Differences Between Sexual and Nonsexual Homicides of Women in the United States: Findings From the National Violent Death Reporting System

Abstract

Sexual homicide (SH) is the most severe outcome of sexual violence and disproportionately affects women. While SH is rare (<1% in the U.S.) and gravely understudied, it is among the most violent, feared, and well publicized forms of murder. Thus, examining predictors is pertinent to identifying targets for prevention and response efforts. Secondary analysis of 2015–2018 National Violent Death Reporting System data on 6461 female homicide victims age 20–64 was conducted to determine if SH represents a unique killing characterized by specific offender, victim, and incident profiles. Law enforcement and coroner/medical examiner narratives were reviewed to identify cases with sexual elements (N=324). Logistic regression estimated odds ratios with 95% confidence intervals. Findings highlight important differences between SH and non-SH. SH victims were more likely to be single (AOR=1.7,p=.006), have a substance abuse problem (AOR=1.4,p=.04), or engaged in prostitution (AOR=10.4,p<.001). SH suspects were more likely to be male (AOR=2.5,p=.04), use an illicit substance in the preceding hours (AOR=1.6,p=.03), or had recent contact with police (AOR=1.6,p=.01). SH was more likely to occur in a hotel/motel (AOR=3.0,p=.002), by asphyxiation (AOR =13.38,p<.001), be perpetrated against an acquaintance (AOR=1.64,p=.007), or be precipitated by another serious crime (AOR=2.1,p<.001). Findings advance our understanding of SH victim, suspect, and incident profiles, which can help to better inform police/investigative practices and crime prevention strategies/interventions as well as to improve how SH cases are managed in correctional programs for offenders who have the opportunity for release back into society.

Keywords

criminology homicide violent offenders

Introduction

Sexual violence is a serious public health issue that disproportionately affects females (Hillis et al., 2016; World Health Organization, 2012). According to the 2015 National Intimate Partner and Sexual Violence Survey, more than 52 million women in the U.S. experienced some form of sexual violence in their lifetime—44% of women compared to 25% of men (Smith et al., 2018). Occasionally, deviant sexual behaviors can trigger life-threatening violence and end in the murder of the victim. These incidents are referred to as sexual homicides (SH). SH is a rare, but often highly publicized, form of murder. Its prevalence is difficult to ascertain due to its infrequency, lack of a universally accepted definition, and insufficient crime details contained in most homicide databases (Karakasi et al., 2017). However, based on studies conducted in several countries, it is estimated that sexual homicides represent between 1 and 4% of all homicides but have been decreasing, including in the U.S. where the prevalence is less than 1% (James & Proulx, 2014; Sturup et al., 2019). Despite its rarity, SH is usually considered one of the most brutal and gruesome criminal acts. For this reason, these violent deaths typically receive substantial media coverage and provoke widespread fear about their occurrence in the community (Beauregard & Martineau, 2017; Smith et al., 2011). Subsequently, further investigation into this unique homicide is warranted and vitally important to the fields of criminal justice and violent crime prevention especially because it is studied far less than other types of murder.

Definition of Sexual Homicide

Lack of a uniform definition of SH makes comparing rates across studies and between countries difficult. Since the 19th century, researchers have made efforts to identify what constitutes SH; however, many have debated the “true” definition (Chan, 2015; Chopin & Beauregard, 2019a; von Krafft-Ebing, 1886). Some studies have defined SH as an intentional killing with evidence of a sexual element, while others believe that sexual gratification from the act of killing should not be included in the definition (Kerr et al., 2013). The most widely accepted operationalized definition of SH in research was developed by the Federal Bureau of Investigation (FBI) (Ressler et al., 1988). It suggests that a case should be considered SH if at least one of the following characteristics is evident: (1) attempted or completed sexual intercourse (vaginal, oral, anal); (2) sexual interest admitted by the offender; (3) attire or lack of attire; (4) exposure of sexual body parts; (5) sexual positioning of the body; (6) insertion of foreign objects into body cavities; (7) substitute sexual activity; (8) semen on or nearby the body; or (9) sadistic fantasies reported by the offender.

Sexual Homicides (SH) Compared to Nonsexual Homicides (NSH)

Despite considerable media coverage of SH, little scientific attention has been devoted to it compared to other homicides (NSH). Nevertheless, previous studies have observed uniqueness in the characteristics and circumstances surrounding SH. For example, demographic characteristics of both the victim and perpetrator were determined to be important predictors of SH. Women are more likely to be victims of SH, whereas men are more often both SH perpetrators and NSH victims (Karakasi et al., 2017; Smith et al., 2011). Additionally, studies conducted in France and Scotland found that, compared to NSH, younger victims and offenders both had higher odds of being involved in SH (Chopin & Beauregard, 2019a; Skott et al., 2018). These findings are supported by descriptive SH studies conducted in Canada and the U.S. (James & Proulx, 2014).

Research shows that the relationship between victim and suspect is a key factor associated with SH perpetrated against women despite inconsistencies among findings. A Finnish study conducted by Hakkanen-Nyholm et al. (2009) revealed that an attack on a stranger was equally uncommon for SH and NSH offenders. Conversely, several recent studies in various countries found that strangers and acquaintances were more likely to perpetrate SH (Chan et al., 2019; Sturup et al., 2019). Other researchers discovered that acquaintances and intimate partners were more common offenders of NSH (Chopin & Beauregard, 2019a; Skott et al., 2018).

Additional unique characteristics of SH include the location of the incident and the weapon used in the killing. Across studies assessing location, there was not a clear consensus about typical incident location; however, the most common locations were the victim’s place of residence and outdoors (Chopin & Beauregard, 2019a; James & Proulx, 2016; Sturup et al., 2019). While SH often occurred in public in other countries, homicides of the same nature in the U.S. often took place inside private residences (James & Proulx, 2014; Skott et al., 2018; Smith et al., 2011). Killing methods have also varied across studies, although most revealed that strangulation and personal weapons are more common in SH and firearms are used more frequently in NSH (Chan et al., 2019; James & Proulx, 2016).

Extant literature highlights prostitution and criminal history as other factors that increase SH risk. Some research has shown that prostitution is an important victim risk factor when comparing SH and NSH (Quinet, 2011) while others found an association at the bivariate level but was attenuated in multivariable analysis (Chopin & Beauregard, 2019a; Skott et al., 2018). In terms of criminal background, investigators have discovered that sexual murderers tend to have criminal histories or have been convicted of a violent crime (Beauregard & Martineau, 2012; Sturup et al., 2019).

Use of alcohol and other substances by the victim or offender can contribute to an increased risk of SH; however, there is no consensus among studies conducted. Sturup et al. (2019) found no differences in intoxication between sexual and nonsexual groups of Swedish victims. French and Chinese investigators learned that the sexual murderers they studied were less likely to be intoxicated during the crime than other homicide offenders (Chan et al., 2019; Chopin & Beauregard, 2019a), whereas a Finnish study showed that sexual murderers have a higher propensity for substance abuse (Hakkanen-Nyholm, 2009).

Purpose of the Current Study

The overall intent of this paper is to present meaningful findings that aid criminal justice, law enforcement, and SH preventionists in improving how SH cases are managed in correctional programs for offenders who have the opportunity for release back into society. Much of the SH literature published in the last 10 years has explored SH versus non-homicidal sex offenses. This study contributes to the existing literature in several ways. First, many of these studies have been limited by their unrepresentativeness and small sample sizes, and there has been less focus on the differences and similarities with other homicides (Beauregard & Marineau, 2012; Smith et al., 2011; Sturup et al., 2019). This exploratory study expounds on prior research by including a nationally representative sample of homicides of U.S. females aged 20–64. Second, to our knowledge, only three studies comparing SH to NSH were published in the past 10 years (Chopin & Beauregard, 2019a; Skott et al., 2018; Sturup et al., 2019). These researchers observed notable differences in their study findings, but all concluded that the characteristics and circumstances precipitating sexual murder are unique from other types of homicide. For these reasons, broadening the knowledge base on SH is critically important. The current project expands the growing literature on SH by building upon prior studies, such as U.S. studies conducted by Chan et al. (2013, 2015, 2019) and the only published study that used data from the National Violent Death Reporting System (NVDRS) to explore SH (Smith et al., 2011), to explore differences between SH and NSH. Specifically, we hypothesized that SH of women in the U.S. is distinct from NSH in terms of offender, victim, and incident characteristics (e.g., weapon choice, victim-offender relationship, location, etc.).

Methods

This study used data collected by the NVDRS. In 2002, the Centers for Disease Control & Prevention (CDC) launched the NVDRS in six states to better understand and prevent suicide and homicide. NVDRS is an ongoing, state-based, active surveillance system that collects data on all violent deaths and their circumstances (see Blair et al., 2016 for detailed description of NVDRS). As of 2018, NVDRS includes all 50 states, the District of Columbia, and Puerto Rico, serving as a source of data for violence research and prevention (Centers for Disease Control and Prevention, 2018). The system links data from death certificates, law enforcement (LE) reports, coroner/medical examiner (CME) records, and toxicology reports using standardized protocols. Homicide cases can involve more than one offender and/or more than one victim.

Study Sample and Procedure

Data collected from 2015 to 2018 for all homicide cases involving female decedents aged 20–64 were extracted from the NVDRS to perform secondary analyses. This sub-group was selected because sexual violence and SH are substantially more common among women than men (Hillis et al., 2016; Karakasi et al., 2017). Moreover, including male homicide victims in the study may not broaden our understanding of distinctions between SH and NSH beyond trivial differences based on gender (Skott et al., 2018). Females under age 20 or over age 65 were not included because sexual victimization of children and seniors represents a distinctive crime process warranting specialized examination beyond the scope of this study (Proulx et al., 2018). Therefore, our findings cannot be generalized to these excluded populations.

This paper focuses on SH, which is defined as homicide with a sexual element. SH is not a measure that is captured in the NVDRS. Therefore, a content analysis was conducted by two researchers, which consisted of reviewing the LE and CME narratives of target homicide cases to identify situational information that explained or described the incident. Since there were 6461 cases that met the inclusion criteria, before the review began, machine learning techniques were employed to narrow the search results. Statistical analysis software (SAS Version 9.4, PROC SQL) was used to perform a text/word search and extract narratives of the cases that contained sexual element keywords from a predetermined list, which was based on criterion from the widely-utilized Ressler et al. (1988) definition of SH. Additionally, a CDC program was used to identify additional words and phrases associated with the sexual element keywords, which expanded the list of search terms to make it as exhaustive as possible (Appendix A). After applying the text search to both narrative types, the sample was reduced to 671 possible SH cases. Narratives of these cases were reviewed by the authors to eliminate false positives and code the new SH variable (detailed in Appendix A). Coding by the two reviewers was compared and all discrepancies were discussed, reconciled, and coded to consensus (inter-rater reliability: 92.1%). This process yielded a final sample of 324 SH and 6137 NSH cases representing diverse backgrounds (socioeconomics, age, race, ethnicity, and geography). Precipitating circumstances were known for 91.7% of SH cases (n = 297) and 86.1% of NSH cases (n = 5282). Cases were stratified by homicide group for victim, suspect, and incident characteristics.

Measures

Sexual homicide is the dependent variable, which was coded as 1 = yes and 0 = no. There are three groups of independent variables for the characteristics and known circumstances related to the victim, suspect, and homicide itself (Table 1). These measures were chosen because previous empirical research has shown that they are important homicide risk factors of and necessary for comparing SH to other forms of murder and non-homicidal sexual offenses (Chopin & Beauregard, 2019a, 2019b; Skott et al., 2018; Sturup et al., 2019). [For precipitating circumstance variables, “no” and “unknown/not available” are grouped together because the NVDRS cannot determine whether the circumstance did not occur or occurred but was not discovered during the LE or CME investigation (National Violent Death Reporting System (NVDRS), 2020).]

Table 1.

Characteristics of Female Homicide Victims Age 20–64, 2015–2018 (N = 6461).

Characteristic	n	%
Victim characteristics
Age in years (standard deviation)	6461	38.3 (12.4)^a
Race/ethnicity
White, non-Hispanic	2645	47.6
Black, non-Hispanic	1845	33.2
Hispanic^b	759	13.7
Other^c	311	5.6
Marital status
Single, never married	3088	48.4
Married	1697	26.6
Separated, divorced, or widowed	1600	25.1
Education level
Less than high school	1194	19.0
High school graduate/General Educational Development (GED)	2681	42.6
Some college/Associate degree (AA) degree	1667	26.5
Bachelor’s	523	8.3
Graduate (Master’s, doctorate, or professional)	228	3.6
Victim had a substance abuse problem	844	15.1
Victim suspected of alcohol use in hours before homicide	679	14.4
Victim had a mental health problem	434	7.8
Victim experienced violence in month before homicide	324	5.8
Victim had an alcohol problem	245	4.4
Prostitution-related activities precipitated homicide	87	1.6
Suspect characteristics
Age in years (standard deviation)	4265	38.0 (13.4)a
Gender
Male	4421	93.8
Female	290	6.2
Race/ethnicity
White, non-Hispanic	1815	45.0
Black, non-Hispanic	1618	40.0
Hispanic^b	452	11.2
Other^c	153	3.8
Suspect recently released from an institution
No evidence of recent release	4046	91.6
Jail, hospital, psychiatric treatment, alcohol and other drug disorders treatment	140	4.2
History of victim abuse by suspect	610	13.3
Suspect had contact with police in past 12 months	548	12.4
Suspected alcohol use in hours before homicide	439	9.9
Suspected substance use in hours before homicide	371	8.4
Homicide was direct result of suspect’s mental illness	344	7.5
Incident characteristics
Location of incident
House/apartment	3754	69.5
Outdoors^d	791	14.6
Motor vehicle	430	8.0
Commercial establishment (bar, bank, etc.)	146	2.7
Hotel/motel	103	1.9
Office or abandoned building, factory, warehouse	73	1.4
Method of killing
Firearm/gun	3861	61.3
Sharp instrument	1108	17.6
Blunt instrument	483	7.7
Hanging, strangulation, suffocation	384	6.1
Personal weapons	218	3.5
Victim–suspect relationship
Spouse/current boyfriend or girlfriend	2150	49.9
Person known/acquaintance	935	21.7
Family member	490	11.4
Ex-spouse, ex-boyfriend, ex-girlfriend	445	10.3
Stranger	289	6.7
Homicide related to intimate partner violence	2902	52.0
Argument/conflict led to homicide	1924	34.5
Homicide was precipitated by another serious crime	1170	21.0
Victim was killed by random act of violence	167	3.0
Homicide precipitated by abuse/neglect by caretaker	84	1.5

^aMean age (standard deviation).

^bPersons of any race.

^cAmerican Indian/Alaska Native, Asian/Pacific Islander, and two or more races.

^dStreet, road, sidewalk, alley, highway, parking lot, park, playground, natural area, railroad tracks, bridge, cemetery, farm.

Victim demographic variables included age, race/ethnicity, marital status, and education level. Six dichotomous victim-specific precipitating circumstances were also assessed to examine potential risk factors that may have facilitated the murder (coded 1 = yes, 0 = no, not available, unknown): (1) substance abuse problem; (2) alcohol problem; (3) used alcohol in the hours before the incident; (4) mental health problem (i.e., depression, bipolar disorder, schizophrenia, anxiety disorder, post-traumatic stress, attention deficit/hyperactivity disorder, eating disorder, or obsessive-compulsive disorder); (5) prostitution or prostitution-related activities played a role; and (6) experienced violence in the month preceding the homicide.

Suspect demographics included age, gender, and race/ethnicity. Whether the suspect had been released from a jail/detention center, hospital, or psychiatric/substance abuse treatment was assessed. Relationship between victim and suspect indicated spouse or current girlfriend, ex-spouse or ex-girlfriend, family member, person known/acquaintance, or stranger. Five dichotomous suspect-specific precipitating circumstances were included (1 = yes, 0 = no, not available, unknown): (1) used alcohol in the hours before the incident; (2) used drugs in the hours preceding the homicide; (3) attack was the direct result of mental illness (defined above); (4) had contact with the police in the past 12 months; and (5) history of abusing the victim.

Several variables related to the homicide incident itself were measured as well. Location where the homicide occurred included: house/apartment; outdoors (street, alley, park, etc.); motor vehicle; office building or factory/warehouse; commercial establishment (bar, bank, etc.); hotel/motel; or other. Type of weapon included: firearm; sharp instrument; blunt instrument; asphyxiation; personal weapon; or other. Five dichotomous incident-specific precipitating circumstances were also measured (1 = yes, 0 = no, not available, unknown): (1) argument or conflict; (2) another serious crime preceded the homicide; (3) victim was killed by a random act of violence; (4) immediate or ongoing conflict or violence between current or former intimate partners; and (5) death related to or precipitated by abuse or neglect by a caretaker.

Statistical Analysis

Descriptive statistics were computed for all outcome and explanatory variables examined. Chi-square tests (and Fisher’s exact when appropriate) and independent samples t-tests assessed differences in victim, suspect, and incident characteristics. Statistical significance was regarded as a p-value below .05. Statistically significant variables at the bivariate level were predictors added by group (victim, offender, incident) into multivariable logistic regression models to estimate adjusted odds ratios with 95% confidence intervals. Analyses were performed using SAS, version 9.4.

Results

Study Sample

Table 1 shows descriptive statistics on all study variables of the sample (N = 6461; SH = 324, NSH = 6137), which represents the diversity of the U.S. females in terms of race/ethnicity, marital status, and education, among other factors. 44,121 homicides occurred in the U.S. between 2015 and 2018. Less than 1% (0.73%) were SH cases that met the inclusion criteria, which is similar to the prevalence found in other U.S. studies (James & Proulx, 2016; Karakasi et al., 2017). Tables 2 –7 only include cases where circumstances related to the homicide were known. Most variables were associated with SH at p < .05 in bivariate analyses.

Table 2

Differences in Victim Characteristics Between Homicide Groups.

Characteristic	Sexual	Nonsexual	χ2 (df)
Characteristic	(n = 297)	(n = 5282)	χ2 (df)
Age in years (M; standard deviation)	297 (38.3;12.2)	5282 (38.3;12.3)	t(5577) = −14^a
Race/Ethnicity			13.1(3)**
White, non-Hispanic	155 (52.4%)	2490 (47.3%)
Black, non-Hispanic	80 (27.0%)	1765 (33.5%)
Hispanic	34 (11.5%)	725 (13.8%)
Other, non-Hispanic	27 (9.1%)	284 (5.4%)
Marital status			12.6(2)**
Single, never married	162 (55.3%)	2414 (46.2%)
Separated, divorced, or widowed	74 (25.3%)	1333 (25.5%)
Married	57 (19.5%)	1476 (28.3%)
Education level			1.39(4)
Less than high school	58 (20.4%)	951 (18.5%)
High school graduate/GED	114 (40.1%)	2184 (42.4%)
Some college/AA degree	76 (26.8%)	1391 (27.0%)
Bachelor’s	27 (9.5%)	436 (8.5%)
Graduate/professional	9 (3.2%)	187 (3.6%)
Had substance abuse problem	79 (26.6%)	765 (14.5%)	32.2(1)***
Suspected alcohol use before homicide	53 (20.8%)	626 (14.0%)	9.0(1)**
Had mental health problem	37 (12.5%)	397 (7.5%)	9.6(1)**
Prostitution precipitated homicide	31 (10.4%)	56 (1.1%)	80.4(1)***
Experienced violence month before	27 (9.1%)	297 (5.6%)	6.2(1)*
Victim had alcohol problem	25 (8.4%)	220 (4.2%)	12.1(1)***

^aPooled independent samples t-test.

* p <.05; ** p<.01; *** p<.0001.

Table 3.

Differences in Suspect Characteristics Between Homicide Groups.

Characteristic	Sexual	Nonsexual	χ2 (df)
Characteristic	(n = 255)	(n = 4456)	χ2 (df)
Age in years, mean	226 (37.1)	3813 (38.3)	t(4037) = 1.32^a
Gender
Male	250 (98.0%)	4171 (93.6%)	8.2(1)**
Female	5 (2.0%)	285 (6.4%)
Race/ethnicity			7.6(3)
White, non-Hispanic	89 (40.5%)	1726 (45.2%)
Black, non-Hispanic	98 (44.6%)	1520 (39.8%)
Hispanic	19 (8.6%)	433 (11.3%)
Other, non-Hispanic	14 (6.4%)	139 (3.6%)
Suspect recently released			3.8(1)
No evidence of recent release	158 (92.9%)	3062 (96.0%)
Jail, hospital, treatment	12 (7.1%)	128 (4.0%)
Contact with police in past 12 months	46 (18.8%)	502 (12.0%)	9.7(1)**
History of victim abuse by suspect	40 (15.9%)	570 (13.1%)	1.6(1)
Alcohol use before homicide	38 (15.5%)	401 (9.6%)	8.8(1)**
Substance use before homicide	29 (12.0%)	342 (8.2%)	4.4(1)*
Direct result of suspect mental illness	24 (9.5%)	320 (7.4%)	1.5(1)

^aPooled independent samples t-test.

* p <.05; ** p<.01; *** p<.0001.

Table 4

Differences in Incident Characteristics Between Homicide Groups.

Characteristic	Sexual	Nonsexual	χ2 (df)
Characteristic	(n = 297)	(n = 5282)	χ2 (df)
Location of incident			50.5(6)***
House/apartment	178 (64.5%)	3576 (69.8%)
Outdoors	47 (17.0%)	744 (14.5%)
Hotel/motel	14 (5.1%)	89 (1.7%)
Building, factory, warehouse	12 (4.4%)	61 (1.2%)
Commercial establishment	9 (3.3%)	137 (2.7%)
Motor vehicle	7 (2.5%)	423 (8.3%)
Other	10 (3.3%)	131 (2.3%)
Method of killing			376.3(5)***
Hanging, strangulation, suffocation	101 (31.9%)	382 (6.4%)
Sharp instrument	70 (22.1%)	1038 (17.4%)
Firearm/gun	58 (18.3%)	3803 (63.6%)
Blunt instrument	44 (13.9%)	340 (5.7%)
Personal weapons	34 (10.7%)	184 (3.1%)
Other	10 (3.2%)	236 (3.9%)
Victim–suspect relationship			35.3(4)***
Spouse, current boy/girlfriend	96 (44.0%)	2052 (51.9%)
Person known/acquaintance	79 (36.2%)	793 (20.0%)
Ex-spouse, ex-boy/girlfriend	17 (7.8%)	428 (10.8%)
Stranger	14 (6.4%)	266 (6.7%)
Family member	12 (5.5%)	417 (10.6%)
Related to intimate partner violence	96 (44.0%)	2052 (51.9%)	3.4(1)
Precipitated by another serious crime	112 (37.7%)	1058 (20.0%)	53.0(1)***
Argument/conflict led to homicide	67 (22.6%)	1857 (35.2%)	19.8(1)***
Precipitated by abuse/neglect by caretaker	11 (3.7%)	73 (1.4%)	7.5(1)**
Homicide was random act of violence	8 (2.7%)	159 (3.0%)	0.10(1)

* p <.05; ** p<.01; *** p<.0001.

Table 5.

Relationship Between Sexual Homicide and Victim Characteristics.

Characteristic	Unadjusted Odds Ratios (95% CI^a)	Adjusted Odds Ratios^b,c (95% CI)
Race/ethnicity
White, non-Hispanic	Ref.^d	Ref
Black, non-Hispanic	0.74 (0.57, 0.96)*	0.75 (0.54, 1.03)
Hispanic	0.82 (0.57, 1.17)	0.90 (0.58, 1.38)
Other, non-Hispanic	1.47 (0.97, 2.24)	1.74 (1.10, 2.76)*
Marital status
Married	Ref	Ref
Single, never married	1.68 (1.25, 2.26)**	1.70 (1.17, 2.48)**
Separated, divorced, or widowed	1.42 (1.01, 2.00)*	1.10 (0.74, 1.63)
Mental health problem	1.75 (1.22, 2.51)**	1.24 (0.82, 1.90)
Substance abuse problem	2.27 (1.74, 2.95)***	1.40 (1.01, 1.95)*
Alcohol problem	2.25 (1.46, 3.46)**	1.44 (0.86, 2.41)
Alcohol use in hours before homicide	1.61 (1.18, 2.21)**	1.34 (0.95, 1.89)
Experienced violence in month before	1.79 (1.19, 2.70)**	1.51 (0.94, 2.41)
Prostitution precipitated homicide	11.49 (7.30, 18.09)***	10.43 (6.23, 17.47)***

^aCI = confidence interval.

^bOnly includes odds ratios for statistically significant variables in bivariate analysis.

^cControlling for age.

^dRef = reference category.

* p <.05; ** p<.01; *** p<.0001.

Table 6.

Relationship Between Sexual Homicide and Suspect Characteristics.

Characteristic	Unadjusted Odds Ratios (95% CI^a)	Adjusted Odds Ratios^b,c (95% CI)
Gender
Male	3.52 (1.44, 8.60)**	2.50 (1.01, 6.15)*
Female	Ref.^d	Ref
Contact with police in past 12 months	1.66 (1.19, 2.31)**	1.60 (1.10, 2.33)*
Alcohol use in hours before homicide	1.72 (1.20, 2.46)**	1.25 (0.80, 1.95)
Substance use in hours before homicide	1.53 (1.02, 2.28)*	1.65 (1.06, 2.56)*

^aCI = confidence interval.

^bOnly includes odds ratios for statistically significant variables in bivariate analysis.

^cControlling for age and race.

^dRef = reference category.

* p <.05; ** p<.01; *** p<.0001.

Table 7.

Relationship Between Sexual Homicide and Incident Characteristics.

Characteristic	Unadjusted Odds Ratios (95% CI^a)	Adjusted Odds Ratios^b (95% CI)
Location of incident
House/apartment	Ref.^c	Ref
Outdoors	1.27 (0.91, 1.77)	1.32 (0.81, 2.15)
Motor vehicle	0.33 (0.16, 0.71)**	0.83 (0.35, 1.97)
Building, factory, warehouse	3.95 (2.09, 7.47)***	1.87 (0.62, 5.71)
Commercial establishment	1.32 (0.66, 2.63)	1.57 (0.59, 4.17)
Hotel/motel	3.16 (1.76, 5.66)**	3.00 (1.50, 5.99)**
Other	1.88 (0.94, 3.79)	1.24 (0.44, 3.52)
Method of killing
Firearm/gun	Ref	Ref
Sharp instrument	4.42 (3.10, 6.30)***	4.70 (3.02, 7.30)***
Blunt instrument	8.49 (5.65, 12.8)***	7.68 (4.51, 13.06)***
Hanging, strangulation, suffocation	17.34 (12.34, 24.35)***	13.38 (8.53, 20.98)***
Personal weapons	12.12 (7.74, 18.97)***	11.25 (6.47, 19.59)***
Other	2.78 (1.40, 5.51)**	2.71 (1.22, 6.03)*
Victim–suspect relationship
Spouse, current boy/girlfriend	Ref	Ref
Ex-spouse, ex-boy/girlfriend	0.85 (0.50, 1.44)	0.99 (0.56, 1.75)
Family member	0.61 (0.33, 1.13)	0.57 (0.31, 1.08)
Person known/acquaintance	2.13 (1.56, 2.90)***	1.64 (1.14, 2.36)**
Stranger	1.13 (0.63, 2.00)	1.11 (0.57, 2.16)
Precipitated by abuse/neglect by caretaker	2.74 (1.44, 5.23)**	1.87 (0.84, 4.17)
Precipitated by another serious crime	2.54 (2.00, 3.22)***	2.08 (1.46, 2.95)***
Argument/conflict led to homicide	0.60 (0.46, 0.79)**	0.64 (0.46, 0.88)**

^aCI = confidence interval.

^bModel only includes odds ratios for statistically significant variables in bivariate analysis.

^cRef = reference category.

* p <.05; ** p<.01; *** p<.0001.

There were several statistically significant differences between SH and NSH that reflect the sample’s diversity. SH victims were more likely to be white, single/never married, and to have had a substance abuse problem, used alcohol in the hours before the homicide, had a mental health problem, been involved in prostitution-related activities, experienced violence in the month before the murder, and had an alcohol problem (Table 2). SH suspects were more likely to be male, have had contact with police in the past 12 months, and used alcohol and/or an illicit drug in the hours before the homicide (Table 3). In terms of the homicide itself (Table 4), SH was more likely to occur outdoors and be precipitated by another serious crime or abuse/neglect by a caregiver. The SH victim was more likely to be an acquaintance of the suspect and killed by asphyxiation (hanging, strangulation, or suffocation). Conversely, an argument or conflict before the incident was less likely in SH than NSH.

Victim Characteristics Associated with Sexual Homicide

In the crude analyses, two demographic characteristics and all six precipitating circumstances were associated with SH. Associations with race/ethnicity, marital status, a substance abuse problem, and prostitution remained significant after adjustment (Table 5). On average, women who were American Indian/Alaskan Native, Asian/Pacific Islander, or two or more races had 74% higher odds of SH than white women (p = .02). Compared to married women, those who were single/never married experienced 70% higher odds of death by SH (p = .006). Having a substance abuse problem was significantly associated with SH (AOR = 1.40, p = .04), and victims engaged in prostitution were over 10 times more likely to die by SH (p < .001).

Suspect Characteristics Associated with Sexual Homicide

Bivariate analyses revealed that gender and three precipitating circumstances were significantly associated with SH (Table 6). Based on multivariable analysis, men were two and a half times more likely than women to commit SH (p = .04). Suspects who had police contact in the preceding 12 months and those who used illegal substances in the hours before the homicide presented higher odds of committing SH (respectively, AOR = 1.60, p = .01; AOR = 1.65, p = .03).

Incident Characteristics Associated with SH

In the crude analyses, all six modus operandi/precipitating circumstances were associated with SH and five remained statistically significant in the multivariable logistic regression. In comparison to a house or apartment, SH was three times more likely to occur in a motel/hotel (p = .002). Compared to firearms, all other killing methods presented greater odds of being used in SH; asphyxiation was 13 times more likely, and a personal weapon was 11 times more likely (p < .001, respectively). The odds of SH were two times higher when another crime preceded the homicide (p < .001). Killing an acquaintance presented higher odds than murdering a wife or girlfriend (AOR = 1.64, p < .007). Conversely, the odds of a killing being motivated by an argument or conflict were lower for SH (AOR = 0.64, p = .007).

Discussion

SH is the most severe outcome of sexual violence and disproportionately affects women. The goal of this study was to determine if SH represents a unique killing characterized by specific offender, victim, and incident profiles. This study extends previous research that has largely focused on describing SH offenders or other forms of homicide. It adds to the literature on key determinants of SH perpetrated against women in all 50 states, including the District of Columbia and Puerto Rico. To the authors’ knowledge, this is one of the first U.S. studies to use qualitative and quantitative data from a nationally representative surveillance system to compare SH with NSH while controlling for victim gender (a potential major confounder based on past studies). The primary contribution of this article is a better understanding of specific risk factors and precipitating circumstances associated with SH.

Victim Characteristics

Consistent with prior research, in the current study, the likelihood of becoming a victim of SH was highest among women who were single/never married, had a substance abuse problem, or were involved in prostitution (James & Proulx, 2016; Karakasi et al., 2017; Smith et al., 2011). Research shows that greater susceptibility to SH among some women may be due to lifestyle behaviors that make them more prone to victimization. The current study revealed that over one-fourth of the SH victims assessed had a substance abuse problem and more than 20% were suspected of using alcohol in the hours before their murder. Studies highlighted by Karakasi et al. (2017) indicate that SH suspects may identify these vulnerabilities prior to selecting them as targets.

Prostitution was another susceptibility factor more prevalent in SH in our sample. Related activities played a precipitating role in 10% of SH cases compared to only 1% of NSH cases. This finding represents a lower proportion than Beauregard and Martineau (2012) observed, where 20% of SH victims were prostitutes. This discrepancy may be because the number of decedents involved in prostitution in the present study (i.e., where circumstances were known) was small (n = 87; 1.6%). Nevertheless, there is consensus among researchers that prostitution is an important variable to consider when comparing SH to NSH (Chan & Beauregard, 2019; Martineau & Beauregard, 2012; Skott et al., 2018; Quinet et al., 2011). The nature of prostitution puts most of these women in high-risk situations (e.g., dangerous locations with few, if any, witnesses). Further, they are specifically targeted by some offenders because prostitutes are generally known to be nomadic and less likely to notify friends or family of their whereabouts (Quinet et al., 2011). These factors increase the likelihood of sexual violence and death.

Suspect Characteristics

Extant research suggests that the sexual murderer is a distinct type of offender distinguishable from other homicide offenders and nonhomicidal violent sexual offenders (Chopin & Beauregard, 2019a; Karakasi et al., 2017). The present study corroborated this assertion and found that SH suspects were most often male, used substances before the homicide, and had a criminal history. However, the SH offenders we assessed did not differ significantly from nonsexual murderers by age, race/ethnicity, history of victim abuse, or mental illness (defined earlier) as has been reported by several other researchers (James & Proulx, 2014, 2016).

Comprehensive systematic reviews conducted by Karakasi et al. (2017) as well as James and Proulx (2014, 2016) revealed that SH offenders tend to have alcohol and other drug (AOD) disorders or used at least one of these substances before committing SH. Our findings support this discovery to an extent. Based on multivariable logistic regression, the odds of SH were statistically higher among offenders who had used an illicit substance in the hours leading up to the homicide compared to those who had not. Further, while only 16% of SH offenders were suspected of using alcohol in the hours before the homicide, less than 10% of NSH murderers had done so; however, there was no statistically significant association observed in the multivariable analysis. One major difference between this and other SH studies is the current sample included unsolved crimes, which meant that important circumstantial information relevant to the murder could not be obtained for all cases. For instance, whether the suspect had used alcohol or drugs prior to the homicide was unknown for 26% of cases. Thus, such data limitations might contribute more to the lower prevalence of AOD risk factors in the present sample compared to prior research rather than to an actual absence or statistical insignificance of these characteristics.

Statistically significant differences were found in the criminal histories of murderers in the current study, with a greater proportion of SH than NSH suspects having contact with police in the past 12 month (19% vs. 12%) and the likelihood of SH being 60% higher among murderers with a criminal background compared to those with no such history. Similarly, general criminality among SH offenders was discovered by other researchers, but with much higher proportions (Beauregard & Martineau, 2012; James & Proulx, 2014, 2016). For instance, 75% of offenders in a Swedish study had been convicted of at least one criminal act (Sturup et al., 2019). The authors attributed this to the detailed offender data collected via their police registries, which might partially explain the discrepancy between studies (i.e., suspect details missing from case narratives in the NVDRS dataset).

Incident Characteristics

Research shows that most of the differences between SH and NSH concern modus operandi (Chopin & Beauregard, 2019a; Skott et al., 2018). The current study is no different. In our sample, SH was associated with location, method of killing, victim–suspect relationship, and the incident being precipitated by another serious crime. Our findings support other researchers’ assertions that incident variables are important for disaggregating SH and NSH.

Consistent with previous research, the largest proportion of SH (and NSH) occurred in a house or apartment (Beauregard & Martineau, 2012; Chopin & Beauregard, 2019a; James & Proulx, 2016; Skott et al., 2018). This finding dispels the belief that women are safer when at home than elsewhere. Beauregard and Martineau (2012) hypothesized that this is because the same features attracting a burglar (e.g., easy access or clear view into the home) probably make the home more appealing to sex offenders as well. However, when compared to a residence, a hotel/motel was a significantly more probable location for SH in the present study. This corroborates Skott et al. (2018), who found that NSH occurred in a private or residential location significantly more often than SH. Collectively, these findings highlight the need to consider location when investigating homicides to determine likelihood of sexual motivation.

In terms of mechanisms used to kill, our study revealed that asphyxiation and personal weapons were significantly more common weapons in the commission of SH. This is consistent with other studies where sexual murderers most often strangled their victims (Häkkänen-Nyholm et al., 2009; James & Proulx, 2014, 2016; Sturup et al., 2019) or used a personal weapon (Chan et al., 2019). The odds of strangulation, suffocation, and hanging in our study were 13 times higher than use of a firearm; the odds of personal weapons were 11 times greater. It is unclear why these methods are used more frequently with SH, but this association could be related to the personal nature of sex crimes as well as the suspect’s anger toward women and desire to continue wielding power and control over the victim until her death (James & Proulx, 2016).

Much of the existing research found that SH offenders frequently targeted strangers (James & Proulx, 2014, 2016; Karakasi et al., 2017). Skott et al. (2018) found that SH was 6.5 times more likely to be committed against a stranger and this variable was one of the strongest predictors of SH in their analysis. In contrast, the present study showed that SH of an acquaintance (e.g., friend, roommate, schoolmate, and co-worker) was more likely, which was the only statistically significant victim–suspect relationship in our multivariable logistic regression model. Differences in how SH was defined could attribute to exhibited distinctions between our sample and other studies. Nevertheless, our findings suggest that focusing police investigations on acquaintances of SH victims more than on strangers or boyfriends/husbands is important in the U.S. compared to other countries where SH of a stranger tends to be more common.

Study Limitations

While this study provides new information on differences between SH and NSH in the U.S. and used a nationally representative dataset, there were several limitations noted. First, these results should be viewed as preliminary given the challenges of identifying SH cases from LE and CME narratives in the NVDRS. The dataset does not always include pertinent victim, suspect, and incident information relevant to assessing cases. For instance, paraphilic disorders (e.g., voyeurism, pedophilia) are under-documented in the NVDRS; therefore, the role these disorders play in SH cannot be estimated from these data. Precipitating circumstance data is limited for homicides due to ongoing investigations and limited findings from the investigation (e.g., body was found but with related circumstances unknown). Our analyses attempted to address this limitation by only comparing cases with known circumstances in both groups but doing so reduced the number of both case types (e.g., from 324 SH cases to 297 or lower).

Second, although it may mostly be at random, some suspect and incident variables had high levels of missing data pertinent to this study. This was primarily because the national clearance rate of homicides in the U.S. is only about 60% according to the Federal Bureau of Investigation, 2017. Therefore, there are many homicide cases in which the offender and precipitating circumstances were not identified.

Third, according to Beauregard (2019), coding homicides as sexual when the only qualifying characteristic is the victim was found naked or partially nude could have led to some false positives. He argues that, instead of selecting cases if they meet only one of the FBI criteria (Ressler et al., 1988), multiple criteria are necessary to reduce false positives and identify true cases of SH. However, several studies have classified SH cases using the same definition used in the current study–that is, based on meeting at least one FBI criterion (Beauregard & Martineau, 2012; Chan, 2015; Ressler et al., 1998; Skott et al., 2018; Smith et al., 2011; Sturup et al., 2019).

Another limitation associated with this study is the potential problem of multiple comparisons/tests. The increase in the familywise error rate (i.e., Type 1 error inflation beyond per analysis alpha) across the reported statistical analyses was not controlled. Nevertheless, most of the group differences and associations were statistically significant at low alpha levels (p < .001). Further, we consider this exploratory research relatively preliminary and encourage replication in future studies.

Conclusion

The research on SH in comparison to NSH in particular has been limited. Using a large, diverse population-based sample, this research advances the field’s understanding of this rare form of violent death and provides more accurate/representative profiles of sexual murder victims, suspects, and incidents in the U.S. It is one of few studies to match female SH victims to female NSH victims, with 6461 adult female homicide decedents from every state, the District of Columbia and Puerto Rico represented. Findings suggest there are important differences that can distinguish SH from NSH. Most notably, the sexual murderer is a unique type of offender distinguishable from other killers. Considering this information with previous research findings that SH perpetrators are more like nonlethal sex offenders warrants the former being treated as sexual perpetrators who escalate to homicide when developing crime prevention programs and conducting future police investigations of homicides. Collectively, this knowledge has important practical implications for correctional rehabilitation as well as for preparation and community management of SH offenders who will be released back into society. Police registries and violent crime surveillance systems such as NVDR should collect more detailed offender data (criminal history, substance use, etc.) and circumstantial information (killing method, victim-offender relationship, etc.) for solved SH cases. This will yield richer SH offender profiles and enable assessment of the characteristics and other factors that could increase the likelihood of a nonlethal sex offender becoming a murderer. Having more comprehensive information is critical to investigating and solving SH cases and for determining potentially greater risk for sexual perpetrators to re-offend and/or escalate to more serious crimes if released from prison and to identify which parolees may require closer monitoring once they are back in the community. Future research should explore effective management strategies and rehabilitation programs for lethal and nonlethal sexual offenders, typologies and subtypes of SH regardless of victim gender and age, match cases on other notable characteristics, compare the U.S. to other countries (Canada, Scotland, China), and use more sophisticated statistical methods such as latent class analysis.

Footnotes

Acknowledgments

The authors would like to thank Carter J. Betz, MS at the Centers for Disease Control and Prevention for his assistance with using SAS Version 9.4 (PROC SQL) to conduct keyword searches of the law enforcement and coroner/medical examiner narratives. The content analysis and identification of sexual homicide cases would have been a much more time-consuming and arduous task without all his hard work and thoughtful contributions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by a 2020 National Violent Death Reporting System New Investigator Research Award (Cooperative Agreement, CFDA # 93.421) from the Centers for Disease Control and Prevention (CDC) in partnership with the American Public Health Association (APHA).

Ethical Approval

Human Subjects (Institutional Review Board) review was not required for the National Violent Death Reporting System as it received a non-research determination.

Disclaimer

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the CDC or APHA.

ORCID iD

Natasha F. De Veauuse Brown

Case Definitions and Keywords Used to Identify Sexual Homicides

The following keywords and inclusion/exclusion criteria were used to search law enforcement and coroner/medical examiner narratives of homicide decedents aged 20–64 who died in the 50 states participating in NVDRS during 2015–2018. The keyword searches were conducted using SAS (SAS Institute, Version 9.4). All cases were independently coded by two reviewers; all discrepancies were discussed and coded to consensus. The following information lists the keywords used as well as the decision process for case ascertainment and for categorizing the cases as sexual homicides.

Please note that some of the keywords are root words, so they would yield other words as well (e.g., “sex” would also catch “sexual”, “expos” would catch “exposed” and “exposing”).

Rape

Raping

Sexual

Sex

Intercourse

Vagina

Molest

Anus

Rectum

Anal

Oral

Penis

Nude

Naked

Unclad

Undress

Unclothed

Unbuttoned

Genital

Breasts

Incest

Sodomy

Semen

Sperm

Masterbat

Condom

Bondage

Bound

Consent

Consensual

Dildo

Porn

Fantasy

Expose

Exposing

Statutory

Underwear

Bra

Panties

Object

Sadism

Sadistic

Inclusion criteria for sexual homicide cases included the following: confirmed rape or molestation; confession by offender indicating sexual contact before, during or after the murder; attempted or completed sexual intercourse between victim and suspect (oral, anal and/or vaginal), before, during, or after the murder (including post mortal activity); exposure of the primary or secondary sexual parts of the body; victim’s attire or lack of attire (e.g., breasts exposed, only wearing underwear); sexual positioning of the victim’s body (e.g., legs spread eagle); insertion of foreign objects into the victim’s cavities (anus, vagina, or mouth); substitute sexual activity (e.g., masturbation; evidence of semen at the scene; porn); sexual interest admitted by the offender; sadistic fantasies reported by the offender; clear forensic evidence of a sexual element; sexual component is suspected; mutilation of victim’s genitals/sexual body parts, including anus; the occurrence of sexual violence or one of aforementioned sexual elements on the same day as homicide (i.e., within preceding 24 hours); and the presence of used or unused condom(s) at the scene of homicide. Some of these criteria consisted of additional circumstances. For instance, attempted or completed sexual intercourse had to involve both the victim and suspect and could be consensual or forced. This is consistent with the definition of sexual homicide by the Federal Bureau of Investigation (Ressler et al., 1988). Sexual intercourse included anal, vaginal and oral sex. The criterion victim’s attire or lack there of focused on exposure of the victim’s body and/or the removal/adjustment of clothing that resulted in exposure of sexual body parts or suggested attempted or completed sexual intercourse occurred.

Exclusion criteria included the following: sexual violence/abuse perpetrated by the homicide victim (i.e., suspect was sexually assaulted/violated by the victim); homicide as the result of vengeance for a prior sexual assault committed by the victim or victim’s spouse/relative; homicide resulting from escalated argument about previous sexual violence/victimization; threats to commit sexual violence/abuse by the homicide victim; homicide that is linked to sexual violence but a bystander or third party is the homicide victim; sexual violence or sexual intercourse (forced/consensual) that occurred more than 24 hours before the homicide; and crimes of passion. Crimes of passion included homicides committed as a result of an event that trigged heartbreak and/or rage. An example is a suspect witnessing his significant other having an affair and killing her as a result.

Author Biography

Natasha DeVeauuse Brown, PhD is a Research Assistant Professor in Center for Leadership in Disability in the Department of Health Policy & Behavioral Sciences in the School of Public Health at Georgia State University. Her research focuses on behavioral health (violence, mental health and substance abuse) in youth and young adults, women’s health, and sexual risk behaviors.

Ashley Watson, MPH is a doctoral student in the Department of Health Policy & Behavioral Sciences in the School of Public Health at Georgia State University. Her research focus is child maltreatment prevention, especially the identification of risk and protective factors of both caregivers and victims. Ms. Watson is also interested in health disparities and women’s health.

References

Beauregard

Martineau

(2012). A descriptive study of sexual homicide in Canada: Implications for police investigation. International Journal of Offender Therapy and Comparative Criminology, 57(12), 1454–1476. https://doi.org/10.1177/0306624X12456682

Beauregard

Martineau

(2017). The sexual murderer: Offender behavior and implications for practice. Routledge.

Beauregard

(2019). Sexual homicide: Need for international examination of the crime scene. International Journal of Offender Therapy & Comparative Criminology, 63(9), 1503–1516. https://doi.org/10.1177/0306624x19839959

Blair

J. M.

Fowler

K. A.

Jack

S. P. D.

Crosby

A. E.

(2016). The national violent death reporting system: Overview and future directions. Injury Prevention, 22(Suppl 1), i6–11. https://doi.org/10.1136/injuryprev-2015-041819

Centers for Disease Control and Prevention . (2018). CDC’s national violent death reporting system now includes all 50 States. https://www.cdc.gov/media/releases/2018/p0905-national-violent-reporting-system.html

Chan

H. C.

(2015). Understanding sexual homicide offenders: An integrated approach. Palgrave Macmillan.

Chan

H. C.

Beauregard

(2019). Prostitute homicides: A 37-year exploratory study of the offender, victim, and offense characteristics. Forensic Science International, 294(January 2019), 196-203. https://doi.org/10.1016/j.forsciint.2018.11.022.

Chan

H. C. O.

Heide

K. M.

Myers

W. C.

(2013). Juvenile and adult offenders arrested for sexual homicide: An analysis of victim-offender relationship and weapon used by race. Journal of Forensic Sciences, 58(1), 85–89. https://doi.org/10.1111/j.1556-4029.2012.02188.x

Chan

H. C. O.

Heide

K. M.

Beauregard

(2019). Male and female single-victim sexual homicide offenders: Distinguishing the types of weapons used in killing their victims. Sexual Abuse: A Journal of Research & Treatment (Sage), 31(2), 127–150. https://doi.org/10.1177/1079063217724765

10.

Chopin

Beauregard

(2019a). Lethal combinations: A conjunctive analysis of crime scene behavior associated with sexual homicide. Behavioral Sciences & the Law, 37(5), 559-578. https://doi.org/10.1002/bsl.2425.

11.

Chopin

Beauregard

(2019b). The sexual murderer is a distinct type of offender. International Journal of Offender Therapy and Comparative Criminology, 63(9), 1597-1620. https://doi.org/10.1177/0306624X18817445.

12.

Federal Bureau of Investigation . (2017). 2017 Crime in the United States. https://ucr.fbi.gov/crime-in-the-u.s/2017/crime-in-the-u.s.-2017/topic-pages/clearances

13.

Häkkänen-Nyholm

Repo-Tiihonen

Lindberg

Salenius

Weizmann-Henelius

(2009). Finnish sexual homicides: Offence and offender characteristics. Forensic Science International, 188(1–3), 125–130. https://doi.org/10.1016/j.forsciint.2009.03.030

14.

Hillis

Mercy

Amobi

Kress

(2016). Global prevalence of past-year violence against children: A systematic review and minimum estimates. Pediatrics, 137(3), e20154079. https://doi.org/10.1542/peds.2015-4079

15.

James

Proulx

(2014). A psychological and developmental profile of sexual murderers: A systematic review. Aggression and Violent Behavior, 19(5), 592–607. https://doi.org/10.1016/j.avb.2014.08.003

16.

James

Proulx

(2016). The modus operandi of serial and nonserial sexual murderers: A systematic review. Aggression and Violent Behavior, 31(November-December 2016), 200-218. https://doi.org/10.1016/j.avb.2016.09.006.

17.

Karakasi

M. V.

Vasilikos

Voultsos

Vlachaki

Pavlidis

(2017). Sexual homicide: Brief review of the literature and case report involving rape, genital mutilation, and human arson. Journal of Forensic and Legal Medicine, 46(February 2017), 1-10. https://doi.org/10.1016/j.jflm.2016.12.005.

18.

Kerr

K. J.

Beech

A. R.

Murphy

(2013). Sexual homicide: Definition, motivation and comparison with other forms of sexual offending. Aggression and Violent Behavior, 18(1), 1–10. https://doi.org/10.1016/j.avb.2012.05.006

19.

National Violent Death Reporting System (NVDRS) (2020, October 2020). Restricted access database (RAD) instructions for users.

20.

Proulx

James

Siwic

Beauregard

(2018). Sexual murderers of children: Psychopathological and modus operandi factors. In Proulx

Beauregard

Carter

A. J.

, et al. (Eds.), Routledge international handbook of sexual homicide studies (pp. 141–156). Routledge.

21.

Quinet

(2011). Prostitutes as victims of serial homicide: Trends and case characteristics, 1970-2009. Homicide Studies, 15(1), 74–100. https://doi.org/10.1177/1088767910397276

22.

Ressler

R. K.

Burgess

A. W.

Douglas

(1988). Sexual homicide: Patterns and motives. Free Press.

23.

Skott

Beauregard

Darjee

(2018). Sexual and nonsexual homicide in Scotland: Is there a difference? Sexual and nonsexual homicide in Scotland: Is there a difference? Journal of Interpersonal Violence, 36(7–8), 3209–3230. https://doi.org/10.1177/0886260518774303

24.

Smith

S. G.

Basile

K. C.

Karch

(2011). Sexual homicide and sexual violence-associated homicide: Findings from the NVDRS. Homicide Studies, 15(2), 132–153. https://doi.org/10.1177/1088767911406236

25.

Smith

S. G.

Zhang

Basile

K. C.

Merrick

M. T.

Wang

Kresnow

Chen

(2018). The national intimate partner and sexual violence survey (NISVS): 2015 data brief–Updated release. Atlanta, GA: National Center for Injury Prevention and Control, CDC.

26.

Sturup

Rodre

Karlberg

von Vogelsang

Rying

Caman

(2019). Male-on-female sexual homicides in Sweden, 1990 to 2013: A population-based controlled study of incidents, victims, and offenders. International Journal of Offender Therapy and Comparative Criminology, 63(9), 1557–1574. https://doi.org/10.1177/0306624X18786609

27.

von Krafft-Ebing

(1886). Psychopathia sexualis. Arcade.

28.

World Health Organization . (2012). Understanding and addressing violence against women. http://apps.who.int/iris/bitstreat/10665/77434/1/WHO_RHR_12.37_eng.pdf