) = 1.59, p = .205, η² = .009, were not significant.

We then conducted a mediation analysis through bootstrapping. Power of the offender was the independent variable (effect-coded −1: high, 1: low in power), utilitarian motives for punishment was the dependent variable, and demonization was the mediator. The total effect of power on utilitarian motives was significant and negative: Observers reported stronger utilitarian motives for punishing an offender when the offender was high as opposed to low in power. When we added demonization as a mediator, both the indirect and the direct effects were significant, suggesting partial mediation. These results support Hypothesis 3 (see, Tables 3 and 4 for the relevant statistics).

Table 3. Mediation results with power as predictor, demonization as mediator, and utilitarian motives for punishment as dependent variable.

Effects of Power on Utilitarian motives	Total effect	Direct effect (c΄)	Unstandardized paths		Indirect effect
			a	b	Estimate	BCA CI
Study 1	−.50 (.07)***	−.32 (.06)***	−.30 (.06)***	.64 (.05)***	−.19 (.04)	−.28 −.11
Study 2	−.30 (.09)**	−.07 (.07)	−.30 (.08)***	.77 (.05)***	−.23 (.06)	−.36 −.11
Study 3	−.26 (.09)**	−.11 (.07)	−.20 (.08)**	.74 (.05)***	−.15 (.06)	−.27 −.03

Notes. Standard errors in parentheses (bootstrap standard errors for the indirect effect estimate); BCA CI: bias-corrected and accelerated bootstrap confidence interval; paths a and b correspond to the prediction coefficients of the independent variable to the mediator (path a) and of the mediator to the dependent variable (path b), see, Figure 1; *p < .05; **p < .01; *** p <.001.

Table 4. Mediation results with power as predictor, demonization as mediator, and restorative motives for punishment as dependent variable.

Effects of Power on Utilitarian motives	Total effect	Direct effect (c΄)	Unstandardized paths		Indirect effect
			a	b	Estimate	BCA CI
Study 1	.26 (.07)**	.21 (.07)**	−.30 (.06)***	.-17 (.05)**	.05 (.02)	.01 .10
Study 2	.17 (.08)*	.12 (.07)	−.30 (.08)***	−.18 (.06)**	.05 (.03)	.01 .11
Study 3	−.06 (.07)	−.69 (.07)	−.20 (.08)**	−.02 (.05)	.004 (.01)	−.03 .03

Notes. Standard errors in parentheses (bootstrap standard errors for the indirect effect estimate); BCA CI: bias-corrected and accelerated bootstrap confidence interval; paths a and b correspond to the prediction coefficients of the independent variable to the mediator (path a) and of the mediator to the dependent variable (path b), see, Figure 1; *p < .05; **p < .01; ***p < .001.

Finally, we conducted a similar mediation analysis with restorative motives as the dependent variable. The total effect of power on restorative motives proved to be significant: Observers displayed stronger restorative motives for punishing a low as opposed to high power offender. The analyses showed that decreased demonization of low-power offenders partially mediates this effect (see, Tables 3 and 4 for the relevant statistics).

Discussion

Supporting Hypothesis 1, results of Study 1 showed that a powerful as opposed to a powerless offender triggers stronger utilitarian motives for punishment. Interestingly, we also found that observers punish a powerless as opposed to a powerful offender with stronger restorative motives. It is worth mentioning that the effect of (low) power on restorative motives was stronger than the effect of (high) power on utilitarian motives which reveals an observer’s stronger desire to correct a powerless offender’s future behavior via restorative practices than to incapacitate a powerful offender. Moreover, as predicted, people perceive powerful offenders as more intrinsically evil than powerless offenders (Hypothesis 2), which in turn mediated the different punishment motives toward high- and low- power offenders (Hypothesis 3). Finally, the results did not support Hypothesis 4 as group membership of the offender did not moderate the effect of power on utilitarian motives for punishment. An unexpected finding in this study is that observers, apart from utilitarian punishments, also assigned stronger retributive punishments to powerful as opposed to powerless offenders. We discuss these results in more detail in the General Discussion.

Study 2 aims to replicate the Study 1 findings using a non-financial offense, namely sexual harassment. Given the similar effects between outgroup offenders and offenders with an ambiguous membership (in line with the ingroup over-exclusion theory; Leyens & Yzerbyt, 1992), Study 2 only included ingroup versus outgroup conditions. Finally, given that power might make the crime itself seem more severe (e.g., power abuse), we included a measure of perceived severity of the harm as control variable.

Study 2

Methods

Participants

A total of 281 American participants (153 females and 128 males; M_age = 37.89, SD = 11.06) living in the United States took part in this study. A sensitivity power analysis with G*Power revealed that this sample yields 80% power to detect an interaction with an effect size of f = .17. The study took place online via Prolific and lasted 10 minutes approximately. Non-American participants could not take part in the survey. No participants were excluded from the study. All participants were paid £1.00 approximately.

Experimental design and procedure

The offender was presented as working in a large American telecommunication company and having a powerful vs. powerless position. The ingroup offender was American whereas the outgroup offender was Canadian. The offender was presented as having sexually harassed a a woman in her mid-twenties. The victim was not a staff member of the company (see online supplemental material for the full vignettes). As a manipulation check, we asked participants to indicate the power of the offender (“The offender is a powerful/powerless member of the company”) and the offender’s ethnic background (“American”/“Canadian”).

Measures

We used the same scales for the assessment of demonization (α =.84) and motives for punishment (α = .95 for utilitarian, .93 for retributive, and .83 for restorative motives for punishment) after adjusting them to the specifics of this study.

Results

Correlations between the study variables, means and standard deviations are presented in Table 5.

Table 5. Pearson correlations coefficients between study variables, means, and standard deviations (Study 2).

	1	2	3	4	5	M (SD)
1. Demonization	1	.69***	.13*	−.20**	.41***	4.33 (1.33)
2. Utilitarian motives		1	.18**	−.05	.58***	4.30 (1.52)
3. Retributive motives			1	.35***	.24***	5.91 (1.15)
4. Restorative motives				1	.11	5.26 (1.29)

Note. * p < .05; ** p < .01; ***p < .001

Punishment Motives

We conducted a 2 (offender’s power: high/low) x 2 (offender’s group membership: ingroup/outgroup) MANOVA with motives for punishment (utilitarian, retributive, restorative) as main dependent variables. The multivariate main effect of power was significant F(3,275) = 5.28, p = .001, η² = .054. In line with Hypothesis 1, observers assigned stronger utilitarian punishments against powerful than powerless offenders F(1,277) = 11.13, p = .001, η² = .039 (see, Figure 1). The univariate effect of power on retributive motives was not significant F(1,277) = .220, p = .640, η² = .001. However, the univariate effect of power on restorative motives proved significant F (1,277) = 5.18, p = .024, η² = .018 and showed that people assigned stronger restorative punishments toward powerless than toward powerful offenders (Means and standard deviations in Table 6, Figure 2). The multivariate effect of group membership was not significant F(3,275) = 1.57, p = .196, η² = .017. Finally, contrary to Hypothesis 4, the interaction effect between power and group membership on punishment motives was not significant F(3,275) = .47, p = .700, η² = .005.⁶

Table 6. Means and standard deviations for the study variables across experimental conditions (Study 2).

	High power		Low power		Ingroup		Outgroup
	M	SD	M	SD	M	SD	M	SD
Demonization	4.66	1.28	4.05	1.31	4.40	1.30	4.25	1.33
Utilitarian motives	4.62	1.39	4.01	1.58	4.42	1.41	4.16	1.64
Retributive motives	5.87	1.16	5.94	1.14	6.00	1.05	5.81	1.25
Restorative motives	5.07	1.12	5.42	1.20	5.37	1.22	5.13	1.37

Note. All ratings were on 7-point scales ranging from 1 = absolutely disagree to 7 = absolutely agree.

The mediating role of demonization

The main effect of power on demonization was significant, F(1,277) = 14.54, p < .001, η² = .05. Observers demonized offenders more when they had high rather than low power, supporting Hypothesis 2. The main effect of group membership, F(1,277) = .91, p = .34, η² = .003, and the interaction effect, F(1,277) = .47, p = .495, η² = .002, were not significant.

As a next step, we ran a mediation analysis through bootstrapping similar to Study 1. The total effect of power on utilitarian motives was significant and negative: Observers reported stronger utilitarian motives for punishing an offender when the offender had high as opposed to low power. When we added demonization as a mediator, only the indirect but not the direct effects were significant, resulting in full mediation. These results replicate Study 1 and support Hypothesis 3 (see, Tables 3 and 4 for the relevant statistics).

Finally, in a similar a mediation analysis with restorative motives as dependent variable, the total effect of power was significant: Observers displayed stronger restorative motives for punishing a low as opposed to high power offender. The analyses showed that low demonization of low power offenders fully mediates this effect (see, Tables 3 and 4 for the relevant statistics).

Discussion

Using a different type of offense (sexual harassment), these results further suggest that a powerful offender triggers stronger utilitarian punishing motives as compared with a powerless offender, whereas a powerless offender triggers stronger restorative motives as opposed to a powerful one. Moreover, we replicated the mediating effect of demonization of an offender in the relationship between power and utilitarian (as well as restorative motives for punishment). Similar to Study 1, and opposite to Hypothesis 4, results did not support the moderating role of an offender’s group membership in the relationship between power and utilitarian motives.

Study 3 aims to replicate these findings with a different manipulation of group membership. Although in Studies 1 and 2 we manipulated group membership via the ethnic background of an offender, in Study 3 we manipulated the offender’s group membership in a vignette (i.e., participants imagined that they were former members of a company with which they identified strongly). We used the same type of offense (e.g., sexual harassment) as in Study 2.

Study 3

Methods

Participants

A total of 283 American participants (146 females and 137 males; M_age = 49.60, SD = 7.74) living in the United States took part in this study. A sensitivity power analysis with G*power based on a MANOVA revealed that this sample yields 80% power to detect an interaction with an effect size of f = .17. The study took place online via Prolific and lasted 10 minutes approximately. No participants were excluded from the study and all participants were paid £1.00 approximately.

Experimental design and procedure

Participants imagined that they were a former employee of a large telecommunication company (“Tele-Talk”) which consisted of two competitive units (“IntraCOM” and “IntraSOFT”). They further imagined that they had spent 15 productive and nice years of their career in the IntraCOM unit and had created a very strong bonding with the unit and the colleagues. IntraSOFT unit was presented as a rival unit with which there were daily conflicts. The offender was presented as currently working in either the IntraCOM (ingroup) or the IntraSOFT unit (outgroup) of the company and had either a powerful position (i.e., making decisions that influence all members of the company) or powerless position (i.e., having to follow decisions made by powerful members of his department). Similarly with Study 2, the offender was presented as having sexually harassed a woman in her mid-twenties (see online supplemental material for the full vignettes). As a manipulation check, we asked participants to indicate the power of the offender (“The offender is a powerful/powerless member of the company”) and the group membership of the offender (“currently working in the IntraCOM”/“intraSOFT” unit of the Tele-Talk company”). As additional manipulation checks we added the following items: “According to the scenario that you read” …: “ … In which unit of the ‘Tele-Talk’ company were you working before you got retired”?: IntraCOM/IntraSOFT; “ … You have a very strong bonding with the IntraCOM unit of the Tele-Talk company and it seems to be part of your identity” Yes/No.

Measures

We used the same scales for the assessment of demonization (α =.84), motives for punishment (α = .94 for utilitarian, .93 for retributive, and .72 for restorative motives for punishment), and perceived severity of the offense (α = .96) after adjusting them to the specifics of the current study.

Results

Correlations between the study variables, means and standard deviations are presented in Table 7.

Table 7. Pearson correlations coefficients between study variables, means, and standard deviations (Study 3).

	1	2	3	4	5	M (SD)
1. Demonization	1	.68***	.24***	−.01	.46***	4.09 (1.33)
2. Utilitarian motives		1	.22**	−.05	.56***	4.23 (1.48)
3. Retributive motives			1	.28***	.39***	6.11 (1.10)
4. Restorative motives				1	.18**	5.33 (1.21)

Note. * p < .05; ** p < .01; ***p < .001

Punishment motives

We performed a 2 (offender’s power: high/low) x 2 (offender’s group membership: ingroup/outgroup) MANOVA with motives for punishment (utilitarian, retributive, restorative) as the main dependent variables. The multivariate effect of offender’s power on motives proved significant F(3,277) = 4.24, p = .006, η² = .044. In line with Hypothesis 1, observers assigned stronger utilitarian punishments to powerful than powerless offenders, F(1,279) = 9.02, p = .003, η² = .031 (see, Figure 1). The univariate effect of power on retributive motives was also significant, F(1,279) = 5.41, p = .021, η² = .019. Observers assigned stronger retributive punishments to a powerful as opposed to powerless offender. However, the univariate effect of power on restorative motives did not come out significant, F(1,279) = 1.21, p = .273, η² = .004 (Means and standard deviations in Table 8). The multivariate effect of group membership was not significant, F(3,277) = .90, p = .441, η² = .01. Finally, contrary to Hypothesis 4, the interaction effect between power and group membership on punishment motives was not significant, F(3,277) = 1.77, p = .152, η² = .019.⁷

Table 8. Means and standard deviations for the study variables across experimental conditions (Study 3).

	High power		Low power		Ingroup		Outgroup
	M	SD	M	SD	M	SD	M	SD
Demonization	4.27	1.23	3.85	1.41	3.96	1.32	4.20	1.31
Utilitarian motives	4.49	1.46	3.96	1.45	4.26	1.56	4.26	1.41
Retributive motives	6.24	1.00	5.94	1.28	6.04	1.14	6.17	1.06
Restorative motives	5.39	1.17	5.26	1.28	5.24	1.26	5.41	1.18

Note. All ratings were on 7-point scales ranging from 1 = absolutely disagree to 7 = absolutely agree.

The mediating role of demonization

The main effect of power on demonization was again significant, F(1,279) = 6.30, p = .013, η² = .022. Providing support for Hypothesis 2, observers demonized offenders more when they had high rather than low power. The main effect of group membership, F(1,279) = 1.97, p = .161, η² = .007, and the interaction effect, F(1,279) = 1.82, p = .177, η² = .007, were not significant.

We then ran a mediation analysis similar to Studies 1 and 2. The total effect of power on utilitarian motives was significant and negative: Observers reported stronger utilitarian motives for punishing a high- as opposed to low-power offender. When we added demonization as a mediator, only the indirect but not the direct effects were significant, indicating full mediation. These results provide support for Hypothesis 3. Finally, for the indirect effect of power on restorative motives through demonization, both the total and the direct effects did not come out significant (see, Tables 3 and 4 for the relevant statistics).

Discussion

These results further support our hypothesis that power of an offender influences the utilitarian punishment motives of an observer (Hypothesis 1) as well as the demonization of an offender (Hypothesis 2). Importantly, demonization mediates the effect of power of an offender on utilitarian motives of an observer (see Hypothesis 3). This suggests that observers feel that power comes with a certain responsibility to behave with high integrity (Sassenberg et al., 2012, 2014), and failing to meet ethical standards may be linked with demonizing perceptions of offenders (Leyens et al., 2000) which elicits a utilitarian motive to restrict the offender’s capacity to harm.

Unexpectedly, as in Study 1, the effect of power on retributive motives also came out significant. Moreover, unlike Studies 1 and 2, the effect of power on restorative motives was not significant. Finally, similarly with Studies 1 and 2 the effect of power on people’s punishment motivation was independent from the offender’s group membership. We discuss these findings in the General Discussion.

General discussion

Previous research suggests that people recommend more severe punishments for powerful offenders (Bowles & Gelfand, 2010; Fragale et al., 2009; Karelaia & Keck, 2013) because they perceive them as self-focused individuals who have higher intentionality to do harm (Dubois et al., 2015; Fragale et al., 2009; Galinsky et al., 2015; Lammers et al., 2015). Moreover, recent research found that people punish powerful suspects of an offense with utilitarian motives with the aim to incapacitate them and stop crime (Fousiani & Van Prooijen, 2022b). In this study, we investigated the motives for which one punishes high vs. low power offenders, whose guilt is beyond any doubt, and further examined the underlying variables that explain this effect. The results obtained in three studies suggest that offender’s power possession influences the punishing motives of an observer. More specifically, across three studies we showed that powerful offenders trigger stronger utilitarian punishments as opposed to powerless offenders (Hypothesis 1). In contrast, people punish powerless offenders, as opposed to powerful ones, with stronger restorative motives (as shown in Studies 1 and 2). However, although unexpected, in Studies 1 and 3, the effect of power on retributive motives for punishment was also significant and showed that observers assign stronger retributive punishment to powerful as opposed to powerless offenders. Taken together, these findings suggest that people are motivated to punish powerful, as opposed to powerless offenders with the aim to incapacitate them and deter them from reoffending (utilitarian motives), but also with the aim to give them their just deserts (retributive motives). In contrast observers give powerless, as opposed to powerful, offenders the opportunity to make up for their mistakes and reintegrate to society.

Furthermore, people demonized a high-power offender more than a low-power offender (Hypothesis 2), which in turn mediated the effect of power on motives for punishment (Hypothesis 3). Contrary to Hypothesis 4, group membership of the offender did not moderate the effects of power on punishment motives in any of our studies.

Powerful people, due to their high position, often have access to resources and therefore control over others (Fiske, 1993; Galinsky et al., 2015), and have ample possibilities to promote self-interested goals at the cost of the collective interest (see, Blader & Yap, 2016; Boles et al., 2000; Gruenfeld et al., 2008; Kipnis, 1972; Piff et al., 2012). Immoral acts of power-holders violate their assigned responsibility to manifest integrity and use their resources in favor of the common interest. Observers therefore demonize powerful offenders and view them as intrinsically evil people who should be incapacitated and excluded from society.

It is noteworthy that the positive effect of high power of an offender on observers’ utilitarian punishment motives is consistent throughout all three studies, as was the mediating role of demonization in these effects. Our exploratory analyses on the effect of power of an offender on restorative punishment and the mediating role of demonization of the offender were consistent throughout Studies 1 and 2 but not 3. This inconsistency might be partly attributed to the different vignette that we used for the manipulation of the variables. For instance, in the vignette of Study 3 participants imagined being retired, meaning that what happens to the company, especially by low power offenders who are less of a threat, does no longer influence them; Hence, reintegrating a powerless offender to the company may not have been experienced as a priority for observers.

Contrary to our predictions (Hypothesis 4) we did not find evidence for the moderating role of an offender’s group membership in the relationship between power and utilitarian motives for punishment. One should be very careful when interpreting this finding as the non-significant moderating role of group membership might have stemmed from the very manipulation of group membership in the vignettes. In Studies 1 and 2 we manipulated group membership via participants’ ethnicity (Study 1: British vs. Indian, Study 2: American vs. Canadian). Although ethnicity is a very common way to manipulate group membership (see also, Fousiani et al., 2019), such a manipulation might not be relevant in the context of the offense that is described in the present vignettes. Differently put, manipulating group membership through ethnicity might have failed to make the observers perceive the offender as either an ingroup or an outgroup member. Study 3 aimed to replicate group identity in an alternative way. Yet, due to the very hypothetical nature of the vignette the manipulation does not seem to have sufficiently produced the intended effects.

Regarding the insignificant moderated effect of group membership on restorative motives in particular, another possible explanation is that restorative justice plays a less important role in intergroup contexts where the involved parties lack a common identity (Wenzel et al., 2008). Future studies should further explore the role of group membership of an offender in shaping people’s punishing reactions when an offender’s likelihood to cause harm is high vs. low.

As regards the unexpected significant effect of power on retributive motives for punishment, this goes against prior research that shows that retributive motivation to punish is not influenced by either an offender’s group identity (Fousiani et al., 2019) or a suspect’s (whose guilt is uncertain) power (Fousiani & Van Prooijen, 2022b). Apparently, at least in some situations powerful offenders elicit stronger fairness concerns in observers, and make them more willing to give them their just desserts. Accordingly, we conclude that power does influence people’s motives to punish past harm-doing and that people are more prone to give powerful as opposed to powerless offenders their just deserts. Future research is required to examine the conditions under which high power drives utilitarian or retributive motives or both and to identify the mediating mechanisms that drive these effects.

Implications of the study

This research has important theoretical and practical implications. Theoretically, these findings provide evidence that people punish powerful versus powerless offenders with different motives. More specifically, power possession seems to play a role in the type of punishment that people administer: Observers punish both powerful and powerless offenders but depending on their power possession, they emphasize either utilitarian (and sometimes retributive) or restorative motives. Furthermore, our findings provide evidence for the psychological mechanisms that mediate the effects of power on motives for punishment. People perceive power as a state that comes with increased responsibility to denote integrity (Sassenberg et al., 2012, 2014). Failing to do so leads observers to perceive powerful offenders as inherently evil people who intentionally cause harm and find pleasure into hurting others, and thus deserve utilitarian punishments designed to incapacitate and restrict them. Interestingly, the opposite holds for low power offenders, who may be perceived as moral patients who, due to their suffering from their own powerless positions, receive less blame (Gray & Wegner, 2011; Gray et al., 2012) and are consequently exempted from demonizing processes.

Our findings might be relevant to the literature on state-corporate punishment as well. Recent research has shown that observers consider the government, as compared to companies, as more blame-worthy and deserving tougher sanctions for corporate wrongdoing (Michel, 2021). It is likely that the government is seen as a more powerful actor than a company and thus triggers stronger demonizing beliefs among observers.

This study has important practical implications as well. Our results suggest that characteristics of the offender (e.g., power) that are unrelated to the offense itself, might shape the motives for which one punishes an offender. Accordingly, it might be the case that legal decision-makers treat offenders of the same misconduct differently depending on their socioeconomic background. We suggest that legal decision-makers be cognigent of people’s biases toward offenders based on their power or social status. Moreover, these findings are useful to organizations as well, where relations are often hierarchical and involve power asymmetries. It is likely that employees overfocus on powerful others’ wrongdoings and are motivated to incapacitate them, while similar wrong-doings are seen with more compassion when carried out by employees of low status positions. We suggest that HR practitioners get aware of people’s attitude about, and motives toward rule-breakers who possess a powerful (e.g., leadership) or a powerless (e.g., followership) position. Finally, this study implies that information that bears a symbolic value, such as group membership of an offender, matters less when combined with information that bears instrumental value, such as an offender’s power. This finding indicates that offender’s power might have a stronger impact on people’s reactions than their perception as a core or a peripheral member of the group to which they belong. In other words, possession of power on the part of an offender seems to carry more weight than their group identity. This information is relevant not only to legal decision-makers and HR practitioners who daily deal with people of various social groups and identities (i.e., members with whom they can identify or not) but also to the broader society, which consists of people of multiple ethnical, religious, political, gender, and many other kinds of identities.

Strengths and limitations

Our study has both strengths and limitations. A strength is that we confronted participants with different offenses throughout our studies. In Study 1 the offense involved a relatively abstract victim, with the offender evading to pay taxes (i.e., white-collar crime). In Studies 2 and 3 the offense involved violent behavior (sexual harassment) toward a victim.⁸ The consistency in results suggests that the findings observed here generalize across different transgression types (e.g., transgressions that are not necessarily associated with powerful offenders) involving different types of victims.

The most important limitation of this work is that group membership was manipulated either via ethnicity (Studies 1 and 2), which might have not been the most relevant concept in the context of this study or via a hypothetical situation (Study 3). Another limitation is that manipulation of power included information about the offender’s status as well for the mere reason that the two concepts, although distinct are interrelated: Power, besides enabling one to exert influence and produce intended outcomes (Weber, 1946), gives one access to control one’s own and others’ resources (Magee & Galinsky, 2008) often resulting in possession of high status. Future research should differentiate the two constructs and replicate the hypothesized effects. Moreover, sexual harassment (see offense in Studies 2 and 3) might have not been the best possible type of offense to use in the current study as it is itself an offense of power and therefore the generalizability of the results of Study 2 might be limited. Some additional limitations of this research are that we did not include a behavioral measure of punishment and that we manipulated the offense through vignettes which are hypothetical in nature. Finally, the participants were either British or Americans and thus we cannot conclude with certainty if these results generalize to populations of different ethnic or cultural backgrounds. This contribution hence is a preliminary step, and future research needs to include field and lab studies in order to better understand the effects of power on people’s reactions to offenders.

Apart from these methodological limitations, this study includes a number of theoretical limitations as well. For instance, prior research has shown that observers’ punishing motives are impacted by the group membership of a victim (Fousiani & Demoulin, 2019). Yet, the current studies did not manipulate the victim’s group membership. Future research may take this important variable into account. Moreover, future research should investigate people’s preferences for the source of punishment (who punishes). For instance, people in higher hierarchical positions (corporal and legal authorities, leaders, etc.) have a stronger punishment power (French & Raven, 1959), and might therefore be perceived as more entitled to utilize utilitarian punishment. Finally, this study took into account the mediating role of one single variable (demonization) and it is plausible that multiple processes are at work. For instance, people may attribute higher intentionality to powerful offenders (see, Kakkar et al., 2020) and attribute their behavior to internal rather than external causes. These possibilities are speculative at this point, and warrant further investigation.

Concluding remarks

The present research was designed to examine people’s punitive motives toward offenders with high versus low power. Powerful offenders trigger stronger utilitarian (but also retributive) motives for punishment whereas powerless, as opposed to powerful offenders trigger stronger restorative punishing motives. Moreover, people are more likely to view powerful (as opposed to powerless) offenders as evil and and demonize them. Interestingly demonization is an underlying mechanism explaining the effect of power of an offender on people’s punishing motives. We conclude that offender’s power position shapes people’s motives to punish them.

Compliance with ethical standards

This research involves human participants. All procedures performed in this study were in

accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. For the current research we got approval from the Ethical Committee of the Free University of Amsterdam; Veste Commissie Wetenschap en Ethiek (Scientific and Ethical Review Board), Nr VCWE-2017-178.

Open practices

Data and Online Supplemental Materials are available from the Open Science Framework at:

https://doi.org/10.17605/OSF.IO/RZT5X

Preregistration of hypotheses and research plan: https://osf.io/ckebt/?view_only=80a2a6af828a4bc8884abc9251c89665

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

1. It is worth mentioning that in this research, no matter what their group membership was, offenders were always presented as part of the broader ingroup society (e.g., in Study 1, the offender was either British, Indian, or British-Indian but the offense was carried out in the UK).

2. In several situations (e.g., in multi-national societies, politics, sports etc.) some members cannot be classified clearly as ingroup or outgroup members because they have characteristics of two or more different groups (e.g., being half Indian and half British), and thus boundaries between the groups are blurred. In such cases, members are ascribed ambiguous group membership. In line with the ingroup over-exclusion theory (Leyens & Yzerbyt, 1992), people treat members with ambiguous membership as outgroup members, aiming to maintain clear group boundaries (Yzerbyt et al., 1995).

3. This study was pre-registered on OSF before data collection (https://osf.io/ckebt/?view_only=80a2a6af828a4bc8884abc9251c89665). It should be mentioned, however, that there were additional predictions and measures that have been excluded from the current work for efficient presentation; the additional measures are available in the online material on OSF.

4. Utilitarian motives for punishment can be distinguished between deterrent (both private and public) and incapacitative motives. For the very reason that the objective of all these motives is to control an offender’s future behavior, they are all included under the umbrella of utilitarian motives for punishment (see, Carlsmith & Darley, 2008). In line with prior research (Fousiani & Demoulin, 2019; Fousiani & Van Prooijen, 2019; Fousiani & van Prooijen, 2022a, 2022b; Fousiani et al., 2019) we calculated a general mean for utilitarian motives for punishment instead of three different means for each of those dimensions separately.

5. In this study, we also measured punishment intent as a proxy for severity of punishment. When controlling for punishment intent, the results are similar to the ones reported in the manuscript. Moreover, in all our studies we included measures of ingroup threat and ingroup typicality. Findings consistently showed that group membership of an offender, but not power, had a significant main effect on these variables. Yet, for the sake of brevity and because the focus of this work lies on punishment motives, we decided to exclude these variables from this study. The data regarding these variables can be accessed in the publicly available datasets.

6. We additionally included a measure for the perceived severity of the offense. When controlling for severity of the offense, the findings are similar to the ones reported in this section.

7. As in Study 2, we ran additional analyses controlling for severity of the offense. These analyses produced similar results to the ones reported in this section.

8. We also conducted two additional studies with a similar vignette but with slightly different offenses (e.g., stealing a wallet out of a colleague’s bag and money embezzlement). The results were largely similar as Study 1: Findings again showed a main effect of power on utilitarian and restorative punishment motives and demonization, as well as the indirect effect of power on punitive motives through demonization. These effects were again independent from group membership. We report the methods and results of the additional studies in the OSF.