Looking at the darker side of the mirror: the impact of CEO ’ s narcissism on corporate social irresponsibility

Purpose – The aim of this article is to highlight the major part played by executives in the escalation of corporate social irresponsibility (CSI). Based on the upper echelons theory, the authors developed a model which showsthe essential role ofCEOs in explaining CSI. The authors proposed that the keypersonality traits of CEOs — narcissism — , as well as their power, could explain the degree of CSI. Design/methodology/approach – DuetothesignificantmethodologicalchallengeswheninvestigatingCSI, theauthorsexploredanovelmethodformeasuringCSIinordertoassessthedegreeofirresponsiblebehaviors. TheauthorsbuildaCSIscalebasedontheperceptionsofkeyinformants,i.e.expertswithdiverseprofessionalbackgrounds.TheauthorsapplyCSIscaleinasampleof84SpanishcompaniesthatwereinvolvedinCSI. Findings – The results of the authors ’ empirical study show the positive and significant influence of CEO narcissism and CEO power on the degree of CSI. Research limitations/implications – On the one hand, corporate irresponsibility scandals have relevant social consequences and practical implications. On the other hand, narcissism is a natural feature of managers in top positions that is increasing in societies. Practicalimplications – Theauthors ’ findingsmayhelpCEOs,TMTsandcorporateboardstoacknowledge potential sources of CSI decreasing its likelihood through counterbalancing CEO ’ s power and considering the dark side of narcissism. Social implications – On the one hand, corporate scandals have relevant social and practical implications. On the other hand, narcissism is a natural feature of managers in top positions that is increasing in societies. Originality/value – Inthispaper,theauthorshighlighttheroleofCEOscharacteristics andtheirfirmsasthe key actors for explaining and understanding the degree of CSI.


Introduction
Corporate social irresponsibility (CSI) and the associated corporate scandals-e.g. the cases of Enron, Petrobras, World.com,Bankia or Volkswagen-have not stopped in this century.CSI, -i.e.corporate acts that intentionally cause harm (Clark et al., 2022;Kemp and Owen, 2022) -has relevant social, environmental and economic consequences impacting companies, communities and people worldwide.As Iborra and Riera (2023) state, there is empirical evidence that CSI provoked enormous consequences, between others, in CEO's narcissism and irresponsible behavior antecedents with CSI behaviors (Iborra and Riera, 2023;Grijalva and Harms, 2014).By focusing on the individual level instead of on institutional variables, we respond to the call made by Ghoshal (2005, p. 79): "when managers, including CEOs, justify their actions by pleading powerlessness in the face of external forces, it is to the dehumanization of practice that they resort.When they claim that competition or capital markets are relentless in their demands, and that individual companies and managers have no scope for choices, it is on the strength of the false premise of determinism that they free themselves from any sense of moral or ethical responsibility for their actions".We argue and provide evidence that CEOs are the key actors influencing CSI.We contribute also to upper echelons theory, specifically analyzing the impact of narcissism, which is one of the traits of upper echelons, on CSI and providing new evidence of the dark side of this trait.Finally, our study draws attention to CSI as a construct that is distinct from corporate social responsibility (CSR) and has its own antecedents [5] that calls for special consideration (Clark et al., 2022;Iborra and Riera, 2023) [6].In this line, we explore a novel method for measuring the degree of CSI as a specific construct.Concretely, we construct a scale of CSI degree based on the perceptions of a panel of experts belonging to different groups of stakeholders.
In the next section, we review the literature and present our hypotheses.After describing our research method, we present our empirical findings, which derive from data on 84 Spanish companies.We conclude with a discussion of the results, together with their implications and issues for further research.

CSI and the role of CEOs
While Armstrong (1977) was a pioneer in introducing the CSI concept in the academic literature, it has been the last 20 years when academia has devoted notable attention to it (Riera and Iborra, 2017;Clark et al., 2022).CSI focuses on the harmful consequences of an act and definitions include acts causing harm, hurting, causing damage or violation with different degrees (Clark et al., 2022) that can go from the loss of human life, to the loss of nonhuman life or to the loss of livelihoods in local communities (Mena et al., 2016;Clark et al., 2022).Additionally, CSI focuses on the organization or the corporation as an actor.Corporations as collective agents because they are capable of intentional actions, having specific decision procedures that help to explain how they reach decisions and how they act.In that sense, Godfrey's (2005, p. 787) states that "bad acts must be accompanied by a bad mind" incorporating intention to the definition and giving place for the role of the two core issues in CSI: damage and intention (Clark et al., 2022).So, we consider that a company is involved in CSI if it intentionally causes damage or harm to others [7].This definition fits with researchers that advocate for considering that CSI is a different construct from CSR deserving individual attention (Strike et al., 2006;Riera and Iborra, 2017;Clark et al., 2022).
In understanding CSI antecedents, scholars have primarily studied antecedents at environmental level.For example, how institutional context and government corruption influences the occurrence of CSI (Ashforth and Anand, 2003;Keig et al., 2015).CSI is seen as a matter of good firms in bad contexts, resulting from poorly regulated settings and weak institutions-external antecedents-rather than from "bad" firms and managers-internal antecedents- (Cuervo-Cazurra et al., 2021).
Contrary to this trend, the upper echelons theory (Hambrick and Mason, 1984) suggests that CEO characteristics, experiences and cognitive values should be studied in order to understand the actions of companies, since CEOs play a key role in decision-making and, therefore, to explain company results, strategic decisions and behaviors.Recently, some articles open this line of research linking the upper echelons theory with CSI.Specifically, they focus on well-known demographic characteristics, such as CEOs' tenure or career horizons, as antecedents of CSI (Lee et al., 2018;Oh et al., 2018).

CEO's narcissism and irresponsible behavior
We delve in this line of research.We argue that given that irresponsibility is an issue closely linked with ethics, values and personal beliefs (Carroll, 1979), certain CEO psychological features may directly influence moral and ethical aspects and hence also irresponsibility issues (Garriga and Mel e, 2004).Psychological features are made up of values, cognitive models and other elements of the personality, with which executives filter and interpret both external and internal stimuli.There is evidence that CEOs' values, together with the attitude and awareness of CEOs concerning these issues, increase the likelihood of having a significant influence on CSR results (Laguir et al., 2016;Waldman and Siegel, 2008).In this regard, Waldman and Siegel (2008) point out that if CEOs have a strong ethical conviction, this contributes to positive results in CSR.In the same vein, Laguir et al. (2016) have shown that one of the key factors for adopting CSR lies in the commitment of CEOs, as well as in their values and culture.
In the case of CSI, we argue that CEOs' psychological features will affect the way in which executives filter and interpret information, demands and interest (Andreoli and Lefkowitz, 2009).Even more, regarding psychological characteristics, irresponsibility is related to causing harm to stakeholders, so it involves a behavior where there is a lack of consideration regarding the interests and expectations of stakeholders, as well as a lack of empathy and selfishness.Consequently, we argue that CEO narcissism, i.e. arrogance, selfishness and feelings of superiority could predict the likelihood of CSI.
CEO narcissism and its link with CSI Upper echelons researchers highlight that CEO narcissism helps to understand decisionmaking processes (Chatterjee and Hambrick, 2007;Cragun et al., 2020;Tang et al., 2018).Ellis (1898) first introduced the concept of narcissism in the field of psychology when alluding to the Greek myth of the young Narcissus, who falls in love with his own reflection in the water.However, this concept grew stronger thanks to the Austrian neurologist, Sigmund Freud (1957).He stated that leaders only need their own love and are usually extremely selfish, self-confident and independent (Freud, 1957).Two decades later, narcissism was considered as a personality disorder (Raskin and Hall, 1979).It was only in 1994, however, when the American Psychiatric Association considered narcissism not only as a clinical disorder but also as a dimension of personality.Therefore, narcissism has been studied under a double perspective, i.e. as a personality disorder (psychiatric feature) and as a personality trait.
In the management literature, Campbell et al. (2011, p. 269) in their review, define narcissism as containing three components, "First, the narcissistic self is characterized by positivity, "specialness" and uniqueness, vanity, a sense of entitlement and a desire for power and esteem.Second, narcissistic relationships contain low levels of empathy and emotional intimacy . . .Third, there are narcissistic strategies for maintaining inflated self-views".For management scholars, narcissistic CEOs are characterized by having traits such as self-admiration, self-absorption, authority, exhibitionism, feelings of superiority, arrogance, exploitation of others, self-sufficiency and extreme vanity (Emmons, 1987;Rijsenbilt and Commandeur, 2013;Kim et al., 2018;Cragun et al., 2020).
In fact, it is considered as a key personality trait to explain corporate outcomes (Campbell et al., 2011;Chatterjee and Hambrick, 2007;Oesterle et al., 2016;Cragun et al., 2020;Seifzadeh et al., 2021) and is a psychological trait of most executives of important global companies (Finkelstein et al., 2009;Kim et al., 2018).In this sense, narcissism has been studied as a feature that lies at the heart of leadership and "anyone who hopes to the rise to the top of an organization should have a solid dose of narcissism" (de Vries, 2004, p. 188).
However, in the debate on the pros and cons associated with narcissism empirical findings are contradictory (Chatterjee and Hambrick, 2007;Maccoby, 2000;Salehi et al., 2022). Cragun et al. (2020) in their meta-analysis of 37 studies found a positive and significative effect of narcissism over financial performance [8], mixed results for innovation and growth and no significant ones for risk-taking.Even more, to date, no consensus has been reached as to how narcissism affects the performance of companies (Anninos, 2018).Seifzadeh et al. (2021) empirically studied that there is a positive and significant relationship between CEO narcissism and overconfidence and real earnings management and managers' myopia and financial statements readability.In their own words, "Managers may use our results to improve their capabilities, such as their accuracy in preparing financial statements, through working on their personal features" (Seifzadeh et al., 2021, p. 123).
In terms of the socially responsible effects, although narcissism has been increasingly considered in the literature as an important factor when developing CSR strategies (Kim et al., 2018;Petrenko et al., 2016), there is also no agreement regarding how narcissism affects CSR.
On the one hand, a few studies have suggested that narcissistic CEOs, due to their charisma and self-esteem, have a positive impact on company results, because they assume more efficient leadership roles and, consequently, obtain greater business benefits (Deutschman, 2005).In this vein, Salehi et al. (2022) found evidence for the impact of CEOs and TMTs narcissism on firms' relative performance.Campbell et al. (2004) highlight that narcissistic personalities are characterized by the use of strategies which improve and preserve their own positive image.They have a need for constant admiration and attention from others, which positively influences the company's results (Bogart et al., 2004).In this sense, narcissistic CEOs can become the center of attention and be socially admired through CSR: as corporate socially responsible activities are likely to involve a facet of the CEO's positive self-image and to bring positive attention to the CEO (Petrenko et al., 2016).Consequently, more narcissistic CEOs are likely to carry out CSR actions since they see CSR as "an opportunity to enhance their own positive self-image by pursuing socially desirable activities" (Kim et al., 2018, p. 206).In this line, Tang et al. (2018) in a sample of 235 USA firms from S&P 1500 found support for the positive effect of narcissism over CSR while moderated by their industry peers' behaviors in terms of higher/lower investment in CSR.They argue and obtain evidence that narcissist CEOs need constant applause and attention to affirm their inflated positive self-view that can be obtained through CSR.
On the other hand, Petrenko et al. (2016) provide a different view of how narcissism impacts social responsibility.Narcissistic CEOs may hide or avoid showing certain behaviors so as not to lose their reputation; i.e. they use strategies to divert attention from harmful or risky behavior (Buss and Chiodo, 1991;Surroca et al., 2013).Thus, although narcissistic CEOs possess good skills to be efficient managers and can even achieve positive results for their companies, their narcissistic personality can cause harm to their businesses in the long term (Lister, 2004).Campbell et al. (2011) also stated that although narcissistic CEOs are self-confident, extroverted, charming and seek attention and applause, they do not feel empathy and show an abusive, arrogant and dominant behavior.Chatterjee and Hambrick (2007) point out that narcissistic executives are also prone to be sensitive to criticism and threats and are highly competitive, which is counterproductive especially for their stakeholders-employees, clients, suppliers, society, etc.-(Anninos, 2018).In this line, O'Reilly et al. (2018) provide evidence that narcissistic CEOs deny others opinion.In their study, narcissistic CEOs are positively related to enter into lawsuits and long litigations and avoiding the opinion of experts about their potential success.
Other scholars have also analyzed how the manipulative and insensitive nature of narcissistic CEOs influences their decision-making, in that materialism is a crucial aspect for understanding the relationship between narcissism and CSI.Materialism is an intrinsic value of narcissistic people, which predisposes them to focus on obtaining personal profits and gains, even transferring company profits to their own benefit (Campbell and Foster, 2007).This suggests that narcissistic top managers do not prioritize stakeholder concerns and CEO's narcissism and irresponsible behavior interests, but concentrate solely on their own expectations (Wales et al., 2013).Narcissistic CEOs are characterized by materialism, making them prone to derive the maximum profit to the detriment of the needs and expectations of their stakeholders (Chatterjee and Pollock, 2017).As Cragun et al. (2020) state in their review, the definitions of narcissism include a general lack of regard for others and it "refers to a lack of empathy toward others and a tendency to exploit situations and persons for personal gain".
In respect to specific stakeholders, several scholars have stressed how narcissism in CEOs causes damage to employees and other organization members (Abbasi and Amran, 2023;Campbell and Siedor, 2016).In fact, Judge et al. (2006) showed that narcissism is positively linked with intentional damage in workplaces.Similarly, Chen et al. (2013) concluded that narcissism exacerbates the effects of incivility in workplaces, while Nevicka et al. (2011) consider that narcissism also influences the information exchange at the group level, which may be detrimental for employee interests and expectations.In this line, Grijalva and Harms (2014) suggest that narcissistic CEOs are related with counterproductive work conduct, aggressiveness and leadership egoism.Therefore, it seems reasonable to believe that narcissistic CEOs lack the capability to socialize and to understand their stakeholders, creating a toxic work environment, which may have a negative effect on the interests and expectations of stakeholders (Grijalva and Harms, 2014).
In summary, we argue that narcissistic CEOs decision-making may take into account solely their own interests and not those of third parties, as well as ignoring the information received from others.Moreover, they lack empathy and disregard the concerns expressed by other stakeholders increasing the likelihood of ignoring the interests of third parties and ignoring the harmful consequences this behavior can cause.Based on the above arguments, we propose the following hypothesis: H1.The higher the degree of CEO narcissism, the greater the degree of CSI.
CEOs' centralized power and CSI CEO characteristics help to understand firms' actions; however, their influence may vary with the centrality of CEOs in decision-making processes.As the upper echelons theory states, CEO power can be centralized at the apex or it can be shared and it can be more or less monitored and controlled by the board of directors (BoD) (Finkelstein et al., 2009).The relationship between CEO characteristics and CSI may change with the degree to which the CEO centralizes power or, conversely, shares power with other members (Pitcher and Smith, 2001).Pearce (1997) stresses that decentralized power mitigates uncivil behaviors in the members of a company.
When decision-making and power is shared between CEO and BoD, there is access to more information and a wider network where the concerns of stakeholders can be made known.Consequently, there is a greater probability of optimally satisfying stakeholder needs.In that sense, Shafeeq Nimr Al-Maliki et al. (2023) evidence the role of the board in providing information and monitoring in relation to CSR.As Pearce and Manz (2011) consider, individualized decision making will not be aligned with the objectives, interests or needs of the rest of the company members.
CEOs who centralize decision-making are characterized by a minimal predisposition to sharing responsibility with other people and this centralized power may favor using their influence to behave corruptly (Pearce et al., 2008).The limited power of the board will not allow monitoring CEOs behavior and controlling its actions and consequences.Likewise, centralization of power means that there are fewer individuals who hold power in decisionmaking, and thus there would be fewer consensuses in a group (Wong et al., 2011).In this type of power structure, one of the problems that companies have to face is the feeling that most employees consider themselves as being "outside" of the decision-making processes, since they do not regard themselves as participants in these processes in the companies where they carry out their professional lives.As a result, they feel demotivated when it comes to sharing the points of view and opinions of the interest groups with which they interact.
We argue that the centralization of decision-making in a powerful CEO would be related to an absence of heterogeneity and diversity regarding the different interests and will not allow taking into account other points of view.In contrast, more diversity helps to understand the needs and desires of the different interest groups, because their characteristics reflect the wishes and preferences of society (Ayuso and Argandoña, 2007).Brammer et al. (2007) consider that decentralization promote consensual decision-making, respect for inclusion of the interests of people with different expectations, which helps to avoid CSI.
Therefore, we consider that power centralization in CEOs will be related to CSI.
H2.The higher the degree of power centralized in a CEO, the greater the degree of CSI.

Sample and data collection
We use secondary sources to select the companies in our study.Through the Factiva database and Google's search tool, Google News, we draw a sample of the Spanish companies [9] that had caused harm to social, economic or/and environmental dimensions and were being investigated, accused of and/or convicted of crimes in the Spanish Penal Code during the period from 2005 to 2012 [10].The key terms [11]: *corruption, *fraud, *bribery, *money laundering, *misappropriation of funds, *false accounting, *false statements *severe labor exploitation, *sexual harassment, *ecological disasters, *environmental disasters, *illegal financing of political parties * urban crime, *tax evasion, * prevarication, * corporate crime; additionally, we also included companies condemned by the Spanish antitrust court for anti-competitive behaviors identified by the Spanish antitrust law [12].The search resulted in a preliminary sample of 256 companies.
We then identify the CEOs of the preliminary sample.The names and surnames of CEOs were obtained from Bureau Van Dijk' Orbis database, company's website or online news and their profile was obtained through LinkedIn.The final sample was made up by 84 companies.
In our sample, 70% of the companies are SMEs [13].With respect to industries, 37% belong to the manufacturing industry, while 63% are service companies.In the sample, 42.9% of the companies are condemned for only one irresponsible behavior, 13.1% are involved in five or more and 44.1% are linked to two to five irresponsible behaviors.In the case of firms condemned for only one irresponsible behavior, in 66.7% of the cases, this behavior lasted for two years or more (in 27.8% for five years or more).

Variables definition
Dependent variable: degree of CSI.The sample selection includes companies involved in CSI: they intentionally caused harm to others in different degrees.In order to evaluate the degree of CSI ,we follow Armstrong (1977) definition and focus on the perceptions of impartial experts/observers.We focus on ranking the offense itself, e.g.sexual harassment or ecological offense and not on ranking the specific firms.We selected a diverse group of experts [14] in management and law who helped us with their informed opinion, knowledge and experience in these fields (Skjong and Wentworth, 2001).Seventeen experts participated in this study (see Table 1), all of whom were asked to rank the degree of irresponsibility of the different criminal offenses stipulated in the Spanish Penal Code.According to their professional background, the experts belong to three different groups, since CSI can be perceived differently [15].In this line, people with similar characteristics evaluate the same issue in a similar way (Tajfel, 2010).

CEO's narcissism and irresponsible behavior
The perceptions from the unbiased experts were collected through a questionnaire, which contained the types of criminal offenses under review with information about their corresponding legal penalties.Each expert assigned a score, from 1 to 5, to each criminal offense according to their perception of its severity, assigning 1 to less severe criminal offenses and 5 to more severe criminal offenses.The average of the scores of each criminal offense by each group of experts was the optimum solution for measuring their answers.
Prior studies (Keig et al., 2015;Strike et al., 2006) have measured CSI through the Kinder, Lydenberg, Domini Research and Analytics database [16].It considers an overall CSI score that results from adding up a set of binary indicators of concerns from a wide range of CSR dimensions [17].In addition, it does not allow measuring properly the degree of harm (concerns are added as each dimension has the same impact).Our proposal is a first step in

EJMBE
evaluating the degree of harm or offense caused by an act and in distinguishing the harm from the firm causing itthat usually will contaminate the evaluation of the act through firm's image and reputation.Independent variables.
Narcissism.The measurement of narcissism through primary sources is difficult, since heads of companies are reluctant to answer questions regarding narcissism (Cragun et al., 2020).In this sense, Chatterjee and Hambrick (2007) were pioneers in creating a narcissism scale based on secondary sources of information.This scale is made up of five indicators and is widely used in the literature (Cragun et al., 2020;Oesterle et al., 2016).Nonetheless, researchers have recently made an effort to adapt Chatterjee and Hambrick's (2007) narcissism scale to private small and medium companies.In this line, Aabo and Eriksen (2018), have adapt the narcissism scale, using indicators from LinkedIn profiles.Since 65% of our samples are small and medium companies, we followed this approach and used LinkedIn as a source of information to obtain data to measure CEO narcissism.So, adapted from Chatterjee and Hambrick's (2007) and Aabo and Eriksen's (2018) our scale relies on four indicators: (1) skills and endorsement sections, (2) previous job positions, (3) LinkedIn photographs and (4) resume sections (see Table 2).
Each indicator is a dichotomous variable, where (0) reflects the absence of the indicator in the LinkedIn profile and (1) reflects the presence of the indicator in the LinkedIn profile.
CEO power.The centralization of CEO power was measured through three indicators which reflect the structure of the board of directors and its distribution of power: (1) the existence of a unique administrator or a solidary administrator versus the existence of a BoD [18], (2) the duality of responsibilities in the BoD, where the roles of CEO and chair of the board are taken on by the same person and (3) ownership control [19] where one individual shareholder has more than 51% of the shares.These three elements are crucial for understanding the relationship between the board's structure and CEO power and discretion (Finkelstein et al., 2009).
Control variables.Several other variables can potentially influence CSI.Following the upper echelons theory, we use three control variables that are related to the behavior of the CEO: the diversity of the board in terms of the percentage of gender diversity and international diversity and the counterbalance role of family ownership in CEO behavior.
Family business scholars agree that family-owned firms are known for their interest in protecting and preserving their socio-emotional wealth (hereinafter "SEW") (G omez-Mej ıa et al., 2014).Furthermore, family-owned firms are driven by the need to reinforce the legacy of the family's SEW and continuously seek the support and approval of their interest groups  (Berrone et al., 2010).Berrone et al. (2010) stated in their empirical study that family-owned firms are less likely to contaminate the environment, their aim being to protect their SEW and lessen the damage to their stakeholders.We also control for two demographic characteristics regarding the diversity in the board composition.Together with board structure, board composition is a key dimension for understanding its decision-making (Finkelstein et al., 2009).For this reason, we control for gender diversity as the percentage of women on the board as well as for international diversity as the presence of different nationalities on the board because these characteristics can be linked to being more open to different points of view, perspectives and interests, which is enriching (Finkelstein et al., 2009).

Data analysis Measurement model validation
The structural model depicted in Figure 1 was estimated by means of the Partial Least Square Path Modeling (PLS-PM) using SmartPLS 4.0 (Ringle et al., 2022).This approach has minimal demands regarding sample size, relaxes the assumption of multivariate normality needed for maximum likelihood-based structural equation modeling (SEM) estimations and is suitable for applications where strong assumptions cannot be fully met (Hair et al., 2012).
Measurement model properties were evaluated according to the recommendations of Hair et al. (2012) for PLS-PM.We analyzed the reliability, convergent validity and discriminant validity of the measurement model.In respect to internal individual consistency, Table 3 shows that all indicators are significantly associated with their respective constructs (p < 0.01) and their individual standardized loadings are greater than 0.70 or their mean is greater than 0.70 [20] (Bagozzi and Yi, 1988;Chin, 1998), which shows that these indicators are highly reliable.Internal consistency reliability was examined via Cronbach's alpha (CA) and composite reliability (CR).All constructs had CA values above 0.7 and their CR values are superior to 0.83 and all were greater than the threshold of 0.7 (Bagozzi and Yi, 1988).

EJMBE
In respect to convergent validity, it allows measuring if all the items measure the same construct.The average variance extracted (AVE) for each construct was higher than the 0.50 threshold (Fornell and Larcker, 1981), which confirms the convergent validity of the measurement model.Through 5,000 bootstrap samples and a number of cases equal to the 84 valid observations of the original sample, we analyzed the size of the standardized loadings and all were significant at p < 0.01.Discriminant validity states that constructs involved in the analysis are measuring different realities.Table 4 shows first criteria of discriminant validity.It was assessed by checking that the correlation between each pair of constructs was smaller than the square root of the AVE of the implied constructs (Fornell and Larcker, 1981).Supporting this conclusion, Table 4 shows the HTMT ratios, proposed by Henseler et al. (2015).They are always lower than 0.90 which provides additional discriminant validity to our measures, i.e. relations between indicators belonging to a same construct (Monotrait heteromethod-MT) are higher than the ones with different constructs (Heterotrait heteromethod-HT).
In sum, the measurement model has reliability, convergent validity and discriminant validity.

Structural model evaluation
We first analyzed internal collinearity of the constructs of our analysis.Table 5 provides the inner VIF values.All of them are lower than five which reveals no collinearity problems (Hair et al., 2012).
To test our model hypotheses, the t-values of the path coefficients used to establish path significance were obtained by applying nonparametric bootstrapping.Following Hair et al.'s (2012) recommendation, we selected 5,000 bootstrap samples and a number of cases equal to the 84 valid observations of the original sample.Regarding the PLS-PM algorithm settings, individual sign changes were allowed, and a uniform value of 1 was set as an initial value for each of the outer weights.
Table 6 summarizes our results.In terms of direct effects, CEO narcissism does have a significant effect on CSI (β 5 0.336; p < 0.05; f 2 5 0.09).Thus, we find evidence to support Hypothesis 1.The standardized path coefficient is higher than 0.3 showing a strong relationship between CEO narcissism and CSI (Chin, 1998).
Additionally, CEO power has a significant effect on CSI (β 5 0.303; p < 0.01; f 2 5 0.08), which confirms Hypothesis 2. The standardized path coefficient is higher than 0.3 showing a strong relationship between CEO power and CSI (Chin, 1998).
None of the control variables were significant.Table 6 shows the hypothesis testing results.
In terms of the total variance explained, the predictive capacity of the model, as measured by adjusted R 2 , was of 10.5%.Predictive relevance of the model was tested using Stone-Geisser's Q 2 statistic (Geisser, 1974;Stone, 1974), which was obtained via blindfolding with an omission distance of 10.According to Stone (1974), this criterion is adequate if Q 2 is positive.Our Q 2 is 0.101 showing predictive relevance of the relationships between the variables.Power analysis was performed using G*Power 3.1 (Faul et al., 2009), to test whether our sample size guaranteed enough power for the R 2 deviation from zero test which was greater than 85%.

Additional analysis
We conducted some checks to confirm the robustness of our results.To assess construct validity in our measurement of the degree of CSI, we first correlated the overall CSI measure (α 5 0.88) with a measure of the degree of harm based on the proposal of Clark et al. (2022) Mena et al. (2016) that consider that CSI degree is related to the harm caused that is best conceptualized in degrees.In that sense, we read the news of our sample and evaluated the harm varying from the maximum of loss of human life, to the loss of nonhuman life or to the minimum, loss of livelihoods in local communities.Our expectation was corroborated by a significant, positive correlation between the degree of CSI and harm (r 5 0.77, p < 0.01), increasing our confidence in the measure's validity.
We repeated the analyses with a new dependent variable.In order to create a new CSI variable with four indicators (three experts and the degree of harm), we first transform each indicator to a scale from zero to one.The scale has a good CA (α 5 0.94).We then run, through SEM estimations and PLS, the basic model regarding hypothesis 1 and 2 without control variables [21] and our results remain stable providing support for both hypotheses.PLS results provide an adjusted R 2 , of 10.4%.CEO narcissism does have a significant effect on the new measure of CSI (β 5 0.633; p < 0.01; f 2 5 0.10).Thus, we find evidence to support Hypothesis 1.Additionally, CEO power has a significant effect on CSI (β 5 0.482; p < 0.05; f 2 5 0.06), which confirms Hypothesis 2. We ran SEM estimations and the results were maintained with the new dependent variable; narcissism remains a clearly significant variable (β 5 0.156; p < 0.05) and CEO power has only a significance of p 5 0.056.

Discussion and conclusions
The recent increase in corporate scandals has fostered concern among managers and researchers, about CSI and its antecedents, opening up many new research questions.In this respect, the purpose of this article is to contribute to the literature by studying CSI antecedents revealing the key role managers play in their companies' irresponsible behaviors.This is the reason why some scholars have suggested the need to study specific individual traits of CEOs as CSI antecedents (Grijalva and Harms, 2014), which is theoretically and empirically attractive.
In respect to CSI, we add new evidence to the underdeveloped line of research that links individual level antecedents with CSI behaviors (Grijalva and Harms, 2014;Iborra and Riera, 2023).By focusing on the individual level instead of on institutional variables, we respond to the call made by Ghoshal (2005) that CEOs are not free from moral or ethical responsibility for their actions and, as shown, we have found empirical support for this argument.We argue and provide evidence that internal factors belonging to the characteristics of the upper echelons of the firm are key antecedents of CSI.In other words, CSI is not just a matter of good firms in bad context: CEOs do matter.
Our research extends the upper echelons theory arguing that CEOs play a key role in explaining CSI.Specifically, we contribute to the literature arguing and providing evidence that CEO narcissism and CEO power, may be considered CSI antecedents.
Regarding narcissism, previous empirical studies have provided contradictory evidence in respect to its positive or negative consequences (Chatterjee and Hambrick, 2007;Cragun et al., 2020;Salehi et al., 2022).In contrast to researchers who highlighted the bright side of narcissism, in this study, we contribute to research on CEO narcissism by analyzing its darker side in explaining CSI which, to date, has only scarcely being examined.For example, Almaleki et al. (2021) found support for the negative impact of CEO narcissism on the quality of financial statements in a sample of 128 Iran firms from 2012 to 2018.They argue that narcissistic CEOs are likely to deliberately distort information, leading to a lack of disclosure of bad news to stakeholders and manipulating them for achieving support, i.e. causing harm to them.Our study provides support for the idea that narcissistic CEOs ignore the interest of third parties, increasing the likelihood of doing harm.This evidence fits with the findings of O' Reilly et al. (2018) for narcissistic CEOs denying the experts' opinion which implies not taking care and be aware of the consequences of the actions.It is also in line with the proposal CEO's narcissism and irresponsible behavior of Salehi et al. (2022) that highlights as negative dimensions of narcissism their lack of cooperation and low tolerance to criticism.In sum, narcissistic CEOs are individuals who lack empathy and disregard potential concerns increasing the likelihood of doing harm.We provide evidence that this specific trait of CEO narcissism is linked to the degree of CSI.
Our evidence is in line with the statement of Cragun et al. (2020) that narcissistic CEOs can cause great harm.This calls for future research that examines which factors might amplify or mitigate these outcomes.
In addition, our evidence supports that power in corporations is a key issue to be taken into account when examining CSI.Power centralization of the CEOs implies that they are not subject to the control and monitoring of others.They can act not taking into account the points of view, interests or demands of others.They can act without a counterbalance.According to Pearce and Manz (2011), centralized decision-making is not aligned with the objectives, interests or needs of the rest of the members of the company.Moreover, powerful CEOs may ignore the need for accountability and may increase the chances for corruptive behaviors to occur.As Cragun et al. (2020) suggest the effect of CEO narcissism could differ by the context: BoD may communicate to a CEO through diverse means that narcissistic behavior is approved or avoided.Our evidence supports this relationship.
Our study also contributes to the empirical research on CSI.The CSI is a complex issue to study.The absence of empirical studies in this field, except for those made on large public firms or case studies, confirm this difficulty (Iborra and Riera, 2023).Using secondary sources, we develop a new approach to this question.We also provide a way to measure the degree of CSI, and not merely its existence or absence, through a panel of experts that allows us to take into account the diversity of views and perceptions of CSI.Hence, bearing in mind Armstrong's (1977: 1) definition in which he states that "an act is irresponsible if a vast majority of unbiased observers would agree that this is so", we propose a novel method for measuring CSI and evaluating different corporate irresponsible behaviors and their degrees based on the perceptions of an unbiased experts' panel.

Managerial implications
Our research shows that the executives' personality is relevant to explain CSI.Increasing their understanding and awareness of CSI may favor to control and prevent it.
We cannot deny that top managers tend to be narcissistic and they are going to be more in the future.First, because researchers evidence the predominance of narcissistic profiles among CEOs (Cragun et al., 2020); second, because narcissistic individuals tend to become CEOs earlier in their careers (O'Reilly et al., 2018); and, lastly, because the next generations of managers have received a parental education and will live in a social context that will increase their trend to narcissism (Young et al., 2016).So, if narcissism will be there, firms have to look for ways that counterbalance narcissism's dark side.
In this study, we found evidence of the relationship between CEO narcissism and power and CSI.Firms should handle power by setting up diverse and pluralistic structures, through the assessment of TMTs and board compositions.In doing so, collaboration and cooperation would emerge, which may compensate for the effect of narcissistic behaviors.
Finally, this study offers CEOs further guidance on making optimal decisions to prevent CSI.The more narcissistic CEOs are, the greater the need to counterbalance their power to avoid the likelihood of CSI.

Further lines of research and limitations
Our paper fosters the development of new lines of research.We reveal that the role played by CEOs, together with power distribution, have an influence on CSI.However, literature dealing with the upper echelons theory (Finkelstein et al., 2009), suggests that other variables related EJMBE to the board composition, as well as TMT diversity, may also have an impact on decision making.As additional future research lines, we propose to carry out further studies regarding CSI antecedents at the individual level.A future line of insight that has provide some light in related areas as CSR is the relationship between CEOs, TMTs and BoD characteristics in explaining CSI (Shafeeq Nimr Al-Maliki et al., 2023).This will provide light on CSI decisionmaking processes by examining CEOs, TMTs, BoD and their interfaces.
Additional research is also needed to improve knowledge of the role of corporate control and prevention tools to mitigate the likelihood of CSI under narcissistic individuals (Young et al., 2016).This may open future research that carries out a more thorough study on antecedents at diverse levels, since there is still no clear understanding of which variables are antecedents of CSI (Zhao et al., 2014).Multilevel analysis would significantly improve CSI areas, both at the firm level and the individual level (Iborra and Riera, 2023).
Additionally, although family ownership has been studied as a control variable in this work, we stress the importance of focusing on specific family firm characteristics.Family firms are a complex reality, which involves more dimensions than merely considering the percentage of shareholder's equity owned by family members.
This study has some methodological limitations.Specifically, the main limitation of our work comes from the sample size.We obtained a sample of only 84 companies due to limited information on narcissism in SMEs.Different reasons may explain it.On the one hand, Spain is the 11th country with more LinkedIn connections (more than 13 million people) but people with ages between 25 and 34 years old, represent more than a half of the accounts in this platform (Statista, 2021) and are unlikely CEOs at these ages.Only 191,416 registered company pages from more than three million of Spanish companies are at LinkedIn decreasing the likelihood of representation.In addition, managers may have removed their LinkedIn profile in order to eliminate any personal information after CSI is known.
Another limitation comes from the sample selection.We focus on understanding the degree of CSI; so, we took into consideration only companies that caused different types of harm to social, economic or/and environmental dimensions and do it in different degrees.We base this option in our conceptualization of CSI as a standalone construct [22].Future research may devise a more thorough approach improving the understanding of the role of CEOs narcissism and power considering both responsible companies and irresponsible companies in their sample.
3. For example, Rovelli et al. (2023) demonstrate that CEO narcissism as a personality trait offers family firm some important business advantages related to innovation.
4. This idea is in line with Salehi et al. (2021) description of the "dark side" of managers when capable managers may misuse their authority, which leads to manipulation.
5. Also, specific consequences call for attention as Feng et al. (2022) evidence for transaction costs.
6. Clark et al. (2022) argue that defining CSR and CSI as opposite constructs produces a lack of clarity between responsible and irresponsible acts.
7. Intentionality is defined in a broad sense including that the company does it knowing the consequences of the act, or being negligent or reckless (Godfrey, 2005;Clark et al., 2022).
8. However, when breaking down the studies by the different measures used for narcissism, only the ones that used indexes remain significant.
CEO's narcissism and irresponsible behavior CEO's narcissism and irresponsible behavior Figure by author

Figure 1 .
Figure 1.Proposed model On the diagonal: square root of AVE.Below the diagonal: correlations between latent variables.Above the diagonal: HTMT ratios Source(s): Table by the authors