Abstract
The present study used a stratified random sample of 332 case files from a sizable, urban police department located in one of the five largest and most diverse U.S. cities to examine the effect of previous stalking on arrest decisions among intimate partner violence (IPV) cases. Relevant extralegal and legal case factors were estimated in a multivariate binary logistic regression model to determine correlates of arrest. Findings revealed that previous stalking behavior was not related to arrest. Evidence and physical injury significantly increased the odds of arrest. Presence of witnesses, alcohol involvement, and history of IPV also significantly increased the odds of arrest. Future research and policy implications are discussed.
Introduction
Intimate partner violence (IPV) and stalking occur with frequency. Findings from the National Intimate Partner and Sexual Violence Survey (NISVS; Smith et al., 2018) indicated more than one third of U.S. women experienced multiple forms of lifetime violence perpetrated by a current or former intimate partner. Moreover, nearly one sixth of U.S. women have reported lifetime stalking victimization (Smith et al., 2018). Stalking has been conceptualized as deliberate conduct characterized by unwanted pursuit behavior that escalates in duration and intensity by an offender and elicits fear in a victim (National Center for Victims of Crime [NCVC], n.d.-b; Spitzberg & Cupach, 2007).
A significant connection exists between IPV and stalking, where women are most often stalked by someone they know. Perpetrators are typically a current or former intimate partner (Catalano et al., 2009; Logan, 2010). More than half (61.5%) of female stalking victims reported being targeted by partners, compared with strangers, acquaintances, or others (Smith et al., 2017). Even so, some scholars have argued that stalking victimization can only occur when the perpetrator and victim are strangers (Douglas & Dutton, 2001). Others have underscored the importance of recognizing the nexus between IPV and stalking (Logan & Walker, 2009), given the escalation of violence, potential harm, even lethality that a stalking perpetrator may inflict on a current or former intimate partner. Partner-perpetrated stalking has been identified as more dangerous, consisting of longer duration, and producing significantly more harm than stranger-perpetrated stalking (Brady & Hayes, 2018; Logan et al., 2007; Logan & Walker, 2010; McFarlane et al., 1999; Mechanic et al., 2000).
Despite the prevalence of IPV and stalking, estimates have demonstrated that both offenses remain largely underreported (Brady & Nobles, 2017; Fisher et al., 2002; Jordan et al., 2007; Morgan & Kena, 2017; Reyns & Englebrecht, 2010). Similarly, low formal reporting has been documented among instances of intimate partner stalking. Augustyn et al.’s (2019) study using a college sample noted approximately 6% of intimate partner-perpetrated stalking incidents were reported to police, though higher rates have been reported among community samples (Brewster, 2001; Logan, Shannon, et al., 2006; Tjaden & Thoennes, 1998). Still, whenever IPV and stalking offenses do come to the attention of police, responses have been lacking. For instance, IPV was not historically perceived as warranting formal police intervention (Martin, 1975), including suspect apprehension and arrest (Lutze & Symons, 2003; Sherman & Berk, 1984). Decades of efforts translated to more progressive police response (Freedman, 2002), though shortcomings have remained, such as the use of discretion in arrest decisions (Dichter et al., 2011) and case attrition (Jordan, 2004).
In addition, police response to stalking has been plagued with similar limitations that impeded IPV case processing. For example, police have reiterated some of the same stereotypes and cultural myths commonly endorsed for IPV (Scott et al., 2013; Weller et al., 2013)—that victims who know their assailants hold some responsibility for precipitating stalking (Lynch & Logan, 2015; Scott et al., 2013; Weller et al., 2013). Police responses have been characterized by the underidentification of stalking, limited knowledge of stalking statutes, and confounding stalking with other offenses (Brady & Nobles, 2017; Jasinski & Mustaine, 2001; Jordan et al., 2003; Lynch & Logan, 2015). This has underscored general deficits in police response to stalking victimization. Much remains unknown, however, regarding the formal processing of IPV incidents that have involved stalking behaviors.
Police Response to IPV and Stalking
IPV
Police response to IPV has evolved to better meet the needs of survivors. Traditionally, IPV was perceived by law enforcement as a “private family matter” (Berk & Loseke, 1980). Police endorsed ambivalent attitudes toward IPV (Belknap, 1995), delayed response to calls for service, and when there was a response, police typically did not arrest the perpetrator (Hirschel et al., 1992; Miller, 2000). This reiterated the lack of seriousness afforded to IPV. The second wave of the women’s movement brought attention in urging police to be more effective in terms of policy and response (Dicker, 2008; Freedman, 2002). Findings from Sherman and Berk’s (1984) Minneapolis Domestic Violence Study, for example, initiated aggressive reform and mandatory arrest policies across the country. This punitive response translated to unintended collateral consequences for IPV survivors, including dual arrests (Lutze & Symons, 2003). Presently, jurisdictions have moved toward primary aggressor laws or proarrest laws (Messing et al., 2015). Despite these changes, the majority of IPV incidents do not result in arrest (Durfee & Fetzer, 2016; Eitle, 2005).
Existing research has attempted to identify legal and extralegal, incident-level factors associated with arrest in IPV incidents. The most robust and consistent predictors of arrest in IPV incidents included the presence of a weapon (Dichter et al., 2011; Durfee & Fetzer, 2016; Eigenberg et al., 1996; Eitle, 2005; Hamilton & Worthen, 2011), victim injury (Bachman & Coker, 1995; Buzawa & Austin, 1993; Dichter et al., 2011; Durfee & Fetzer, 2016; Eitle, 2005; Hamilton & Worthen, 2011; Robinson & Chandek, 2000; Tatum & Pence, 2015), presence of witnesses (Buzawa & Austin, 1993; Eigenberg et al., 1996; Hamilton & Worthen, 2011; Robinson & Chandek, 2000), and victim preference (Buzawa & Austin, 1993; Eigenberg et al., 1996). Extralegal factors, such as alcohol involvement, have also predicted arrest (Dichter et al., 2011; Durfee & Fetzer, 2016; Hamilton & Worthen, 2011; Roark, 2016). In terms of race/ethnicity, some studies have demonstrated that arrest has increased in IPV incidents involving White victims when compared with other racial/ethnic groups (Dichter et al., 2011; Eitle, 2005; Hamilton & Worthen, 2011; McCormack & Hirschel, 2018).
Stalking
Only recently have police been tasked with handling stalking in the context of stranger- and partner-perpetrated behavior. The first stalking statute was created in California in 1990 and by 1995, 50 states, the District of Columbia, and the federal government had laws to guide criminal justice processing of stalking offenses (NCVC, n.d.-a; Spitzberg & Cupach, 2007). Each jurisdiction has varied in terms of the law, but there are several consistent components: stalking must involve repeated attempts at unwanted contact and the victim must acknowledge a sense of “reasonable person” fear from the pursuit behavior (Fox et al., 2011). Advances in technology have created avenues for perpetrators to maintain contact with victims through electronic communication or internet-capable devices (Reyns et al., 2011; Spitzberg, 2016). Stalking is further complicated where the police and victim are working in concert during the investigative process because victims are often relied upon to save and/or maintain evidence (e.g., letters, text messages, voice mail, and other electronic or technological communication) to display a pattern of contact (Lynch & Logan, 2015). This translates to multiple interactions with police, which may frustrate police and inhibit victim cooperation.
Limited studies have examined police response to stalking (Brady & Nobles, 2017; Jasinski & Mustaine, 2001; Jordan et al., 2003; Storey & Hart, 2011; Tjaden & Thoennes, 2000) and have noted shortcomings. Police may not recognize stalking behavior or may not fully understand the statute guiding their jurisdiction (Lynch & Logan, 2015). Officers may use alternative offense codes to produce case dispositions that can be applied to a single incident, like harassment or terroristic threat, rather than statutes that capture unwanted pursuit behavior over time (Brady & Nobles, 2017). In addition, police may not recognize the gravity of a stalking situation or see the need for justice or social service intervention (Logan, Cole, et al., 2006). Finally, taken alone, some of these contact behaviors are typically not criminal and must be evaluated in the context of a relationship (Logan, 2010). Consequently, officers are tasked with delineating events in the current incident from the broader context of pursuit behavior. This may be difficult for police who have not had extensive experience in responding to stalking incidents and they may be unwilling or unable to devote time, energy, and resources (Horwitz et al., 2011; Lynch & Logan, 2015).
Compared with the literature on arrest in IPV incidents, to the best of our knowledge, one study has examined police decision making in stalking incidents. Ngo (2019) examined data from the 2006 Stalking Victimization Supplement (SVS) of the National Crime Victimization Survey (NCVS) where 1,599 respondents met the criteria for stalking victimization. Overall, 348 respondents reported they or someone they knew had contacted police about the stalking incident. Officers were significantly more likely to arrest the suspect if the victim sustained a physical injury.
The Nexus Between IPV and Stalking
The important connection between IPV and stalking cannot be understated. A meta-analysis of 108 stalking studies, representing nearly 69,000 participants, identified about half of the stalking incidents involved perpetrators who had a romantic history with the victim, whereas approximately 18% of incidents involved stranger perpetrators (Spitzberg, 2002). Among the most robust findings in the stalking literature have been the dangerousness involved when the perpetrator is a former or current intimate partner (Logan, 2010; Spitzberg & Cupach, 2007). Logan and Walker (2009) found that victims of partner-perpetrated stalking were more likely than victims of nonintimate partner stalking to report protection order violations and faced increased risk of physical, sexual, and psychological abuse (Brewster, 2003; Logan et al., 2007).
Furthermore, partner-perpetrated stalking behavior escalates in both frequency and duration (Mohandie et al., 2006; Nicastro et al., 2000), and draws upon the prior relationship context to elicit fear. Partner-perpetrators are privy to personal information regarding the victim’s weaknesses and can exploit this information to punish and humiliate a victim. In addition, partner- or ex-partner-perpetrators have intimate knowledge surrounding a victim’s routine activities, friends, schedules, and details surrounding their occupations, hobbies, and family members (Logan, Cole, et al., 2006). All of this can be used by partner-perpetrators to intimidate, coerce, and threaten victims. Despite these findings and consistent evidence regarding the severity and dangerousness of stalking offenses, limited analyses have examined police response to stalking behaviors within IPV incidents.
Tjaden and Thoennes (2000) were among the first to examine police response to stalking in the context of IPV. Using a sample of 1,785 incident reports from the Colorado Springs Police Department, Tjaden and Thoennes (2000) found 16.5% of incidents mentioned behaviors that qualified as stalking. A closer look revealed that only a single case was disposed through formal stalking charges. Results demonstrated that stalking reports were significantly less likely to involve physical abuse or victim injury, alcohol consumption, or the presence of children compared with reports of no stalking behaviors. Stalking incidents were significantly more likely to involve cooperative victims, victims who showed behavioral signs of emotional upset, and had an active restraining order in place. Despite differences in presenting conditions, law enforcement outcomes did not differ substantially across cases with and without stalking behaviors, including tagging evidence, suspect arrest, or felony charge. More recently, Melton (2012) used 1,140 police-classified heterosexual IPV incident reports to assess the extent of stalking and predictors of arrest. Overall, 27.9% of incidents had evidence of at least one stalking behavior. In addition, police officers were significantly more likely to make an arrest in IPV cases if stalking behaviors were existent, the suspect was on scene, either party used drugs or alcohol, and there was evidence of prior relationship violence.
Purpose of the Present Study
Literature has focused on identifying correlates of arrest within IPV incidents and attention to the formal processing of stalking offenses is growing. Presently, only two studies have focused on police response to stalking within the context of IPV, despite the nexus between IPV and stalking. Much remains unknown in terms of identifying factors associated with the arrest decision for case clearance. The present study used a stratified random sample of 332 redacted case files from a sizable, urban police department located in one of the five largest and most diverse U.S. cities to examine the effect of previous stalking behavior on arrest in IPV incidents. The study addressed the following research questions:
Method
It is important to note the process by which an incident number is generated and a case file is created. First, a complainant or victim contacts police, activating the criminal justice response. In this jurisdiction, a call to the 911 dispatcher sends a responding officer to the scene of the incident. Once on scene, the responding officer will identify the nature of the presenting incident, noting characteristics such as the location, victim demeanor, physical appearance, injury, suspect(s), suspect demeanor, physical appearance, injury, and the presence of witnesses. During this initial report, the officer will record information obtained from any person on scene. This typically involves an interview with the victim to record the details of the family violence offense. Depending on the circumstances, the officer may make an on-scene arrest, provide service referral, gather contact information, take initial statements, and record any notable injury or weapons present at the incident location. Once this report is taken, the case may be assigned to an investigator from the specialized Family Violence Unit for follow-up.
When an investigator has been assigned, tasks involve securing sworn statements; making contact with victims, witnesses, and suspects; evidence retrieval into police custody; arrest; and presentation to the district attorney (DA). Each time an officer/investigator engages with a case, a new document or supplement is produced and added to the case file that is identified by the initial incident number from the presenting incident to which the responding officer was called. Case files for the current study were obtained from family violence incidents. This means that all incidents where the criminal justice system was activated in the present study involved a family violence offense (Special Victims’ Division Administrative Lieutenant, Personal Communication, September 19, 2017).
Sample and Procedure
Data
Data were derived from a larger federally funded study by the Office on Violence Against Women that involved a stratified random sample of redacted family violence case files collected from a sizable municipal police agency located in one of the five largest and most diverse U.S. cities. A population of 52,058 family violence cases were generated from January 1, 2014, through August 31, 2016, by the police agency and provided to the principal investigator (PI), who then created a stratified random sample of 619 cases. This included 125 cases in each category by disposition: (a) cleared—arrest, (b) cleared—other, (c) open, and (d) inactive. In addition, the population of 119 unfounded cases were added to the sample. A total of 619 family violence incident numbers were provided to the police agency and personnel collected, printed, and redacted each case file so that victim, suspect, and officer information were de-identified. Once that information was obtained by the PI from the police agency, 618 files were coded and entered for quantifiable data into an SPSS spreadsheet. One case file was labeled confidential, precluding access. From there, 428 cases were selected as current or former intimate partners. Among these, one incident had missing data on suspect sex for a total subsample of 427 cases. Incidents involving adult female complainants and male suspects were selected for a final sample of 332 cases.
Following training on coding case files, including extensive and repeated meetings with Special Victims’ Division personnel at the police partner agency, the research team entered data directly into SPSS 25.0. Data entry involved extensive training with six graduate research assistants, a detailed codebook, and regular contact via in-person and telephone meetings and email communication with police personnel to first navigate the police case file format and subsequently discuss issues that arose. The PI supervised and checked data entry to ensure intercoder reliability. Case files were entered in an iterative process where the research team first coded together for a total of approximately 80 hours and then met routinely to discuss discrepancies and reach consensus throughout the data entry process. Each case file allowed for the extraction of detailed quantifiable information including the responding officer’s initial incident report; victim, suspect, and witness information; accounts of the incident; whether the case was assigned to an investigator and subsequent reports; description of evidence; injury; victim cooperation; criminal history; and whether the incident was accepted for charges by the DA. The quantitative coding and data entry of each case file took approximately 35 to 60 min, depending on the complexity of each incident and length of each case. From this data spreadsheet, variables for the present study were operationalized and relied upon affirmative, explicit mention of particular measures of interest. This was done because the information that has been documented in the case file by the responding officer or investigator represents decision making by police in terms of which details were meaningful for that criminal justice actor to record and deemed pertinent from their perspective at the time of the incident. This information, once recorded by the responding officer or investigator, follows the case through multiple formal processing stages and may have an influence on the final disposition.
Measures
Dependent Variable
Arrest was captured from information contained in the case file and measured as a binary variable for the presenting IPV incident (no = 0, yes = 1).
Independent Variable
Previous stalking behavior was captured based on existing research. Previous stalking behavior was conceptualized as an indicator of severity and coded based on behavioral descriptions of a repeated, unwanted course of pursuit conduct identified in the case file. Nine items from Fisher et al. (2000, 2002) captured stalking and included whether the suspect had previously “telephoned,” “waited outside or inside places,” “watched from afar,” “followed,” “sent letters,” “emailed,” “sent gifts,” or “showed up uninvited.” In addition to the eight pursuit contact behaviors, an additional option captured “other” unique pursuit behaviors. For example, one complainant reported that the suspect had previously sent friends to drive by the complainant’s home multiple times. Coding affirmative to any of these pursuit behaviors was based on explicit mention in the case file of at least two pursuit behaviors (e.g., followed and watched from afar, texted and followed, or multiple instances of the same pursuit behavior) prior to the presenting IPV incident. Nine items were collapsed into one binary measure and titled previous stalking behavior (no = 0, yes = 1).
Extralegal Factors
Seven variables were included to represent extralegal factors in police arrest decisions. Victim race was captured as a binary variable (White = 0, People of Color = 1). Unfortunately, due to the nature of case file data and the discretion that police officers have when reporting information, Latinx ethnicity was not reliably or systematically captured. Victim race and suspect race were highly correlated (rs = .768); therefore, the authors chose to retain victim race. Victim age was captured as a continuous variable. Victim age and suspect age were highly correlated (rs = .751); therefore the authors chose to retain victim age. History of IPV captured any explicit mention in the case file of whether the suspect had: (a) been a suspect in a previous incident prior to the current case, (b) had been previously arrested for domestic violence, or (c) had been previously indicted for domestic violence–related charges. Any affirmative, explicit history was coded “yes” (no = 0, yes = 1). Alcohol involvement was captured as a binary variable that measured if one or both parties had consumed alcohol and was explicitly mentioned in the case file as part of the presenting IPV incident by the responding officer, complainant, suspect, or witnesses (no = 0, yes = 1). While alcohol consumption has often deterred survivors from contacting police, here, alcohol was used as a proxy for victim and/or suspect demeanor where research has demonstrated that arrest is more likely when suspects behaved in disrespectful or belligerent ways (e.g., Klinger, 1994).
In addition, presence of children was captured using a single binary variable to identify the affirmative and explicit mention in the case file of whether children were on scene during the presenting IPV incident (no = 0, yes = 1). Threats to kill captured suspect threats to kill the complainant and/or children in the past or presenting IPV incident and was measured through affirmative and explicit mention in the case file (no = 0, yes = 1). Victim preference for formal intervention was captured using three variables of victim explicit preference for criminal justice intervention in the presenting IPV incident, “arresting the suspect,” “charging the suspect,” or “prosecuting the suspect.” Affirmative mention of any preference for formal intervention were coded “yes” (no = 0, yes = 1).
Legal Factors
Four variables were included in the present study to represent legal factors on police arrest decisions. Evidence was captured using eight binary items that reflected tangible, physical evidence described by the responding officer or investigator in the case file. These included the presence of “biological evidence,” “photographs,” “weapons,” “complainant sworn statement,” “witness sworn statement,” “electronic text messages/voice mails/calls,” “destroyed personal property” and “other” evidence, to capture unique evidence that may not have been regularly present in each incident. A complainant’s debit card was evidence in one case file and a suspect’s birth certificate was evidence in a separate case file. The eight items were summed and ranged from 0 to 8 where higher numbers represented increased quantity of physical evidence for the presenting IPV incident. Evidence such as bed sheets, underwear, condoms, clothes, and surveillance footage was part of the codebook; however, none of the incidents retained for analysis in this study described these specific types of evidence.
Physical injury was captured using 10 binary items that reflected injuries inflicted on the victim as described in the case file for the presenting IPV incident, including “bruises,” “cuts,” “scratches,” “redness,” “swelling,” “broken bones,” “bleeding,” “head injury,” or “strangulation,” and “other” injury, such as bite marks. The 10 items were summed and ranged from 0 to 10 where higher numbers represented increased injury to the survivor in the presenting incident. Injuries resulting from being shot, stabbed, and being burned were part of the codebook; however, none of the incidents retained for analysis in this study described these specific types of injuries. Weapon used was measured as a binary variable that captured explicit mention of weapon use in the case file (no = 0, yes = 1). This excluded bodily force. Presence of witnesses captured explicit mention of witnesses (e.g., adults) at the scene of the presenting IPV incident (no = 0, yes = 1).
Analytic Strategy
The analyses proceeded in two stages. First, bivariate correlations, means, and standard deviations were calculated to identify significant relations between variables. Given the binary nature of the dependent variable, a sequential multivariate binary logistic regression model was estimated to assess arrest (Long, 1997). Model 1 included the dependent variable regressed on the extralegal variables. Model 2 added the legally relevant factors. Finally, Model 3 contained the primary variable of interest: previous stalking behavior. SPSS, version 25.0, was used to analyze the data. Prior to estimating the statistical models, the data were screened for skewness and kurtosis. Estimates fell within the acceptable range and did not exceed recommended cut off values of 3.0 and 8.0, respectively (Kline, 2005). Multicollinearity diagnostics were also evaluated; tolerances ranged from .707 to .970, and variance inflation factors (VIFs) ranged from 1.03 to 1.19, indicating that multicollinearity was not a problem (Belsey et al., 1980).
Results
Descriptive Statistics
Table 1 presents the descriptive statistics for the variables addressing RQ1. Specifically, 19.3% of 332 IPV cases involved previous stalking behavior. As displayed in Table 1, approximately one third of IPV incidents resulted in arrest. Just more than half of the victims were People of Color and victims reported a mean age of 30.77 years. Among IPV cases, limited evidence was available and victims presented with at least one form of physical injury. In addition, the vast majority of cases did not involve the use of a weapon, nor did they involve alcohol, although about a quarter of incidents did involve a witness. More than half of the victims did not have a preference for formal intervention, and children were not present in most of these IPV incidents. Finally, the majority of suspects did not have a history of IPV and in the majority of incidents, suspects did not make threats to kill the victim and/or children.
Descriptive Statistics of Study Variables (N = 332)
Note. IPV = intimate partner violence.
Bivariate Correlation Matrix
Table 2 presents the Spearman bivariate correlation matrix and demonstrates several significant and substantively important relations between the dependent and independent variables.
Bivariate Correlation Matrix of Study Variables (N = 332)
Note. IPV = intimate partner violence.
p < .05. **p < .01.
Binary Logistic Regression Model Assessing the Arrest Decision
Table 3 presents the results of the sequential multivariate binary logistic regression model, addressing RQ2. The regression equation for Model 1, where the dependent variable was regressed on extralegal factors, was statistically significant, χ2(7) = 35.83, p < .001. First, history of IPV was significant and positively correlated with arrest, B = 0.87, odds ratio (OR) = 2.39, p < .001, where incidents involving a history of IPV were 2.39 times more likely to result in arrest compared with counterparts. Alcohol involvement emerged as the strongest correlate of arrest, B = 1.34, OR = 3.82, p < .001, with alcohol-involved incidents nearly four times more likely to result in arrest as those without alcohol consumption.
Sequential Multivariate Binary Logistic Regression Models Examining the Arrest Decision in IPV cases (N = 332)
Note. IPV = intimate partner violence; CI = confidence interval.
p < .05. **p < .01.
The regression equation for Model 2, where the dependent variable was regressed on legal and extra legal factors, was statistically significant, χ2(11) = 101.95, p < .001. In Model 2, history of IPV remained a significant, positive correlate of arrest, B = 0.65, OR = 1.92, p = .030, as did alcohol involvement, B = 1.14, OR = 3.12, p = .002. Physical evidence was also a significant, positive correlate of arrest, B = 0.95, OR = 2.60, p < .001, where a one-unit increase in evidence increased the odds of arrest by 2.60 times. Physical injury was also a significant, positive correlate of arrest, B = 0.26, OR = 1.29, p = .014, where a one-unit increase in the physical injury index increased the odds of arrest by 1.29 times. Finally, incidents involving witnesses were significantly more likely to end in arrest, B = 0.69, OR = 1.99, p = .031, compared to incidents with no witnesses.
The regression equation for Model 3, where the dependent variable was regressed on extralegal and legal factors and the independent variable of interest, was statistically significant, χ2(12) = 102.82, p < .001. First, findings revealed previous stalking behavior was not a statistically significant correlate of arrest, B = −0.40, OR = 0.67, p = .358. Evidence remained significant, B = 0.94, OR = 2.57, p < .001, where a one-unit increase in the evidence index increased arrest by 2.57 times. Physical injury was also a significant, positive correlate of arrest, B = 0.23, OR = 1.26, p = .028, increasing the odds of arrest by 1.26 times. Presence of witnesses remained significant, B = 0.68, OR = 1.98, p = .032, where cases involving witnesses were 1.98 times more likely to result in arrest. Alcohol involvement also remained significant, B = 1.11, OR = 3.02, p = .003, where cases with alcohol consumption were 3.02 times more likely to end in arrest. Finally, history of IPV was a significant, positive correlate of arrest, B = 0.70, OR = 1.99, p; = .023 incidents with a history of IPV were nearly two times more likely to end in arrest. The remaining variables were not significant.
Discussion
This study used a stratified random sample of 332 case files from a sizable police department located in one of the five largest and most diverse U.S. cities to address: (a) the extent of previous stalking behavior among IPV cases, and (b) the effect of previous stalking behavior on arrest decisions in IPV cases, while considering extralegal and legal factors. Several findings warrant discussion. First, results demonstrated that nearly one in five incidents, or 19.3% of the total incidents, included previous stalking (RQ1). This estimate falls in line with existing research among family violence incident reports from other police agencies (Melton, 2012; Tjaden & Thoennes, 2000). The IPV case file data used for the current study did not have a formal indicator for stalking behaviors. From a policy standpoint, the police agency partner may benefit from including a stalking assessment or supplement for use when responding to IPV cases to document the life course of abusive relationships and highlight the seriousness of partner-perpetrated stalking as a risk factor for future justice involvement.
Second, previous stalking behavior was not a correlate of arrest (RQ2). This finding contradicts Melton’s (2012) study findings. One explanation may be the difference in measures used to operationalize stalking. This study employed items from Fisher et al. (2000, 2002) to capture stalking behaviors prior to the presenting IPV incident detailed in the case file, while Melton (2012) used the Stalking Behavior Checklist (Coleman, 1997) to capture stalking within domestic violence–related incident reports. This reiterates the need for general consensus on how to capture stalking victimization (Fox et al., 2011).
Another possible explanation could be that officers in this agency have difficulty identifying stalking behaviors, particularly those within the context of IPV, as potentially dangerous for victim safety, and may not take this into consideration when deciding to make an arrest. Research has demonstrated limitations in police response to stalking, including underidentification and the confounding of stalking with other misdemeanor offenses (Brady & Nobles, 2017; Farrell et al., 2000; Lynch & Logan, 2015; Tjaden & Thoennes, 2000). Educational programming that targets stalking identification may better equip officers in addressing stalking that may present in IPV incidents (The Stalking Prevention, Awareness, and Resource Center [SPARC], n.d.). Recent years have seen the implementation of specialized police training to develop best practices concerning gender violence (Franklin et al., 2020; Oehme et al., 2016; Renzetti et al., 2015; Smith et al., 2016).). Stalking may be among those gendered offenses in which specialized training enhances police response and subsequently, successful case processing.
Third, results indicated that evidence and physical injury significantly increased the odds of arrest. These findings reiterate existing research on IPV case processing that has demonstrated the role of legally relevant factors, such as evidence (Franklin et al., 2019) and physical injury (Durfee & Fetzer, 2016; Eitle, 2005; Hamilton & Worthen, 2011; Ngo, 2019; Tatum & Pence, 2015) in explaining arrest decisions. Where police have traditionally operated as reactive in crime response (Bayley, 1994), demonstrable physical injuries and the quantity of corroborating evidence can enhance the strength of a case and limit discretion used by officers when responding to IPV. Future research should consider the role of these factors in terms of other police decision making including the decision to prescreen a case with the DA or the decision to refer victims to social services.
Fourth, history of IPV was a significant correlate of arrest in incidents, reiterating existing research (Melton, 2012) and underscoring the salience of documenting the life course of violent relationships. The presence of witnesses was also an influential factor in the arrest decision. This replicates existing literature where on-scene witnesses have contributed to formal arrest decisions in IPV incidents (Buzawa & Austin, 1993; Eigenberg et al., 1996; Hamilton & Worthen, 2011; Robinson & Chandek, 2000), suggesting that corroboration of partner abuse allegations is a salient factor in police decision making. Finally, alcohol involvement was the strongest correlate of arrest. This also reiterates previous research where alcohol consumption has been correlated with arrest (Dichter et al., 2011; Durfee & Fetzer, 2016; Eitle, 2005; Hamilton & Worthen, 2011; Melton, 2012; Roark, 2016). This study did not account for demeanor while intoxicated, alcohol consumption functioned as a proxy for demeanor and compliance with police. Future research should continue to include this consideration.
Findings from this study are instructive, though not without limitations. First, the conceptualization of stalking did not include the element of fear as the research team quantified explicit mention of pursuit behaviors in the case file using items taken directly from Fisher et al. (2000, 2002). The majority of stalking statutes require a pattern of pursuit conduct that invokes fear for the victim (Fox et al., 2011; NCVC, n.d.-a), though scholars have raised several issues regarding the fear standard, including the invalidation and undercounting of victims who internalize emotions differently (Dietz & Martin, 2007). These individuals may not approach police with behavioral displays of fear commensurate with their trauma (Franklin et al., 2020). Future research should employ methodological designs to measure fear within the context of partner-perpetrated stalking to capture the way fear operates for survivors and how this affects police decisions. In addition, the measure does not capture stalking behaviors that began after the initial report to police. In these case files, there were limited incidents where following the initial report, the suspect began to call or harass her, follow her, show up uninvited to her house, or engaged in some other type of stalking behavior. This was not captured in the prior stalking behavior measure and is a point for future research.
Second, this study examined the arrest decision in heteronormative IPV cases with a female complainant and male suspect in a current or former intimate relationship. Findings cannot be generalized to police interactions with female suspects and male complainants and other intimate relationship dyads such as same-sex, transgender, or gender-nonconforming couples. Cases involving same-sex couples or female suspects were excluded from analyses for a couple of reasons. Prior research has reported female-perpetrated cross-sex stalking has accounted for approximately 9% of stalking incidents (Spitzberg, 2002). In this sample, only six cases involved female-perpetrated, cross-sex stalking. In addition, inherent problems exist when applying the gender paradigm and heteronormative language designed to captured heterosexual IPV to LGBTQ+ (lesbian, gay, bisexual, transgender, queer) couples because it hinders theoretical understanding of the IPV dynamics and experiences in this community (Cannon & Buttell, 2015; Sanger & Lynch, 2017). Future research should assess police response to female-perpetrated cross-sex stalking within the context of IPV and incidents involving LGBTQ+ couples to determine if police decision making functions similarly.
Finally, redacted case file data used for the present study were collected from a sizable, municipal police department located in one of the five largest and most diverse U.S. cities. Results should be interpreted accordingly. While case files are rich in detail, it is noteworthy that each is limited in terms of what is reported by the individuals involved and how it gets recorded by the responding officer and investigators responsible for the case. Researchers can only extract available data per the report. This means that in some instances, if an officer or investigator did not record information, it does not mean the information does not exist, just that it was not explicitly mentioned in the case file and could not be captured quantitatively. In addition to interpreting these results with caution, future endeavors should examine case file data from police agencies located in smaller or rural jurisdictions with homogeneous populations to consider legal and extralegal factors that influence arrest, while considering previous stalking behaviors in the context of IPV.
The present study contributed to the existing literature and examined the effect of previous partner-perpetrated stalking on police arrest decisions among reported IPV incidents, while accounting for extralegal and legal factors. Findings underscore the important contribution that legally relevant case characteristics and corroborating evidence have had on police decision making within IPV cases. In addition, while partner-perpetrated stalking behaviors have been documented as a measure of IPV severity, results of the current study suggest they were not a significant correlate of arrest decisions in this police agency. Educational programming and training modules for law enforcement that address the dynamics of stalking within the broader context of IPV have the potential to increase stalking identification, enhance victim participation, facilitate suspect apprehension, and increase public safety. Future endeavors should continue to examine the effect of IPV stalking on justice outcomes from multiple data sources to better understand formal decision making among criminal justice actors.
Footnotes
Authors’ Note:
The authors would like to thank the police partner agency for their cooperation in providing the data and the supervisory staff at the Special Victims’ Division for their assistance, support, and instructive feedback with this project; Dr. Alex Updegrove, Jessica Fleming, and Matthew Bills for their assistance with data entry; and Drs. Patrick Brady and Brittany Hayes for their helpful comments on earlier drafts. The authors have no conflicts of interest. This project was supported by Grant No. 2016-SI-AX-0005 awarded by the Office on Violence Against Women, U.S. Department of Justice. The opinions, findings, conclusions, and recommendations expressed in this presentation are those of the authors and do not necessarily reflect the views of the Department of Justice, Office on Violence Against Women.
