Hiding in plain sight: identifying partner stalking in intimate partner violence episodes reported to New Zealand Police

ABSTRACT

In relationships characterised by current or previous intimate partner violence (IPV), partner stalking is a commonly occurring phenomenon. In this study, we examined police-recorded partner stalking in IPV episode reports across 1150 cases to (a) consider the overlap between IPV and partner stalking, and the relevance of the aggressor and victim’s relationship phase to defining partner stalking; and (b) contribute empirical evidence about partner stalking prevalence rates and identification rates by police and victims. A secondary aim involved exploring possible differences between IPV cases with and without partner stalking. Although few police or victims explicitly used the label ‘stalking', we identified evidence of partner stalking within one in every seven IPV cases reported to New Zealand Police. Further, this study contributed novel analysis about relationship dynamics by adding an on–off relationship phase to the typically dichotomised categories of intact and separated relationships. In turn, we identified partner stalking relatively rarely within intact relationships; and significantly more frequently, and at similar rates, across both the separated and on–off relationship phases. Finally, we discuss implications for defining the overlap between IPV and partner stalking, police practice, and future research.

KEYWORDS:

• Domestic violence
• partner abuse
• stalking acknowledgement
• harassment
• unwanted pursuit behaviours

Partner stalking is an all-too-common experience (Breiding et al., 2014; Thorburn & Jury, 2019). When combined with previous or current intimate partner violence (IPV) between the aggressor and victim, the effects may be more harmful and dangerous overall than victimisation only by IPV or partner stalking (Ferreira & Matos, 2013; Logan et al., 2007; Logan & Walker, 2010; Spencer & Stith, 2020). But the conceptual overlap between IPV and partner stalking remains unclear (Douglas & Dutton, 2001; Flowers et al., 2020; Logan & Walker, 2009; McEwan et al., 2017) and further research is needed on IPV and partner stalking within a police practice context (Backes et al., 2020; Dreke et al., 2020; Thorburn & Jury, 2019). Therefore, the current study examines partner stalking between aggressors and victims with at least one IPV episode reported to police, to (a) consider the overlap between partner stalking and IPV, and the relevance of the aggressor’s and victim’s relationship phase to defining partner stalking; and (b) contribute empirical evidence about partner stalking prevalence rates and identification rates by police and victims. A secondary aim involves exploring possible differences between IPV cases with and without partner stalking.

Defining stalking

Stalking – whether of strangers, acquaintances, or intimates – refers to an aggressor’s intentional pattern of repeated, unwanted stalking behaviour¹ towards a victim over time, that may cause the victim fear (Spitzberg & Cupach, 2007). In other words, this definition contains three key components. The repeated component refers to a recurring pattern of behaviours (Spitzberg & Cupach, 2007). However, the operationalisation of this component varies, with most studies either stating the behaviour was repeated without specifying further details or requiring a minimum of two incidents involving stalking behaviour (Fox et al., 2010). Few studies apply strict requirements (Fox et al., 2010), such as a minimum of five incidents (e.g. McEwan et al., 2020) or a minimum duration of two weeks (e.g. Purcell et al., 2004). The unwanted component refers to these behaviours being experienced as unwelcome by the victim (Spitzberg & Cupach, 2007). Finally, the victim fear component, henceforth labelled the fear standard (Dietz & Martin, 2007), requires that the victim explicitly expresses fearfulness or that a reasonable person in the same situation would feel fearful (Spitzberg & Cupach, 2007). But the measurement of these three components – and, correspondingly, of stalking overall – remains inconsistent across individual studies (see Fox et al., 2010).

In particular, several concerns persist about the fear standard. Evidence suggests that requiring the fear standard underestimates stalking prevalence rates (e.g. Owens, 2016). Factors such as the victim’s gender, ethnicity, the victim-aggressor relationship, the specific victimisation experiences, and the timing of fear assessment may influence fear expression or relate to under-detection by third parties (Dietz & Martin, 2007; Owens, 2017; Reyns & Englebrecht, 2013; Sheridan et al., 2014). And many emotions other than fear, such as anxiety, helplessness or anger, can result from stalking victimisation (e.g. Englebrecht & Reyns, 2011). So, arguably, requiring victims to demonstrate only one type of emotion (i.e. fear) is overly prescriptive and restrictive (Dietz & Martin, 2007; Fissel et al., 2022). Finally, fear-related information may not have been systematically recorded; making research using archival records unable to meet this standard.

Defining partner stalking and the conceptual overlap with IPV

Additional complexities are specific to partner stalking. Partner stalking occurs between people who (at some point) have been in an intimate relationship. However, the literature appears divided on (a) the relevance of the aggressor’s and victim’s relationship phase (i.e. an intact versus separated relationship) at the time of stalking for this definition and (b) the nature of the overlap between IPV and partner stalking. We consider two main approaches to conceptualising, defining, and measuring the overlap between IPV and partner stalking. In some studies, IPV and partner stalking are conceptualised as related – but distinct – phenomena (e.g. Ferreira & Matos, 2013; McEwan et al., 2017; Senkans et al., 2021). This approach appears grounded in the general stalking literature. In this way of thinking, partner stalking occurs between people who are not in a current relationship. So, although these researchers note that IPV during the relationship and partner stalking post-relationship may be characterised by similar behaviours (Ferreira & Matos, 2013; McEwan et al., 2017), whether the couple is currently ‘together’ distinguishes IPV from stalking. Here, researchers define a pattern of abusive behaviours during the relationship as IPV and following the relationship as stalking (Senkans et al., 2021). The relationship phase between the aggressor and victim is thus a highly important criterion for defining partner stalking in this approach.

In other research, partner stalking has been defined as a subtype of IPV (e.g. Basile & Hall, 2011; Breiding et al., 2014; Cunha et al., 2022; Logan & Walker, 2010; Mechanic et al., 2000). This approach is grounded in the IPV literature and views partner stalking as just another component of abuse, alongside physical, sexual, and psychological violence (Breiding et al., 2014). Here, distinctions between partner stalking and other forms of IPV are based on the pattern of behaviours, regardless of whether they occur while the couple’s relationship is ongoing, or post-separation. Behaviours labelled as partner stalking primarily focus on surveillance and unwanted contact (e.g. Truman & Morgan, 2021), whereas behaviours labelled as other forms of IPV focus on physical violence; sexual violence; or tactics relating to intimidation, minimisation, victim-blaming, isolation, and financial abuse under the umbrella of psychological violence (e.g. Lehman et al., 2012). Thus, the relationship phase between the aggressor and victim is not part of the criteria for defining partner stalking in this approach.

But regardless of the definitional perspective used, another issue relates to defining exactly what phase the victim and aggressor’s relationship is in when partner stalking occurs. Relationship instability (i.e. on–off relationships, Dailey et al., 2009; relationship churning, Halpern-Meekin et al., 2013; Halpern-Meekin & Turney, 2021; or relationship cycling, Monk et al., 2018) represents a dynamic relationship phase. Henceforth labelled on–off relationships, this term refers to any couple who breaks up and reconciles at least once (Dailey et al., 2009). Despite the high prevalence of on–off relationships in the general population (Dailey, 2019) and the high average number of on–off phases experienced by IPV aggressors and victims (Logan et al., 2008), this phenomenon has received relatively little attention across the partner stalking, IPV and wider relationship-related research literature. Instead, most researchers adopt a binary view and dichotomise relationship phase categories as intact or separated only (Dailey et al., 2009) – in turn, making an on–off relationship invisible for many couples. For partner stalking research, this dichotomisation represents a potential source of inaccuracy; comparisons of partner stalking rates across different relationship phases may be contaminated by researchers labelling some couples’ current relationship phase simply as intact or separated when, in fact, they are cycling in and out of the relationship.

Partner stalking in cases of IPV reported to police

Notwithstanding these issues in defining and measuring partner stalking, we now synthesise the most relevant empirical evidence relating to (a) the prevalence and identification rates of partner stalking in cases of IPV and (b) any differences between IPV cases with and without partner stalking. Here, IPV cases refer to relationships with physical, psychological or sexual violence, irrespective of the aggressor’s and victim’s relationship phase classification.

Prevalence rates

One review estimated that 50% of IPV victims experienced partner stalking by the same aggressor, based on data self-reported from victims or aggressors in forensic, community or student samples (McEwan et al., 2017). However, using police data and the methods outlined in Table 1, partner stalking was identified in 16.5% (Tjaden & Thoennes, 2000), 19.3% (Garza et al., 2020) and 27.9% (Melton, 2012) of IPV cases reported to police. Given the methodological differences, the latter prevalence rates are understandably lower than in McEwan and colleagues’ (2017) review; but these figures suggest police episode reports for IPV still commonly include statements that indicate the presence of partner stalking.

Table 1. Overview of key studies’ methods.

	Garza et al. (2020)	Melton (2012)	Tjaden and Thoennes (2000)
Data source and sample	Police IPV case files; N = 332 from the United States	Police IPV episode reports: one episode examined per case; N = 1440 from the United States	Police IPV episode reports: one episode examined per case; N = 1785 from the United States
Stalking behavioural items	Nine items from Fisher et al. (2002): Telephone; Waited outside or inside places; Watched from afar; Followed; Sent letters; Emailed; Sent gifts; Showed up uninvited; ‘Other’ stalking behaviour	Coleman’s (1997) Stalking Behavior Checklist: Broke into home, car; Stole, read mail; Came unwanted; Threatened, harmed new partner; Unwanted calls; Unwanted messages; Unwanted email; Unwanted gifts; Checked up on; Followed, watched; ‘Other’ stalking behaviour	The ‘antistalking code for states developed by the National Criminal Justice Association for the National Institute of Justice’ (p. 430): Following; Face-to-face confrontations; Unwanted communications by phone, page, letter, fax, e-mail
Repeated component of stalking	Two or more different stalking behaviours OR one stalking behaviour must occur on two or more occasions	Not included	Same or different stalking behaviours, must occur on two or more occasions
Unwanted component of stalking	Required evidence of ‘unwanted course of pursuit’ (p. 1020) but no further details about operationalisation	Relied on wording for some behavioural items, e.g. unwanted telephone calls	Relied on wording for some behavioural items, e.g. unwanted communications
Fear standard for stalking	Not included	Not included	All cases coded with victim fear: ‘it was assumed that persons who were identified as victims in a domestic violence crime report had experienced fear as a result of behaviors perpetrated against them by the suspect’ (p. 428)

Police- and victim-identified partner stalking

Of note, these figures hide the observation that – despite including statements that indicate the presence of partner stalking – police rarely identify and explicitly label a pattern of partner stalking. Tjaden and Thoennes (2000) found police-identified partner stalking in one in every 14 IPV cases that researchers later determined met the United States National Institute of Justice’s definition for stalking. Other studies also demonstrated low rates; police identified partner stalking in 6.8% (Klein et al., 2009) and 40.9% (Brady et al., 2020) of researcher-identified partner stalking cases. Overall, these findings suggest that police may ‘grossly under-identify stalking cases from among domestic violence cases' (Klein et al., 2009, p. 40).

Similar patterns are evident for victims’ self-identification of partner stalking. Tjaden and Thoennes (2000) demonstrated that, despite articulating experiences that constituted partner stalking, victims (were recorded by police as having) labelled their experiences as stalking in only one in every 20 researcher-identified partner stalking cases. Klein and colleagues (2009) found similarly low rates. The general literature on (all types of) stalking shows this same common disconnect between victims describing experiences that meet the criteria for stalking but not self-identifying as a stalking victim (e.g. Englebrecht & Reyns, 2011; McNamara & Marsil, 2012; Ménard & Cox, 2016).

Differences between IPV cases with and without partner stalking

Several other differences are evident between the recorded characteristics of IPV cases with and without partner stalking. Threats of harm (Garza et al., 2020; Melton, 2012), a separated relationship status (Melton, 2012; Tjaden & Thoennes, 2000), and the victim self-reporting the episode to police (Tjaden & Thoennes, 2000) were more likely in IPV cases with partner stalking than without. Further, the availability of evidence (e.g. aggressor at scene, electronic or physical evidence; Garza et al., 2020; Melton, 2012), substance use (Garza et al., 2020; Melton, 2012; Tjaden & Thoennes, 2000), co-occurring physical violence (Melton, 2012; Tjaden & Thoennes, 2000), victim injury (Garza et al., 2020; Tjaden & Thoennes, 2000), the presence of a witness during the episode (Garza et al., 2020), the victim and aggressor sharing children (Tjaden & Thoennes, 2000) and victim fear (Melton, 2012) were less likely in IPV cases with partner stalking than without.

Policing partner stalking and partner stalking legislation

Research using police data must be understood in the context of the policies and legislation of the jurisdictions in which the research is conducted. For example, in some jurisdictions, all types of stalking (i.e. partner, stranger, acquaintance) are a crime – enabling arrest, prosecution and sentencing of IPV aggressors that stalk their victims (e.g. Bouffard et al., 2021; Brady et al., 2020; Brady & Nobles, 2017; Garza et al., 2020). In other jurisdictions, partner stalking may not be criminalised; and instead covered by civil orders, such as in New Zealand (Thorburn & Jury, 2019). Partner stalking – as one type of psychologically abusive behaviour – can then form part of the grounds for granting a civil Protection Order, the breaching of which is a criminal offence ([NZ] Family Violence Act, 2018).

It may be assumed that jurisdictions that have criminalised partner stalking provide better sites for this research. However, evidence suggests that increasingly comprehensive stalking legislation has not yet translated into consistent recording of, or responses to, stalking in criminal justice settings (Bouffard et al., 2021; Brady & Nobles, 2017). Experts argue that ‘legislation alone does little to guide frontline officers and investigators in the proper handling of stalking investigations' (Brady & Nobles, 2017, p. 3154) so local policies may be more important than the legislative context. In New Zealand, addressing IPV is one of New Zealand Police’s (NZP) most important strategic aims, alongside a shift away from prioritising prosecution to providing more holistic support (New Zealand Police, 2022a, 2022b). Much of what police respond to, and correspondingly record, falls outside of criminalised behaviour; NZP do not record criminal offences in the majority of reported IPV episodes (Jolliffe Simpson et al., 2021; New Zealand Police, 2022b). Regardless of whether offending is detected or charged, NZP still record extensive information about IPV episodes and respond in varied ways to prevent ongoing harm (New Zealand Police, 2022b). With this practice context in mind, this study will explore information recorded about partner stalking in IPV episodes reported to NZP.

The current study

The current study aims to build on previous research in two main ways. First, inconsistencies persist over how to conceptualise the overlap between IPV and partner stalking, and the relevance of the aggressor’s and victim’s relationship phase to defining partner stalking. As such, we adopt an intentionally broad definition of partner stalking that covers stalking behaviours occurring in any relationship phase and enables us to examine prevalence rates across different relationship phase categories. Additionally, research largely overlooks the dynamics of on–off relationships when considering how to define and measure partner stalking. So, as per recommendations to comprehensively describe the nature of the aggressor-victim relationship (McEwan et al., 2020), we seek to explore partner stalking across three categories that represent different relationship phases: namely, intact, on–off, and separated. Notwithstanding the limitation that involved parties, practitioners and researchers alike cannot know for certain whether a currently intact or separated couple may experience an on–off dynamic in the future (Dailey et al., 2009), we propose it is worthwhile attempting to account for this additional relationship phase. Such efforts may improve the accuracy of partner stalking prevalence estimates across different relationship phases, using whatever sufficiently detailed data is available to researchers. Second, only a small body of empirical research has sampled IPV cases reported to police to provide information about partner stalking prevalence rates and identification practices by police and victims, or differences between cases with and without partner stalking. Accordingly, this study will add to this modest evidence base. We posed the following main research questions (1–3) and a secondary research question (4):

1. How frequently is partner stalking² evident in cases of IPV reported to NZP?

2. Focusing on the relationship phase between aggressors and victims, how frequently is evidence of partner stalking recorded across the intact, on–off, and separated relationship categories?

3. How frequently do police and victims explicitly identify and label partner stalking?

4. Are there any differences between IPV cases with and without partner stalking, across other recorded variables?

Method

Data source

This study used a cross-sectional design based on archival data from NZP. Drawing from the national police database, a NZP employee compiled the initial dataset. This dataset contained information about 2259 unique aggressor and victim pairings with at least one IPV episode reported to NZP in the Waikato and Christchurch regions between 1 November and 31 December 2018 (although we do not assume that this episode was the first ever reported to NZP for each case). These two regions include urban and rural communities with varied socio-demographic profiles (Mossman et al., 2017).

The first author cleaned and organised the data. We removed a total of 67 cases where: the aggressor’s or victim’s details were unknown (n = 12), police misclassified other types of family violence as IPV (n = 35), aggressors or victims were aged under 16 years (n = 9), or episodes were ‘locked for security reasons' (n = 11). From the remaining 2192 cases, we used SPSS (version 28) to randomly select a final sample of 1150 cases. Next, we examined any subsequent IPV episodes reported to NZP during the six-month study period for each case (i.e. ending between 1 May and 30 June 2019, depending on the first sampled episode date). We only retained episodes where police identified the aggressor and victim in the same roles as in the first sampled episode. Accordingly, we examined a total of 2168 individual episode reports across the 1150 cases. Most cases had one reported episode recorded (59.9%) during the study period, but this count ranged from 1 to 23 per case (M = 1.9 episodes).

Measures

NZP officers completed all episode reports used in this study. Our first data type, police-coded variables, was taken from answers coded by police, based on prompts with a series of pre-determined options. These variables included the aggressor’s and victim’s age, gender, and ethnicity; as well as episode location types (private dwelling or public place), the overall risk level assigned by police at the first sampled episode, and lastly, whether police dichotomously recorded the presence of verbal harm, physical harm, sexual harm, threats of harm, property damage, alcohol use, or drug use during the first sampled episode (i.e. present = 1, absent = 0).

For the second data type, researcher-coded variables, the first author coded variables from the free-text section of all sampled episode reports across the six-month study period. This section detailed information about the aggressor, victim, their relationship and family situation, witnesses, evidence, and what happened during the episode. We describe these variables below, which were mostly dichotomously recorded (i.e. present = 1, absent = 0).

Partner stalking

To code partner stalking as present, we required evidence of the following:

1. At least one behavioural item from the United States National Crime Victimization Survey’s Supplemental Victimization Survey on stalking; henceforth named NCVS/SVS stalking behaviours (Truman & Morgan, 2021). These NCVS/SVS stalking behaviours include Followed: the aggressor followed the victim around and watched them; Sneaked (we named Forced/covert entry): the aggressor sneaked into the victim’s home, car or other location and did unwanted things to indicate their presence; Waited: the aggressor waited at the victim’s home, work, school or other location where they were unwelcome; Showed up: the aggressor showed up, rode or drove by places where the victim was when they had no business being there; Items (we named Inappropriate gifts): the aggressor left or sent cards, letters, presents, flowers or any other unwanted items; Third parties: the aggressor harassed or repeatedly asked the victim’s friends or family for information about the victim; Telephone (we named Unwanted phone contact): the aggressor made unwanted phone calls to the victim, left voice messages, sent text messages or used the phone excessively to contact them; Spyware: the aggressor spied on the victim or monitored their activities using technology (e.g. listening device, camera, computer/phone monitoring software); Location tracking: the aggressor tracked the victim’s whereabouts with an electronic device or app (e.g. GPS); Reputation damage: the aggressor posted or threatened to post inappropriate information about the victim online (e.g. photographs, videos, rumours); Other messages (we named Unwanted cyber contact): the aggressor sent the victim unwanted online messages (e.g. via email, Instagram, Twitter, Facebook); or Social media: the aggressor monitored the victim’s activities using social media apps.

2. Representing the repeated component, either (a) two types of NCVS/SVS stalking behaviours over one or more episodes (e.g. ‘Followed’ and ‘Unwanted gifts’) or (b) the same type of NCVS/SVS stalking behaviour across either two or more episodes or discrete dates within a single episode (e.g. ‘Waited’ today and two weeks ago).

3. Representing the unwanted component, at least one of the following co-occurred with an NCVS/SVS stalking behaviour: (a) the aggressor and victim were broken up; (b) an active court order (e.g. Protection Order, Trespass Order) prevented contact between the aggressor and victim, irrespective of the current relationship phase; or (c) the victim reported the episode to police themselves, irrespective of the current relationship phase.

Partner stalking with fear standard

To provide supplementary insights about operationalising partner stalking, we separately examined a modified version of the partner stalking measure above. In addition to the three points from our main partner stalking measure, we additionally required explicit evidence of recorded victim fear (see Other Variables below).

Relationship phase

To categorise the relationship phase for each case, we used a combination of police-coded relationship status (‘Violence Between Partners’ or ‘Violence Between Ex Partners’) and researcher-coded variables (On–Off Relationship; whereby police described in-text that the aggressor and victim had an on-again/off-again relationship or had broken up and resumed the relationship at least once) during the study period . The term intact refers to cases where police consistently labelled sampled episodes as involving Partners, with no researcher-coded evidence of an on–off relationship or separation; separated refers to cases where police consistently labelled sampled episodes as involving Ex-partners, with no researcher-coded evidence of an on–off or intact relationship; and on–off refers to cases where police labelled sampled episodes as involving Partners and Ex-Partners at different times across the study period or there was researcher-coded evidence of an on–off relationship.

Simplified relationship phase

To provide supplementary insights about operationalising relationship phase, we separately examined a simplified version of the main relationship phase measure above. Instead, here, we only used the police-coded relationship status variable (‘Violence Between Partners’ or ‘Violence Between Ex Partners’) from the first sampled episode. In other words, any further relationship phase-related information from across the study period was ignored; this measure represented a static, single point in time. The term simplified intact refers to cases where police labelled the first sampled episode as involving Partners. The term simplified separated refers to cases where police labelled the first sampled episode as involving Ex-Partners.

Police- and victim-identified partner stalking

We separately coded whether police and victims, respectively, used the words ‘stalker’, ‘stalked’, ‘harassed’, ‘stalking’ or ‘harassment’ to describe the aggressor or their behaviour.

Other recorded variables

We coded the presence of several other variables, such as victim fear, victim injuries, weapon use, strangulation, and whether the victim and aggressor had shared children.

Data preparation

Several steps ensured the inter-rater reliability of researcher-coded variables. The first author trained a second coder using the coding protocol described below and the data dictionary available in the Supplemental Online Material. We then independently coded 10% (n = 115) of the full sample to test inter-rater reliability. This coding protocol involved two phases. In phase one, we coded all researcher-coded variables apart from overall partner stalking (but including individual NCVS/SVS stalking behaviours). Based on Koo and Li’s (2016) criteria, we removed three types of NCVS/SVS stalking behaviours (Third parties, Spyware, Social media) with poor reliability (i.e. intraclass correlation coefficients [ICC] < .50). We resolved any disagreements through discussion. In phase two, we focused exclusively on cases with at least one NCVS/SVS stalking behaviour coded during phase one (n = 39) to independently code overall partner stalking. Based on the inter-rater reliability results for retained variables across both phases³, the first author independently coded all remaining cases (n = 1035).

Sample

Most sampled aggressors were men (n = 957, 83.2%) and most victims were women (n = 966, 84.0%). Aggressors and victims were of similar ages (M_agg = 35.3 years, SD = 11.7 years; M_vic = 33.5 years, SD = 11.3 years; range 16 to 77 years). Aggressors and victims were commonly identified by police as European (n_agg = 539, 46.9%; n_vic = 544, 47.3%) or New Zealand Māori (n_agg = 482, 41.9%; n_vic = 452, 39.3%). Less commonly, aggressors and victims were identified as African (n_agg = 1, 0.1%; n_vic = 3, 0.2%), Asian (n_agg = 31, 2.7%; n_vic = 25, 2.2%), Latin American (n_agg = 2, 0.2%; n_vic = 0, 0.0%), Middle Eastern (n_agg = 2, 0.2%; n_vic = 1, 0.1%), Pasifika (n_agg = 40, 3.5%; n_vic = 34, 3.0%) or labelled ‘Other Ethnicity’ by NZP (n_agg = 53, 4.6%; n_vic = 91, 7.9%). Cases predominantly involved aggressors and victims with different genders (n = 1128, 98.2%) and the same ethnicity (e.g. when aggressors were identified as European, victims were also identified as European in 75.5% of cases; when aggressors were identified as Māori, victims were identified as Māori in 71.8% of cases).

Plan for analyses

Police-coded sexual harm was rarely present (n = 5, 0.4%), so we removed this variable prior to analysis. At the other end of the spectrum, police-coded verbal harm was ubiquitous (n = 1113, 96.8%) so we removed this variable prior to analysis. Instead, we created and analysed a related variable called ‘Verbal harm only’, which captured when verbal harm, but no other police-coded harm type, was identified.

We conducted all analyses in IBM SPSS Statistics version 28. We used descriptive statistics to summarise the prevalence of partner stalking overall and across different relationship phases, the rates of police and victim-identified partner stalking, and the presence of other recorded variables. We also used descriptive statistics to provide insights about the modified versions of partner stalking and relationship phase, respectively, but we did not use these supplementary variables in any of the statistical analyses. Next, we examined the differences between IPV cases with and without partner stalking using phi coefficients, Chi-square tests, and Independent Samples t-tests. Finally, we conducted a Binary Logistic Regression to predict (researcher-identified) partner stalking using other recorded variables. Because the IPV reported episode count variable violated the linearity assumption for the binary logistic regression (Tabachnick & Fidell, 2013), we transformed it from a continuous variable into four categorical ‘bins’ (i.e. 1, 2, 3–4, ≥5 episodes).

Results

Prevalence rates

Overall, we identified partner stalking in 13.8% (n = 159) of all sampled cases. In other words, approximately one in every seven IPV cases reported to NZP involved partner stalking. The most common NCVS/SVS stalking behaviour type, evident in almost one fifth of cases (n = 217, 18.9%), involved the aggressor showing up to the victim’s house, work or another location without a legitimate reason and often forcing some sort of interaction with the victim. The next most frequently recorded type of NCVS/SVS stalking behaviour involved the aggressor making unwanted contact with the victim via calls or texts (n = 122, 10.6%). Other types of NCVS/SVS stalking behaviours were rarely recorded (Followed n = 38, 3.3%; Forced/covert entry n = 58, 5.0%; Waited n = 9, 0.8%; Inappropriate gifts n = 10, 0.9%; Location tracking n = 5, 0.4%; Reputation damage n = 28, 2.4%; Unwanted cyber contact n = 26, 2.3%). Additionally, the partner stalking prevalence rate with the fear standard dropped to 6.7% (n = 77) of all sampled cases; approximately one in every 15 IPV cases reported to NZP involved partner stalking with the fear standard.

Relationship phase

As per Table 2, intact relationships were the most common category overall, but we rarely identified partner stalking during this relationship phase. Compared to IPV cases without partner stalking, IPV cases with partner stalking were significantly less likely to involve an intact relationship between the aggressor and victim. On the other hand, we frequently identified partner stalking across both the on–off and separated relationship phases; these two relationship categories accounted for over 90% of all partner stalking cases. Compared to IPV cases without partner stalking, IPV cases with partner stalking were significantly more likely to involve an on–off or separated relationship between the aggressor and victim.

Table 2. Differences between IPV cases with and without partner stalking.

Variable	Overall	Cases without stalking	Cases with stalking	Stalking-related differences
	n (%)	n (%)	n (%)	χ^2 (df)	φ	p
Relationship phase				142.390 (2)	.352	<.001
Intact	601 (52.3)	587 (59.2)	14 (8.8)
Separated	225 (19.6)	159 (16.0)	66 (41.5)
On–off	324 (28.2)	245 (24.7)	79 (49.7)
Shared children	536 (46.6)	449 (45.3)	87 (54.7)	4.875 (1)	.065	.017
Alcohol use	440 (38.3)	387 (39.1)	53 (33.3)	1.897 (1)	-.041	.168
Drug use	175 (15.2)	125 (12.6)	50 (31.4)	37.667 (1)	.181	<.001
Verbal harm only	414 (36.0)	368 (37.1)	46 (28.9)	4.002 (1)	-.059	.45
Physical harm	486 (42.3)	428 (43.2)	58 (36.5)	2.529 (1)	-.047	.112
Property damage	239 (20.8)	188 (19.0)	51 (32.1)	14.293 (1)	.111	<.001
Threats of harm	245 (21.3)	177 (17.9)	68 (42.8)	50.697 (1)	.210	<.001
Victim injury	203 (17.7)	178 (18.0)	25 (15.7)	0.472 (1)	-.020	.492
Strangulation	99 (8.6)	87 (8.8)	12 (7.5)	0.264 (1)	-.015	.607
Victim fear	246 (21.4)	169 (17.1)	77 (48.3)	80.207 (1)	.264	<.001
Weapon use	144 (12.5)	109 (11.0)	35 (22.0)	15.173 (1)	.115	<.001
Episode location/s				9.543 (1)	.091	.002
Dwelling only	903 (78.5)	793 (80.0)	110 (69.2)
Public place (at least one episode)	247 (21.5)	198 (20.0)	49 (30.8)
Risk level assigned at index episode^a				37.624 (2)	.182	<.001
High	673 (58.5)	544 (55.6)	129 (81.1)
Medium	212 (18.4)	195 (19.9)	17 (10.7)
Low	252 (21.9)	239 (24.4)	13 (8.2)

^a 13 cases with missing data on this variable.

Additionally, we examined partner stalking prevalence across relationship categories as per the simplified relationship phase measure. Using the dichotomous police-coded relationship status from the first sampled episode only, 33 previously on–off cases would have been reclassified as (simplified) intact and 42 previously on–off cases would have been reclassified as (simplified) separated. Overall, this simplified relationship measure would have classified 29.6% (n = 47) of all partner stalking cases as intact, 67.9% (n = 108) as separated, and 2.5% (n = 4) as missing relationship phase data.

Police- and victim-identified partner stalking

Despite commonly including statements that indicated the presence of partner stalking (and met our definitional criteria), few police or victims explicitly identified that stalking was occurring. Police identified stalking in only 8.2% (n = 13) of the cases that met our main criteria for partner stalking (i.e. not the Partner Stalking with Fear Standard measure). Victims were recorded by police as having identified stalking in 9.4% (n = 15) of partner stalking cases.

Differences between IPV cases with and without partner stalking

Table 2 presents the overall rates of other recorded variables. Comparison of these variables across cases with and without partner stalking revealed several differences; significant effect sizes ranged from 0.07 to 0.35, reflecting weak to moderate relationships (Cohen, 2013). Threats, property damage, drug use, victim fear (recall that we did not require the fear standard to define partner stalking in this study), weapon use, shared children between the victim and aggressor, a high NZP risk assessment rating at the first sampled episode, and at least one reported episode occurring in a public place were all recorded significantly more often in IPV cases with partner stalking than without. European aggressors were significantly more likely to perpetrate partner stalking than Māori aggressors. No between-group differences were found for aggressor age (t [1,147] = −1.12, p = .264, two-tailed; stalking M = 35.7 years, SD = 10.1; non-stalking M = 34.6, SD = 11.6), but IPV cases with partner stalking had a higher number of IPV episodes reported to NZP (during the study period) than IPV cases without partner stalking (t [1,148] = −7.00, p < .001, two-tailed; stalking M = 3.2, SD = 2.8; non-stalking M = 1.7, SD = 1.3).

As shown in Table 3, we estimated a logistic regression model to determine which variables predicted partner stalking. This model was significant overall, and the Pseudo R² estimate suggested a good fit. Several of the earlier differences remained significant predictors of partner stalking in this model. Victim fear, threats, a higher number of reported IPV episodes, verbal harm only, a high-risk rating at the first sampled episode (compared to a low-risk rating), and a separated or on–off relationship phase (compared to intact relationships) were each uniquely associated with increased odds of partner stalking. Finally, ethnicity was significantly associated with partner stalking: Māori aggressors had decreased odds of partner stalking, compared to European aggressors.

Table 3. Binary logistic regression using other variables to predict partner stalking.

Variable	B	SE	Wald χ^2	p	OR [95% CI]
Aggressor age	0.02	0.01	2.10	.148	1.02 [0.99–1.04]
Woman as aggressor^a	0.30	0.39	0.73	.394	1.36 [0.67–2.72]
Aggressor ethnicity: European			19.87	<.001
Māori	−1.15	0.27	18.42	<.001	0.32 [0.19–0.54]
Pasifika	0.24	0.66	0.13	.716	1.27 [0.35–4.59]
Asian	0.22	0.62	0.13	.722	1.25 [0.37–4.20]
Relationship phase: Intact			64.63	<.001
Separated	3.13	0.39	64.60	<.001	22.97 [10.70–49.32]
On–off	1.82	0.36	25.72	<.001	6.17 [3.05–12.47]
Shared children	0.06	0.24	0.06	.800	1.06 [0.67–1.70]
Alcohol use	−0.10	0.25	0.16	.692	0.91 [0.55–1.49]
Drug use	0.25	0.28	0.81	.367	1.29 [0.74–2.23]
Verbal harm only	0.97	0.39	6.16	.013	2.65 [1.23–5.71]
Physical harm	−0.42	0.34	1.55	.214	0.66 [0.34–1.27]
Property damage	0.54	0.30	3.24	.072	1.71 [0.95–3.06]
Threats	0.79	0.32	6.36	.012	2.21 [1.19–4.10]
Victim injury	0.15	0.37	0.17	.683	1.16 [0.56–2.41]
Strangulation	−1.05	0.50	4.54	.033	0.35 [0.13–0.92]
Victim fear	1.20	0.25	23.46	<.001	3.30 [2.04–5.36]
Weapon use	0.41	0.33	1.51	.219	1.51 [0.78–2.90]
Public episode/s	−0.04	0.26	0.02	.886	0.96 [0.58–1.61]
Episode count: 1 only			55.86	<.001
2 episodes	1.36	0.32	18.48	<.001	3.90 [2.10–7.24]
3–4 episodes	2.48	0.39	40.56	<.001	11.93 [5.56–25.60]
5 or more episodes	3.26	0.48	45.44	<.001	25.93 [10.06–66.83]
Risk level: Low			6.46	.004
High	0.79	0.40	3.86	.049	2.20 [1.00–4.81]
Medium	0.17	0.48	0.13	.716	1.19 [0.47–3.05]
Constant	−6.53	0.77	71.89	<.001	0.00
Model	Pseudo R^2 = .47, χ^2 (24) = 316.27, p < .001

Note. Pseudo R² is Nagelkerke. SE, standard error, OR, odds ratio, CI, confidence interval.

^a Men as reference category.

Discussion

To understand more about the overlap between IPV and partner stalking within a police practice context, we analysed 1150 cases of IPV reported to NZP. Building on three foundational studies (Garza et al., 2020; Melton, 2012; Tjaden & Thoennes, 2000), the current study adds to a small body of empirical evidence using police data to establish partner stalking prevalence rates in reported IPV cases, highlight rates of police – and victim-identified partner stalking, and explore differences between IPV cases with and without partner stalking. To our knowledge, this study is the first to include three categories – representing dynamics for intact, on–off and separated relationships – when considering the relevance of the victim’s and aggressor’s relationship phase in defining partner stalking. We discuss each of these contributions in turn.

Focusing on our first main research question, this study confirmed that partner stalking commonly occurs in IPV cases reported to NZP; we identified partner stalking in one in every seven cases sampled. This rate aligned with findings from prior United States-based studies that used similar methods (Garza et al., 2020; Melton, 2012; Tjaden & Thoennes, 2000). Together, these findings suggest that partner stalking in IPV cases presents at similar rates to police across multiple jurisdictions, irrespective of local stalking legislation.

We briefly note other details relating to partner stalking prevalence in this study. The most frequently recorded types of NCVS/SVS stalking behaviour involved the aggressor showing up to the victim’s house without a legitimate reason or making unwanted phone contact. In comparison, other types of NCVS/SVS stalking behaviours were rarely recorded; but the covert nature of several of these behaviours (e.g. the aggressor tracking the victim via technology) may have reduced victim awareness and third-party detection (Ménard & Cox, 2016). Additionally, the partner stalking prevalence rate lowered when we required evidence of the fear standard, which aligns with fear standard-related findings from the general stalking literature (e.g. Owens, 2016). This finding highlights how different ways of defining and measuring partner stalking – even while using the same data – produce different prevalence estimates.

Addressing our next main research question, we rarely identified partner stalking within intact relationships. Instead, we identified partner stalking significantly more frequently, and at similar rates, across both the separated and on–off relationship phases. This finding is interesting when considering the inconsistencies around conceptualising the overlap between IPV and partner stalking, and the relevance of relationship phases to defining partner stalking. Recall that one approach conceptualised partner stalking as a behavioural subtype of IPV, irrespective of the relationship phase in which stalking occurred (e.g. Basile & Hall, 2011; Breiding et al., 2014; Cunha et al., 2022; Logan & Walker, 2010; Mechanic et al., 2000); whereas the other approach notes that IPV during the relationship and stalking post-relationship may be characterised by similar behaviours, so primarily distinguished the two phenomena using the aggressor’s and victim’s relationship status (e.g. Ferreira & Matos, 2013; McEwan et al., 2017; Senkans et al., 2021). While our definition and measurement of partner stalking aligned with the former approach, the findings suggest that partner stalking may be most directly related to relationship breakdown in the on–off and separated relationship phases. Thus, relationship phase appears to be an important criterion for defining partner stalking and conceptualising the overlap between partner stalking and IPV.

In comparing different ways of measuring relationship phase, our findings also provided empirical evidence about the benefits of examining three relationship phase categories. Without including the on–off category to our analysis, several stalking cases were artificially labelled as intact when, in fact, information about the aggressor’s and victim’s relationship instability or breakdown was available in the data. As such, we demonstrated that dichotomous, static measures of relationship phase can contaminate comparisons of partner stalking prevalence between intact and separated groups. To the extent that such analysis is possible (i.e. depending on the level of relationship phase-related information recorded in any given dataset), this finding highlights the value of accounting for the complexities of relationship dynamics across time, rather than relying on dichotomised relationship information from a single point in time.

Turning to our last main research question, this study demonstrates that few police or victims explicitly identified partner stalking, even though information that met the criteria for partner stalking was disclosed by involved parties and recorded by police. Police-identified stalking in just one in every 12 partner stalking cases. Similarly, victims identified stalking in one in every nine partner stalking cases. Overall, our police and victim identified partner stalking rates aligned with other research (e.g. Klein et al., 2009; Tjaden & Thoennes, 2000). Together, the current and previous findings suggest that even when police capture partner stalking-related information in their records, a disparity exists between police and victims reporting this information and explicitly identifying this phenomenon as stalking.

Across criminal justice system processes, under-identification and mislabelling is a challenge for all types of stalking (Bouffard et al., 2021; Brady & Nobles, 2017). Within the context of IPV, what remains unknown is whether low partner stalking identification rates mean that police (and other practitioners) might be missing stalking behaviours in their appraisals of harm and corresponding risk assessments, or recognising these behaviours but instead labelling them as another form of IPV (e.g. psychological IPV). We were not able to examine these possible explanations directly in our study. But either may be problematic for victim safety, with researchers and practitioners arguing that under-identification and mislabelling of partner stalking could lead to stalking-specific risks and support needs being overlooked (Dreke et al., 2020; Thorburn & Jury, 2019).

Finally, regarding the secondary research question, this study compared IPV cases with and without partner stalking to examine differences across other recorded variables. Few differences were found that were consistent across bivariate and multivariate analyses in this study, and also aligned with findings from the key studies summarised in Table 1. Here, we focus only on the differences that did align. Compared to IPV cases without partner stalking, partner stalking cases were more likely to involve threats (Garza et al., 2020; Melton, 2012) and verbal harm (Melton, 2012), which likely points to an overlap between partner stalking and (other forms of) psychological violence in IPV cases (see Logan & Walker, 2009). Next, and consistent with the relationship status findings and conclusions above, a separated or on–off relationship between the victim and aggressor was more likely (than an intact relationship) in IPV cases with partner stalking than without (Melton, 2012; Tjaden & Thoennes, 2000).

However, because most of the examined variables yielded non-significant or inconsistent relationships in analyses across the current and previous studies (Garza et al., 2020; Melton, 2012; Tjaden & Thoennes, 2000), this analysis only represents a preliminary building block. More robust empirical research is needed to quantify factors that systematically differ between IPV cases with and without partner stalking before clear conclusions can be drawn. Practically, such knowledge could then inform targeted interventions for IPV aggressors that also stalk their victims. Theoretically, the empirical relationships between partner stalking and other factors could be organised into a theoretical framework in the first instance; and a multifactor theory as explanations develop about the causal mechanisms underpinning these empirical relationships (Ward & Hudson, 1998).

Limitations

We remain mindful of general limitations relating to studies that use police data. The dark figure of crime is an ever-present issue; estimates suggest that two thirds of IPV (and non-IPV family violence) is never reported to NZP (New Zealand Crime and Victims Survey, 2022). Further, and consistent with the low number of same-gender couples, male victims and gender diverse individuals sampled in this study, concerns about under-reporting are disproportionately heightened for these groups (Brooks et al., 2021; Edwards et al., 2022). Thus, the findings may not generalise to IPV cases that do not come to the attention of police.

More specifically, even for IPV cases that are reported to police, the level of information disclosed by involved parties and recorded by police varies. Of note, NZP officers responding to IPV are prompted to record information about other harm types (e.g. physical violence, property damage, strangulation), but no specific prompts currently exist for systematically gathering and recording partner stalking-related information. This issue has been raised in other police jurisdictions (e.g. Garza et al., 2020). Therefore, our stalking findings may be conservative relative to other forms of IPV reported to police and are limited by the inherent uncertainty around how victims’ disclosure patterns and police recording practices relate to each other. Further, although increasingly comprehensive stalking legislation in other jurisdictions has not translated into meaningful improvements in policing stalking (Bouffard et al., 2021; Brady & Nobles, 2017), it remains unclear whether this study’s results would differ if New Zealand laws explicitly criminalised partner stalking.

We briefly note several other limitations. First, given the wide variability in defining and measuring partner (and other types of) stalking across the literature, other researchers may challenge the definition and measurement decisions made in this study. However, we have attempted to comprehensively describe both the rationale for them and the decisions themselves, to enable robust comparisons with future studies. Second, we measured the relationship phase categories using data available for the six-month study period. As such, we could not determine how many cases that were categorised as intact or separated (using our main relationship phase measure) would have been more correctly identified as ‘on–off’ if we had access to (a) data captured by police before and after the study period or (b) interview data from the victim or aggressor directly. Our frequency estimates for the on–off relationship category therefore remain conservative. However, even with this caveat, the focus on on–off relationships represents a novel – and, we argue, worthwhile – approach to understanding relationship phases within the context of IPV and partner stalking, and lends credence to arguments that stalking is a function of relationship breakdown and separation (McEwan et al., 2017; Senkans et al., 2021). Third, we note there may be specific dynamics that made sampled victims less likely to disclose, or know about, stalking behaviours that occurred during an intact relationship; but this issue could not be examined using the current study’s data. Finally, based on the number of IPV episodes reported to NZP for each couple during the study period, the level of available data for each case varied considerably. Partner stalking was more likely to be identified in cases with a higher number of reported IPV episodes during the study period than in cases with fewer reported IPV episodes; but we could not determine if partner stalking involved more complex dynamics than (other forms of) IPV alone and so required a higher level of ongoing police contact, or whether this finding was an artefact of having more police data available in which to find stalking-related information.

Practical implications

Keeping in mind that partner stalking in IPV cases reported to police appears relatively common, but few police and victims explicitly identify partner stalking, we briefly outline some practical implications related to improving partner stalking awareness. Focusing on victims, these findings suggest that educational interventions could be valuable. For example, material aimed at improving victim awareness of what constitutes partner stalking could be disseminated via several channels: including specialist IPV support services, criminal justice and social support services, and to the wider community through public health campaigns.

Focusing next on police, these findings could be used to highlight the potential value of partner stalking identification training for frontline staff. Researchers have cited stalking-specific training as an important endeavour for police internationally (Backes et al., 2020; Brady et al., 2020; Brady & Nobles, 2017; Garza et al., 2020; Klein et al., 2009; Melton, 2012; Taylor-Dunn et al., 2021; Thorburn & Jury, 2019; Tjaden & Thoennes, 2000). However, as flagged, this study (and, arguably, other research to date) cannot determine whether the low rates of police-identified partner stalking mean that police overlook partner stalking in their appraisals of harm and assessments of risk, or recognise these patterns of behaviour but instead label them as other forms of IPV (e.g. psychological IPV). As such, we more conservatively suggest a practical implication relating to relationship phases. Training could emphasise the value of systematically capturing on–off relationships when responding to IPV cases with and without partner stalking, and police could consider adding this third relationship phase category to data collection prompts. Finally, and more generally, monitoring of stalking by internal police research teams could be valuable in further developing policy and best practice guidelines for police when responding to IPV cases with partner stalking.

Future research

Ongoing research is necessary to better inform police practices around partner stalking (Backes et al., 2020) and several possible directions could build on the current study’s findings. First, one such direction could involve researchers analysing the relationship between partner stalking and psychological violence in IPV cases reported to police. Second, and to extend insights relating to relationship phases, researchers could explore if rates of police – and victim-identified partner stalking differ based on whether the victim’s and aggressor’s relationship was categorised as intact, on–off or separated when partner stalking occurred. Using this more refined specification of relationship phases in future research may also highlight different levels of risk and case management implications for stalking that occurs during on–off phases versus longer-term, stable separation. A third direction could involve researchers focusing on partner stalking, risk assessments and IPV-related outcomes. Regarding the former, future studies could analyse whether IPV cases with partner stalking (or, separately, cases with police and victim identified partner stalking) predict higher subsequent police-rated risk ratings than cases without partner stalking. Regarding the latter, limited empirical evidence exists about the predictive utility of partner stalking for ongoing IPV (e.g. Jung et al., 2021). Thus, future studies could analyse whether partner stalking (or specific types of NCVS/SVS stalking behaviour) predict various longitudinal IPV outcomes, such as IPV recurrence and the presence or severity of specific harm types in IPV recurrence episodes reported to police.

More broadly, another aim of future research should be theory development. Such research could usefully use qualitative methods with aggressor or victim self-reported data to better describe the occurrence and function of partner stalking across different relationship phases. Even preliminary research of this nature would be valuable in advancing insights about the potential mechanisms underpinning partner stalking during different relationship phases, especially the on–off versus separated phases. Such research could provide a good foundation for addressing theoretical gaps within the partner stalking literature, including why partner stalking occurs in some relationships but not others, and the conceptual and behavioural overlaps between IPV – especially psychological IPV – and partner stalking.

Conclusion

We identified partner stalking within one in every seven cases reported to NZP, yet police and victims rarely explicitly identified and labelled this pattern of behaviour as stalking. Overall, these findings corroborate previous empirical findings about partner stalking in cases of IPV reported to police, across different police jurisdictions and legislative contexts (Garza et al., 2020; Klein et al., 2009; Melton, 2012; Tjaden & Thoennes, 2000). Further, and to our knowledge, this study was the first to include three relationship phase categories in the investigation of partner stalking identification. We added the on–off phase to the typically dichotomised categories of intact and separated relationships; and, consequently, avoided artificially labelling many cases as intact or separated and contaminating partner stalking prevalence estimates between the intact and separated groups. Accordingly, we then rarely identified partner stalking in intact relationships; and compared to IPV cases without partner stalking, IPV cases with partner stalking were significantly more likely to involve victims and aggressors with an on–off or separated relationship dynamic, which lends credence to the idea that relationship breakdown and separation represents an important criterion for defining partner stalking. This study represents a building block towards improving collective understandings of the overlap between IPV and partner stalking, and the relevance of the aggressor and victim’s relationship phase to defining partner stalking.

Acknowledgments

We would like to thank Josephine Ryan, the Integrated Safety Response, and the New Zealand Police for supplying the data for this manuscript. We would also like to thank Samantha Taaka, a PhD student at the University of Waikato, for coding support.

Disclosure statement

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

Data Availability Statement

Data not available due to ethical restrictions.

Additional information

Funding

This work was supported in part by funding from a University of Waikato Doctoral Scholarship, awarded to the first author.

Notes

1 Types of stalking behaviour include the aggressor following the victim, showing up at the victim’s home uninvited, making unwanted phone contact, posting inappropriate information about the victim online, monitoring the victim via tracking devices, and so on (Truman & Morgan, 2021).

2 Our operational definition of partner stalking is similar to previous research (e.g., Garza et al., 2020; Tjaden & Thoennes, 2000) and is detailed comprehensively in the method section.

3 As per Koo and Li’s (2016) criteria, four variables showed moderate reliability (ICC = .50-.74): Reputation damage, Unwanted cyber contact, Victim-identified partner stalking, Unwanted phone contact; seven variables showed good reliability (ICC = .75-.90): Shared children, Victim fear, Weapon use, Followed, Forced/covert entry, Showed up, Partner stalking; and seven variables showed excellent reliability (ICC > .90): Relationship instability, Victim injury, Strangulation, Waited, Inappropriate gifts, Location tracking, Police-identified partner stalking.