The measurement trap: a meta-analytic review of measures in CEO narcissism research

The influence of CEOs on organizational outcomes has long intrigued scholars and practitioners. Growing concerns about the complex relationship between CEO personality traits and firm performance, particularly narcissism, stem from its ambiguous effects (Kelleci et al., 2018; O’Reilly et al., 2018; Palmer et al., 2020). However, efforts to synthesize findings (Cragun et al., 2020; Van Scotter, 2020) reveal a key issue: wide variability in CEO narcissism measurement, raising concerns about research reliability. To address this, we conduct a meta-analytic review examining the impact of CEO narcissism on organizational and CEO correlates, with a particular focus on methodological challenges. Our study expands the scope by evaluating previously unexamined relationships, assesses measurement reliability and validity, highlighting increasing methodological divergence (Koch and Biemann, 2014; Carey et al., 2015; Van Scotter, 2020), and advocates for standardized measurement to improve cross-study comparability and research consistency.

While our study is inspired by Cragun et al. (2020), who provided a broad meta-analytic overview of CEO narcissism, we take a more focused approach by systematically examining one issue only briefly acknowledged in their review: how different measurement methods of CEO narcissism may shape both the magnitude and direction of empirical findings. We argue that methodological choices are not mere technicalities, but pivotal factors in interpreting the effects of CEO narcissism.

Narcissism, marked by self-focus, grandiosity, superiority, and low empathy (WHO, 2016; APA, 2013), presents psychological challenges for individuals and those around them. Research links narcissism to problematic interpersonal relationships (Wright et al., 2017) and negative social effects (Wurst et al., 2017), extending into the workplace, where it shapes employee relations, leadership style, and organizational dynamics (Chatterjee and Hambrick, 2007; Wang et al., 2023). Narcissism exists on a spectrum, from a personality trait to a clinical disorder (Miller and Maples, 2011). As a trait, it varies in self-centeredness, admiration-seeking, entitlement, and lack of empathy, while Narcissistic Personality Disorder (NPD) is a clinical diagnosis in the APA (2013) and WHO (2016). The ICD defines narcissism as persistent grandiosity, arrogance, excessive self-love, egocentrism, and hypersensitivity to criticism, while the DSM describes NPD as grandiosity, admiration-seeking, and low empathy, requiring at least five of nine criteria, including preoccupation with success, belief in being special, entitlement, exploitation, envy, and arrogance.

Narcissism can also be categorized into two distinct forms: grandiose (overt) narcissism, characterized by an open display of superiority and a desire for admiration, and vulnerable (covert) narcissism, marked by defensiveness, hypersensitivity to criticism, and feelings of inadequacy (Luchner et al., 2011). Grandiose narcissism typically involves assertiveness and high self-esteem, while vulnerable narcissism is associated with emotional instability and psychological distress (Miller et al., 2021).

Narcissism is widely discussed in organizational contexts as narcissistic individuals often pursue high-level leadership roles, including CEO positions (Brunell et al., 2008; Brown et al., 2013). Their self-centered behavior facilitates this trajectory (Rovelli and Curnis, 2021), driven by a need for admiration, attention, and validation, sometimes called “narcissistic supply” (Aabo et al., 2020). This craving for recognition propels them into leadership, where they can exert control and enhance their status. Narcissists are also perceived as strong leaders due to the association between power and leadership (Nevicka et al., 2018; Brunell et al., 2008). This is because narcissists typically exhibit grandiosity, a heightened need for admiration, and a sense of superiority (DSM-5), which are often perceived by others as signs of powerful and confident leadership. However, these outward traits do not necessarily translate into actual leadership effectiveness. Therefore, it is important to distinguish between leadership emergence and leadership effectiveness. (Campbell and Miller, 2011). While narcissistic leaders advance through ambition, confidence, and self-presentation, their long-term impact can be problematic. They may belittle employees, avoid close working relationships, and impose grandiose, erratic demands, harming employee well-being (Campbell and Miller, 2011). Prioritizing personal status over organizational health, they may also manipulate financial reports for self-benefit (Peterson et al., 2012; Braun et al., 2016, 2017; Buchholz et al., 2018). Despite these drawbacks, some research suggests that narcissistic CEOs can foster innovation, enhance organizational performance (Kraft, 2022; Steinberg et al., 2022; Tuwey and Ngeno, 2019), and may be more effective in crises (Patel and Cooper, 2014; King, 2007; Sedikides and Campbell, 2017). In family firms, highly narcissistic CEOs may leverage innovation to guide strategic decisions (Rovelli et al., 2023). Narcissists are also more likely to engage in entrepreneurial activity (Yadav, 2023).

CEO narcissism is measured through various methods. Cragun et al. (2020) classify these into five types: the CEO Narcissism Index (CNI), psychometric self-reports, third-party reports, pronoun usage, and signature size. Apart from self-reports, these methods are unobtrusive, reducing respondent burden (Auriacombe, 2007), which is beneficial given CEOs’ time constraints and reluctance to participate in surveys (Matthews et al., 2023). However, concerns remain regarding reliability, validity, and ethical implications (Cragun et al., 2020; Koch-Bayram and Biemann, 2020; Van Scotter, 2020). For instance, can narcissism truly be assessed based on signature size?

Self-report tools require CEOs to assess their own narcissistic traits. The Narcissistic Personality Inventory (NPI) (Raskin and Hall, 1981) is the most widely used, measuring grandiosity, entitlement, and admiration-seeking. Shorter versions, such as NPI-16 (Ames et al., 2006) and NPI-13 (Gentile et al., 2013), focus on leadership and exhibitionism. Despite criticism for potentially conflating self-esteem with narcissism (Rosenthal et al., 2011), the NPI demonstrates strong validity (Miller et al., 2011, 2013). However, concerns remain regarding its suitability as a measure of leadership behavior. Due to its strong emphasis on leadership-oriented traits, the NPI may blur the distinction between adaptive leadership qualities and maladaptive narcissistic tendencies, potentially oversimplifying the relationship between narcissism and leadership outcomes. Other validated self-report measures include the Grandiose Narcissism Scale (GNS) (Foster et al., 2015), Five-Factor Narcissism Inventory (FFNI) (Glover et al., 2012), and Narcissistic Admiration and Rivalry Questionnaire (NARQ) (Back et al., 2013), which assess admiration, rivalry, and entitlement (Foster et al., 2015; Back et al., 2013; Leckelt et al., 2018). The Hogan Development Survey (HDS) is used in CEO studies but mainly identifies traits that can limit professional success rather than fully capturing narcissistic behavior (Wang et al., 2023). Some Dark Triad assessments, like the Dirty Dozen (Jonason and Webster, 2010) and Short Dark Triad (SD3) (Jones, 2020), provide brief measures of narcissism, psychopathy, and Machiavellianism to reduce participant fatigue. Self-reports are in general vulnerable to biases in question wording, response format, and social desirability (Fox and Schwartz, 2002). Moderate correlations between self-reports and clinical diagnoses suggest that self-assessments may only partially capture Narcissistic Personality Disorder (Baggio et al., 2022). Among existing tools, FFNI and NPI are most aligned with expert consensus on grandiose narcissism (Miller et al., 2014).

Narcissism can also be measured through third-person reports (peers, family members, or trained observers). Clinical assessments use structured interviews and standardized tools like the Pathological Narcissism Inventory (PNI) (Miller et al., 2013), while third-person reports offer insights into behaviors self-reports may overlook (Jauk et al., 2022). Studies show that third-person reports correlate significantly with clinical assessments, reinforcing their validity (Miller et al., 2013). However, both approaches have limitations: clinical assessments may be affected by clinician subjectivity, while third-person reports can be biased by the observer’s personal experiences with the CEO (Kaufman et al., 2020).

Unobtrusive methods estimate narcissism without CEO involvement, using publicly available data such as signature size, pronoun use, and CEO photo prominence. The widely used CEO Narcissism Index (CNI) (Chatterjee and Hambrick, 2007) aggregates indicators including CEO cash and noncash compensation relative to the next-highest paid executive, CEO photo size in annual reports, CEO mentions in company press releases, and pronoun usage in interviews (the latter was omitted in Chatterjee and Hambrick, 2011). The prominence of the CNI in CEO narcissism research is evident in the fact that many subsequent measurement approaches build upon it. For example, Olsen et al. (2014) proposed an abridged version of the CNI consisting of three indicators. The abridged version retained two indicators from the original CNI (CEO cash and noncash compensation relative to the next-highest-paid executive) and included one adjusted indicator: the CEO’s photograph in the annual report, which was modified by adding one point to the original scale. Expanding on the CNI, the CEO Narcissism Score (Rijsenbilt, 2011, Rijsenbilt and Commandeur, 2013) incorporates 15 indicators across four dimensions: compensation, exposure, power, and acquisition behavior. Studies applying these measures (e.g. Buchholz et al., 2020; Kim, 2018) highlight their ability to capture broader narcissistic behaviors. But the reliability of narcissism indices remains debated. Van Scotter (2020) raised concerns about reliability, with some studies lacking sufficient validation. Koch-Bayram and Biemann (2020) argued that measures like signature size and pronoun use may reflect company culture or marketing strategies rather than narcissistic tendencies.

Singular-item measures face even greater criticism. While signature size has been linked to narcissism (Ham et al., 2017, 2018), Mailhos et al. (2016) found significance only in female CEOs, limiting generalizability. Pronoun use, once assumed to indicate narcissism, has been debunked. Carey et al. (2015) found near-zero correlation between first-person pronouns and narcissistic traits.

These methodological inconsistencies raise a key question: To what extent do differences in measurement strategy explain the inconsistent empirical findings on CEO narcissism and firm outcomes, such as innovation, risk-taking, or organizational performance?

To address our research question, we conducted meta-analyses of the bivariate relationships involving CEO narcissism. Our analytic procedures follow the Meta-Analysis Reporting Standards (MARS) and practical implementation guidelines (Kepes et al., 2013) to ensure transparency and replicability. We compiled an exhaustive sample of CEO narcissism studies available as of March 2025 using four search strategies:

1. Database search: Web of Science, Scopus, and ABI/INFORM, with keywords CEO narcissism, executive narcissism, destructive leadership, and abusive supervision.

2. Journal review: Manual screening of leading journals in leadership and ethics (such as The Leadership Quarterly, Leadership or Journal of Business Ethics).

3. Meta-analyses and literature reviews: Identification of additional studies from recent meta-analyses on CEO narcissism (Cragun et al., 2020), entrepreneurial narcissism (Liu et al., 2021), leadership narcissism (Braun, 2017; Grijalva et al., 2015), destructive leadership (Mackey et al., 2021), and general narcissism (Ismail et al., 2022).

4. Unpublished research: Searches in ProQuest Dissertations and Theses and SSRN for theses, dissertations, and working papers.

We applied three inclusion criteria. Only English-language studies were involved, they had exclusive focus on narcissistic CEOs (not general leadership or managerial narcissism), and they reported bivariate effect sizes in correlation tables (or we obtained it directly from authors). The final dataset comprises 124 studies (119 published, 5 unpublished) spanning 2007–2025 (a list of the studies in the Supplementary Material).

In the next step, we developed a coding form to ensure consistency and reliability. We first coded general publication details (e.g. type, year) and sample characteristics (size, origin, firm size, type). Next, we coded variable types and measurement methods, as many variables were assessed using different approaches. The operationalization and definitions of all variables are presented in Table 1.

To compare our findings with prior meta-analyses, we adopted and extended the categorization from Cragun et al. (2020) and considered ongoing discussions on the strengths and weaknesses of unobtrusive measures (Matthews et al., 2023; Cragun et al., 2020; Koch-Bayram and Biemann, 2020; Van Scotter, 2020). CEO narcissism measures were classified as: (1) psychometric self-report, (2) psychometric third-party, (3) the 2011 version of the CEO Narcissism Index (CNI), which excluded the original fifth item (first-person singular pronoun usage) due to reduced internal consistency, (4) Altered CEO Narcissism Index (e.g. abridged 3-item CNI by Olsen et al., 2014, original Chatterjee and Hambrick’s CEO Narcissism Index (Chatterjee and Hambrick, 2007) and its variations), (5) single-item measures (e.g. first-person singular pronoun use, signature size, photo size), and (6) other constructs (e.g. CEO narcissism score (Rijsenbilt, 2011, Rijsenbilt and Commandeur, 2013), or LinkedIn-based measure (Junge et al., 2024)). Other variables were grouped into firm-level and CEO-level attributes. All effect sizes were recorded as Pearson correlation coefficients.

Two co-authors independently coded the sample. Each coded a subset of 20 studies to assess consistency, achieving over 95% inter-rater agreement based on percent agreement (McHugh, 2012). Discrepancies, mainly related to general study details, were resolved through discussion. No protocol modifications were needed, and the remaining studies were coded accordingly.

For the analysis, we applied the meta-analysis method to examine bivariate relationships between CEO narcissism and all variables of interest. Effect sizes were extracted as Pearson’s correlation coefficients and transformed using Fisher’s Z to correct for skewness. If a study contained multiple measures of the same variable (e.g. two CEO narcissism assessments), all effect sizes were included to explore measurement differences rather than derive a single overall effect. Studies were weighted by the number of CEOs; if unavailable, the number of firms was used instead. To address differences between cross-sectional and panel data designs, if a study only reported firm-year observations, we divided the total number of firm-year observations by the number of years to approximate the number of unique CEOs/firms. This allowed us to harmonize the unit of analysis across studies. As a sensitivity check (reported in the Supplementary Material), we reran the analysis using the original firm-year observations, which yielded similar results.

Although we coded a range of bivariate relationships between variables and CEO narcissism, we included in the analysis only those with at least three effect sizes per CEO narcissism measure subgroup to ensure meaningful representation. The final list of variables includes a combination of three firm-level and three CEO-level attributes. The relationship between CEO narcissism and financial performance has been a central topic in the literature and is examined in most studies within our sample. Firm size and firm age are commonly used (often as control variables in performance models) and prior research suggests they may confound unobtrusive measures of CEO narcissism (Cragun et al., 2020; Koch-Bayram and Biemann, 2020; Van Scotter, 2020). Regarding CEO-level attributes, there is no strong theoretical basis to expect systematic differences in how narcissism relates to gender or age across measurement types. However, the relationship with CEO tenure may vary depending on how narcissism is operationalized. For example, third-party assessments may reflect perceptions shaped by a CEO’s visibility or long-standing public image, which tends to increase with tenure. In contrast, self-report measures are less likely to capture these externally reinforced traits and may show weaker or no association.

A list of omitted relationships and study counts is available upon request. To account for variability across studies, we used a random-effects model (Borenstein et al., 2010) rather than a fixed-effects model, which assumes a single true effect size. Effect sizes were weighted using the inverse variance method (Hedges and Olkin, 1985).

Between-study heterogeneity was assessed using τ² (REML estimator) (Hardy and Thompson, 1996), Cochran’s Q test (Cochran, 1954), where a low p-value suggests heterogeneity, though its low power with small samples limits interpretation, and I² statistic (Higgins et al., 2003), which quantifies heterogeneity as low (25%), moderate (50%), or high (75%). While I² values above 25% justify moderator analysis, they may overestimate heterogeneity in small samples (von Hippel, 2015) and should be interpreted alongside credibility intervals. Subgroup difference tests were performed to evaluate variability across CEO narcissism measures.

We conducted post-hoc tests to assess the robustness of results, including sensitivity analyses for influential cases, overlapping samples, and publication bias. All analyses were performed in R using the metafor package (Viechtbauer, 2010).

Results of the bivariate meta-analyses are presented in Table 2. Looking at the overall effect sizes of CEO narcissism and firm attributes, firm size (r = 0.100, 95% CI = [0.064, 0.137]) and firm age (r = 0.052, 95% CI = [0.020, 0.083]) are statistically significant, while financial performance (r = 0.010, 95% CI = [−0.005, 0.024]) is positive but nonsignificant. These findings align with prior research, suggesting that narcissistic CEOs are more likely to lead larger and older firms.

For CEO demographics, tenure is not significantly linked to narcissism (r = 0.006, 95% CI = [-0.025, 0.038]). Similarly, age shows no significant effect (r = 0.023, 95% CI = [−0.004, 0.050]). Therefore, the only overall significant effect size for women (r = −0.019, 95% CI = [−0.036, −0.002], p < 0.05) suggest that female CEOs might be less likely narcissistic. However, as already discussed in the theoretical section, previous research indicates that some of the relationships may be influenced by the method of measurement.

Table 2 further presents between-study heterogeneity measures, showing high heterogeneity, suggesting that effect size differences may be influenced by moderators, particularly how CEO narcissism is measured. All overall effect size distribution, except for CEO gender, show high heterogeneity (measured by Q-statistics and I²). As discussed in the theoretical section, methodological variations across studies likely contribute to this variability. To investigate, we conducted subgroup analyses based on the six coded measurement methods. While subgroup analysis requires a sufficient number of studies for statistical power (Cuijpers et al., 2021), we opted to report all six categories individually to highlight differences in effect sizes. Similar to within group heterogeneity, all subgroup difference tests (measured by Q statistic between groups), except for CEO age, indicate that measurement method significantly affects effect sizes.

Our analysis revealed notable discrepancies between psychometric measures and other methods, particularly in the firm attributes. At the same time, we see that psychometric self-report and third-party evaluations break from each other in most cases, even in opposite directionality, raising concerns about how narcissism is perceived vs self-reported. Third-party measures consistently indicate stronger and more significant positive associations, such as for firm performance r = 0.053, p ≤ 0.01 or CEO tenure r = 0.131, p ≤ 0.10. In contrast, effect sizes measured by self-report evaluation are often weaker and negative, such as for firm age r = −0.057, p < 0.05 or CEO age r = −0.011, n.s. Finally, the Chatterjee and Hambrick (2011) version of the CNI remains widely used but often appears in modified forms, such as the abridged 3-item variant or other versions with additional indicators. Our results indicate that these modified variants typically yield stronger and more statistically significant correlations with organizational variables than the original four-item CNI. For example, in relation to firm size, the altered CNI reports r = 0.156, compared to r = 0.094 for the original; similarly, for firm age, the altered version shows r = 0.117 versus r = 0.078 for the original. However, these modified versions also introduce greater heterogeneity across studies, suggesting that the standardized CNI remains the most consistent among the three construct categories. These differences are discussed in more detail in the discussion section.

To test the robustness of our findings, we conducted several post hoc analyses addressing alternative weighting, overlapping samples, outliers, potential publication bias, and comparison with previous meta-analysis by Cragun et al. (2020). Although minor differences arose due to changes in weighting, sample overlap, or publication bias, the overall pattern of results remained stable. Crucially, these comparisons reinforce our broader conclusion: variation in measurement approaches is a key driver of inconsistency in CEO narcissism research, highlighting the need for more rigorous, standardized, and theoretically grounded methodologies. A description of the methods and recalculated results are provided in the Supplementary Material.

This study addresses the inconsistency in CEO narcissism measurement, which affects both academic research and corporate decision-making. Without reliable and standardized measures, conclusions remain questionable, potentially leading to misguided leadership selection and governance decisions. Our meta-analysis highlights that CEO narcissism’s impact on firm performance is inconsistent, with different measurement methods yielding contradictory results. Our analysis diverges from Cragun et al. (2020) by identifying a growing trend in modifying the CNI, reflecting concerns over its limited reliability. Despite widespread use, the CNI undergoes frequent adaptations that fail to address its fundamental weaknesses, leading to greater inconsistency in findings. Different CEO narcissism measures continue to produce varying and sometimes contradictory results. Self-reports typically show weak or negative correlations with financial performance, whereas third-party assessments suggest a small positive correlation, pointing to differences between self-perceived and externally observed narcissism. The CNI, while common, often produces moderate correlations that lack consistent statistical significance, raising doubts about its robustness. Measures such as pronoun use and signature size, serving as indirect proxies, show minimal correlations with wide confidence intervals, further supporting concerns over their limited reliability (Carey et al., 2015).

These findings suggest that the choice of measurement method is a key determinant of the observed relationships between CEO narcissism and organizational outcomes. For instance, third-party assessments tend to show stronger associations with demographic variables such as tenure and gender. This pattern may reflect the performance cue effect (Lord et al., 1978), where successful CEOs are perceived as more dominant or confident (and thus more narcissistic) regardless of their actual traits. This aligns with the Romance of Leadership theory (Meindl et al., 1985), which suggests that leadership attributes are often inferred from organizational performance rather than observed directly.

Moreover, the CEO Narcissism Index (CNI) and similar constructs include indicators such as compensation and media visibility, which may themselves be influenced by firm characteristics like size and age. For example, CEO compensation tends to increase with firm size, meaning that part of the observed relationship may stem from structural overlap between the proxies used to measure narcissism and the firm-level variables, rather than from narcissism itself (Cragun et al., 2020). This confounding effect may explain why studies using archival data often report stronger correlations with firm size and age than those using psychometric measures.

Another issue concerns the continued use of single-item measures (e.g. signature size, pronoun use), which produce effect sizes that are often inconsistent with those derived from multi-item or psychometric instruments. Notably, some of these single-item indicators, such as the CEO’s photograph in the annual report, are also used as components within broader constructs like the CNI or CEO Narcissism Score. Interestingly, single-item measures are more likely to show negative associations with financial performance, diverging from the findings associated with other measurement approaches. Despite their variability, these measures tend to exhibit comparatively lower between-study heterogeneity.

To sum this up, each of the measurement approaches comes with its own set of limitations. Psychometric self-report measures, for instance, are susceptible to social desirability bias, where individuals might tailor their responses to align with perceived social norms rather than provide genuine answers (Deshields et al., 1995). Additionally, narcissistic individuals may struggle with self-awareness, particularly in recognizing their communal qualities, further limiting the accuracy of self-reports (Grijalva and Zhang, 2015). Third-party assessments, while offering an external perspective, are not immune to subjectivity, as they depend on the observers’ interpretations of narcissistic traits (Connelly, 2022). The CEO Narcissism Index (CNI), despite its popularity, may fail to capture the full spectrum of narcissistic traits, with some components, such as pronoun use, having questionable validity (Van Scotter, 2020). For example, extensive research, including Carey et al. (2015), has shown a negligible correlation between narcissism and first-person singular pronoun use, problematising the assumption that language patterns can reliably indicate narcissism. Similarly, the validity of using signature size as a proxy for narcissism is also questionable, as it may be influenced by factors unrelated to narcissism, such as individual writing style or cultural norms (Klimosky and Rafaeli, 1983; Thiry, 2009).

In clinical settings, the most reliable assessments of narcissism typically involve a combination of self-report measures, observer ratings, and clinical interviews. Instruments like the Narcissistic Personality Inventory (NPI) (Ames et al., 2006) and the Pathological Narcissism Inventory (PNI) (Pincus et al., 2009) are well-established tools in clinical contexts, offering structured frameworks for assessing narcissistic traits. Observer ratings, conducted by trained clinicians or knowledgeable informants decrease self-report bias (John and Robins, 1994). Clinical interviews allow for a deeper exploration of narcissistic traits and their impact on functioning (Gunderson et al., 1990).

Comprehensive clinical approaches in CEO research are rare due to several factors. First, they require extensive resources, both time and financial investment, making them impractical for large-scale studies. Second, reliable execution demands expertise and specialized training, which is not always available in research settings. Finally, coordinating clinical interviews with high-profile individuals, such as CEOs with demanding schedules, presents logistical challenges. Consequently, more resource-efficient tools such as surveys and indices are preferred despite their limitations in capturing the full complexity of narcissistic traits. Among these, the Narcissistic Personality Inventory (NPI) remains the most widely used and validated tool. Its applicability across diverse populations, including the general public, students, and nonclinical groups, highlights its broad utility and relevance (Twenge et al., 2008; Pincus et al., 2009). The recent validation of a shorter version of the NPI by Schmalbach et al. (2020) suggests its potential for efficient use in organizational settings and is promising for future research.

Given the risks of self-serving leadership behaviors, unethical decision-making, and weakened organizational culture, firms should adopt rigorous assessment processes, including third-party evaluations and strong governance, to balance potential benefits with long-term risks (Petrenko et al., 2016; Fuller et al., 2018; Pauknerova et al., 2022; Palmer et al., 2022). Involving clinical experts for external assessments, though resource-intensive, could offer deeper insight. A promising development highlighted by Matthews et al. (2023) is the use of video clips and voice recordings, analyzed manually or with computer-aided tools. It is equally important to acknowledge the ethical challenges involved, particularly those related to diagnosis. The Goldwater Rule, issued by the American Psychiatric Association, advises against offering psychological diagnoses of public figures without a direct evaluation and informed consent. One promising alternative is to focus on observable leadership behaviors rather than risk problematic diagnostic labeling.

Another emerging trend is machine learning, which could assess CEO narcissism by analyzing textual data (e.g. speeches, social media, shareholder letters). Machine learning presents challenges like algorithmic bias, which can reinforce stereotypes, and the black box problem, making predictions hard to interpret. Longitudinal studies could track how CEO narcissism changes over time and its impact on organizations, especially during crises or growth periods.

While our meta-analysis focused on linear relationships, future studies should consider the potential for curvilinear effects. Moderate levels of narcissism may benefit organizations by fostering confidence and strategic vision, whereas excessive narcissism may result in impulsive or self-serving actions that harm firm performance. Also, although our analysis primarily reflects grandiose narcissism (as most existing measurement tools such as the NPI, CNI are designed to capture this dimension) future research should explicitly examine vulnerable narcissism, which remains understudied in executive contexts but may have distinct implications for leadership behavior and firm outcomes.

Our study has some limitations. The first limitation is related to the applied meta-analytic techniques. While subgroup analysis and meta-regression are valuable tools for exploring heterogeneity and investigating potential moderators, one must be cautious in the interpretation of the results. First, two methods cannot prove causality as the results are observational in nature explaining differences between studies (Borenstein and Higgins, 2013). Second, even if we did not find a difference in subgroups’ effect sizes, it does not mean that there is none. The subgroup analysis often lacks statistical power for a small number of studies and the significant level of heterogeneity common in the field of social sciences (Cuijpers et al., 2021).

This meta-analysis reviews research on CEO narcissism and how different ways of measuring it affect the results. We highlight the need for clearer and more consistent measurement, especially given the challenges narcissistic leaders can pose for employee well-being and organizational health.

The supplementary material for this article can be found online.