Narcissism and Intimate Partner Violence: A Systematic Review and Meta-Analysis

Abstract

This review aimed to examine the relationship between intimate partner violence (IPV) perpetration and trait narcissism, and whether the strength of this relationship differs depending on narcissism type (grandiose or vulnerable), the type of violence perpetrated, or the perpetrator’s gender. Scopus, Medline, PsycInfo, and Academic Search Complete databases were searched on August 11, 2022. Studies were included if they were in English, measured IPV perpetration and trait narcissism, and examined the relationship between trait narcissism and IPV perpetration. Studies were excluded if they were review papers, conference extracts, book chapters, or if the data was not specific to trait narcissism. The AXIS tool was used to assess the quality and risk of bias of the studies. Twenty-two studies (N = 11,520 participants) were included in the random effects meta-analysis revealing a significant, weak, positive relationship between trait narcissism and IPV perpetration, r = .15. Subgroup analyses revealed physical IPV perpetration was not significantly related to trait narcissism while cyber and psychological IPV perpetration were significantly, positively, weakly related to trait narcissism. No significant difference in the strength of the relationship with IPV perpetration was found between males and females. The relationship between trait narcissism and IPV perpetration was significantly greater for vulnerable narcissism than grandiose narcissism. Overall, the quality of the included studies was high, and risk of bias was low. All measures were self-report and underreporting could be present given both narcissistic traits and IPV perpetration are considered socially undesirable. Future research examining these relationships should specify IPV and narcissism types.

Intimate partner violence (IPV; frequently referred to as domestic violence) is a global health crisis characterized by behaviors causing sexual, physical, or psychological harm to a partner (World Health Organization, 2021). Estimates vary across countries but globally 30% of women have reported physical or sexual partner violence (World Health Organization, 2021). While IPV affects all genders, women experience disproportionately higher victimization compared to other genders (World Health Organization & Pan American Health Organization, 2012).

IPV has adverse effects on a victim’s physical and mental health and has been linked to an increased risk of alcohol and drug abuse, eating disorders, sleep disorders, post-traumatic stress disorder, smoking, self-harm, anxiety, depression, and unsafe sexual behavior (Dillon et al., 2013; World Health Organization & Pan American Health Organization, 2012). These consequences often continue after the IPV has ceased (Bonomi et al., 2009). IPV contributes to the burden of disease and declining economic productivity through ongoing health costs, reduced capacity to work, and economic abuse (McLean & Bocinski, 2017). Research has demonstrated that individuals exposed to one incident of IPV are more likely to be exposed again and at an increased severity (Aldarondo, 1996). Therefore, understanding the risk factors for IPV is critical to the task of developing prevention strategies to reduce the negative impact of IPV on individuals, families, and the economy.

Narcissism

Trait theory indicates that personality traits are understood to be stable (Cervone & Pervin, 2015). Therefore, examining personality traits that are predictive of IPV perpetration, is vital to understanding how IPV is perpetrated. One such personality trait that has shown to be predictive of IPV perpetration is narcissism (Green et al., 2020). Narcissism is conceptualized as a dimensional personality construct characterized by extreme self-love, a strong sense of entitlement, and a need for attention, admiration, and approval (Miller et al., 2011). The Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association, 2022) identifies narcissistic personality disorder, which is an extreme presentation of narcissistic traits that adversely impact relationships (Skodol et al., 2014). Collison and Lynam (2021) conducted a meta-analysis examining the relationship between Narcissistic Personality Disorder (NPD) and IPV perpetration and found that a significant weak positive relationship exists. However, they suggested that because personality disorders are heterogeneous and comorbid in many samples, to determine specific personality correlates of IPV perpetrators, individual personality traits should be examined. Additionally, categorical diagnoses of personality disorders have been demonstrated to have limited utility in clinical practice and in recent years there has been a push for a dimensional model of personality (Hopwood et al., 2018). Understanding trait narcissism’s relationship to IPV perpetration is important for developing recommendations that inform policy and clinical guidelines to provide better treatment for victims that is specific and tailored.

Studies have demonstrated that IPV perpetration is positively associated with trait narcissism; however, because of divergent findings, the strength of this relationship is unclear (Gormley, 2003; Turner, 2014). Thus, this systematic review and meta-analysis investigated the relationship between trait narcissism (hereafter referred to as narcissism), and IPV perpetration.

Potential Moderators of the Relationship Between IPV and Narcissism

Studies that evaluate the relationship between IPV perpetration and narcissism employ a number of different methodologies which may account for divergent findings. Below we discuss potential moderators of the relationships between narcissism and IPV perpetration.

Grandiose and Vulnerable Narcissism

There are two main subtypes of trait narcissism: grandiose and vulnerable (Miller et al., 2012). Recent models indicate that antagonism (e.g., arrogance, entitlement) is common to both grandiose and vulnerable narcissism. Differentially, grandiose narcissism is associated with exhibitionism, authoritativeness, excitement-seeking and an exaggerated sense of one’s own importance (Miller et al., 2021; Rohmann et al., 2012). Whereas, vulnerable narcissism is characterized by fragile self-esteem, anger, shame, hypersensitivity to rejection, and elevated self-consciousness (Miller et al., 2021; Rohmann et al., 2012). Given the hypersensitivity to rejection, present in vulnerable narcissism, it is likely that these traits may play out in romantic relationships (Rohmann et al., 2012). In contrast, the exaggeration of one’s importance, present in grandiose narcissism, may not be as closely linked to violence in intimate relationships as this relies less on the validation of others (Rohmann et al., 2012).

To date, the evidence supporting these contentions has been mixed. One study found a significant relationship between vulnerable narcissism and IPV perpetration but not grandiose narcissism (Ménard et al., 2021), while another study found a significant relationship between both types of narcissism and IPV perpetration (Green et al., 2020). Thus, there could be a difference in the strength of the relationship between IPV and the two types of narcissism. This meta-analysis examined whether the strength of the relationship between IPV perpetration differs between grandiose and vulnerable narcissism.

Type of IPV Perpetration

In Collison and Lynman’s (2021) meta-analysis, the authors separated out the type of IPV (sexual, physical, and psychological) and measured their relationships with NPD. They recorded non-significant differences between physical, psychological, and sexual perpetration (Collison & Lynam, 2021). However, given that narcissism has been linked to psychological IPV perpetration in some research (Green et al., 2020) but not physical or sexual IPV perpetration in other studies (Plouffe et al., 2022; Ryan et al., 2008), it could be that narcissism is associated with psychological IPV perpetration rather than physical or sexual IPV perpetration. Hence, this review examined the relationship between narcissism and the different types of IPV perpetration (sexual, physical, psychological, and cyber).

Gender of Perpetrators

Males are far more likely to be perpetrators of IPV than females (World Health Organization & Pan American Health Organization, 2012), however, no review has compared the relationship between IPV perpetration and narcissism for males and females. Some research suggests a stronger relationship between IPV perpetration and narcissism for males (Gormley, 2003) whereas other research found a stronger relationship exists between IPV perpetration and females (Tetreault et al., 2021). Thus, this meta-analysis examined if a difference in gender exists across the studies included in this review.

Objectives

Despite research suggesting a need to focus on dimensional traits to understand the individual facets of personality that contribute to violent behavior, no meta-analysis has examined the relationship between narcissism and IPV perpetration. To address this gap and to synthesize findings across published and grey literature, a systematic literature review of studies was conducted and a meta-analysis was performed to determine the strength of the relationship between IPV perpetration and narcissism. Secondary aims of the review were to determine if the strength of the relationship differs based on narcissism types (grandiose, vulnerable), IPV types (psychological, physical, cyber, and sexual) or gender of the perpetrator (male, female).

Methods

Search Strategy

In accordance with the PRISMA 2020 guidelines for systematic reviews (Page et al., 2021) on August 11, 2022, a comprehensive literature search was completed. PsycInfo, MedLine, Scopus, and Academic Search Complete databases were searched for two concepts: (a) Narcissism (Narcissism OR “dark triad” OR “dark tetrad” OR “dark personality”) and (b) IPV (“intimate partner” OR “relationship quality” OR “sexual violence” OR “sexual abuse” OR “sexual assault” OR “dating violence” OR “dating assault” OR “dating abuse” OR “partner abuse” OR “partner assault” OR “partner violence” OR “interpersonal violence” OR “interpersonal abuse” OR “interpersonal assault” OR “domestic violence” OR “domestic abuse” OR “domestic assault” OR “spousal abuse” OR “spousal assault” OR “spousal violence” OR “marital violence” OR “marital abuse” OR “marital assault” OR “intimate assault” OR “intimate violence” OR “intimate abuse”). See Appendix A for the complete search strategy. Additional records were identified by a manual search of the included papers’ reference lists.

Eligibility Criteria

Studies were included if they; (a) were written in English; (b) contained a measure of trait narcissism; (c) included a measure of perpetration of IPV; and (d) reported a relationship between narcissism and IPV perpetration. As per the World Health Organization definition of IPV, this could be any behaviors that cause psychological (including economic and cyber), physical, or sexual harm, or any combination of these (World Health Organization, 2021). Studies were excluded if they (a) were review papers, conference extracts, or book chapters; or (b) if the data was not specific to narcissism (i.e., papers measuring all dark triad traits without isolating results of narcissism).

The review included data from community, university, and forensic populations. It was understood that this data would likely be heterogeneous across these populations. However, it was reasoned that information about convicted offenders was important to include as they contribute to the perpetration of IPV.

Study Screening

After duplicates were removed using EndNote (version 20.3, Clarivate), titles and abstracts were screened based on the pre-determined inclusion/exclusion criteria using Covidence Systematic Review Software (Innovation, 2018). The full texts of the retained studies were then screened using the inclusion/exclusion criteria. Two reviewers (EO and AC) completed the screening independently and any discrepancies were resolved between the two.

Data Extraction

Data was extracted using Covidence Systematic Review Software and downloaded into a Microsoft Excel (version 16.62) spreadsheet. Information on study characteristics was extracted for the following: (a) publication year; (b) publication country; (c) sample size; (d) age means and standard deviations of sample; (e) gender breakdown of sample; (f) population source (community, forensic, and university); (g) measure(s) of narcissism; (h) type(s) of narcissism; (i) measure(s) of IPV perpetration; and (j) domain(s) of IPV perpetration. Additionally, (k) all correlation coefficients examining IPV and its subgroups, narcissism and its types, and gender-specific samples were extracted for the meta-analysis. One study (Meier, 2003) reported t-tests comparing narcissism between IPV perpetrators and non-perpetrators, these were converted to correlations to use in the meta-analysis using procedures outlined by (Borenstein et al., 2014). One study (Kanemasa et al., 2022) provided longitudinal data and measured the correlation between IPV and narcissism at four time points, the first time point was used in the meta-analysis. Two authors (Plouffe et al., 2020; Smoker & March, 2017) were contacted for additional correlational data between narcissism and IPV perpetration. This data was received for only one of the two studies (Plouffe et al., 2020).

Quality of Studies and Risk of Bias

The Appraisal Tool for Cross-Sectional Studies (AXIS) was used to assess the quality and risk of bias of included studies (Downes et al., 2016). Two independent raters (EO and AC) assessed each study and then resolved discrepancies in their ratings. The AXIS tool is comprised of 20 questions including “Were the methods (including statistical methods) sufficiently described to enable them to be repeated?” and “yes,” “no,” or “don’t know.” Answers are required for each item. These questions assess sample appropriateness and adequacy, ethics, conflicts of interest, response rate reporting, aims, appropriateness of population, design, measures, and discussion, as well as consistency and description of results and significance reporting, description of data and methods, and limitations. Notably, the AXIS tool does not come with a scoring template, so a scoring method employed in previous systematic literature reviews was applied (Casale & Banchi, 2020; Moor & Anderson, 2019) where yes or not applicable (N/A) = 1 point, no or don’t know = 0 points. Including N/A as one point mark was decided to avoid penalizing studies not required to disclose information about missing data where studies had complete datasets. Two items were reverse scored, “Does the response rate raise concerns about non-response bias?” and “Were there any funding sources or conflicts of interest that may affect the author's interpretation of the results?” as a “yes” response would be considered poorer methodology. The quality of each study was given a score out of 20. Using interpretation employed by other researchers (Casale & Banchi, 2020; Moor & Anderson, 2019) low quality scores range from 1 to 7, medium quality from 8 to 14, and high quality scores range from 15 to 20.

Statistical Analysis and Publication Bias

A random effects meta-analysis was completed using Comprehensive Meta-Analysis Version 3 (Borenstein et al., 2014). The effect size of interest was correlation coefficients. Effect sizes were interpreted using guidelines by Schober et al. (2018) where coefficients smaller than ±0.10 are considered negligible, ±0.10 to ±0.39 are weak/small, ±0.4 to ±0.69 are moderate, and values greater than ±0.70 are strong. Q and I² statistics were examined to assess heterogeneity. Q statistics with corresponding p-values of less than .05 indicated heterogeneity was present. I² statistics of 25% indicated low heterogeneity, 50% indicated medium heterogeneity and over 75% indicated high heterogeneity (Higgins et al., 2003).

Results were coded prior to running the overall meta-analysis. For studies that included more than one effect size, the mean effect sizes were calculated to give an overall effect size for that study. Type of narcissism was coded based on the measure used in the studies, as guided by Crowe and colleagues’ (2019) review of trait narcissism measures. We determined this was the most appropriate given inconsistencies in the labels used to refer to types of narcissism, particularly in early literature. When the measure assessed subfactors of narcissism that did not map directly onto grandiose or vulnerable types, or the statistics reported did not discriminate between grandiose and vulnerable subtypes, the narcissism type was coded other narcissism. IPV was coded as multidomain if no violence domain was specified or if the type of violence was multidomain (i.e., did not clearly fit into a specific domain of IPV). Verbal assault and dominance were coded as psychological IPV, and sexual violence was coded as physical as there were insufficient studies reporting this to conduct a subgroup analysis.

Publication bias was assessed for the overall analyses. Funnel plots were examined visually to assess if there was evidence of asymmetry in the graph which might indicate the presence of outliers. Duval and Tweedie (2000) trim and fill analyses were conducted which balance out any asymmetry present in the funnel plots through the addition or removal of any studies and present bias-adjusted results. These bias-adjusted results were then compared to the strength of the examined relations to assess for the effects of publication bias. The significance of the publication bias was declared using Egger’s test with a p-value cut-off of .05 (Egger et al., 1997). Additionally, classic fail-safe N tests were completed to determine how many studies would need to be present for a null effect.

Results

Study Selection

Figure 1 displays the PRISMA flow diagram for the search strategy. The search identified 582 records, 316 remained after duplicates were removed. Two hundred and twenty-five records were excluded after title and abstract screening, the full-texts of 90 articles were reviewed. In total, 21 articles met eligibility criteria and were included in the review. One article (Plouffe et al., 2020) included two studies that met the inclusion criteria, bringing the total number of studies included to 22.

Figure 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram for search strategy of databases and other sources.

Sample Characteristics

Sample characteristics are presented in Table 1. Included in this review, are 22 studies comprising a total of 11,520 participants who were recruited from community (k = 12), university (k = 7), and forensic settings (k = 1), or a combination of these (k = 2). Research was conducted in the United States (k = 6), Australia (k = 5), the United Kingdom (k = 3), Canada (k = 3), Japan (k = 3), and Spain (k = 1). One study was conducted in both the United Kingdom and Sweden. Eighteen of the included studies were published peer-reviewed research and four were doctoral dissertations. All of the studies were conducted between 2003 and 2022.

Table 1.

Sample Characteristics of Included Studies.

Study	Country of Study	Sample Source(s)	Sample Size (N)	Mean Age (SD)	% Male
Bhogal and Wallace (2022)	UK	Community	124	30.68 (5.92)	21.0
Branson and March (2021)	Australia	Community	817	28.16 (10.49)	21.8
Carton and Egan (2017)	UK	Community	128	NR	18.0
Gormley (2003)	USA	University	127	20.00 (NR)	48.0
Green et al. (2020)	UK	Community	328	27.93 (9.09)	46.3
Kanemasa et al. (2022)	Japan	Community	1,392	29.73 (5.92)	43.7
Kiire (2017)	Japan	University	344	19.00 (1.25)	47.1
Kiire (2019)	Japan	University	380	18.90 (1.21)	44.21
March et al. (2020)	Australia	Community	405	24.67 (7.33)	30.4
March et al. (2021)	Australia	Community	415	32.68 (10.62)	49.0
March et al. (2022)	Australia	Community	405	41.19 (11.35)	55.6
Meier (2003)	USA	Combined	118	35.38 (NR)	100.0
Ménard et al. (2021)	USA	University	2,441	19.40 (2.18)	45.2
Pineda et al. (2021)	Spain	Community	1,189	29.36 (10.46)	22.0
Plouffe et al. (2020) - 1	Canada	University	399	18.74 (1.84)	27.3
Plouffe et al. (2020) - 2	Canada	Community	360	34.39 (10.96)	42.5
Plouffe et al. (2022)	Canada	Community	399	33.50 (10.26)	38.3
Rinker (2010)	USA	Forensic	104	33.48 (10.78)	100.0
Ryan et al. (2008)	USA	University	126	NR	50.0
Smoker and March (2017)	Australia	Community	689	26.00 (10.21)	30.0
Tetreault et al. (2021)	UK and Sweden	University	342	26.54 (7.73)	40.4
Turner (2014)	USA	Combined	488	28.03 (10.07)	100.0

Note. NR = not reported.

Study Characteristics

Study characteristics are presented in Table 2. Twelve studies included only a measure of grandiose narcissism (Bhogal & Wallace, 2021; Carton & Egan, 2017; Gormley, 2003; Kiire, 2017, 2019; March et al., 2022; Pineda et al., 2021; Plouffe et al., 2020, 2022; Smoker & March, 2017; Tetreault et al., 2021). Six studies assessed both vulnerable and grandiose narcissism (Branson & March, 2021; Green et al., 2020; March et al., 2020; Meier, 2003; Ménard et al., 2021; Rinker, 2010), one included a measure of vulnerable narcissism only (March et al., 2021). Two studies measured other types of narcissism that did not meet classification for vulnerable or grandiose narcissism (Kanemasa et al., 2022; Turner, 2014) and Ryan et al., (2008) measured both vulnerable narcissism and another form of narcissism.

Table 2.

Characteristics of Included Studies.

Study	Narcissism Measure (s)	Subgroup of Narcissism	IPV Measure(s)	Domain(s) of IPV
Bhogal and Wallace (2022)	SD3	GN	CDAQ	Cyber
Branson and March (2021)	B-PNI	GN	CDAQ	Cyber
		VN
Carton and Egan (2017)	SD3	GN	MMEA	Psychological Multidomain
			CTS-2S
Gormley (2003)	NPI - 40	GN	TDS	Psychological
		MMEA
Green et al. (2020)	PNI	GN	CTS-2	Multidomain Psychological
		VN	MMEA
Kanemasa et al. (2022)	DTDD	Other	IPIPVS (perpetration subscale)	Psychological
Kiire (2017)	SD3(J)	GN	IPV scale (perpetration only)	Physical, Psychological, Multidomain
Kiire (2019)	SD3(J)	GN	IPV Scale (perpetration only)	Physical, Psychological, Multidomain
March et al. (2020)	NPI -16	GN	CBS-R	Psychological Cyber
	HNS	VN	IPCS
March et al. (2021)	HNS	VN	CARS	Cyber
March et al. (2022)	SD4	GN	Cyberstalking assessment	Cyber
Meier (2003)	NPI-40 HNS	GN VN	IPV Conviction Y/N	Multidomain
Menard et al. (2021)	PNI	GN	CTS-2	Multidomain
		VN
Pineda et al. (2021)	SD3	GN	CDAQ	Cyber
Plouffe et al. (2020) - 1	NPI - 40	GN	CTS- 2 (psychological and physical domains only)	Psychological Physical
Plouffe et al. (2020) - 2	NPI - 40	GN	IPV perpetration assessment adapted from Graham et al.	Physical
Plouffe et al. (2022)	NPI - 40	GN	IPV perpetration assessment adapted from Graham et al.	Physical
Rinker (2010)	NPI - 40	GN	CBS	Multidomain Psychological Physical
	HNS	VN	CTS SOPAS
			Severity of violence frequency (two questions)
Ryan et al. (2008)	NPI (E/E items only)	Other VN	CTS-2 (physical and sexual items only)	Physical
	HNS
	HISN
Smoker and March (2017)	SD3	GN	IPCS	Cyber
Tetreault et al. (2021)	SD3	GN	CTS (Verbal, explosive, and displaced aggression only)	Multidomain Psychological Physical
			RCRQ (two items only)
Turner (2014)	NPI - 40	Other	CTS-2	Multidomain

Note. (B)-PNI = (Brief) Pathological Narcissism Index; CARS = Cyber Aggression in Relationships Scale; CBS-(R) = Controlling Behavior Scale-(Revised); CDAQ = Cyber Dating Abuse Questionnaire; CTS-(2)(S) = Conflict Tactics Scale-(Revised) (Shortform); DTDD = Dark Triad Dirty Dozen Scale; GN = grandiose narcissism; HISN = Hurlbert Index of Sexual Narcissism; HNS = Hypersensitive Narcissism Scale; IPCS = Intimate Partner Cyberstalking Scale; IPIPVS = Indirect and Psychological Intimate Partner Violence Scale; MMEA = Multidimensional Measure of Emotional Abuse; NPI = (Brief) Narcissistic Personality Inventory 16 or 40 item; PDQ-R = Personality Disorder Questionnaire-Revised; RCRQ = Richardson Conflict Response Questionnaire; SD3(J) = Short Dark Triad (Japanese version); SD3 = Short Dark Triad; SD4 = Short Dark Tetrad; SOPAS = Subtle-Overt Psychological Abuse Scale; TDS = The Dominance Scale; VN = vulnerable narcissism.

Across the studies, nine different measures were used to assess narcissism. Five studies assessed grandiose narcissism with the Short Dark Triad (SD3; Bhogal & Wallace, 2022; Carton & Egan, 2017; Pineda et al., 2021; Smoker & March, 2017; Tetreault et al., 2021), Kiire (2017, 2019) used a Japanese version of the SD3 and March et al. (2022) used the SD4. Seven studies used the 40-item Narcissism Personality Inventory (NPI-40; Gormley, 2003; Meier, 2003; Plouffe et al., 2020, 2022; Rinker, 2010; Turner, 2014) and one used the 16-item NPI (NPI-16; March et al., 2020), Ryan et al. (2008) used only the Exploitative/Entitlement subscale of the NPI. Five studies measured vulnerable narcissism using the Hypersensitive Narcissism Scale (HNS; March et al., 2021; March et al., 2020; Meier, 2003; Rinker, 2010; Ryan et al., 2008). Two studies administered the Pathological Narcissism Index (PNI; Green et al., 2020; Ménard et al., 2021) and one utilized the Brief PNI (B-PNI; Branson & March, 2021). Both the Hurlbert Index of Sexual Narcissism (HISN), and the Dark Triad Dirty Dozen (DTDD) were used once only by Ryan et al. (2008) and Kanemasa et al. (2022), respectively. Four studies used more than one measure of narcissism (March et al., 2020; Meier, 2003; Rinker, 2010; Ryan et al., 2008).

Nine studies measured multidomain IPV perpetration, (Carton & Egan, 2017; Green et al., 2020; Kiire, 2017, 2019; Meier, 2003; Ménard et al., 2021; Rinker, 2010; Tetreault et al., 2021; Turner, 2014), seven assessed cyber IPV perpetration (Bhogal & Wallace, 2022; Branson & March, 2021; March et al., 2020, 2021, 2022; Pineda et al., 2021; Smoker & March, 2017), 10 measured psychological IPV perpetration (Carton & Egan, 2017; Gormley, 2003; Green et al., 2020; Kanemasa et al., 2022; Kiire, 2017, 2019; March et al., 2020; Plouffe et al., 2020 -1; Rinker, 2010; Tetreault et al., 2021) and eight assessed physical IPV perpetration (Kiire, 2017, 2019; Plouffe et al., 2020, 2022; Rinker, 2010; Ryan et al., 2008; Tetreault et al., 2021). Eight of the studies reported outcomes for more than one type of IPV perpetration (Carton & Egan, 2017; Green et al., 2020; Kiire, 2017, 2019; March et al., 2020; Plouffe et al., 2020 -1; Rinker, 2010; Tetreault et al., 2021).

IPV perpetration was assessed using the Conflict Tactics Scale-Revised (CTS-2) in five studies (Green et al., 2020; Ménard et al., 2021; Plouffe et al., 2020-1; Ryan et al., 2008; Turner, 2014) and two used the original CTS (Rinker, 2010; Tetreault et al., 2021) whereas Carton and Egan (2017) used a short form of the CTS-2. The Cyber Dating Abuse Questionnaire (CDAQ) was utilized in three studies (Bhogal & Wallace, 2022; Branson & March, 2021; Pineda et al., 2021) as was the Multidimensional Measure of Emotional Abuse (MMEA; Carton & Egan, 2017; Gormley, 2003; Green et al., 2020). Intimate Partner Cyber Stalking Scale (IPCS) was used in two studies (March et al., 2020; Smoker & March, 2017), and March et al. (2022) devised their own cyberstalking assessment. The Controlling Behavior Scale (CBS) was used by one study (Rinker, 2010) and another study used the revised version (CBS-R; March et al., 2021). Each of the following scales were used once; Richardson Conflict Response Questionnaire (RCRQ; Tetreault et al., 2021), Subtle-Overt Psychological Abuse Scale (SOPAS; Rinker, 2010), Cyber Aggression in Relationships Scale (CARS; March et al., 2021), The Dominance Scale (TDS; Gormley, 2003), Indirect and Psychological IPV Scale (IPIPVS; Kanemasa et al., 2022). One study assessed physical IPV using two questions which asked convicted perpetrators to indicate how frequently in their past relationship they had perpetrated different types of violence (Rinker, 2010). Two studies used an adapted version of an assessment by Graham et al. (2012) which asked participants to number the frequency of violent experiences that they had perpetrated to a partner (Plouffe et al., 2020-2; 2022). Two studies used the IPV scale (Kiire, 2017, 2019). One study (Meier; 2003) measured IPV categorically where participants were either convicted of IPV perpetration or they were not.

Quality of Studies and Risk of Bias

Assessment of the quality of studies using the AXIS tool is presented in Table 3. Two independent raters assessed the quality of studies with an overall consensus of 93%. Discrepancies in ratings were resolved by agreement leading to a mean total score on the AXIS tool of 18.27. All 22 studies scored between 15 and 20, indicating they are of high quality and low risk of bias (Casale & Banchi, 2020; Moor & Anderson, 2019).

Table 3.

Risk of Bias Assessment Using AXIS Tool for Cross-Sectional Design Risk of Bias.

	Intro		Methods									Results					Discussion		Other		Total Score
	1	2	3	4	5	6	7	8	9	10	11	12	13*	14	15	16	17	18	19*	20
Bhogal and Wallace (2022)	✓	✓	X	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	✓	✓	19
Branson and March (2021)	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	X	✓	✓	✓	✓	✓	✓	19
Carton and Egan (2017)	✓	✓	✓	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	DK	DK	18
Gormley (2003)	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	X	✓	✓	✓	✓	✓	✓	19
Green et al. (2020)	✓	✓	✓	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	✓	DK	19
Kanemasa (2022)	✓	✓	✓	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	✓	✓	20
Kiire (2017)	✓	✓	X	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	DK	✓	18
Kiire (2019)	✓	✓	X	✓	✓	✓	✓	✓	✓	✓	✓	✓	X	X	✓	✓	✓	✓	DK	✓	16
March et al. (2020)	✓	✓	✓	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	DK	✓	19
March et al. (2021)	✓	✓	X	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	✓	✓	19
March et al. (2022)	✓	✓	✓	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	✓	✓	20
Meier (2003)	✓	✓	X	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	✓	✓	19
Menard et al. (2021)	✓	✓	X	✓	✓	✓	✓	✓	✓	✓	✓	✓	DK	✓	✓	✓	✓	✓	✓	✓	18
Pineda et al. (2021)	✓	✓	X	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	✓	✓	19
Plouffe et al. (2020) - 1	✓	✓	X	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	N/A	✓	✓	✓	✓	✓	✓	19
Plouffe et al. (2020) - 2	✓	✓	X	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	N/A	✓	✓	✓	✓	✓	✓	19
Plouffe et al. (2022)	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	20
Rinker (2010)	✓	✓	X	✓	✓	✓	✓	✓	✓	✓	✓	✓	DK	X	✓	✓	✓	✓	DK	✓	16
Ryan et al. (2008)	✓	✓	X	✓	✓	✓	✓	✓	✓	✓	✓	✓	DK	X	✓	✓	✓	✓	✓	DK	15
Smoker and March (2017)	✓	✓	X	✓	✓	✓	N/A	✓	✓	✓	✓	✓	DK	DK	✓	✓	✓	✓	DK	DK	15
Tetreault et al. (2021)	✓	✓	X	✓	✓	✓	N/A	✓	✓	✓	✓	✓	N/A	N/A	✓	✓	✓	✓	✓	✓	19
Turner (2014)	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	DK	X	✓	✓	✓	✓	DK	✓	17

Note. ✓ = Yes; X = No, * Item is reverse scored, DK = Don’t know, N/A = not applicable.

Items: 1. Were the aims/objectives of the study clear? 2. Was the study design appropriate for the stated aim(s)? 3. Was the sample size justified? 4. Was the target population clearly defined? 5. Was the sample frame taken from an appropriate population base so that it closely represented the target population under investigation? 6. Was the selection process likely to select participants that were representative of the target population under investigation? 7. Were measures undertaken to address and categorize non-responders? 8. Were the outcome variables measured appropriate to the aims of the study? 9. Were the outcome variables measured correctly using measures that had been trialled, piloted, or published previously? 10. Is it clear what was used to determine statistical significance and/or precision estimates? (e.g., p values, CIs) 11. Were the methods (including statistical methods) sufficiently described to enable them to be repeated? 12. Were the basic data described? 13. Does the response rate raise concerns about nonresponse bias? 14. If appropriate, was information about non-responders described? 15. Were the results internally consistent? 16. Were the results for the analyses described in the methods presented? 17. Were the authors’ discussions and conclusions justified by the results? 18. Were the limitations of the study discussed? 19. Were there any funding sources or conflicts of interest that may affect the authors’ interpretation of the results? 20. Was ethical approval or consent of participants attained?

All of the studies employed an appropriate design for their aims, clearly defined their target population, represented their target sample, used appropriate and validated measures, clearly described significance with p-values, described data and methods adequately, presented internally consistent results, and matched descriptions outlined in the methods. In all studies, the discussions and conclusions were justified by the results of the studies. The majority of studies (k = 13, 59%) did not justify their sample size using a power analysis.

Meta-Analysis

All 22 studies were included in the random effects meta-analysis; however, where independent samples were included (by gender) these were entered as separate samples in the meta-analysis which brought the total number of samples to 33. Figure 2 shows the forest plot presenting the effect sizes for each sample and the total effect. Overall, there was a significant, weak, positive relationship between narcissism and IPV perpetration, (r = .15, 95% CI [0.12, 0.19], p < .001, T² = 0.01). High heterogeneity was present between the studies, Q = 235.69, p < .001, I² = 86.42; as such, we proceeded with our intended subgroup analysis between IPV types, and narcissism types, and gender.

Figure 2.

Random effects meta-analysis, correlations, confidence intervals, Z-values, p-values, and Forrest plot for included studies.

Individual subgroup effect sizes for subgroup analyses and Q-statistics are reported in Table 4. Except for physical IPV perpetration, all of the subgroups showed significant, high levels of between-study heterogeneity (I² > 75%). For IPV type, there were significant, weak, positive relationships between narcissism and cyber IPV perpetration (r = .22, p < .001), and psychological IPV perpetration (r = .18, p < .001). Physical IPV perpetration was not related to narcissism (r = .06, p = .057). The strength of the relationship differed between IPV types, p = .001. Both cyber IPV perpetration and psychological IPV perpetration had significantly larger effect sizes than physical IPV perpetration.

Table 4.

Pearson’s r, 95% Confidence Intervals, p-values, Heterogeneity Statistics; Q-Statistics for Heterogeneity and Associated p-values, Tau-Squared and I-Squared for Overall Effect and Subgroups.

Group	Samples in Analysis	Effect Size	95% CI	p	Q	p	I 2	T 2
		Correlation Coefficient	[Lower, Upper]
Overall effect	33	.15	[.12, .19]	<.001	235.69	<.001	86.42	0.01
Cyber IPV	9	.22	[.16, .29]	<.001	59.16	<.001	86.48	0.01
Physical IPV	14	.06	[-.00, .12]	.057	21.97	.056	40.82	0.01
Psychological IPV	15	.18	[.12, .23]	<.001	59.50	<.001	76.47	0.01
Grandiose Narcissism	27	.11	[.07, .15]	<.001	127.76	<.001	79.65	0.01
Vulnerable Narcissism	12	.25	[.19, .31]	<.001	90.16	<.001	87.80	0.01
Females	11	.14	[.07, .22]	<.001	96.55	<.001	89.64	0.02
Males	14	.15	[.08, .22]	<.001	104.91	<.001	87.61	0.02

Note. Df = number of samples-1, all analyses were random effects analyses.

Grandiose narcissism was significantly, positively, and weakly related to IPV perpetration (r = .11, p < .001), as was vulnerable narcissism (r = .25, p < .001). There was a significantly stronger relationship between IPV perpetration and vulnerable narcissism, compared to grandiose narcissism, p < .001.

Analysis of gender subgroups revealed IPV perpetration was significantly, positively, and weakly related to narcissism for females (r = .14, p < .001) and males (r = .15, p < .001). There was no significant difference between the strength of the relationship between IPV perpetration and narcissism for males and females, p = .851.

Publication Bias

To assess risk of bias between studies, the funnel plot of the overall effect was inspected visually, see Figure 3. There was evidence of general symmetry across the studies. This was analyzed further using a leave-one-out analysis which demonstrated when each study was removed, the overall effect ranged from r = .14 to r = .16. Additionally, Egger’s intercept was 0.47 and nonsignificant, p = .646, suggesting no bias was detected. Furthermore, a Duval and Tweedie (2000) trim and fill analysis revealed that no studies were required to be trimmed from the left side and six from the right side of the funnel plot. This would result in an adjusted significant effect size of r = .16 (95% CI = [0.15, 0.18]) which also represents a positive, weak relationship, suggesting no bias detected (Shi & Lin, 2019). Finally, using the classic failsafe N method, 3,798 non-significant studies would be needed for there to be a null effect, a number deemed extremely unlikely given only 33 samples were included in this meta-analysis.

Figure 3.

Funnel plot showing publication bias of included studies.

Discussion

Is Narcissism Related to IPV Perpetration?

This study aimed to measure the relationship between narcissism and IPV perpetration. Of the 22 studies investigated using random effects meta-analysis, the overall effect size between narcissism and total IPV perpetration was significant, positive, and weak, r = .15. This is consistent with research examining the relationship between IPV perpetration and NPD, which also found a correlation of r = .15 (Collison & Lynam, 2021). Moreover, Kjærvik and Bushman’s (2021) meta-analysis did not find a difference between narcissism and pathological narcissism (i.e., NPD) and the strength of their relationships with aggression and violence, suggesting that narcissism need not meet diagnostic criteria for there to be a risk of violent perpetration. Overall, the relationship between narcissism and IPV perpetration is small but consistent.

Is There a Difference in the Relationship Between Grandiose and Vulnerable Narcissism and IPV Perpetration?

A secondary goal of the present study was to ascertain if there is a difference in the strength of the relationship between IPV perpetration and grandiose and vulnerable narcissism. Subgroup analyses demonstrated weak, positive relationships, between IPV perpetration and both grandiose narcissism (r = .11) and vulnerable narcissism (r = .25). A significant difference between grandiose and vulnerable narcissism was evident. However, the high level of heterogeneity present in the studies suggests that moderators could be affecting the strength of the relationship. A recent qualitative study (Green & Charles, 2019) found that while both grandiose and vulnerable narcissists perpetrated violence, the violence was triggered by different factors. Grandiose narcissists were violent when triggered by threats to their self-esteem, while for those high in vulnerable narcissism, explosive anger episodes were triggered by fears of abandonment (Green & Charles, 2019). Moreover, a meta-analysis examining differences in narcissism types in forensic rehabilitation found that grandiose narcissism was correlated to proactive violence whereas vulnerable narcissism was correlated to reactive violence mediated by impulsivity and distressing emotions (Keune et al., 2021). Together this research aligns with this review’s finding that the strength of the relationship between IPV perpetration differs significantly between grandiose and vulnerable narcissism but this is likely moderated by different precipitating factors of the IPV perpetrated by grandiose and vulnerable narcissists; future research should explore this.

Is Narcissism Related to the Type of Violence Perpetrated Against Intimate Partners?

A third goal of this study was to examine if the strength of the relationship between narcissism differed based on the type of IPV perpetrated. There were insufficient studies assessing sexual IPV perpetration to analyze this relationship with narcissism on its own. The analysis revealed that narcissism was significantly, positively, and weakly related to cyber IPV perpetration (r = .22) and psychological IPV perpetration (r = .18). Notably, no significant relationship was found between physical IPV perpetration and narcissism. The relationship between cyber IPV perpetration and narcissism was significantly stronger than the relationship between physical IPV perpetration and narcissism. Additionally, the relationship between psychological IPV perpetration and narcissism was significantly stronger than the relationship between physical IPV perpetration and narcissism. Given this, violence perpetrated by people high on narcissism may be more likely to be cyber or psychological in nature. As such, future research should specify the type of violence that is perpetrated and focus on cyber and psychological forms of IPV.

Does the Relationship Between IPV Perpetration Differ in Strength for Males and Females?

A further goal of this study was to investigate if the relationship between IPV perpetration and narcissism differed in strength between males and females. Analysis of gender subgroups found no such difference exists. However, it may be that while the overall strength of the relationships between IPV perpetration and narcissism may not differ, these effects could be moderated by gender in combination with other factors. For example, Study 1 from Plouffe et al. (2020) found that the combination of being female, being a victim of IPV perpetration, and having higher levels of secondary psychopathy significantly increased the odds of more severe IPV perpetration. As such, it is imperative that researchers analyze data from different genders separately to better understand how IPV is perpetrated and its relationship with personality traits.

Limitations of the Studies

A limitation of the included studies is that all used self-report questionnaires to assess both narcissism and IPV perpetration. Given that some types of IPV perpetration are illegal and most are socially undesirable in most countries, participants may have been inclined to underreport this data and as such, results should be interpreted with caution. Moreover, while those high on grandiose narcissism are likely to believe in their own superiority, those with more moderate levels of narcissism may underreport narcissistic traits due to social undesirability of regarding oneself with such high esteem.

Self-reporting of these traits is also likely to be affected by cultural norms. Of the studies included, the majority (82%) were conducted in English and all were conducted in democratic and industrialized countries. The prevalence of IPV in these countries is similar and thus, is not representative of global IPV perpetration which varies greatly (Sardinha et al., 2022). Therefore, future research is needed to assess these relationships in non-democratic, industrialized countries.

The most frequently used measure of IPV was the CTS-2 which traditionally gives a total score and is not typically used to measure different types of IPV. Additionally, the NPI (n = 9) and a version of the SD3/SD4 (n = 8) were used most frequently. These measures are understood to measure grandiose narcissism and consequently most studies did not measure vulnerable narcissism which has been demonstrated in this meta-analysis to have a significantly stronger relationship with IPV perpetration than grandiose narcissism. Additionally, the systematic review identified variability in the measures used to assess narcissism and IPV perpetration which likely contributes to heterogeneity.

Additionally, the AXIS tool assessment identified that 8% of responses were either “no” or “don’t know.” This information was missing in regard to non-responders and their data, ethics and consent, conflicts of interest, aims, and sample size justification which reduced the overall interpretability of these studies’ results. It is therefore important that future studies follow reporting guidelines to assess that quality and risk of bias can be assessed comprehensively in the future.

Limitations of the Current Review and Meta-Analysis

To focus on the overall relationship between IPV perpetration and narcissism, no analysis was conducted on mediators, moderators, or other variables present in some of the studies. Reviewing these factors would be useful to provide insight into potential patterns of behavior and mediators that may account for IPV perpetration indirectly. These may include attachment style, which was assessed in two of the included studies, the different types of IPV perpetrated by different genders, the sexual orientation of the perpetrator, or cultural expectations and tolerance toward violence. Additionally, there was an insufficient number of studies that analyzed sexual IPV perpetration on its own to run a subgroup analysis, and so its relationship with narcissism remains unclear.

Implications for Further Research

While some of the included studies (k = 8) evaluated physical types of IPV perpetration, no significant relationship was found between physical IPV and narcissism. Thus, further research should use assessment tools that classify types of IPV perpetration so that the relationship between narcissism and IPV can be classified and reported with respect to its type of perpetration; in particular, there is a lack of studies looking at narcissism and sexual IPV.

Future research should assess both grandiose and vulnerable narcissism. Assessing these constructs under the two domains ensures their relationships to other variables are being accurately measured. The majority of studies (55%) measured only grandiose narcissism; thus there is a need for additional research to examine the relationship between vulnerable narcissism and IPV perpetration. In particular, the relationship between type of IPV and type of narcissism warrants further investigation. As it may be the case that different types of narcissism are related to different types of IPV perpetration and these relationships could be moderated by gender, but to date there are an insufficient number of studies that have examined this.

Moreover, future research should examine mediators and moderators of the relationships between grandiose as well as vulnerable narcissism and IPV perpetration. Given the different unique and complex traits of these types of narcissism, violence perpetrated by those with each trait may be influenced by a third variable such as impulsivity (Keune et al., 2021), or attachment (Buck et al., 2014). Rinker (2010) argued that physical IPV perpetration and controlling behaviors are mediated by the subtypes of narcissism and thus, it is necessary for research to differentiate between these traits. However, in the 13 years since this was articulated, most researchers continue to use a grandiose measure of narcissism and so, understanding how IPV is perpetrated differently between those with trait grandiose or vulnerable narcissism remains unclear.

Lastly, given the wide variation in IPV rates globally, and the limited diversity in our sample, future research should assess the relationship between IPV and narcissism in non-democratic, industrialized countries and consider how the relationship between IPV perpetration and narcissism may differ depending on culture and nationality. Our review did not look at a diversity of sexual orientations which may also alter the relationship between narcissism and IPV perpetration and future research should look to examine this in more diverse samples.

Conclusions

The current review and meta-analysis revealed the relationship between IPV perpetration and narcissism to be significant, weak, and positive; this relationship is consistent with findings between IPV perpetration and NPD suggesting that narcissism need not reach pathological levels for there to be a risk of IPV. The strength of this relationship did not differ significantly when assessing the gender of the preparator but the relationship between vulnerable narcissism and IPV perpetration was significantly stronger than that of grandiose narcissism and IPV perpetration. Subgroup analyses found both cyber IPV and psychological IPV were significantly, positively, and weakly related to narcissism; however, no relationship was found between narcissism and physical IPV. The critical findings from this systematic review and meta-analysis are summarized in Table 5. Implications for practice, policy, and research are summarized in Table 6. Future research should address limitations by ensuring assessments include both grandiose and vulnerable narcissism subgroups and specify the type of IPV perpetration being measured as well as exploring moderating variables and conducting research across more diverse nations. This will ensure that findings in this area are accurate and representative and by doing so, clinical and policy recommendations can be made specific for victims of IPV by narcissistic perpetrators including with a focus on psychological and cyber abuse prevention and treatment.

Table 5.

Critical Findings.

Narcissism and IPV perpetration  • There is a significant, positive, weak relationship between IPV perpetration and narcissism.  • This represents the same effect size as the relationship between IPV perpetration and NPD.  • Narcissism does not need to reach pathological levels for there to be a risk of IPV.

Grandiose and vulnerable narcissism and IPV perpetration  • The relationship between vulnerable narcissism and IPV perpetration is significantly stronger than the relationship between grandiose narcissism and IPV perpetration.

Narcissism and type of IPV perpetrated  • Narcissism is significantly, positively, weakly related to psychological IPV perpetration and cyber IPV perpetration.  • Narcissism was not related to physical IPV perpetration.  • The relationships between narcissism and cyber IPV perpetration and narcissism and psychological IPV perpetration were significantly stronger than the relationship between physical IPV perpetration and narcissism.  • Insufficient studies reported sexual IPV to run a subgroup analysis.

Narcissism and IPV perpetration for males and females  • There was no significant difference in the relationship between IPV perpetration and narcissism between males and females.

Note. IPV (Intimate Partner Violence), NPD (Narcissistic Personality Disorder).

Table 6.

Implications for Practice, Policy, and Research.

Practice  • Given that physical IPV showed no significant relationship with narcissism, clinicians should work with clients to identify other types of IPV that are less overt such as psychological and cyber abuse.

Policy  • National and international policies criminalizing IPV should include cyber and psychological violence as well as physical, in cases where this has not yet been adopted.

Research  • Research should use tools that measure both grandiose and vulnerable narcissism.  • Research should use tools that measure the separate domains of IPV.  • Future research should explore the role of variables that potentially moderate the relationship between IPV and narcissism.

Author Biographies

Eliza Oliver is Clinical Psychology Masters student at the Australian Catholic University. She also holds a Bachelor of Psychological Sciences (Hons) and a Bachelor of Creative Arts (Hons). Her research interests include understanding the intersection between personality and intimate partner violence perpetration and researching clinical interventions to reduce intimate partner violence perpetration and victimization.

Alexander Coates is a Clinical Psychology Masters student and PhD candidate at the Australian Catholic University. He holds a Bachelor of Psychological Sciences (Hons) and Bachelor of Property Economics (Hons). His research interests include psychological abuse in adult intimate partner relationships, particularly subtle abusive behaviors that are not recognized as abusive, and in relationships involving those who identify as LGBTQIA+.

Joanne Bennett, PhD, is a Lecturer in Psychology within the School of Behavioural and Health Sciences at the Australian Catholic University. She specializes in teaching research design and statistics, including advanced statistical methods such as meta-analysis and structural equation modeling. Her area of expertise is in applied neuropsychology, with a focus on research into road safety.

Megan Willis, PhD, is a Senior Lecturer in the School of Behavioural and Health Sciences at the Australian Catholic University. Her research interests include understanding and identifying risk and protective factors for intimate partner violence perpetration and victimization. She has a particular interest in the early identification of behaviors that may indicate that an individual is at risk of intimate partner violence.