Risk factors for intimate partner violence among native and immigrant male partners in Sweden

ABSTRACT

This study aimed to examine the importance of differentiating between partner violent men in terms of immigrant or native Swedish background, when assessing risk for intimate partner violence (IPV) recidivism. A quantitative design was used with a sample of 1263 alleged male perpetrators reported for IPV-related crimes to the Swedish police. Data consisted of police officers’ risk assessments using the Swedish version of the Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER). Results showed that risk factors such as 'Violent threats or thoughts', and 'Violent attitudes', were more common for perpetrators with an immigrant background. In contrast, risk factors such as 'General criminality' and 'Substance use problems' were more common among native perpetrators. Furthermore, in relation to the importance of specific risk factors for an elevated risk of IPV recidivism, such risk was elevated for perpetrators with a native background when ‘Violent threats or thoughts’, ‘General criminality’ and ‘Relationship problems’ were present. Corresponding factors for immigrant perpetrators consisted of ‘Escalation’ and ‘General criminality’. These results are essential for the police to recognize and understand since only victims in IPV cases with an elevated assessed risk will be eligible for risk management and protective actions.

KEYWORDS:

• Intimate partner violence
• risk factors
• risk assessment
• integrated ecological framework
• immigrant background

Introduction

Intimate partner violence (IPV) is a global public health issue that occurs across all socioeconomic, religious and cultural groups. In support of this notion, the World Health Organization (WHO) estimates that every third woman worldwide will be subjected to IPV sometime during her lifetime (World Health Organization, 2013). However, such rates have also been found to vary quite substantially between regions, where women in high-income regions such as Northern Europe are victimized to a lower degree (21–23%) compared to women in low-income regions such as the Middle East (24–46%; World Health Organization, 2021). In Sweden, where the present study was conducted, the corresponding figure amounts to 25% of women aged 16–79 (National Council for Crime Prevention, 2014).

Although most IPV research is conducted in Western societies, results from studies using multi-ethnic samples show that IPV perpetrators are a heterogeneous group (Dixon & Browne, 2003; Petersson & Strand, 2020; Tran et al., 2016). Among other things, this heterogeneity is reflected by the fact that such perpetrators demonstrate different risk factors for IPV. To this end, the current research on IPV perpetrators and heterogeneity of risk factors has focused primarily on behavioural and psychosocial differences (e.g. Petersson & Strand, 2020). However, drawing on the fact that various risk factors might have different relevance for using IPV depending on the perpetrator’s cultural background, it is important to advance the research on this topic. For instance, in relation to risk for using IPV, risk factors like substance abuse and mental health problems may be more important to consider in countries that lack resources to provide proper healthcare treatment (Heise, 2011). Moreover, attitudes and beliefs about IPV vary across cultural context, where several countries still lack legislation on IPV (e.g. Graham-Kevan, 2007; Waltermaurer, 2012). Thus, attitudes that justify or condone the use of such violence are, on a societal level, not evenly distributed and may therefore hold greater importance for some perpetrators’ use of IPV but not for others.

Drawing on the fact that IPV is defined and perceived differently across countries and cultural groups (e.g. Heise, 1998; Waltermaurer, 2012), it is important that research identify risk factors for IPV among perpetrators with different backgrounds. As such, this endeavour could ultimately help to understand and prevent further IPV. Thus, the aim of this study was to examine the presence of risk factors for IPV, as well as the relevance of specific risk factors for IPV recidivism, among perpetrators with a native background (i.e. perpetrators born to Swedish parents) and perpetrators with an immigrant background.

Risk factors for IPV

To gain a deeper understanding of why and how IPV occurs, the identification of risk factors for such violence is central. To this end, the most widely adopted model used to understand the complexity and heterogeneity of IPV is the integrated ecological model (World Health Organization, 2012). In contrast to previous single-factor theories, the ecological model assumes that IPV best can be explained as an interaction between risk factors on four different levels: society, community, relationship and individual (Heise, 1998, 2012). Additionally, according to Heise, this model can also serve as a framework for understanding how IPV best can be prevented (i.e. risk management). As such, the identification of risk factors at various levels within the model provides both an indication of why IPV was used, as well as on which levels risk management should be targeted. Thus, examining the presence of risk factors in an IPV case helps to decide about the level of risk for a perpetrator’s recidivism in such violence, as well as inform the risk assessor about the nature of risk management strategies necessary to mitigate this risk. The previously mentioned variability of IPV prevalence across the world (World Health Organization, 2021) indicates that risk factors on different levels within the ecological model also might be of different importance depending on a perpetrator’s cultural background. Thus, it is important to examine if risk factors in different groups of perpetrators influence the risk for IPV recidivism differently. In order to better understand how risk factors on different levels within the ecological model might be manifested in different parts of the world, examples of such factors will be described for each level this model (Heise, 1998, 2012).

The societal level. This level relates to factors such as gender equality, laws and attitudes against IPV. In societies strongly influenced by patriarchal and masculine values and norms, women are traditionally seen as subordinate to men and legislation on IPV that can protect women is usually lacking (e.g. Tran et al., 2016). As such, a strive towards greater gender equality in patriarchal societies could challenge men’s notions of power and control, which could result in a higher risk for IPV in order to control liberated women (also known as the ‘backlash effect’: e.g. Dugan et al., 2003; Vyas & Watts, 2009).

In contrast, countries with specific laws aiming to protect women’s rights (e.g. the Swedish law on gross violation of a woman’s integrity) take a stance against IPV by viewing such violence as a matter for the criminal justice system to handle (Weldon, 2002). However, paradoxically, crime statistics show that gender-based violence is more commonly reported in the Nordic countries than in other European countries, although the former countries have higher levels of gender equality in general (European Institute for Gender Equality, 2021; Eurostat, 2021). Meanwhile, countries that emphasize patriarchal values and norms, such as Jordan, Iraq and Tunisia, have a weaker structure to protect women from IPV (Klugman, 2017; Sardinha & Nájera Catalán, 2018; World Health Organization, 2021). Still, official statistics in such countries show that the rates of IPV victimization is either lower or similar to that of Sweden and Norway. However, such similarities between the aforementioned countries are cancelled out when studying IPV victimization rates using self-reported data, where the former countries, in general, demonstrate a higher prevalence of IPV (Tavares & Wodon, 2017). It is not clear why the official statistics are higher in the Nordic countries compared to countries with mainly patriarchal societies. However, some explanations refer to actual higher prevalence rates, a higher willingness to report or that increased gender equality also have a backlash on attitudes towards IPV resulting in increased violence (Gracia & Merlo, 2016). Meanwhile, the validity of crime statistics in countries where victims are blamed for their victimization and IPV seen as legitimate is questionable (e.g. World Health Organization, 2021).

The community level. This level involves risk factors such as low socioeconomic status, unemployment, general criminality and lack of integration. For example, unemployment is a strong risk factor for IPV as this can create financial stress and marital conflicts, which can escalate into physical IPV (e.g. Salmi & Danielsson, 2014). This may be especially pertinent in relationships where both partners are unemployed since spending more time with each other can increase the risk for marital conflict and IPV (e.g. Caetano et al., 2008).

Other risk factors on the community level concern how norms that legitimize violence are distributed in social networks and in the neighbourhood where the couple lives (Heise, 2012). For instance, drawing on the strong influence that peers can have on an individual’s behaviour, men who associate with peers who display antisocial values and attitudes that support the use of IPV are at an increased risk for using IPV themselves (e.g. Williams et al., 2008).

Finally, social isolation is also a risk factor for IPV as this makes help and support from family, friends and neighbours less accessible. A common feature among IPV perpetrators is to use controlling and coercive behaviour to restrict their partners contacts with family and friends, thereby also limiting their possibilities, as well as inclination, to seek help from others (e.g. Kropp et al., 2008). Relatedly, living in neighborhoods with low levels of informal social control is also a risk factor for IPV, as neighbors may be less inclined to intervene on behalf of the victim (e.g. Benson et al., 2004). This might be explained by norms and traditions of considering IPV as a private matter and in combination with low confidence in the authorities it decreases the probability that someone else will call the police when violence occurs (e.g. Lanier & Maume, 2009).

The relational level. This level mainly concerns risk factors associated with the relationship between the perpetrator and the victim. For instance, the inability to handle conflicts or changes in the relationship, such as pregnancy and having children, increases the risk for IPV (Heise, 2012). As such, poor conflict resolution skills correlate positively with the escalation of marital conflicts from verbal arguments to physical violence (e.g. Abrahams & Jewkes, 2005). Furthermore, a household’s low socioeconomic status can increase the risk of IPV as financial stress can lead to frustration and a feeling of inadequacy in men, as cultural norms designate them as the household provider (Abramsky et al., 2011; Heise, 1998). Other risk factors within the relationship level include inequality between the partners, for example, related to financial decision-making, which can increase the risk for IPV. In contrast, relationships where both partners have an equal role in decision-making and a shared responsibility for the household’s resources have been shown to report lower levels of marital conflict and IPV (e.g. Heise, 2012).

The individual level. This level concerns the perpetrator’s personal history and behavior. It consists of biological, psychological and psychosocial risk factors. For example, witnessing IPV as a child or being abused during childhood have been found to increase the risk for IPV (Heise, 2012). However, proximal risk factors on the individual level have garnered stronger empirical support, including substance abuse and mental health problems (e.g. Corvo & Johnson, 2013). As such, substance abuse is common among perpetrators of IPV and a strong risk factor for such violence (e.g. Kropp et al., 2008). However, the risk for IPV is even higher in relationships where both partners have an ongoing substance abuse (Fals-Stewart et al., 2005; Walton-Moss et al., 2005; World Health Organization, 2013). Furthermore, mental health problems as well as the comorbidity of substance abuse and such problems have been shown to increase the risk of IPV (Fals-Stewart et al., 2005; Yu et al., 2019). For instance, mental health problems, such as ADHD, anxiety disorder, or bipolar disorder, have been found to increase the risk of IPV (Yu et al., 2019).

Intimate partner violence in the Nordic countries

Previous research in the Nordic countries has identified differences in terms of risk factors for IPV among perpetrators with a native background and perpetrators with an immigrant background. For instance, studies have shown that substance abuse was more present among perpetrators with a native Norwegian background compared to perpetrators with an immigrant background (e.g. Lund, 2014; Vatnar et al., 2017). More specifically, Lund (2014) reported that in 57.5% of the IPV cases involving a perpetrator with a native background, the violent act was committed while being under the influence of substances. In contrast, 25.7% of perpetrators with immigrant background committed IPV while being intoxicated. These results are in line with other studies in a Nordic context (see, e.g. Vatnar et al., 2017). Thus, substance abuse is identified as a stronger risk factor for increased risk for Nordic perpetrators than immigrant perpetrators. In sum, these results may be explained by differences in alcohol consumption among perpetrators with different backgrounds. For example, previous research has shown that heavy drinking is more typical in the Nordic countries (Bye & Rossow, 2010).

However, in relation to comparing perpetrators with a native background and perpetrators with an immigrant background, previous research in the Nordic context has focused on crime rates in general and not specifically in relation to IPV risk factors (e.g. Skardhamar et al., 2014). Moreover, among the few studies that do focus on IPV in the Nordic context, focus has generally been directed to victimization rather than perpetration (e.g. Salmi & Danielsson, 2014; Tanskanen & Kivivouri, 2021). Thus, less is known about how IPV perpetrators with a native and an immigrant background, respectively, differ in terms of risk factors and risk for IPV recidivism.

The present study

Overall, it is difficult to identify which risk factors significantly impact the risk of IPV in a specific case. The integrated ecological model shows that risk factors might be of different importance depending on wherein the world IPV occurs. Thus, risk factors can vary in their relevance for perpetrating IPV depending on the cultural context. For example, substance abuse might not be as relevant in countries that are heavily influenced by religious laws as compared to countries, which are more secularized, since substance abuse is illegal in some religions. Also, countries with higher healthcare standards might be more likely to, at an early stage, detect and adequately treat individuals with, for example, mental health problems or substance abuse. In turn, this could help prevent these individuals from using IPV.

Taken together, it would be wrong to assume that all perpetrators have the same risk factors for IPV and that all such risk factors carry the same weight or relevance for IPV perpetration. Risk factors can be something that an individual have, is exposed to, or develops over time, depending on the context. To this end, new environments and events can present new risk factors or consolidate and reinforce existing ones, and research shows that the violence might start or escalate when individuals migrate (e.g. Gulesci, 2018). As such, migrating to a new country is often associated with several challenges, including language barriers, difficulties finding employment and accustoming to a new culture and way of living. These cumulative challenges may create stressful situations and therefore increase immigrant families’ vulnerability for IPV exposure (Sullivan et al., 2005). Another challenge can be the result of dissonant acculturation, meaning that the partners in a relationship acculturates to a new context at different paces, which can increase the risk for IPV (Samuel, 2009). Furthermore, migrating from a patriarchal cultural context to a more gender equal context may present several challenges. As such, men may find it difficult to share power with their wives and expect that the patriarchal norms and values should be retained, including maintaining wives’ dependency on their husbands (Ayyub, 2000; Sullivan et al., 2005). Additionally, a new country and culture may expose immigrant couples to new risk factors (e.g. unemployment), which in turn can result in increased risk for IPV (Heise, 1998).

This study aimed to examine differences between partner violent men, in terms of immigrant or native Swedish background, related to risk assessment for IPV. More specifically, this paper aimed to compare these two groups in terms of (a) the presence of risk factors for IPV recidivism and (b) the association and relevance of specific risk factors for an elevated assessed risk for IPV recidivism.

Method and material

Study design

This was a cross-sectional quantitative study, where data were retrieved from a prospective research project conducted in four Swedish police districts between 2009 and 2016 aiming to implement and evaluate the use of structured violence risk assessment methods for IPV, stalking and honour-based violence (Strand et al., 2016). The four police districts are from different parts of Sweden, constituting one urban district, two rural districts and one remote district. Thus, the police districts are a convenience sample chosen for the overall purpose of the project.

Sample

In total, 1573 alleged male perpetrators were reported to any of the four Swedish police districts for male-to-female perpetrated IPV during the study period. However, 310 (19.7%) perpetrators were excluded since information about native or immigrant background was missing, resulting in a final sample of 1263 partner violent men used for the first set of analyses comparing perpetrators with an immigrant background to perpetrators with a native Swedish background. Immigrant background was defined as perpetrators who were either born or had at least one parent born in another country than Sweden (n = 485). The remaining perpetrators were defined as having a native Swedish background (n = 778).

Previous definitions of immigrant background vary between studies, where some define immigrant background as having two parents born abroad and some define it as the perpetrator being born abroad (Skardhamar et al., 2014). This study relied on data reported to the police, where immigrant background at the time of data collection was defined as either the perpetrator, or at least one of the perpetrator’s parents, was born outside of Sweden. This is also the definition used in other Nordic studies of IPV (e.g. Lund, 2014).

To enable more detailed analyses, perpetrators were classified according to the geographical region from where they, or their parents, originated. However, 33 (2.6%) perpetrators were missing such information, and another 25 (2.0%) perpetrators were born in countries that could not be grouped for statistical reasons (i.e. creating too small groups) and were thereby excluded from these analyses. The sample used for the geographical analyses consisted of 1205 perpetrators, where the reduced group of perpetrators having an immigrant background (n = 427) was categorized into the following five groups: Africa (n = 94), Western Europe (n = 81), Eastern Europe (n = 78), the Middle East (n = 150) and South/Central America (n = 24). We adopted this grouping strategy, which has also been used in previous research (e.g. Skardhamar et al., 2014), so as not to obscure any differences within the group of immigrant perpetrators. Thus, only comparing perpetrators with a native background and perpetrators with an immigrant background would ignore the heterogeneity among perpetrators in the latter group.

Sample characteristics were calculated on the final sample of 1263 male perpetrators. Native perpetrators (M = 40, SD = 12.9, range = 17–86) were slightly older than perpetrators with an immigrant background (M = 38, SD = 10.6, range = 18–75), t(1167) = −4.35, p = .000, and less likely than immigrant perpetrators to have children younger than 18 years together with their female victim (46.2% vs. 72.7%), χ²(1, 1213) = 81.8, p = .000, phi = .26.

Moreover, differences in reported index crimes were compared between the two groups. Index crimes were defined as the most severe IPV crime reported to the police, where physically violent crimes were deemed more severe than psychological violence (e.g. threats) due to its potential to cause bodily injuries to the victim. The index crimes were categorized as either performed or attempts to commit the following crimes; murder/manslaughter (n = 5, 0.4%), severe violation of a woman’s integrity (n = 331, 26.2%), assault (n = 594, 47.0%), sexual assault (n = 19, 1.5%), illegal threats (n = 161, 12.7%), other crimes (n = 61, 4.8%) and for 92 perpetrators (7.3%) data on index crime was missing. There were no significant differences between the two groups in terms of reported index crimes, χ²(5, 1171) = 7.05, p = .217, φ_c = .08. Severe violation of a woman’s integrity is an internationally unique crime for Swedish legislation. A male perpetrator who is found guilty of repeated harassments (i.e. on more than one occasion and towards the same female partner) aimed at violating his female partner’s integrity (e.g. through assault, sexual violence, or threats) can be charged with this crime (Law 2013:367).

Material

The material consisted of risk assessments performed by police officers in the four police districts, using the Swedish version of the Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER: Kropp et al., 2008). The B-SAFER is a risk assessment instrument that consists of ten perpetrator risk factors and five victim vulnerability factors. The first five risk factors are related to the perpetrator’s history of IPV. They include violent acts (including both physical and sexual violence), violent threats or thoughts, escalation of violence or threats, violation of court orders (e.g. restraining orders) and violent attitudes (e.g. justifying the use of IPV or sexual jealousy). The remaining five risk factors relate to the perpetrator’s psychosocial adjustment and consist of general criminality, relationship problems (e.g. numerous marital conflicts or separations), employment problems (e.g. long periods of unemployment and related financial issues), substance use problems and mental health problems. Moreover, the five victim vulnerability factors consist of inconsistent attitudes or behaviour (e.g. contacting the perpetrator despite a restraining order in place or excusing the perpetrator’s use of IPV), extreme fear of the perpetrator, inadequate support or resources (e.g. lacking access to legal and informal social support), unsafe living situation and health problems (e.g. problems with physical or mental health).

Using the B-SAFER, the assessor rates the risk and victim vulnerability factors on a three-point scale as either present, partially present or absent (Kropp et al., 2008). This information should be acquired from multiple sources, preferably from hearings with the victim, the alleged perpetrator and potential witnesses, as well as registered information (e.g. previous convictions of the alleged perpetrator). Based on the presence and absence of the risk and victim vulnerability factors and drawing on the risk assessor’s professional experience and knowledge of working with IPV, the assessor makes two summary risk ratings: risk for acute (i.e. imminent) IPV and risk for severe or lethal IPV, which is rated on a three-point scale as low, moderate or high risk. Importantly, being based on the structural professional judgement (SPJ) approach to violence risk assessment, the summary risk ratings are not calculated algorithmically. Instead, the summary risk is generated by the evaluator’s professional discretion and expertise of IPV. Thus, the B-SAFER risk factors are intended to aid in this rating procedure, not to be summed up into a numerical score. For this study, the risk factors in the B-SAFER and the summary risk rating of severe or lethal IPV were used.

The tool is reported to have good-to-excellent interrater reliability and predictive validity (e.g. Loinaz, 2014; Storey et al., 2014). With few exceptions (see, Svalin et al., 2018), the B-SAFER has proven to be a useful instrument within the Swedish police, showing valid and reliable results (Belfrage & Strand, 2008; Storey et al., 2014).

Outcome. This study’s outcome variable was the summary risk rating for severe or lethal IPV, as assessed by the police using the B-SAFER. The rationale for using this outcome, and not actual recidivism in IPV, was mainly because this study did not seek to validate the B-SAFER. Importantly, in line with the SPJ approach, the main goal of assessing the risk for future violence is to recommend and implement adequate risk management strategies. Therefore, it is of interest to investigate what risk factors are present among perpetrators who have elevated risk, i.e. moderate or high-risk, since only victims in cases with elevated risk will be eligible for risk management. Thus, the summary risk rating was chosen as the outcome variable for this study.

Procedure

A case was found to be eligible for a B-SAFER assessment when identified by police officers, either by inspection of the police records of reported crimes or by crime investigators’ requests. Reported cases where such a risk was identified were referred to a unit of police officers with specialized training in carrying out structured violence risk assessments using the B-SAFER.

Once eligible cases had been identified, the police officer tasked with performing the risk assessment followed four steps. First, background information about the case was collected (e.g. demographics of the victim and perpetrator), including a brief description of the IPV incident. Second, the presence of the ten perpetrator risk factors in the B-SAFER was assessed. Third, the five victim vulnerability factors in the B-SAFER were assessed. Fourth, based on the presence (or absence) of the risk and victim vulnerability factors, the police officer assessed the summary risk ratings for the risk for acute IPV and the risk for severe or lethal IPV.

Statistical analysis

Statistical comparisons of risk factors between native and immigrant perpetrators were carried out using chi-square tests, supplemented with odds ratios (OR) with 95% confidence intervals (CI) and phi for effect size. To enable these analyses, risk factors were dichotomized as either present (i.e. combining original B-SAFER ratings of present and partially present) or absent. The same analytic procedure was used to compare each of the five geographical groups of immigrant perpetrators against native perpetrators.

The association between the B-SAFER risk factors and the summary risk rating for IPV recidivism was examined separately for the two groups (i.e. native and immigrant perpetrators), using direct binary logistic regression analyses. Thus, these analyses aimed to examine if there were any specific risk factors that were more strongly related to risk for further IPV (i.e. ‘red flags’). The results of these analyses could indicate if those tasked with assessing and managing risk for IPV need to be especially attentive to certain risk factors. To this end, the summary risk rating of severe or lethal IPV was dichotomized as either elevated risk (i.e. combining ratings of moderate and high risk) or low risk. The dichotomized B-SAFER risk factors were entered as predictors in the logistic regression analyses, using the dichotomized summary risk rating for severe or lethal IPV as the dependent variable. The rationale for excluding the summary risk ratings for imminent violence was that this risk rating is primarily used by the police to assess the need of imminent risk management. As this was not the focus of this study, only risk for severe or lethal violence was used. No multicollinearity issues were identified in the logistic regression analyses.

The significance level used in this study was p < .05. All statistical analyses were computed using IBM SPSS (version 25). This study received ethical approval from the Swedish Ethical Review Board.

Results

Prevalence of risk factors

Chi-square tests and odds ratios were carried out to compare the presence of risk factors for IPV between native and immigrant perpetrators (see, Table 1).

Table 1. Differences in risk factors between IPV perpetrators with a native background and perpetrators with an immigrant background, calculated with chi-square tests (N = 1263).

	Present risk factors
	Native background (n = 778)			Immigrant background (n = 485)
B-SAFER item	n^a	%		n^a	%	χ^2	p	OR	95% CI	phi
Perpetrator risk factors
Violent acts	762	92.4		475	91.4	0.41	.520	0.9	[0.6, 1.3]	−.02
Violent threats or thoughts	722	70.4		459	79.1	11.0	.001	0.6	[0.5, 0.8]	.09
Escalation	630	61.3		394	64.0	0.75	.388	1.1	[0.8, 1.5]	.03
Violation of court orders	669	10.6		370	11.4	0.13	.714	1.1	[0.7, 1.6]	.01
Violent attitudes	585	72.6		375	81.9	10.7	.001	0.6	[0.4, 0.8]	.11
Psychosocial adjustment
General criminality	742	55.5		439	43.7	15.3	.000	1.6	[1.3, 2.0]	−.11
Relationship problems	495	75.6		306	73.2	0.55	.457	0.8	[0.6, 1.2]	−.03
Employment problems	513	58.9		293	64.2	2.19	.139	1.2	[0.9, 1.7]	.05
Substance use problems	622	78.1		304	60.2	32.77	.000	2.4	[1.8, 3.2]	−.19
Mental health problems	440	66.4		217	60.8	1.95	.163	0.8	[0.6, 1.1]	−.05

IPV = Intimate partner violence; B-SAFER = Brief Spousal Assault Form for the Evaluation of Risk (Kropp et al., 2008); OR = odds ratios; CI = confidence intervals.

^aNumber of assessments with complete coding

Native perpetrators were less likely than immigrant perpetrators to display the risk factors ‘Violent threats or thoughts’ (OR = 0.6) and ‘Violent attitudes’ (OR = 0.6). In contrast, native perpetrators were more likely to display the psychosocial risk factors ‘General criminality’ (OR = 1.6) and ‘Substance use problems’ (OR = 2.4).

Next, native perpetrators were compared to immigrant perpetrators grouped into the five different geographical regions (see, Table 2). For perpetrators with an Eastern European background, the odds of displaying the risk factor ‘Violent threats or thoughts’ were twice as high as for native perpetrators (83.3% vs 70.4%), χ²(1, 800) = 5.21, p = .022, phi = .08 (OR = 2.1; 95% CI = [1.1, 3.9]). Eastern European perpetrators were also less likely to display the risk factor ‘General criminality’ compared to native perpetrators (38.6% vs 55.5%), χ²(1, 812) = 6.74, p = .009, phi = −.09 (OR = 0.5; 95% CI = [0.3, 0.8]).

Table 2. Differences in risk factors between perpetrators with a native background and perpetrators with an immigrant background divided according to the geographical region of origin (N = 1205).

B-SAFER^2 item	Native^a background (n = 778)	Eastern European background (n = 78)	Middle Eastern background (n = 150)	African background (n = 94)	Western European background (n = 81)	South and Central American background (n = 24)
Violent acts	92.4%	98.7%	89.7%	89.0%	91.3%	-
Violent threats or thoughts	70.4%	83.3%*	85.6%***	75.6%	77.0%	70.8%
Escalation	61.3%	69.2%	60.0%	68.0%	57.4%	77.3%
Violation of court orders	10.6%	7.7%	12.2%	9.7%	15.5%	-
Violent attitudes	72.6%	83.9%	85.8%***	80.3%	73.8%	-
General criminality	55.5%	38.6%*	40.9%***	38.1%***	57.3%	52.2%*
Relationship problems	75.6%	77.1%	66.7%	70.5%	83.0%	-
Employment problems	58.9%	73.3%	63.8%	70.4%	66.1%	-
Substance use problems	78.1%	75.0%	39.5%***	51.9%***	78.5%	-
Mental health problems	66.4%	63.0%	61.4%	45.0%**	63.4%	72.7%

^aThe native group is reference group in all statistical analysis. ²B-SAFER = Brief Spousal Assault Form for the Evaluation of Risk (Kropp et al., 2008). * p < .05; ** p < .01; *** p < .001, – = expected value is below 5 and therefore cannot be calculated.

For perpetrators with a Middle Eastern background, the odds of displaying ‘Violent threats or thoughts’ (85.6% vs 70.4%), χ² (1, 868) = 13.55, p = .000, phi = .13 (OR = 2.5; 95% CI = [1.5, 4.1]) and ‘Violent attitudes’ (85.8% vs 72.6%), χ² (1, 705) = 8.52, p = .004, phi = .11 (OR = 2.3; 95% CI = [1.3, 3.9]) were twice as high compared to perpetrators with a native background. Moreover, perpetrators with a Middle Eastern background were less likely to display ‘General criminality’ (40.9% vs 55.5%), χ² (1, 879) = 9.39, p = .002, phi = −.11 (OR = 0.5; 95% CI = [0.4, 0.8]) and ‘Substance use problems' (39.5% vs 78.1%), χ² (1, 708) = 55.76, p = .000, phi = −.29 (OR = 0.2; 95% CI = [0.1, 0.3]) than native perpetrators.

Perpetrators with an African background were less likely to display the risk factors ‘General criminality’ (38.1% vs 55.5%), χ² (1, 826) = 8.53, p = .003, phi = −.11 (OR = 0.5; 95% CI = [0.3, 0.8]), ‘Substance use problems’ (51.9% vs 78.1%), χ² (1, 674) = 16.72, p = .001, phi = −.16 (OR = 0.3; 95% CI = [0.2, 0.5]) and ‘Mental health problems’ (45.0% vs 66.4%), χ² (1, 480) = 6.41, p = .011, phi = −.12 (OR = 0.4; 95% CI = [0.2, 0.8]), compared to native perpetrators.

Perpetrators with a South and Central American background were less likely than native perpetrators to demonstrate ‘General criminality’ (14.3% vs 58.9%), χ² (1, 520) = 3.96, p = .047 and phi = −.10 (OR = 0.1; 95% CI = [0.0, 1.0]). The results showed no significant differences when comparing native perpetrators to perpetrators with a Western European background.

The importance of risk factors for elevated risk

An initial analysis demonstrated no difference between native perpetrators and immigrant perpetrators in terms of being assessed with an elevated risk for severe or lethal IPV (37.9% vs. 39.2%), χ²(1, 1209) = 0.21, ns.

To determine the association between the B-SAFER risk factors and the summary risk rating for severe or lethal violence, binary logistic regressions were carried out separately for immigrant perpetrators and native perpetrators (see, Table 3). Since logistic regression analysis excludes cases with missing information, sample sizes were reduced. For native perpetrators, 215 (27.7%) perpetrators were eligible for the logistic regression analyses. Furthermore, the group of immigrant perpetrators was reduced for two reasons. First, since only perpetrators originating from Eastern Europe, the Middle East and Africa differed significantly from the native perpetrators, only perpetrators from these three groups were included in the analysis. As perpetrators from South and Central America only differed from native perpetrators on general criminality, we decided to exclude the former group, together with perpetrators from Western Europe, from the logistic regression analyses. Second, in this reduced group of immigrant perpetrators, 13 (4%) perpetrators were missing information about assessed risk and were therefore excluded. Thus, 61 (18.9%) immigrant perpetrators were eligible for the logistic regression analysis.

Table 3. Logistic regression models of the B-SAFER risk factors’ predictability of elevated risk for IPV among perpetrators.

Perpetrators with a native background (n = 215)	β	Wald	p	OR	95% CI
Model
Violent acts	0.4	0.4	.508	1.4	[0.5, 4.3]
Violent threats or thoughts	1.1	5.3	.022	2.9	[1.2, 7.2]
Escalation	0.7	3.6	.058	2.1	[1.0, 4.6]
Violation of court orders	−0.6	1.9	.174	0.5	[0.2, 1.3]
Violent attitudes	0.6	1.9	.171	1.9	[0.8, 4.7]
General criminality	0.8	5.0	.025	2.3	[1.1, 4.8]
Relationship problems	1.3	7.0	.008	3.6	[1.4, 9.5]
Employment problems	0.1	0.0	.877	1.1	[0.5, 2.2]
Substance use problems	0.4	0.4	.349	1.5	[0.6, 3.4]
Mental health problems	0.4	0.4	.334	1.5	[0.7, 3.2]
Constant	−4.6	28.1	.000
Perpetrators with an immigrant background (n = 61)	β	Wald	p	OR	95% CI
Model
Violent acts	1.3	0.9	.342	3.5	[0.3, 47.6]
Violent threats or thoughts	0.9	0.6	.423	2.6	[0.3, 25.6]
Escalation	2.0	4.6	.032	7.5	[1.2, 46.7]
Violation of court orders	0.2	0.0	.894	1.2	[0.1, 20.6]
Violent attitudes	0.2	0.0	.854	1.2	[0.1, 10.3]
General criminality	1.7	4.6	.031	5.6	[1.2, 26.9]
Relationship problems	0.8	0.5	.498	2.1	[0.2, 18.9]
Employment problems	1.2	2.0	.158	3.4	[0.6, 18.4]
Substance use problems	1.1	2.1	.148	3.0	[0.7, 13.4]
Mental health problems	−0.9	0.8	.368	0.4	[0.1, 2.8]
Constant	−5.9	7.2	.007

B-SAFER = Brief Spousal Assault Form for the Evaluation of Risk (Kropp et al., 2008); IPV = Intimate partner violence; OR = Odds ratios; CI = confidence interval.

To control for systematic bias, a missing value analysis was conducted by comparing those included with those excluded on descriptive variables, risk factors and assessed risk. The only difference between excluded and included perpetrators was substance use problems. For immigrant perpetrators, those who were not included in the analysis displayed more problems with substance use (58.1% vs 41.0%), χ² (1, 190) = 4.89, p = .027, phi = .16 (OR = 2.0; 95% CI = [1.1, 3.7]. Similar results were found for native perpetrators, as those excluded had more substance use problems (81.1% vs 72.6%), χ² (1, 622) = 5.98, p = .014, phi = .10 (OR = 1.6; 95% CI = [1.1, 2.4]).

Native perpetrators. The binary logistic regression model containing all ten risk factors was statistically significant, χ² (10, 215) = 67.86, p < .001. This model explained between 27.1% (Cox & Snell R square) and 36.6% (Nagelkerke R squared) of the variance in risk for severe or lethal IPV and correctly classified 75.3% of the cases with either low or elevated risk for such violence. Of the perpetrators included in the analysis, 130 (60.5%) had been assessed with low risk for severe or lethal IPV, and 85 (39.5%) had been assessed with an elevated risk for such violence. The risk factors ‘Violent threats or thoughts’, ‘General criminality’ and ‘Relationship problems’ had a uniquely significant association with an elevated assessed risk for IPV. Among these risk factors, ‘Relationship problems’ had the strongest association with an elevated assessed risk for IPV (OR = 3.7).

Immigrant perpetrators. The binary logistic regression model containing all ten risk factors was statistically significant, χ²(10, 61) = 31.89, p < .001. This model explained between 40.7% (Cox & Snell R square) and 54.5% (Nagelkerke R squared) of the variance in assessed risk for severe or lethal IPV and correctly classified 78.7% of the cases with either low or elevated risk for such violence. Of the individuals included in the analysis, 34 (55.7%) had been assessed with a low risk of severe or lethal IPV and 27 (44.3%) had been assessed with an elevated risk for such violence. The risk factors ‘Escalation’ and ‘General criminality’ had a uniquely significant association with the assessed risk for severe or lethal IPV. Among these risk factors, ‘Escalation’ had the strongest association with an elevated assessed risk for IPV (OR = 7.5).

Discussion

The aim of this study was to compare the presence and importance of risk factors for further IPV, as well as the relevance of specific risk factors for elevated assessed risk for further IPV, between perpetrators with a native Swedish background and perpetrators with an immigrant background. Noteworthy, before discussing the results of the comparisons between the groups, more than 90% of perpetrators in the overall sample demonstrated physically violent acts. This is not surprising, however, considering that IPV reported to the criminal justice system usually is more severe in nature compared to IPV within community samples (National Council for Crime Prevention (NCCP), 2014; Sartin et al., 2006). Meanwhile, the nature of the sample used in this study means that the results herein cannot be generalized to other forms of IPV in general (Heise, 2012; World Health Organization, 2013).

The results showed that immigrant perpetrators were more likely than native perpetrators to display the risk factors ‘Violent threats or thoughts’ and ‘Violent attitudes’, which especially held true for perpetrators with Eastern European and Middle Eastern backgrounds. These results are in line with previous research (Tran et al., 2016; World Health Organization, 2012), where societies in the Middle East generally have a more permissive view of violence towards women (Boy & Kulczycki, 2008). Applying Heise’s (1998, 2012) ecological framework, a possible explanation for these results could be that such societies transmit values and norms (often institutionalized in terms of laws) on a societal and community level, advocating men’s right to chastise their female partners. Thus, originating from gender unequal societies could make threats and violent attitudes seen as more acceptable forms of behaviour towards women. In support of this notion, seven countries from the Middle East (five represented in this study) place themselves among the ten least gender equated countries in the world (World Economic Forum, 2018). In fact, it is more common with attitudes condoning IPV in countries with low gender equality than in countries that rank higher in gender equality (e.g. Sardinha & Nájera Catalán, 2018).

In contrast, Sweden is ranked among the top ten most gender-equated countries, which could explain the lower prevalence of violent threats and attitudes among perpetrators with a Swedish native background. Although, findings show that high gender equality is not a guarantee towards IPV, there is still less IPV in the Nordic countries compared to the ten countries that are least gender-equated (Gracia & Merlo, 2016). In sum, the differences in relation to violent threats or thoughts and attitudes between immigrant and native perpetrators could be linked to the societal level in the form of patriarchal societies. In such societies, the prevailing norm is usually that men are superior to women and, therefore, attitudes that support or excuse IPV are not seen as deviant. Considering that attitudes are difficult to change, these are likely maintained when immigrating to Sweden and might also be embedded in the parental upbringing through traditional cultural values and hence influence the children (Klugman, 2017; Zakar et al., 2012).

Problematically, a quarter of all countries in the world do not have specific laws against IPV, where two-thirds of countries in the Middle East and North Africa have no laws at all against IPV (Tavares & Wodon, 2017). Moreover, police in Eastern European countries often refuse to register an IPV crime report and instead tries to convince the victim to mediate or talk with the perpetrator (United Nations Population Fund, 2015). Additionally, convictions are rare, and if the perpetrator is convicted, it is also common that the punishment will be disproportionately low since mediation or similar actions are preferred. Therefore, these structural issues serve to uphold and reinforce violent attitudes that condone IPV throughout large parts of the world, only making it a deviant behaviour when the individual migrates to a country with different sets of laws and norms. Therefore, when a person migrates from a country with a more accepting view of IPV compared to Sweden their behaviour is seen as deviant and, in some cases, illegal.

Sweden has laws specifically targeting IPV, which indicates that Sweden, at the societal level, is a country that does not accept this form of violence. Therefore, IPV perpetrators with backgrounds in, for instance, Eastern European countries or the Middle East might have less experience in the judiciary system and thus have less understanding of IPV as a crime. Furthermore, immigrating to Sweden might make the assimilation to the Swedish society regarding both laws and norms much more complicated if the laws and norms differ from their own country of origin. Thereby, risk factors must be considered to a larger extent when dealing with perpetrators who have immigrated from cultures with a permissive view of IPV since the structural and societal values might be of more importance than relational or individual level factors. On the other hand, police cannot generalize that this would be true for all perpetrators having an immigrant background. As such, it is the task of the police to assess if risk factors that appear on societal and community levels are relevant for the individual perpetrator. From an IPV risk assessment perspective, this highlights the importance of the summary risk ratings being made based on the police officer’s professional judgement.

The results also showed that ‘General criminality’ and ‘Substance use problems’ were more common among native perpetrators than among immigrant perpetrators. In terms of substance abuse, these results are in line with previous Nordic research (Lund, 2014; Vatnar et al., 2017) and could be explained by differences in drinking patterns. In Sweden, binge drinking is common and a known risk factor for IPV (e.g. Andreasson et al., 2015; Bye & Rossow, 2010). Moreover, Sweden is a secularized country, which could also explain the higher presence of substance abuse among native perpetrators than among perpetrators with an immigrant background. In many countries in the Middle East and Africa, religion prohibits the use of alcohol (Amodeo & Jones, 1997). Relatedly, problematic alcohol consumption has been linked to antisocial behaviour (Walton-Moss et al., 2005), and the higher presence of substance abuse among native perpetrators could thereby partly explain the differences in ‘General criminality’. However, criminal history among perpetrators with an immigrant background could have been underestimated since the Swedish police did not always have access to crime registers outside of Sweden.

In relation to the importance of specific risk factors for an elevated risk of IPV recidivism, such risk was elevated for perpetrators with a native background when ‘Violent threats or thoughts’, ‘General criminality’ and ‘Relationship problems’ were present. Corresponding factors for immigrant perpetrators consisted of ‘Escalation’ and ‘General criminality’. However, ‘Escalation’ was almost significant (p = .058) also for the native perpetrators. As such, ‘General criminality’ and ‘Escalation’ are important risk factors for both groups’ risk for further IPV. Men who are generally violent or have a criminal lifestyle have been found guilty of committing more frequent and severe IPV, which could explain that police perceive general criminality as a general risk factor and consider its presence as an increased risk for such violence regardless of the perpetrator’s native or immigrant background (Kropp et al., 2008).

Moreover, the results of the logistic regressions also demonstrated that ‘Relationship problems’ was a unique contributing risk factor for elevated risk for further IPV for native perpetrators but not for immigrant perpetrators. One explanation for this may be that the police do not have enough knowledge or insight into other cultures and how relational problems would manifest themselves from such a perspective. Identifying various risk factors may be based solely on the assessing police officer’s professional judgement and experience of IPV, rather than the risk assessment instrument itself (Svalin et al., 2018). However, the results from the logistic regressions pertaining to the group of immigrant perpetrators are based on a relatively small sample size. Thus, these results should be considered as preliminary and need to be replicated in studies with a larger sample.

Limitations

A major limitation of this study was the large number of missing values for risk factors in the B-SAFER, which directly reduced the number of perpetrators eligible for the logistic regression analyses. Although a limitation, our missing value analyses demonstrated almost no differences between the included and excluded perpetrators regarding risk factors, assessed risk and demographic information. However, a larger number of cases are always preferable when conducting logistic regressions with several predictors (Tabachnick & Fidell, 2013).

As previously mentioned, it is important to remember that the sample in this study comprised perpetrators reported to the police for IPV. As such, samples of partner violent men reported to the criminal justice sample have generally committed more severe IPV compared to other samples (e.g. community samples: National Council for Crime Prevention (NCCP), 2014; Sartin et al., 2006). Thus, the generalizability of the results presented herein is primarily related to male-to-female IPV reported to the police in Sweden. Generalizing beyond these settings should be made with caution. Furthermore, since the risk assessments were performed by the police, it was not possible to control whether there have been any biases regarding group-based attributions. If the police have troubles to identify relevant risk factors on individual level it might be possible that risk factors such as ‘General criminality’ or ‘Substance use problems’, which are more common for the native group, are set as the standard risk factors on group level for high-risk IPV perpetrators. Since immigrant perpetrators have less of these risk factors, other relevant risk factors might be missed when assessing the overall risk, and, thus, the overall risk will be assessed as lower.

Another limitation of this study is the, rather broad, definition regarding immigrant background, which for this study included either being born abroad or having at least one parent with an immigrant background. However, we argue that cultural upbringing might influence one’s attitudes and therefore the definition might not be a problem in explaining the results on a group level. Moreover, although we attempted to minimize the heterogeneity in the immigrant group by categorizing them into different geographical regions (e.g. the Middle East, Africa etc.), we realize that there is still great heterogeneity within these groups, which were based on entire continents. Furthermore, some perpetrators with an immigrant background were not included in the logistic regression analyses, meaning that the result from these analyses is only applicable to perpetrators from Eastern Europe, the Middle East and Africa. Future research could benefit from more detailed analyzes of immigrant background, categorizing perpetrators in groups based on their country of origin.

Practical implications

This study’s findings support the notion that there are differences in the relevance of risk factors for the risk of IPV recidivism between native and immigrant perpetrators. These differences included both the presence of risk factors as well as differences in which factors uniquely contribute significantly to the elevated risk for IPV. These results demonstrate the importance of taking a perpetrator’s immigrant background into consideration when assessing the relevance of risk factors such as violent thoughts, threats and attitudes, escalation, general criminality and substance use problems when assessing and managing risk for severe IPV.

Overall, the results support previous claims that IPV perpetrators are a heterogeneous group of men (e.g. Dixon & Browne, 2003; Petersson & Strand, 2020) and that a one-size-fits-all approach to understanding and responding to IPV is outdated. To this end, risk assessment instruments, such as the B-SAFER, focus on individual-level risk factors, but it is important also to increase the understanding and awareness that these risk factors can have different causes depending on a perpetrator’s origin and thereby different relevance for the overall risk. For example, it is important to understand that not all risk factors stem from an individual’s poor psychosocial adjustment but could also be a result of the sociocultural context in one’s country of birth and upbringing. The results of this study, pertaining to the importance of also considering risk factors on societal and community levels, can be used in risk assessment training for police and other professions who are assigned with assessing and managing risk for future IPV.

Conclusion

Native perpetrators presented with more risk factors related to psychosocial adjustment, while immigrant perpetrators presented with more violent threats or thoughts and violent attitudes. This indicates that some risk factors on societal and community levels are more common for immigrant perpetrators. The greater extent of violent threats or thoughts and violent attitudes can arguably be seen as expressions of, or stemming from, societies and cultures with a more permissive view of IPV. In contrast, risk factors on the relational and individual levels are more present among native perpetrators. Surprisingly, when identifying which risk factors were more important for the elevated risk for further IPV, it was almost the same in both groups, where risk factors were mainly on the individual or relational level.

In sum, risk factors identified at the societal and community levels are not assessed as important or relevant for the summary risk rating for further IPV to the same extent as risk factors on relationship and individual levels. This might be due to the police lacking knowledge of the relevance of risk factors in different cultural contexts and need more education to this end. Thus, to prevent further IPV, the police must be able to acknowledge risk factors on a structural level and their importance when assessing risk for IPV among perpetrators with an immigrant background. It can therefore be argued, in general, that the training and knowledge required to identify important and relevant risk factors in all levels of Heise’s (2012) is insufficient. Risk assessments are instead made mostly from risk factors at the individual and relationship levels, indicating that risk factors from societal and community levels that potentially could contribute to an elevated risk might be overlooked when assessing the summary risk rating. These results are essential for the police to recognize and consider, since only victims facing an elevated risk for further IPV will be eligible for risk management and protective actions. By not considering this, victims who need protection might not be identified.

Moreover, since our results found that risk factors differed between native perpetrators and perpetrators with an immigrant background, future research should also investigate the impact of victim vulnerability factors in relation to the assessed risk of re-victimization. This would enhance the possibilities to get a better understanding of how risk factors and victim vulnerability factors influence the overall risk for further IPV in different groups of perpetrators and victims.

Acknowledgments

The authors would like to thank the police officers who participated in this longitudinal work for their valuable help in this research.

Disclosure statement

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

Additional information

Funding

This research project was partly funded by the Swedish Crime Victim Compensation and Support Authority (Brottsoffermyndigheten).