Risk Factors of Female Intimate Partner and Non-Intimate Partner Homicides

Violence against women is a major public concern whose prevention has become a target for policy makers, social and health care professionals, as well as researchers. In the context of this complex phenomenon the prevention of intimate partner violence (IPV) is a high priority, with efforts being especially directed at reducing the number of women who are killed by their partner. The homicide of a female intimate partner (the killing, and the man who kills) is also known as uxoricide (literally, the killing of the wife; uxor means wife in Latin). According to a review covering 66 countries (Stöckl et al., 2013) the worldwide prevalence of intimate partner homicides (IPH) among all female homicides is 38.55% (range 30.84-45.31), with the rate in high-income countries, including Spain, being 41.19% (range 30.84-44.45). Despite greater social awareness, the development of new policies, and changes to the law, the number of female victims of IPH remains relatively constant in Spain, with an average of 65 women killed by their current or former partner each year over the last decade (range 52-73).

One question considered by Moffitt, Krueger, Caspi, and Fagan (2000) was whether people who abuse their partners are the same as those who engage in other criminal behavior. In short, they found that “partner abuse and general crime represent different constructs that are moderately related; they are not merely two expressions of the same underlying antisocial propensity.” (p. 219). Recently, however, it has been argued that there is significant overlap between criminal violence and IPV (Piquero, Theobald, & Farrington, 2014). Research suggests that the specialization (commission of only a specific type of crime; in this case, only IPV) of partner violent men (PVM) depends on the sample, being less common in the prison population (Klein & Tobin, 2008; Loinaz, 2014).

Research by Campbell indicates that the risk factors related to IPH include a documented history of battering, estrangement (in the form of physically leaving or starting legal separation procedures), stalking, substance abuse, mental illness, and access to or availability of firearms (Campbell, 2012; Campbell & Glass, 2009; Campbell, Webster, & Glass, 2009; Campbell et al., 2003). It has also been documented that the major risk factor for IPH is prior domestic violence (Campbell, Glass, Sharps, Laughon, & Bloom, 2007). Jealousy, the woman’s desire to leave the man, and control seem to be the most common motivations for female IPH (Elisha, Idisis, Timor, & Addad 2010).

Research in the field suggests that people who kill their intimate partners may differ from other murderers, for example, by being “more conventional in terms of the levels of education, employment, persistent criminal behavior, and general use of physical violence.” (Dobash, Dobash, Cavanagh, & Lewis, 2004, p. 599). A comparison of intimate partner murders (IPM) with other murders showed that the perpetrators of IPM are more socially bonded (married and employed), more likely to have mental health problems, and more likely to use violence to meet emotional rather than instrumental needs (Thomas, Dichter, & Matejkowski, 2011). Recently, Juodis, Starzomski, Porter, and Woodworth (2014a) compared the correctional files of 37 males convicted of domestic homicide (DH) (including homicides of children, women’s new partners, family members, friends, and third-party interveners) with those of 78 non-domestic homicide (NDH) perpetrators (involving the killing of 34 women, nine children, and 29 men). Their results indicated a number of differences: NDH perpetrators were younger and more affected by psychopathic traits, DH males’ violence was more reactive than instrumental (emotional arousal in response to conflict or provocation), and NDHs were motivated by external gain (i.e., money or drugs). For their part, Belfrage and Rying (2004) studied all spousal homicides committed in Sweden between 1990 and 1999 (n =164). They found that psychiatric disorders were more prevalent among IPH offenders than in a control group of all other perpetrators of homicide in Sweden during the same period. They also found a four-fold higher rate of suicide among IPH offenders, and concluded that 80% of them could be characterized as mentally disordered. Kivivuori and Lehti (2012) examined 836 homicides committed in Finland between 2002 and 2012, comparing different homicide subtypes such as IPH against women and men, male-to-male homicides, or filicides. They concluded that people from socially disadvantaged groups are overrepresented among male IPH offenders, with crimes being closely linked to alcohol consumption and with most offenders being unemployed, having a long history of violence, and a prior conviction.

Dobash and Dobash (2015) analyzed three types of male-perpetrated homicides against women: IPH, sexual violence related homicides, and older women homicide, comparing the event and the perpetrators characteristics. The vast majority of men in all four groups had similar life experiences (i.e., serious antisocial and criminal behavior and families with social assistance). The summary of childhood experiences did not differentiate the group of men who murder intimate partner. There were also many similarities in the adult lives of four types of murders. Dobash and Dobash’s (2015) work consistently shows that “the backgrounds of men who murder women are in many aspects similar to those of men who murder other men. However, there are important differences, and this is specially the case of IPMurders.” (p. 247). Regarding their life course, they are regularly employed, have higher education and fewer alcohol or drug problems, and are less likely to engage in criminal behavior. One quarter of IPH offenders had conventional personal and demographic characteristics. Weizmann-Henelius et al. (2010) supported the conventionality of the IPH ofenders’ profile.

A recent work by Caman, Howner, Kristiansson, and Sturup (2017) concluded that IPH offenders (n = 46) were older, more likely to be employed, less likely to have been convicted, and had fewer persistent criminal histories compared to non-IPH offenders (n = 165) in a Swedish context. Perpetrators of IPH were also less likely to be intoxicated at the time of the offense although intoxication was a common feature among victims and perpetrators in both groups. These results demonstrate that perpetrators of IPH constitute a separate subtype of offender and, conceptually, ought to be treated separately. In a previous paper (Caman, Howner, Kristiansson, & Sturup, 2016), the same group compared all female (n = 9) and male (n = 36) IPHs committed in Sweden between 2007 and 2009, suggesting that female perpetrators are more likely to be unemployed, have more substance abuse problems, have minor differences in their criminal history, and are more often victimized by the male victim.

Differences in recidivism among types of homicide offenders have been also analyzed. Roberts, Zgoba, and Shahidullah (2007) found that the rate of recidivism (new violent or drug crime) was lowest (10%) among domestic violence IPH offenders, the figure being even lower than that for the accidental homicides group (17%). Domestic offenders had the highest proportion of arrests for technical parole violations (24%). It is important to point out that none of the different types of homicide offenders in this study committed another murder during the follow-up of a minimum of 5 years, and that the majority (65%) of domestic violence homicide offenders in this study were females. It should also be noted, as a recent review (Liem, 2013) shows, that recidivism among homicide offenders is rarely assessed and very little is known about this group.

Despite of all this knowledge regarding IPH, there is certain discrepancy regarding the profile of the IPH offenders and the possibility of predicting their crimes. As Dutton and Kerry (1999) assert, “there is some evidence that spousal homicide may not necessarily be predictable on the basis of nonlethal violence” (p. 298). Regarding criminal history, Eke, Hilton, Harris, Rice, and Houghton (2011) analyzed 146 men who attempted or committed an actual IPH and 42% had prior criminal charges, 15% had a psychiatric history, and 18% had both. Only 24% had no formal contact with either criminal justice or mental health systems before the crime. The authors concluded, on the basis of a retrospective assessment of 30 of these cases, that IPH perpetrators could potentially be identified as high-risk offenders prior to the homicide or attempted homicide because they scored in the highest category of risk on the ODARA (an actuarial IPV risk assessment tool). Therefore, they were not as conventional as found in other studies. Juodis et al. (2014a) coded the Danger Assessment (DA; Campbell et al., 2009) retrospectively, taking into account risk factors preceding the IPH. They asserted that, using the most liberal cutoff for the revised DA, 86.5% of domestic homicides (DHs) would have been identified as a homicide risk, suggesting that DHs are potentially preventable crimes. Moreover, 82.9% of cases showed elements of planning, such that most did not occur “out of the blue” as has been hypothesized by others (e.g., Dobash, Dobash, & Cavanagh, 2009). Elisha et al. (2010) also found that “contrary to the impression, some of the participants attempted to generate during the interviews, most of the homicide incidents were not impulsive, unplanned, or uncontrolled” (p. 512). However, these authors also note that not all cases included previous violence or signs of disturbed behavior and, therefore, considering other risk factors than those present in tools such the DA is needed (i.e., desire for general control, expressions of despair, depression, alienation, and isolation), an aspect also highlighted by Nicolaidis et al. (2003), that proposed consider the lethality risk of women anywhere along the spectrum of abusive relationships, not only the most severe abuse.

The performance of some risk assessment tools has been found to depend on whether the IPV is non-severe or life threatening (Watt, Storey, & Hart, 2013). A recent examination of the DA has found an unrealistically high estimate of risk for lethal IPV (Storey & Hart, 2014), such that these estimates should be treated with caution. If, as already noted, uxoricides are ‘ordinary’ people who often have no prior convictions or complaints against them, risk assessment becomes almost impossible. Consequently, the more we can learn about these people, the better we will be able to intervene or manage the risk. In light of these complexities a comprehensive approach to the assessment of IPH has emerged, namely what are known as fatality review teams. These are a kind of community partnership in which an alliance is formed among people and organizations from multiple sectors with the aim of achieving the common goal of preventing IPH (see Watt, 2003, 2008). Teams of this kind may be an alternative way of improving our knowledge about IPH, crimes, and offenders.

There remains a lack of consensus in the literature about IPH offenders. Some research indicates that this crime is more common among antisocial or pathological PVM (basically, those with borderline features) (Dixon, Hamilton-Giachritsis, & Browne, 2008) and that socially disadvantaged groups are overrepresented among male IPH offenders; for instance, “permanently unemployed male alcoholics with a long history of violence” (Kivivuori & Lehti, 2012, p. 73). Others, however, describe socially bonded and non-pathological offenders among these IPH (Dobash & Dobash, 2011; Dobash et al., 2009; Dobash et al., 2004; Dobash, Dobash, Cavanagh, & Medina-Ariza, 2007; Dutton & Kerry, 1999; Thomas et al., 2011). Some authors who have examined differences between IPH and non-IPH offenders argue that little has been done about male perpetrators (Juodis et al., 2014a), while others note that most research has focused on victims and incidents (Thomas et al., 2011). In addition, previous comparisons among homicide offenders have not focused on a group of female IPH versus female non-intimate-partner homicides (non-IPH). In our view, a comparison of this kind should be made with a similar type of victims and offenders (male offenders and female victims in our case).

The purpose of the present study is to explore whether men who kill their partner present different risk factors to those who kill women outside of a relationship. We analyze data from all cases of men convicted of partner murder (or attempted murder) in Barcelona between 2004 and 2009, comparing them with a sample of men convicted of killing women outside of an intimate relationship. The specific objectives are: a) to describe the characteristics of IPH sentenced in Barcelona, b) to analyze the risk factors present in these IPH cases, and c) to compare risk factors with those present in a sample of female homicides not related to intimate partner violence.

Sample

The sample consisted of 30 IPHs (murders) or attempted homicides committed between 2004 and 2009 in Catalonia (Spain) (21 convicted of murdering their female partners and 9 of attempted murder – tried to kill their intimate partner but did not reach the target). All perpetrators were serving a prison sentence and their mean age at the time of the crime was 42.40 years (SD = 15.52, range = 22-81). Sixty percent of offenders were from Spain, 10% from other European countries, and 30% from other countries. As a control group, 20 (all these cases available at this time) cases of non-IPH were selected from the same prison database. The mean age of this group was 38.85 (SD = 9.58, range = 26-57). Most of these men were Spanish (75%).

Measures

Data were gathered from different sources, mainly prison databases and sentencing files (summary of conviction). The conviction summaries of IPH perpetrators provided a comprehensive description of the offense, including police reports and forensic evidence, among others. Risk factors from IPH and non-IPH cases were gathered from the prison risk assessment protocol, known as RisCanvi (Andrés-Pueyo, Arbach-Lucioni, & Redondo, 2010). This is a risk assessment tool coded by the correctional staff (the psychologist, lawyer/criminologist, and social worker each scored a set of risk factors specific to their perspective). The RisCanvi provides an actuarial estimate of risk based on 43 risk factors grouped in five categories: criminal and penitentiary, biographical and personal, social, clinical, and personality factors. Risk factors are coded in three categories of response: presence (yes), absence (no), and probable presence (?).

Through a computer program (eRisCanvi), technicians evaluate the inmate every six months to manage their progression through the prison and rehabilitation programs to which they must participate in accord with the risk for violence level obtained in the RisCanvi. Assessments with the RisCanvi are carried out in an iterative way and allow the evaluation of the risk by means of an actuarial algorithm and a structured professional assessment of each inmate, violent recidivism, and other antinormative behavior inside the correctional facilities (Andrés-Pueyo, Arbach-Lucioni, & Redondo, 2018).

Procedure

Information on IPH offenders (including attempted homicide) was obtained through the criminal justice system. After receiving official approval from the Correctional Services of Catalonia, a file review was carried out for cases of interest, and risk factors for both those committing homicide and the control group were retrieved from the prison database. The control group consisted of all cases listed in the prison database that met the inclusion criteria: male perpetrators of homicide or murder involving a female victim with whom they had no intimate relationship. The definition of homicide, criminal act of violence by one or more human beings resulting in the death of one or more human beings (Caman et al., 2016) included murder but did not include manslaughter.

Data Analysis

Data analyses were carried out with SPSS v22 for Windows. The analysis was mainly descriptive, with bivariate comparisons of groups being performed using independent samples t-tests for continuous variables and chi-square tests for categorical variables (with Fisher’s exact test when the expected frequencies are too low).

Characteristics of Intimate Partner Homicides

Table 1 shows the mitigating and aggravating circumstances of completed and attempted IPH, as well as the type of weapon used, the place of the crime, and the victims involved.

Table 1

Characteristics of Attempted and Completed IPH Crimes (n = 30)

Only 23.3% of IPH offenders had a prior conviction and 9.5% prior imprisonment. The use of a sharp instrument (knife or similar) was the most common means of committing the offense (present in 50% of cases), followed by punching/hitting or similar (e.g., physical contact with a blunt object) (23.3%). The family home (53.3%) and the victim’s house (23%) were the most common locations for the crime. In 6.7% (2) of cases the victims included a child. The mean age of victims was 40.5 years (SD = 14.67, range = 19-76), and 32% of them were foreigners.

Risk Factors: IPH (n = 21) vs. Non-IPH (n = 20)

Table 2 shows the comparison of risk factors between the two groups of offenders: completed intimate partner homicides (IPH) and the control group of perpetrators of female homicide with no relationship to the victim (non-IPH). Offender age at the time of the crime differed significantly between the two groups (χ² = 8.409, p = .013), there being a higher proportion of older offenders among IPH cases (age > 28 = 85.7%, age 23-28 = 14.3%), and a higher proportion of younger offenders among cases of non-IPH (age > 28 = 50%, age 23-28 = 20%, age < 22 = 30%).

Table 2

IPH and Non-IPH Offenders Risk Factors Prevalence

As can be observed, both groups showed very similar prevalence across all risk factors. The only risk factors associated with significant differences were the presence of a criminal role and recklessness (as a temperamental disposition related to a chronic need to carry out risky activities, to have new experiences and to reject routine or isolated activity), both of which were more prevalent among the non-IPH group. Child adjustment problems were more prevalent among non-IPH (60% vs. 28.6 %), as well as personality disorder with anger (45% vs. 19%) but there were no statistical differences. In other risk factors both groups had similar prevalence.

This study has described the crime characteristics of IPH and has compared the associated risk factors with those presented by femicides not related to IPV. As in the study by Juodis et al. (2014a), offenders in the non-IPH group were younger at the time of the crime. There were also differences in the presence of a criminal role (Dobash et al., 2004) and recklessness, both of which were more prevalent among perpetrators of non-IPH. In contrast to some previous research (Campbell, 2012; Kivivuori & Lehti, 2012) we found no differences in levels of education, and perpetrators of IPH were not more socially disadvantaged. Furthermore, although employment problems were more prevalent in the non-IPH group, the difference was not significant. In contrast to previous findings (Belfrage & Rying, 2004; Thomas et al., 2011), there were no differences in psychological problems, although it should be noted that we were only able to consider a dichotomous risk factor rather than specific diagnoses or scores. Nonetheless, a recent work with serial homicide offenders (assessed with the MCMI-III) concluded that they were not different from single murders, or even just violent offenders, regarding the prevalence of mental disorders, and, therefore, these types of crimes share many features (Culhane, Hildebrand, Mullings, & Klemm, 2016). Men who assaulted or killed their partner were as likely as their non-IPH counterparts to be intoxicated during the crime and have alcohol problems, as Felson and Lane (2010) found. Problems related to substance abuse have been reported to be a risk factor for partner homicide (Campbell, 2012; Campbell & Glass, 2009; Campbell et al., 2009; Campbell et al., 2003), although they were not highly prevalent in our samples.

Although some research conclude IPH offenders can be identified as high-risk prior to the crime (Campbell et al., 2009; Eke et al, 201; Juodis et al., 2014), recidivism prediction is a complex target, and sometimes IPH offenders do not have records or complains. In a recent work with police risk assessment and management in 6,613 cases, López-Ossorio, González, Buquerín, García, and Buela-Casal (2017) found up to 46 indicators significantly related to the police recidivism. Even more, they concluded that some of the risk factors included in risk assessment tools may be related to a punctual episode, but do not work as good recidivism indicators. Further, relevant indicators as can be the perception of risk of the victim, may not be available in homicide cases if there are not previous complains. Even recent developments with a short version of DA (Messing, Campbell, & Snider, in press) show few improvements in prediction, adding specific risk factors, like strangulation, present in some cases. According to Dobash and Dobash’s (2015) work, IPH offenders are very similar to other homicide offenders but usually have conventional characteristics and backgrounds. Caman et al. (2017) also concluded that IPH offenders are more conventional than non-IPH offenders (more likely employed, less likely convicted, less likely intoxicated), something that we can only confirm regarding the criminal activity. Accordingly, it seems that the prediction of these cases can be very complicated.

Regarding the criminal role discrepancy, we must take into account that different features of the criminological profile of PVM depend on the subtype of offender. Only the antisocial subtype is characterized by generalized violence (non-specialization), criminal record, prior convictions, and a criminal career, all of these associated with a higher risk of recidivism (Carbajosa, Catalá-Miñana, Lila & Gracia, 2017; Cunha & Gonçalvez, 2013; Hilton, & Eke, 2016; Llor-Esteban, García-Jiménez, Ruiz-Hernández, & Godoy-Fernández, 2016; Loinaz, 2014; Petersson, Strand, & Selenius, 2016; Ruiz-Hernández, García-Jiménez, Llor-Esteban, & Godoy-Fernández, 2015). Subtyping IPV offenders and matching risk to management strategies is still an ongoing research and practice target, mainly between antisocial and family-only offenders (Loinaz, 2014; Petersson et al., 2016), although available research suggest that typologies are not used in practice because of their complexity (Ali, Dhingra, & McGarry, 2016).

In the UK, Dixon et al. (2008), following the typological proposal of Holtzworth-Munroe and Stuart (1994), concluded that 15% of the IPH offenders met the low criminality/low psychopathology profile, 36% the moderate-high criminality/high psychopathology profile, and 49% the high criminality/low-moderate psychopathology profile. Elisha et al., (2010) also identified three types of IPH through the qualitative assessment of 15 inmates in Israel (10 were convicted of murder, 3 of manslaughter, and 2 of attempted murder). In their typology, the betrayed husband killed his partner after discovering her ongoing sexual betrayal, which led to their family breakup; the abandoned obsessive lover murdered his intimate partner following her intention to leave him; and the tyrant murdered his intimate partner following an ongoing confrontation with her, which gradually escalated until he decided to kill her as a way of triumphing over her. In fact, leaving a violent relationship can be a risk situation, with offenders developing different strategies to gain access to women, restart the relationship, and so on (Gonzalez-Mendez & Santana-Hernandez, 2014).

The hypothesis that IPH are similar to other types of murderers (Cunha & Gonçalvez, 2013; Dobash & Dobash, 2015; Kivivuori & Lehti, 2012) is confirmed in this paper as far as both types of offenders were similar in their risk factors prevalence. In the same sense, Dobash and Dobash (2015) indicate that there were also many similarities in the adult lives of different types of murders and show that the backgrounds of men who murder women are in many aspects like those of men who murder other men. We can add, moreover, that IPH offenders are like men who kill adult women without intimate relationship. The implications of these results for the prevention of IPV murders indicate that in addition to considering the risk factors inherent to IPV, those that are also characteristic of violent offenders and homicides should be added.

This study has certain limitations that need to be addressed in future research. First, unlike some previous research based on all cases of spouse homicide in a country (Belfrage & Rying, 2004), our sample is small to allow for generalization of results, although it includes all cases in Catalonia in a period of 10 years. It can nonetheless be considered a first step in relation to this topic and as an important addition to the literature insofar as a comparison is made between intimate partner homicides and other femicides (female homicides), not general homicides or domestic homicides. Second, we did not differentiate between IPH offenders and IPV non-homicide offenders to see if there are differences in the progression from violence to homicide, something done by others (Cunha & Gonçalves, in press). Moreover, only risk factors available in the prison protocol were compared. In this respect, it would be interesting to analyze other variables in greater detail, for instance, the psychological profile by means of specific psychometric tools (e.g., for personality disorders) (Loinaz, Ortiz-Tallo, & Ferragut, 2012). These assessment limitations prevented us from analyzing typological proposals, as others have done (Dixon et al., 2008; Elisha et al., 2010). Recent studies (Debowska, Boduszek, Dhingra, Kola, & Meller-Prunska, 2015; Debowska, Boduszek, & Willmott, in press) also emphasize the weight of childhood exposure to violence in the development of cognitive distortions, and dating sexual violence-supportive attitudes among different types of offenders, including homicide offenders. Although child adjustment and problematic socialization items may include such kind of child abuse, it could be interesting to compare the specific prevalence of victimization in the same type of sample.

In conclusion, and as Thomas et al. (2011) indicate, “there remains disagreement in the literature about whether IPH offenders need to be treated as a unique group rather than as a member of the larger group of homicide offenders” (p. 293). Even though some authors argue that it is possible to identify IPH offenders prior to the crime, we agree with others that homicide may not be predictable in the same way as non-lethal violence (Dutton & Kerry, 1999). In terms of prevention, therefore, there is a need to develop empirically-validated risk assessment tools for identifying high-risk IPH perpetrators, as well as specific programs for treatment-resistant men (Day, Richardson, Bowen, & Bernardi, 2014; Juodis, Starzomski, Porter, & Woodworth, 2014b). A long criminal career and a high number of prior imprisonments are common among some partner-violent men in prison (Loinaz, 2014), over whom protection orders have little deterrent effect (Frantzen, Miguel, & Kwak, 2011; Strand, 2012). This highlights the need to intervene early and with alternatives to criminal proceedings (e.g., with risk management strategies). Others, however, as may be the case of some IPH, do not have the same risk profile, so prediction and management may be almost impossible. Moreover, it is important to assess specific variables related to IPH offenders, such as previous reports by the woman of violence, the perception of dangerousness, the availability of resources, and other variables related to victim protection (Gonzalez-Mendez & Santana-Hernandez, 2012). According to the structured approaches to violence risk assessment, the formulation paradigm is helpful to understand the complexity of risk factors, and the most appropriate steps to mitigate the level of risk (Cook, Murray, Amat, & Hart, 2014). Therefore, we should not focus only on the presence or absence of some risk factors but on the relevance of the factors that have been involved in homicide. One helpful tool to achieve this goal may be the already mentioned fatality review teams (see Watt, 2008).