Reported Sexual-offense Incidents in the United States: Arrest Disparities between Women and Men

Abstract

This study examined the effect of an offender’s sex (male/female) on whether sexual-offense incidents reported to law enforcement culminated in an arrest. Two hypotheses, chivalry and evil woman, are relied upon and suggest that the probability of arrest differs for women and men, yet in differing directions. The chivalry hypothesis suggests women are treated more leniently than men and, therefore, less likely to be arrested. The evil woman hypothesis, however, suggests the opposite: Women are treated more harshly than men and, therefore, more likely to be arrested. Seven years of National Incident-based Reporting System [NIBRS] data were relied upon (National Archive of Criminal Justice Data, 2010–2016, National Incident-based Reporting System: Extract Files); all of the reported sexual-offense incidents committed by women were included, along with a matched sample of reported sexual-offense incidents committed by men, culminating in a sample of 22,744. Overall, women were 42% significantly less likely than men to be arrested when controlling for other known offense, offender, and victim characteristics. The odds for women to be arrested increased, however, when specific offender demographics, offense characteristics, and victim characteristics were taken into account. The implications of these findings are discussed in regard to their application of the chivalry and evil woman hypotheses.

Keywords

sexual offense gender arrest chivalry hypothesis evil woman hypothesis

Introduction

Historically, women have been viewed as incapable of committing violent crime, particularly sexual offenses, due to their innate feminine qualities (Daly, 1989). Women are perceived as more communal, nurturing, empathetic, and caring than men, who are perceived as aggressive, assertive, and competitive (Daly, 1989). The gendered assumption concerning women conflicts with recent crime reports and research findings that show that women do commit sexual offenses. In 2018, 407 women were arrested for rape in the United States (Federal Bureau of Investigation [FBI], 2019). Not including prostitution and rape, an additional 1,527 women were arrested for committing other sexual offenses (FBI, 2019). Although the number of sexual-offense incidents committed by women pales in comparison to those committed by men, these numbers warrant further inquiry—especially regarding how such offenses are perceived and responded to by law enforcement.

Women who commit sexual offenses continue to be under-researched (Denov, 2003; Hassett-Walker et al., 2014), particularly with regard to assessing disparity between men and women that occurs via the criminal justice system’s response to sexual offenses (Hassett-Walker et al., 2014; Vandiver, 2010). Men are more likely than women to commit sexual offenses (Vandiver, 2010), but research assessing whether this results in varying arrest probability risks has only begun to emerge. The question that deserves attention is whether there is a gap in law enforcement’s response to reported sexual-offense incidents, based on the sex of the offender. The focus of this research, therefore, is to identify law enforcement’s response to those reported sexual-offense incidents by relying upon a national sample of such incidents. Additionally, the present study examines the extent to which offender, victim, and offense characteristics affect the odds of an arrest occurring among all reported sexual-offense incidents committed by women and a comparable group of men. Here, particular attention is given to determining how the offender’s sex influences the odds of arrest.

Women Who Have Committed a Sexual Offense

More is known today about women who have committed a sexual offense than just a few decades ago; there has been more empirical research on these women published within the past decade than all of the studies published in the previous 30 years (Cortoni et al., 2017). Researchers have found that women who commit a sexual offense are typically in their late 20s or early 30s, Caucasian, with victims who are approximately 12 years old, on average (Sandler & Freeman, 2009; Vandiver & Kercher, 2004). Researchers, however, need to continue assessing the role of the offender’s sex, particularly in the context of offenses that are traditionally committed by men.

Women who do sexually offend commit different types of sexual offenses at varying rates than men who sexually offend (Vandiver & Kercher, 2004). Women are less likely than men to have intercourse and oral-genital contact with their victims (Kaufman et al., 1995). Moreover, women are more likely than men to use foreign objects in the commission of the sex offense (Kaufman et al., 1995). Women who have committed a sexual offense have fewer prior arrests for violent crimes, such as assault, rape, and murder, when compared to their male counterparts (Schwartz et al., 2009).

Judicial Response to Men and Women Who Have Committed a Sexual Offense

The “gender gap” refers to the empirical finding that women commit less crime than men, as evidenced by official reports (i.e., police reports, victim reports, crime surveys; Stolzenberg & D’Alessio, 2004). This gender gap is widely considered within criminological research as an uncontested “fact” (Barnes et al., 2015; Vaughan et al., 2015). For women who commit a sexual offense, a large gender gap exists, with official data indicating that women commit less than 10% of all sexual offenses (Vandiver, 2010).

Researchers have identified disparate responses by criminal justice officials to women and men who have committed a sexual offense (Goulette et al., 2015). Many researchers have examined the much more visible stage of the criminal justice system—sentencing, while fewer have explored sex-based disparities at pretrial stages, such as arrest (Katz & Spohn, 1995; Kruttschnitt, 1984). Embry and Lyons (2012) found that even after sentencing guidelines were implemented to ensure equality in sentencing, men still received disproportionately more severe sentences than women who commit sexual offenses. Sandler and Freeman (2011) found that among 138,000 arrestees in New York state, women arrested for a sexual offense were not significantly different from men who were arrested for a sexual offense with regard to whether the offense resulted in a conviction, yet women were significantly more likely than men to receive probation as compared to an incarceration sentence. Likewise, Spivak et al. (2014) reported that female juvenile status offenders were more likely than male juvenile status offenders to have their petitions filed for review and less likely to be adjudicated guilty. These studies, therefore, suggest support for the notion that women are generally treated more leniently than men in the criminal justice system, at least with regard to sentencing type and length of sentence.

One study reported that the probability of arrest for women who were suspected of committing rape and forcible fondling was lower than men who were also suspected of committing the same sexual offenses (Stolzenberg & D’Alessio, 2004); this study, however, was limited to assessing two types of sexual offenses and relied on only one year of data. Women who commit sexual offenses may be charged less severely when they victimize adolescents for whom they are not caretakers, possibly because of the gendered perception that the victims are consenting participants (Hassett-Walker et al., 2014). These findings reflect an assumption—perhaps a wrong one—that women who have committed a sexual offense pose a lower risk to society than men who engage in the same behavior.

Relevant Theoretical Constructs

Researchers have explained the disparity between male and female offenders at the various stages of the criminal justice process by relying on the chivalry and evil woman hypotheses (Chesney-Lind, 1978; Curran, 1983; Daly, 1989; Romain & Freiburger, 2016; Rodriguez et al., 2006; Steffensmeier, 1980; Tillyer et al., 2015). To address whether law enforcement responds to women who commit sexual offenses differently than men, we compare the likelihood of arrests based on several offenses, offenders, and victim characteristics. The arrest likelihood patterns are then assessed in the lens of the chivalry and evil woman hypotheses.

The chivalry hypothesis is defined as a code of conduct in which women are given priority by men who believe that women should be afforded special benefits, such as protection and leniency (Steffensmeier, 1980). This hypothesis is based on the belief that men should not inflict harm on women (Daly, 1989), and that women should conform to traditional gender roles. A premise of the chivalry hypothesis is that the criminal justice system is dominated by men, and as a result, judges may be more inclined to connect female defendants with the women in their own lives, ultimately treating them more leniently than similarly situated men (Goulette et al., 2015; Visher, 1983). Applying the chivalry hypothesis to women and men who commit a sexual offense, one would expect that women would receive more lenient legal sanctions than men to reduce the amount of harm that is inflicted upon them.

In recent decades, researchers have reported little or only partial support for the chivalry hypotheses (Curran, 1983; Daly, 1987; Edwards, 1984). Demuth and Steffensmeier (2004) found that women are much more likely than men to receive pretrial release. Women with children have been shown to receive more lenient treatment than women who do not have children (Daly, 1987). In an assessment of 45,060 sentences (37,104 cases involving male offenders and 7,956 cases involving female offenders) from 1996 to 2002, within the United States Sentencing Commission’s data collection, Sarnikar et al. (2007) found that, after controlling for past criminal history and offense severity, women receive a prison sentence that is, on average, two years shorter than those given to men who commit the same crime and have the same prior criminal record. Likewise, Griffin and Wooldredge (2006) found that women were sentenced more leniently than men for felony offenses both before and after sentencing reforms in Ohio. This finding corroborates an earlier study that found this was also true in Pennsylvania, where sentencing guidelines had been implemented (Blackwell et al., 2008). These findings suggest that a possible underlying and false assumption exists—women who commit a sexual offense pose a lower risk than men who commit a sexual offense.

Conversely, the evil woman hypothesis, also referred to as selective chivalry, argues that women who commit a sexual offense will be treated more harshly when they step outside prescribed gender roles (Chesney-Lind, 1978; Goulette et al., 2015; Rodriguez et al., 2006; Tillyer et al., 2015; Vaughan et al., 2015). It is assumed that a woman will receive a longer sentence the more she acts outside of traditional gender roles (Embry & Lyons, 2012). In line with the evil woman hypothesis, women who commit characteristically unfeminine acts, such as a violent crime and sexual abuse, maybe “treated much more severely than their counterparts whose illegal activity conforms to the standards of womanhood” (Grabe et al., 2006, p. 139). Women who commit a sexual offense are typically seen as evil women who have “lost … the women nurturing gene,” while men who commit a sexual offense may also be evil and monstrous but not perceived as unnatural in the same way as women (Hayes & Carpenter, 2013, p. 163).

There is support for the idea that women are treated more harshly than men by the criminal justice system (Crew, 1991; Feld, 2009; Tracy et al., 2009; Wilbanks, 1986; Zingraff & Thomson, 1984). Crew (1991) found that women who committed offenses that violated traditional gender roles received harsher punishments than men. In that connection, Zingraff and Thomson (1984) reported that women receive longer sentences than men for child abandonment and simple assaults. Likewise, Wilbanks (1986) found that women who committed a sexual offense and charged with fondling a child and abduction (two offenses that violate gender role expectations) were treated more harshly by decision-makers in the criminal justice system than men charged with the same offenses. More recent elaborations have focused on the effect of the female gender role on sentencing decisions. Gathings and Parrotta (2013), for example, found that women who draw on gendered narratives in their appeals for leniency ultimately receive lighter sentences. These studies demonstrate that when women engage in criminal behavior that violates traditional gender stereotypes, they will be treated more harshly than their male criminal counterparts.

The evil woman hypothesis suggests that women who commit crimes are treated more harshly by criminal justice officials, whereas the chivalry hypothesis suggests more lenient treatment of women by criminal justice officials. Although often viewed as conflicting hypotheses, they share the idea that women are treated differently from men within the criminal justice system, depending on whether they conform to traditional gender roles (Tillyer et al., 2015).

Methods

This research included two research questions. First, we assessed whether reported sexual-offense incidents committed by women are more likely than those committed by men to result in an arrest, after controlling for other known offense, offender, and victim characteristics. Notably, the term “committed” offense does not indicate a completed offense; rather, it includes allegations and/or reports to law enforcement. The dichotomous nature of the outcome (arrest/no arrest) necessitates a binary logistic regression model for multivariate analyses. The second research question assesses how other known offense, offender, and victim characteristics (defined later) differentially affect the odds of arrest among reported sexual-offense incidents committed by women and men for the purpose of determining what conditions affect women being treated either more harshly or less harshly than their male counterparts.

Data Source

As of 2012, NIBRS included information reported by more than 6,115 agencies spread across 32 states, with 15 states submitting all their crime data (Justice Research and Statistics Association, 2012). NIBRS participating agencies covered an estimated 29.3% of the total U.S. population and differed only slightly from Uniform Crime Report (UCR) arrest data across demographic measures, yet NIBRS offered additional, detailed incident-level crime data (Pattavina et al., 2017).

The structure of NIBRS allows researchers to assess and choose different units of analysis depending on their focus. The data can be retrieved with a focus on incidents, offenses, offenders, or victims. Here, incident-level data are relied upon; yet only incidents involving a lone adult offender (17 years old and older) and a lone victim are included. Incidents with multiple offenders and/or victims were excluded, as the offender’s sex was our primary interest and multiple offenders confound the examination of the relationship between the offender’s sex and resulting arrest or lack thereof. This approach is in line with multiple studies that excluded incidents with multiple offenders and victims because of the complex dynamics involved (Pattavina et al., 2007; Vaughan et al., 2015). NIBRS data are examined from 2010 to 2016 (National Archive of Criminal Justice Data, 2010–2016). Incidents involving the following reported sexual-offense incidents were included: rape, sodomy, fondling, sex assault with an object, incest, and statutory rape against a single victim.

Case-control Matching

Given the dissimilarities among offense types committed by men and women who sexually offend (refer to generally: Lauritsen et al., 2009; Sandler & Freeman, 2011; Stolzenberg & D’Alessio, 2004; Vandiver, 2010; Williams & Bierie, 2015), along with the knowledge that such differences may influence the outcome variable (arrest versus no arrest), case-control design is employed. Case-control matching allows groups to be compared as closely as possible based on observed factors of interest (Shadish et al., 2002). This method removes bias due to an imbalance in any confounding factors (Cologne & Shibata, 1995). The resulting balance reduces the variance in the parameters of interest, which improves statistical efficiency (Rose & van der Laan, 2009). In short, the matching approach creates “equal groups” through pairing variables that are known to affect the outcome.

The original dataset included all reported sexual-offenses incidents from 2010 and through 2016. Among those, 11,372 cases involved reports of female offenders; those were matched to incidents committed by men who were the same age and involved the same sexual offense, culminating in 22,744 incidents. Both sexual offense and age have been identified as important factors in previous literature. For example, evidence to date has shown that women who commit violent offenses tend to cluster at the lower ends of offense severity and criminal history, indicating that men commit more severe offenses more often (Griffin & Wooldredge, 2006; Steffensmeier et al., 1993). Age is also an important characteristic, as prior literature reveals this to be a critical factor concerning criminal patterns, with younger individuals more likely than those who are older to commit crimes (Blumstein et al., 1986; DeLisi et al., 2013; Farrington et al., 2003; Moffitt, 2017; Sampson & Laub, 2003).

The ages of the female offenders ranged from 17 to 98+¹

Forty-seven of the male offenders and 47 of the female offenders were categorized as “over 98 years old” in NIBRS; for the purpose of computing an average and standard deviation, 98+ was counted as 98 years old.

years old, with an average of 25.47 (SD = 14.60). Among this sample, the majority of the reported offenses included fondling (n = 6,118; 53.8%), followed by rape (n = 1,674; 14.7%), statutory rape (n = 1,534; 13.5%), sodomy (n = 1,102; 9.7%), sexual assault with an object (n = 726; 6.4%), and incest (n = 218; 1.9%). The age and types of offenses among the reported sexual-offense incidents committed by men mirrored that of the female offenders, as they were matched on these two variables.

Measurement of Variables

The dependent variable was whether law enforcement made an arrest for the given incident (arrest: yes/no). The independent variables included the offense, offender, and victim characteristics. With regard to the offender and victim characteristics associated with the incident, sex, age, race, and ethnicity are provided in NIBRS. The offender’s and victim’s ages was kept as continuous variables for descriptive purposes, yet for the logistic regression models, victim’s age was dichotomized (yes = 1).²

“Yes” was coded as 1, and all other cases were coded as 0; this method has been employed by others (refer to generally: Stolzenberg et al., 2004).

Although NIBRS data include variables for race and ethnicity, ethnicity is not relied upon in these analyses. More than 90% of the information for ethnicity is missing or undetermined.

As noted throughout the literature, race affects the likelihood of many of the outcomes in the criminal justice system. More specifically, Blacks are more likely than other races to be treated harshly by criminal justice officials (D’Alessio & Stolzenberg, 2003; Kamalu, 2016; Ousey & Lee, 2008; Ridgeway, 2006; Spohn, 2013). Thus, for these analyses, race is measured by whether one is Black (yes = 1). NIBRS data include the relationship between the victim and the offender at the time of the incident at issue. There are 25 victim–offender relationship types contained in NIBRS. These were collapsed into three new dichotomized variables: stranger, acquaintance, and related (yes = 1).³

The first variable, “related,” includes 13 relationship types: spouse, common-law spouse, parent, sibling, child, grandparent, grandchild, in-law, stepparent, stepsibling, or another family member. The second variable, “acquaintance,” includes 11 relationship types: acquaintance, friend, neighbor, babysitter (the child being cared for), boyfriend/girlfriend, child of boyfriend/girlfriend, homosexual relationship, ex-spouse, employee, employer, or otherwise known to the offender. The third variable, “stranger,” consists of only one relationship type—stranger.

Regarding offense characteristics, the sexual offenses included rape, sodomy, fondling, sex assault with an object, incest, and statutory rape. The seven types of violent offenses (murder, negligent manslaughter, justifiable manslaughter, aggravated assault, simple assault, intimidation, and kidnapping/abduction) were collapsed and dummy coded into whether a violent offense was also reported as part of the incident (yes = 1). The 11 types of property offenses (burglary/breaking and entering, pocket-picking, purse snatching, shoplifting, theft from building, theft from coin-operated machine, theft from car, all other larcenies, motor vehicle theft, stolen property offenses, and destruction of property) were collapsed into whether a property offense was also reported as part of the incident (yes = 1). These two categories of additional offenses, violent and property, were further collapsed into another variable, “additional offense” (yes = 1). The extract file also contained information on whether the offender was suspected of being under the influence of alcohol and/or a controlled substance. In line with the Williams and Bierie (2015) study, whether the offender was under the control of alcohol and/or a controlled substance at the time of the incident was combined and dichotomized (yes = 1).

The level of injury incurred by the victim is assessed after aggregating across several injury categories in the victim segment. Law enforcement can report up to four injuries incurred by each victim. Additionally, victim injury is categorized by type, including whether the victim sustained no injury, an apparent minor injury, apparent broken bone(s), another major injury, possible internal injury, loss of teeth, severe laceration, and unconsciousness. Relying upon the method described by Gennarelli and Wodzin (2006), the injury types were collapsed into three categories, including victims who did not sustain any minor or major physical (coded as “0”), victims who sustained apparent minor injuries (coded as “1”), and victims who sustained a major injury, including apparent broken bones, another major injury, possible internal injury, loss of teeth, severe laceration or unconsciousness (coded “2”). For logistic regression, all injury categories were collapsed into one variable, any injury (yes = 1).

Whether a weapon was used by the offender is also recorded in NIBRS. The variable contained 22 different types of weapons.⁴

Including whether the offender used a firearm, automatic firearm, handgun, automatic handgun, rifle, automatic rifle, shotgun, automatic shotgun, another firearm, another automatic firearm, knife/cutting instrument, blunt object, a motor vehicle, personal weapons—including one’s hands, feet, or teeth, poison (including gas), explosives, fire/incendiary device, drugs/narcotics/sleeping pills, asphyxiation, another type of weapon.

For this study, the weapon types were collapsed into whether any weapon was used (yes = 1). The location where the offense took place was included in this study. NIBRS included 57 different locations,⁵

Residential locations included a private residence/home and coded as 1. Locations that were non-residential were coded 0, and included the following locations: farm facility, military installation, college, school—elementary, tribal land, and any type of correctional facility, air/bus/train terminal, bank, bar, church, commercial, construction site, convenience store, department store, drug store, field/woods, government or public building, grocery, highway/road/alley/street/sidewalk, hotel/motel, lake/waterway or beach, liquor store, parking lot, rental store, restaurant, school, gas station, specialty store, other or unknown, abandoned condemned structure, amusement park, arena, ATM separate from a bank, car dealership, camp/campground, daycare, dock/wharf, gambling facility, industrial site, park/playground, rest area, school, shelter/mission, shopping mall, or a community center.

and these were collapsed into one variable, if the reported incident occurred at a residence (yes = 1), given that most sexual victimizations occur at a residence. For example, Calkins et al. (2015) found that 67% of all sex offenders met their victims in private settings, such as a residence, and offenders who victimized family members were 98% more likely to offend in a private location, such as a residence, as opposed to a non-residential location.

Analytic Strategy

In addition to descriptive analysis, three logistic regression models are estimated. Here, the dependent variable is whether the offender was arrested. The independent variables relied upon for the regression analysis include the offender’s sex (first model only), age, race (i.e., Black), the reported sexual offense committed (rape, sodomy, sexual assault with an object, incest, statutory rape⁶

Fondling was treated as the reference category and omitted in the analyses.

), whether the reported offender was suspected of committing an additional violent and/or property offense, whether the reported offender was suspected of alcohol and/or drug use, weapon use, whether the offense was completed (as opposed to attempted), whether the victim was injured, whether the offense was committed at a residence, the victim’s sex, the victim’s age, whether the victim was Black, and whether the victim–offender relationship was unknown to law enforcement, whether the victim and offender were related, and whether the victim and offender were strangers.⁷

Related was treated as the reference category and omitted in the analyses.

To assess the effects of an offender’s sex (male/female) regarding whether the offender was arrested, three logistic regression models were estimated: (1) all incidents (with female and male offenders), (2) men-only incidents, and (3) women-only incidents. The dependent variable for each model was whether the offender had been arrested. In addition, the independent variables, all models utilized cluster robust standard errors to account for clustering of offenses within police agencies, and the Originating Agency Identifier was used as the cluster variable. This will take into account variations that occur between and among different police departments.

The first model sought to address the first research question: What is the effect of sex on arrest while controlling for other known offense, offender, and victim characteristics. The second and third models allowed us to assess the second research question: How the other known offense, offender, and victim characteristics differentially affect the probability of arrest for men and women committing sexual offenses. These models inform whether patterns in the arrest outcome among reported sexual-offense incidents committed by women were treated more or less leniently than incidents committed by their male counterparts and which factors affected their odds of arrest.

Two of the variables of interest included missing and/or unknown data, victim’s age (n = 229; 1.0%) and victim–offender relationship (n = 3,664; 10.8%). The 229 cases with missing victim age were imputed with the mean (17.17). The relationship to the offender was handled by creating a category called “relationship missing/unknown” and including it in the analyses; this method has been employed by others (refer to generally: Thompson et al., 1999). Additionally, collinearity diagnostics revealed a high degree of collinearity between offender suspected of alcohol/drug use and offender committed an additional offense (VIFs > 4.0). The former variable, “offender suspected of alcohol/drug use” was retained, while whether the “offender committed an additional offense” was excluded in all logistic regression models.

Results

Although the reported sexual-offense incidents committed by women were matched to those committed by men on age and type of sexual offense, we examined any additional differences (refer to Table 1). Overall, the phi coefficients revealed weak and/or negligible differences between reported incidents committed by women when compared to those committed by men (i.e., phi values less than .30). The only exception to this was victim sex, where men were more likely than women to have a female victim (88% compared to 40%, phi = .50). This difference, in addition to the variables that differ slightly between men and women, may affect whether an arrest occurs and, therefore, was relied upon as control variables to answer the research questions.

Table 1.

Characteristics of Offenses, Offenders, and Victims.

	Men (n = 11,372)		Women (N = 11,372)
	n	%	n	%
Offense characteristics
Committed additional offense, property offense	131	1.2	59	.5
Committed additional offense (non-sexual) violent offense	21	.2	12	.1
Committed an additional offense, property and/or violent	1,020	9.0	690	6.1
Weapon use (yes)	106	.9	110	1.0
Offense committed (as opposed to attempted)	11,092	97.5	11,050	97.2
Offense occurred at a residence	7,918	69.6	8,355	73.5
Offender suspected of alcohol use	759	6.7	502	4.4
Offender suspected of drug use	139	1.2	136	1.2
Offender suspected of alcohol and/or drug use
Victim—minor injury	988	8.7	677	6.0
Victim—major injury	183	1.6	212	1.9
Victim—minor and/or major injury	1171	10.3	889	7.8
Offender race
Black	1,911	16.8	1,799	15.8
Victim characteristics
Female victim	10,038	88.3	4,567	40.2
Black	1,462	12.9	2,024	17.8
Younger than 14	5,158	45.4	5,485	48.2
Victim and offender were strangers	663	5.8	272	2.4
Victim and offender were acquaintances	5,606	49.3	6,913	60.8
Victim and offender were related	3,144	27.6	3,337	29.3
Victim and offender’s relationship unknown/missing	1,959	17.2	850	7.5

Table 2 includes a logistic regression model that examines the effect of an offender’s sex (male/female) and the variables that strongly contributed to predicting arrest while controlling for the other variables in the model. The first logistic regression model includes all reported sexual-offense incidents with female and male offenders. The second model (women only) and third model (men only) in Table 3 allow for comparisons among the known offense, offender, and victim characteristics and how they differentially affect women’s and men’s arrest probabilities.

Table 2.

Logistic Regression Model of All Incidents Involving a Sexual Offense, Dependent Variable: Arrested (Yes/No).

	Exp(B)	Cluster Robust SE
Offense characteristics
Rape	1.001	.015
Sodomy	1.017	.011
Sexual assault with an object	.994	.014
Incest	1.020	.020
Statutory rape	1.140	.015
Committed any additional offense	1.237***	.049
Suspected of alcohol/drug use	.997***	.001
Weapon use	1.015	.029
Offense committed (as opposed to attempted)	1.072***	.020
Offense committed at a residence	.982*	.008
Any injury	1.001	.011
Offender demographics
Female	.923***	.011
Age	1.000***	.000
Black	.981	.012
Victim characteristics
Female	.923	.011
Younger than 14	1.010	.007
Black	.981	.010
Victim and offender were acquaintances	1.048***	.008
Victim and offender were strangers	1.134***	.022
Victim and offender’s relationship unknown	.987	.013
R ²	.04

Notes. *p ≤ .05 ***p ≤ .001.

Reference categories are fondling for sex offense and related to victim–offender relationships.

Table 3.

Logistic Regression Models of Sexual-offense Incidents Committed by Women and Men, Dependent Variable: Arrested (Yes/No).

	Model 1:Female Offenders Only(N = 11,372)		Model 2:Male Offenders Only(n = 11,372)
	Exp(B)	Cluster Robust SE	Exp(B)	Cluster Robust SE
Offense characteristics
Rape	1.006	.012	1.000	.027
Sodomy	1.047***	.014	1.005	.016
Sexual assault with an object	1.009	.152	.984	.022
Incest	1.007	.022	1.022	.032
Statutory rape	1.123***	.017	1.125***	.022
Committed any additional offense	1.261***	.086	1.229***	.060
Suspected of alcohol/drug use	.997	.001	.997*	.011
Weapon use	1.040	.388	.994	.043
Offense committed (as opposed to attempted)	1.077***	.017	1.066*	.035
Offense committed at a residence	.986	.010	.974*	.012
Any injury	1.025	.015	.988	.014
Offender demographics
Age	1.000***	.000	1.000*	.035
Black	.981	.012	.987	.019
Victim characteristics
Female	.970 ***	.007	1.050**	.016
Younger than 14	.982	.008	1.043***	.012
Black	.986	.012	1.023	.018
Victim and offender were acquaintances	1.066***	.009	1.026*	.012
Victim and offender were strangers	1.082***	.030	1.143***	.029
Victim and offender’s relationship unknown	1.036	.017	.956*	.020
R ²	.043		.030

Note. *p ≤ .05 ** p ≤ .01 *** p ≤ .001.

Women: More or Less Likely to Be Arrested than Men?

The first research question assessed whether reported sexual-offense incidents committed by women are more likely than those committed by men to result in an arrest, after controlling for other known offense, offender, and victim characteristics. Among the reported sexual-offense incidents committed, women were less likely than men to be arrested (women = 1,853; 16.3%; men = 2,752; 24.2%). This was further explored in the first logistic regression model (refer to Table 2), which indicated that women were 42% less likely to be arrested than men (1 – exp(b) × 100; 1 – .582 × 100 = 42%) while controlling for other known offense, offender, and victim characteristics.

Differential Arrest Risks Factors

The second research question addresses how the other known offense, offender, and victim characteristics differentially affect the odds of arrest among reported sexual-offense incidents committed by women and men. Here, the results in Table 3, which include all reported sexual-offense incidents committed by women (Model 1) and all reported sexual-offense incidents committed by men (Model 2), were compared. Although an assessment of differences in coefficients between these two groups, women and men, lends itself to assessing interactions, such analyses are discouraged with logistical regression (refer to generally: Mustillo et al., 2018).⁸

Mustillo et al. (2018) in their guidelines for quantitative submissions for the American Sociological Review cite problems recognized by others (Allison, 1999; Williams, 2009; and others) of interpreting interactions with nonlinear models with categorical dependent variables and posit, “The case is closed: don’t use the coefficient of the interaction term to draw conclusions about statistical interaction in categorical models such as logit, probit, Poisson, and so on.” (p. 1282).

These results, therefore, are discussed in an exploratory and descriptive manner by assessing variables that: (a) increased the odds of arrest for women, yet decreased the odds of arrest for men; (b) decreased the odds of arrest for women, yet increased the odds of arrest for men; (c) decreased the odds for both, but with substantively large differences in the effects; and (d) increased odds for both, but with substantively large differences in the effects.

For many of the offense, offender, and victim characteristics assessed, the odds of arrest for women increased, and they decreased for men. This pattern suggests women were treated harsher than men with regard to an arrest decision. For example, when the victim–offender relationship was unknown or not entered, the odds of arrest increased by 52% for women, yet they decreased by 27% for men. Also, when women used a weapon during the commission of the offense, the odds of arrest increased by 34% for women and decreased by 3% for men. Likewise, when it was recorded that the victim had an injury, the odds of arrest increased by 24% for women and decreased by 5% for men. Moreover, when women reportedly committed sexual assault with an object, the odds of arrest increased by less than 1%; for men, the odds of arrest decreased by 11%. When women reportedly committed rape, the odds of arrest increased by less than 1% for women and decreased for men by 5%. Victim age also had different implications for men and women with regard to their odds of arrest. For each additional year of the victim’s age, the odds of arrest increased for women by approximately 1%, but for men, it decreased by .5%. Although the difference between men and women’s likelihood of arrest is relatively small, it must be considered in the context that this variable is continuous, and each additional year of a victim’s age has an additive effect on the probability of arrest, with older victims being associated with an increased risk of arrest for women and a decreased risk of arrest for men.

Two victim characteristics decreased the odds of arrest for men, yet increased the odds for women—suggesting leniency for women. This occurred when the offender had a female victim, and when the offender had a Black victim. Men who had a female victim had 34% increased odds of arrest, whereas women who had a female victim had a 24% decreased odds of arrest. Similarly, the odds of arrest increased by 15% when men had a Black victim, yet it decreased by 12% when a woman had a Black victim.

Several of the offender and offense characteristics were associated with a decrease in the odds of arrest for both men and women, yet the amount of risk reduction was more for women than men—suggesting leniency for women. This occurred when the offender was Black, and for each additional year of age of the offender. Women who were Black had a 13% reduction in odds of arrest, compared to men who had only a 7% reduction in odds of arrest. For each additional year of age, women were 1.2% less likely to be arrested, while men were only .5% less likely to be arrested. Although the odds of arrested decreased 11% for men and 10% for women when the offense occurred at a residence, the difference between the two is likely not substantial enough to draw any defensible conclusions.

Many of the offense characteristics were associated with an increase in arrest odds for both men and women, yet with women much more likely than men to be arrested. For example, at the time of the reported sexual-offense incident, when the offense was reported as merely “attempted” as opposed to “completed,” women’s odds of arrest increased by 129%, while men’s odds increased by only 40%. Additionally, women also had an increased odds of arrest when compared to men when the victim was someone they knew (84% compared to 1%), committed sodomy (44% compared to 1%), the victim was a stranger (112% compared to 83%), and committed incest (11% compared to 1%).

Discussion

The first aim of this study was to assess whether reported sexual-offense incidents committed by women were more likely than men to result in an arrest. All reported sexual-offense incidents committed by women and a matched sample of incidents committed by men over seven years, from 2010 to 2016, were relied upon. Women were 42% less likely than men to be arrested for a sexual offense when it comes to the attention of law enforcement, supporting much of the empirical research on female offenders that have found women receive more lenient treatment than men (Hasset-Walker et al., 2014; Romain & Freiburger, 2016; Sarnikar et al., 2007). This overarching finding points to support for the chivalry hypothesis (i.e., women who commit sexual offenses are less likely to be arrested than their male counterparts), at least at the arrest stage. Moreover, these findings support what others have found at the sentencing stage, where women often receive more lenient sentences than men (Romain & Freiburger, 2016).

A perceived assumption about sexual offenses committed by women—albeit an inaccurate one—is that they are not serious, and much damage could not be caused by women who sexually offend (Vandiver et al., 2016). This has been echoed by many researchers who state, “women don’t do such things” (Wijkman et al., 2010) when referring to whether women commit a sexual offense. Similarly, statements such as “what harm can be done without a penis” suggest an inability of women to commit sexual offenses (Hislop, 2001). This false assumption is likely to be perceived by law enforcement officers and affect arrest decisions. These types of assumptions lead to the chivalry hypothesis in that when a sexual offense is committed by a woman, she is treated less harshly than a man.

The second aim of this study considered how the other known offense, offender, and victim characteristics differentially affect the probability of arrest among reported sexual-offense incidents committed by women and men. More specifically, it was hypothesized that women may be treated less harshly or more harshly, depending on whether the variables could be associated with gendered assumptions, such as women cannot do much harm and/or the offense is not that serious when committed by a woman. Although methodological constraints prevent us from indicating whether statistically significant differences occur between men and women with regard to how specific offense, offender, and victim characteristics contribute to an arrest/no arrest decision, we highlight instances in which each variable contributed in substantially different ways for women compared to men (i.e., variables increased odds of arrest for women, but decreased for men). We also address when the odds of arrest increased and/or decreased for men and women (same direction of impact) but varied between men and women by at least 20%. These factors are discussed in terms of when women are more likely than men to be arrested (i.e., treated more harshly) and when women are less likely than men to be arrested (i.e., treated more leniently). These are presented in Table 4.

Table 4.

Summary of Differences in Odds of Arrest for Women and Men.

Women’s Odds of Arrest Increase (More than Men) When:	Difference in Odds of Arrest(%)
Offense committed (as opposed to attempted)	89
Victim–offender relationship: Acquaintances	83
Victim–offender relationship: Unknown	79
Offense committed: Sodomy	44
Used a weapon	37
Caused the victim minor and/or major injury	29
Victim–offender relationship: Strangers	29
Offense committed: Statutory rape	20

Women’s Odds of Arrest Decrease (More than Men) When:
Female victim	58
Black victim	27

Note. The difference in odds of arrest was computed by calculating the difference in odds between women and men. Only those with at least 20% difference are noted here.

Women were more likely to be arrested than men when the victim–offender relationship was unknown and/or not recorded; the difference in odds of arrest was 79%. Little can be drawn from this, given that it includes missing data. It is unclear if the officer did not know the victim–offender relationship, it was not entered, and/or, was simply missing data.

The remaining variables, especially when considered together, provide a profile of when women are more likely than men to be arrested while controlling for other factors (refer to Table 4). When these factors are assessed through the lens of gendered assumptions, it is important to recall that the evil woman hypothesis suggests that women will be treated harshly when they commit sexual offenses, uncharacteristically violent offenses, and/or offenses against children—as that is a violation of their nurturing qualities (Curran, 1983; Hayes & Carpenter, 2013). Several of the factors that increase the odds of arrest for women (as compared to their male counterparts) lend themselves to the concept of uncharacteristically violent; these include committing the offense (as opposed to attempting), using a weapon during the offense, causing the victim minor and/or major injury, committing sodomy, rape, and statutory rape (in comparison to committing fondling). Thus, when women commit a sexual offense with characteristics that deviate substantially outside the gendered norms of women being nurturing, caring, and non-violent, she is more likely to be arrested than a male counterpart who has a committed an offense with the same characteristics.

Furthermore, when the victim is either Black and/or female, she is treated more leniently—less likely to be arrested. These variables, however, may have little to do with either the chivalry or evil woman hypotheses, as the victim’s race may not pertain to assumptions about gendered norms.

Overall, this research reveals that women are treated less harshly when the characteristics of the offense target victims who are often associated with more marginalized and/or vulnerable groups—Black and/or female. Further, when woman do not deviate too substantially from characteristics associated with more typical sexual-abuse scenarios, she will be treated more leniently. Yet, when she commits offenses that involve more aggressive behaviors (committing the offense, as opposed to simply attempting), using a weapon, committing the offense against someone who is not related (i.e., acquaintance, stranger, and/or the relationship is unknown), commits sodomy, and/or statutory rape—she is more likely than her male counterpart to be arrested. She is more likely to be treated as an evil woman.

Limitations and Implications

Quantitative studies that take into account diversity are instrumental in informing inclusive policy decisions. Here we address some of the limitations of the data broadly, as well as the diversity-related data specifically. While the sample is generally consistent with the U.S. population in terms of race, ethnicity was not considered because more than 90% of the information for ethnicity was missing or undetermined, possibly limiting the ability to account for differences in arrest.

The results of this study should be interpreted with caution. It is acknowledged that many do not report sexual offense victimization. Truman and Morgan (2016) estimated that only 32% of sexual assaults were reported to police, compared to 62% of aggravated assaults and robberies. While this may not be problematic in itself with regard to the research questions in this study, it can impact the results if sexual offenses known to law enforcement are reported differently among female and male offenders. Prior research suggests this phenomenon likely occurs, as victims of sexual offenders are less likely to report when the offender is female, compared to male (Denov, 2001). This research, therefore, likely underestimates and possibly presents a non-representative sample of women who have committed a sexual offense.

Additionally, NIBRS suffers from measurement issues regarding substance use (Pattavina et al., 2013). Substance use, in particular, may increase the chance of arrest; however, the strength of the association may be weaker than current evidence suggests when considering police reliability recording (Pattavina et al., 2013). Also, NIBRS data, though rich in detail compared to other sources, do not include all information about the context, evidence availability, or the circumstances of an offense. This could directly affect whether an arrest was made. For example, if witnesses or substantial corroborating evidence existed, this is likely to affect the arrest decision, more so than other known factors. Even though the sample is evenly divided between women and men, incidents that were missing the offender’s sex were excluded from this study, removing cases where law enforcement had little to no evidence available to make an arrest.

In terms of the implications of this research, these findings should be interpreted with caution, as the overall variation explained is quite small (as indicated by the Nagelkerke R² in all of the models), suggesting many other factors not included in these analyses account for the arrest decision. The findings should also be couched in terms of the overall problems associated with the sexual offenses, which are reported to law enforcement at a lower rate than other offenses (Bureau of Justice Statistics, 2017). Additionally, once reported, a small portion of those cases lead to prosecution; estimates vary among different samples, but Bouffard (2000) reported only 18% of one sample resulted in an arrest. This is likely due to prosecutors taking only cases that have enough evidence to lead to a successful conviction (Miller et al., 2011). This research suggests that law enforcement officers may believe when the offender is a woman, only the cases that involve substantial deviations from gendered norms (i.e., involve serious violence) are worthy of making an arrest. All law enforcement training should include an increased awareness of all personal biases, but especially gender biases, with regard to arrest decisions. In terms of future research, it is suggested that rather than focus on individual factors that increase the likelihood of arrest, additional analyses should assess interactions among these factors and profiles of women who are most and least likely to be arrested to identify consequences of gendered assumptions.

Footnotes

Declaration of Conflicting Interests

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

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Callie Dara Shaw

Author Biographies

Callie Shaw, JD, LLM, is currently an instructor and a Doctoral Student in the School of Criminal Justice and Criminology at Texas State University. Her research interests include female criminal behavior, green criminology, and white-collar crimes.

Tyler J. Vaughan, PhD, is an assistant professor in the Department of Sociology and Corrections at Minnesota State University Mankato. His research interests center on criminological theory, interactional theory, theory testing, mediation & moderation, and measurement.

Donna M. Vandiver, PhD, is a professor at Texas State University in the School of Criminal Justice and Criminology. Her research focuses on sexual offenders, recidivism, and co-offending patterns of interpersonal violence.

References

Allison

P. D.

(1999). Comparing logit and probit coefficients across groups. Sociological Methods & Research , 28(2), 186–208.

Barnes

J. C.

, Jorgensen

, Beaver

K. M.

, Boutwell

B. B.

, & Wright

J. P.

(2015). Arrest prevalence in a national sample of adults: The role of sex and race/ethnicity. American Journal of Criminal Justice , 40(3), 457–465.

Blackwell

B. S.

, Holleran

, & Finn

M. A.

(2008). The impact of Pennsylvania sentencing guidelines on sex differences in sentencing. Journal of Contemporary Criminal Justice , 24(4), 399–418.

Blumstein

, Cohen

, Roth

, & Visher

C. A.

(Eds.). (1986). Criminal careers and “career criminals” (Vol. 1). National Research Council.

Bouffard

(2000). Predicting type of sexual assault case closure from victim, suspect, and case characteristics. Journal of Criminal Justice , 28(6), 527–542.

Bureau of Justice Statistics. (2017). National Crime Victimization Survey , 2010–2016. GPO.

Calkins

, Colombino

, Matsuura

, & Jeglic

(2015). Where do sex crimes occur? How an examination of sexual offense location can inform policy and prevention. International Journal of Comparative and Applied Criminal Justice , 39(2), 99–112.

Chesney-Lind

(1978). Chivalry reexamined: Women and the criminal justice system. In Women, Crime and the Criminal Justice System. Lexington.

Cologne

J. B.

, & Shibata

(1995). Optimal case-control matching in practice. Epidemiology , 6(3), 271–275.

10.

Cortoni

, Babchishin

K. M.

, & Rat

(2017). The proportion of sexual offenders who are women is higher than thought: A meta-analysis. Criminal Justice & Behavior , 44(2), 145–162.

11.

Crew

(1991). Sex differences in criminal sentencing: Chivalry or patriarchy? Justice Quarterly , 8(1), 59–83.

12.

Curran

(1983). Judicial discretion and defendant’s sex. Criminology , 21(1), 41–58.

13.

D’Alessio

S. J.

, & Stolzenberg

(2003). Race and the probability of arrest. Social Forces , 81(4), 1381–1397.

14.

Daly

(1987). Discrimination in the criminal courts: Family, gender, and the problem of equal treatment. Social Forces , 66(1), 152–175.

15.

Daly

(1989). Rethinking judicial paternalism: Gender, work–family relations, and sentencing. Gender and Society , 3(1), 9–36.

16.

DeLisi

, Neppl

T. K.

, Lohman

B. J.

, Vaughn

M. G.

, & Shook

J. J.

(2013). Early starters: Which type of criminal onset matters most for delinquent careers? Journal of Criminal Justice , 41(1), 12–17.

17.

Demuth

, & Steffensmeier

(2004). The impact of gender and race-ethnicity in the pretrial release process. Social Problems , 51(2), 222–242.

18.

Denov

M. S.

(2001). A culture of denial. Exploring professional perspectives on female sex offending. Canadian Journal of Criminology , 43(3), 303–329.

19.

Denov

M. S.

(2003). The myth of innocence: Sexual scripts and the recognition of child sexual abuse by women perpetrators. Journal of Sex Research , 40(3), 303–314.

20.

Edwards

(1984). Women on trial: A study of the women suspect, defendant, and offender in the criminal law and criminal justice system . Manchester University Press.

21.

Embry

, & Lyons

P. M.

(2012). Sex-based sentencing: Sentencing discrepancies between men and women sex offenders. Feminist Criminology , 7(2), 146–162.

22.

Farrington

D. P.

, Jolliffe

, Hawkins

D. J.

, Catalano

R. F.

, Hill

K. G.

, & Kosterman

(2003). Comparing delinquency careers in court records and self-reports. Criminology , 41(3), 933–958.

23.

Federal Bureau of Investigation (FBI). (2019). Crime in the United States, 2018 . https://ucr.fbi.gov/crime-in-the-u.s/2017/crime-in-the-u.s.-2018/topic-pages/tables/table-42

24.

Feld

(2009). Violent girls or relabeled status offenders? Crime Delinquency , 55(2), 241–265.

25.

Gathings

M. J.

, & Parrotta

(2013). The use of gendered narratives in the courtroom: Constructing an identity worthy of leniency. Journal of Contemporary Ethnography , 42(6), 668–689.

26.

Gennarelli

T. A.

, & Wodzin

(2006). AIS 2005: A contemporary injury scale. Injury , 37(12), 1083–1091.

27.

Goulette

, Wooldredge

, Frank

, & Travis III

(2015). From initial appearance to sentencing: Do women defendants experience disparate treatment? Journal of Criminal Justice , 43(5), 406–417.

28.

Grabe

M. E.

, Trager

K. D.

, Lear

, & Rauch

(2006). Gender in crime news: A case study test of the Chivalry hypothesis, mass communication & society. Mass Communicaton & Society , 9(2), 137–163.

29.

Griffin

, & Wooldredge

(2006). Sex-based disparities in felony dispositions before versus after sentencing reform in Ohio. Criminology , 44(4), 893–923.

30.

Hassett-Walker

, Lateano

, & Di Benedetto

(2014). Do women sex offenders receive preferential treatment in criminal charging and sentencing? Justice System Journal , 35(1), 62–86.

31.

Hayes

, & Carpenter

(2013). Social moralities and discursive constructions of women sex offenders. Sexualities , 16(1/2), 159–179.

32.

Hislop

(2001). Female sex offenders: What therapists, law enforcement and child protective services need to know. Idyll Arbor.

33.

Justice Research and Statistics Association. (2012). Status of NIBRS in the United States . Retrieved April 1, 2017, from http://www.jrsa.org/ibrrc/background-status/nibrs-states.html

34.

Kamalu

C. N.

(2016). African Americans and racial profiling by U.S. law enforcement: An analysis of police traffic stops and searches of motorists in Nebraska, 2002–2007. African Journal of Criminology & Justice Studies , 9(1), 187–206.

35.

Katz

C. M.

, & Spohn

C. C.

(1995). The effect of race and gender on bail outcomes: A test of an interactive model. American Journal of Criminal Justice , 19(2), 161–184.

36.

Kaufman

K. L.

, Wallace

A. M.

, Johnson

C. F.

, & Reeder

M. L.

(1995). Comparing women and men perpetrators’ modus operandi: Victims’ reports of sexual abuse. Journal of Interpersonal Violence , 10(3), 322–333.

37.

Kruttschnitt

(1984). Sex and criminal court dispositions: The unresolved controversy. Journal of Research in Crime and Delinquency , 21(3), 213–232.

38.

Lauritsen

J. L.

, Heimer

, & Lynch

J. P.

(2009). Trends in the gender gap in violent offending: New evidence from the National Crime Victimization Survey. Criminology , 47(2), 361–399.

39.

Miller

S. L.

, Iovanni

, & Kelley

K. D.

(2011). Violence against women and the criminal justice response. In Renzetti

C. M.

, Edleson

J. L.

& Bergen

R. K.

(Eds.), Sourcebook on violence against women (2nd ed., pp. 267–287). Sage Publications.

40.

Moffitt

T. E.

(2017). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. In Beaver

(Ed.), Biosocial theories of crime (pp. 69–96). Routledge.

41.

Mustillo

S. A.

, Lizardo

O. A.

, & McVeigh

R. M.

(2018). Editors’ comment A few guidelines for quantitative submissions. American Sociological Review , 83(6), 1281–1283.

42.

National Archive of Criminal Justice Data. (2010–2016). National Incident-based Reporting System: Extract Files (Inter-university Consortium for Political and Social Research [distributor]). https://doi.org/10.3886/ICPSR33601.v2

43.

Ousey

G. C.

, & Lee

M. R.

(2008). Racial disparity in formal social control: An investigation of alternative explanations of arrest rate inequality. Journal of Research in Crime and Delinquency , 45(3), 322–355.

44.

Pattavina

, Carkin

D. M.

, & Tracy

P. E.

(2017). Assessing the representativeness of NIBRS arrest data. Crime & Delinquency , 63(12), 1626–1652.

45.

Pattavina

, Hirschel

, Buzawa

, Faggiani

, & Bentley

(2007). A comparison of the police response to heterosexual versus same-sex intimate partner violence. Violence Against Women , 13(4), 374–394.

46.

Pattavina

, Hirschel

, & Scearbo

(2013). Reliability in NIBRS reporting of substance use in incidents of intimate partner violence. Justice Research and Policy , 15(12), 21–42.

47.

Ridgeway

(2006). Assessing the effect of race bias in post-traffic stop outcomes using propensity scores. Journal of Quantitative Criminology , 22(1), 1–29.

48.

Rodriguez

, Curry

T. R.

, & Lee

(2006). Gender differences in criminal sentencing: Do effects vary across violent, property, and drug offenses? Social Science Quarterly , 87(2), 318–339.

49.

Romain

D. M.

, & Freiburger

T. L.

(2016). Chivalry revisited: Gender, race/ethnicity, and offense type on domestic violence charge reduction. Feminist Criminology , 11(2), 191–222.

50.

Rose

, & der Laan

M. J. van.

(2009). Why match? Investigating matched case-control study designs with causal effect estimation. The International Journal of Biostatistics , 5(1). https.//doi.org/10.2202/1557-4679.1127

51.

Sampson

R. J.

, & Laub

J. H.

(2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology , 41(3), 555–592.

52.

Sandler

J. C.

, & Freeman

N. J.

(2009). Women sex offender recidivism: A large-scale empirical analysis. Sexual Abuse: A Journal of Research and Treatment , 21(4), 455–473.

53.

Sandler

J. C.

, & Freeman

N. J.

(2011). Women sex offenders and the criminal justice system: A comparison of arrests and outcomes. Journal of Sexual Aggression , 17(1), 61–76.

54.

Sarnikar

, Sorensen

, & Oaxaca

R. L.

(2007, June). Do you receive a lighter prison sentence because you are a woman? An economic analysis of federal criminal sentencing guidelines (IZA Discussion Paper No. 2870). An Economic Analysis of Federal Criminal Sentencing Guidelines. http://ftp.iza.org/dp2870.pdf

55.

Schwartz

, Steffensmeier

, Zhong

, & Ackerman

(2009). Trends in the gender gap in violence: Reevaluating NCVS and other evidence. Criminology , 47(2), 401–425.

56.

Shadish

W. R.

, Cook

T. D.

, & Campbell

D. T.

(2002). Experimental and quasi-experimental designs for generalized causal inference . Houghton Mifflin.

57.

Spivak

A. L.

, Wagner

B. M.

, Whitmer

J. M.

, & Charish

C. L.

(2014). Gender and status offending: Judicial paternalism in juvenile justice processing. Feminist Criminology , 9(3), 224–248.

58.

Spohn

C. C.

(2013). Racial disparities in prosecution, sentencing, and punishment. In M. Tonry & S. M. Bucerius (Eds.), The oxford handbook of ethnicity, crime, and immigration . https://doi-org.libproxy.txstate.edu/10.1093/oxfordhb/9780199859016.001.0001

59.

Steffensmeier

(1980). Assessing the impact of the women’s movement on sex-based differences in the handling of adult criminal defendants. Crime & Delinquency , 26(3), 344–357.

60.

Steffensmeier

, Kramer

, & Streifel

(1993). Gender and imprisonment decisions. Criminology , 31(3), 411–446.

61.

Stolzenberg

, & D’Alessio

S. J.

(2004). Sex differences in the likelihood of arrest. Journal of Criminal Justice , 32(5), 443–454.

62.

Stolzenberg

, D’Alessio

S. J.

, & Eitle

(2004). A multilevel test of racial threat theory. Criminology , 42(3), 673–698.

63.

Thompson

M. P.

, Saltzman

L. E.

, & Bibel

(1999). Applying NIBRS data to the study of intimate partner violence: Massachusetts as a case study. Journal of Quantitative Criminology , 15(2), 163–180.

64.

Tillyer

, Hartley

R. D.

, & Ward

J. T.

(2015). Differential treatment of women defendants: Does criminal history moderate the effect of gender on sentence length in federal narcotics cases? Criminal Justice & Behavior , 42(7), 703–721.

65.

Tracy

, Kempf

, & Abramoske

J. S.

(2009). Gender differences in delinquency and juvenile justice processing: Evidence from national data. Crime Delinquency , 55(2), 171–215.

66.

Truman

, & Morgan

(2016). Criminal Victimization, 2015 . U.S. Department of Justice, Bureau of Justice Statistics.

67.

Vandiver

D. M.

(2010). Assessing gender differences and co-offending patterns of a predominantly “male-oriented” crime: A comparison of a cross-national sample of juvenile boys and girls arrested for a sexual offense. Violence and Victims , 25(2), 243.

68.

Vandiver

, Braithwaite

, & Stafford

(2016). Sex crimes and sex offenders: Research, and realities . Routledge; Taylor & Francis Group.

69.

Vandiver

D. M.

, & Kercher

(2004). Offender and victim characteristics of registered women sexual offenders in Texas: A proposed typology of women sexual offenders. Sexual Abuse: A Journal of Research and Treatment , 16(2), 121–137.

70.

Vaughan

T. J.

, Pollock

, & Vandiver

D. M.

(2015). Sex differences in arrest for juvenile assaults. Violence and Gender , 2(1), 24–34.

71.

Visher

C. A.

(1983). Gender, police arrest decisions, and notions of chivalry. Criminology , 21(1), 5–28.

72.

Wijkman

, Bijleveld

, & Hendriks

(2010). Women don’t do such things! Characteristics of female sex offenders and offender types. Sexual Abuse: A Journal of Research and Treatment , 22(10), 135–156.

73.

Wilbanks

(1986). Are female felons treated more leniently by the criminal justice system? Justice Quarterly , 3(4), 517–529.

74.

Williams

(2009). Using heterogeneous choice models to compare logit and probit coefficients across groups. Sociological Methods & Research , 37(4), 531–559.

75.

Williams

K. S.

, & Bierie

D. M.

(2015). An incident-based comparison of women and men sexual offenders. Sexual Abuse: A Journal of Research & Treatment , 27(3), 235–257.

76.

Zingraff

, & Thomson

(1984). Differential sentencing of women and men in the USA. International Journal of the Sociology of Law , 12(4), 401–441.