Introduction
How, and to what extent, a CEO’s personal characteristics influence organizational performance is a topic of interest in the business press (Adams, Reference Adams2012; Maccoby, Reference Maccoby2003; Sahadi, Reference Sahadi2018) as well as in academic fields such as accounting (Olsen, Dworkis, & Young, Reference Olsen, Dworkis and Young2014), cross-cultural psychology (Acker & Duck, Reference Acker and Duck2008; Meisel, Ning, Campbell, & Goodie, Reference Meisel, Ning, Campbell and Goodie2016), entrepreneurship (Koellinger, Minniti, & Schade, Reference Koellinger, Minniti and Schade2007), finance (Nelson, Reference Nelson2005), human resource development (HRD) (HRD is defined as “a process of developing and/or unleashing human expertise through organization development and personnel training and development for the purpose of improving performance” [Swanson, Reference Swanson1995, p. 208]) (Godkin & Allcorn, Reference Godkin and Allcorn2009), and strategic management (Hambrick & Mason, Reference Hambrick and Mason1984; Nadkarni & Herrmann, Reference Nadkarni and Herrmann2010). Researchers have traditionally examined CEO demographics such as age, educational level, and functional background experience (Datta & Guthrie, Reference Datta and Guthrie1994; Hambrick & Mason, Reference Hambrick and Mason1984) as this data is relatively easy to collect. More recently, however, researchers have begun to examine the “Black Box” of CEO psychological characteristics such as overconfidence (Brown & Sarma, Reference Brown and Sarma2007; Chen, Crossland, & Luo, Reference Chen, Crossland and Luo2015; Hirshleifer, Low, & Teoh, Reference Hirshleifer, Low and Teoh2012) and its relationship with organizational performance.
Overconfidence refers to one’s tendency to overestimate one’s abilities, knowledge, and certainty about being right (Chen et al., Reference Chen, Crossland and Luo2015; Hiller & Hambrick, Reference Hiller and Hambrick2005). Fischhoff, Koriat, and Lichtenstein (Reference Fischhoff, Koriat and Lichtenstein1980) suggest that overconfident individuals may have “an unwarranted belief in the correctness of one’s answers” (p. 108). A growing stream of literature has linked overconfidence to senior executives and entrepreneurs (e.g., Hirshleifer et al., Reference Hirshleifer, Low and Teoh2012; Koellinger, Minniti, & Schade, Reference Koellinger, Minniti and Schade2007; Malmendier & Tate, Reference Malmendier and Tate2005; Simon & Houghton, Reference Simon and Houghton2003). Research findings suggest that overconfident executives irrationally narrow “error bars” around a prediction, feel that they are “better than average,” and underestimate a situation’s degree of uncertainty (Grinblatt & Keloharju, Reference Grinblatt and Keloharju2009). Overconfident CEOs tend to attribute their decision-making errors to external factors such as bad luck rather than internal factors such as incompetence (Chen et al., Reference Chen, Crossland and Luo2015). Similar to executives at larger organizations, overconfident entrepreneurs may undercapitalize their new venture (Hayward, Shepherd, & Griffin, Reference Hayward, Shepherd and Griffin2006), make overly optimistic forecasts (Invernizzi, Menozzi, Passarani, Patton, & Viglia, Reference Invernizzi, Menozzi, Passarani, Patton and Viglia2017), rely on their perceptions of a situation rather than objective probabilities (Koellinger, Minniti, & Schade, Reference Koellinger, Minniti and Schade2007), and remain unaware of poor performance (Shepherd, Wiklund, & Haynie, Reference Shepherd, Wiklund and Haynie2009).
Despite these seemingly negative attributes, empirical results are not as consistent. Researchers have reported a positive relationship between CEO overconfidence and firm innovation performance measures such as patents citations, citations per patent (Galasso & Simcoe, Reference Galasso and Simcoe2011), and patent counts (Hirshleifer et al., Reference Hirshleifer, Low and Teoh2012). Some researchers have produced negative relationships between CEO overconfidence and firm performance measured as earnings forecast accuracy (Chen et al., Reference Chen, Crossland and Luo2015) and shareholder losses following acquisitions (Hayward & Hambrick, Reference Hayward and Hambrick1997). Adding additional uncertainty are researchers who have reported nonsignificant relationships between CEO overconfidence and forecast accuracy improvement in the presence of greater managerial discretion (Chen et al., Reference Chen, Crossland and Luo2015), and conflicting (Simon & Houghton, Reference Simon and Houghton2003) relationships between CEO overconfidence and pioneering product introductions. Addressing additional issues pertaining to culture, industry, and construct measurement may serve to clarify what we can conclude from the CEO overconfidence–firm performance literature.
Based on our prior research review, we believe there are several research questions in need of attention, which if answered, make significant contributions to the current literatures. First, what is the overall relationship between CEO overconfidence and firm performance? Building on upper echelons theory (Hambrick, Reference Hambrick2007; Hambrick & Mason, Reference Hambrick and Mason1984) and personality theories (Grijalva, Harms, Newman, Gaddis, & Fraley, Reference Grijalva, Harms, Newman, Gaddis and Fraley2015), we aim to resolve conflicting findings and explore how and why CEO overconfidence is associated with firm performance. Our research findings will also be of both theoretical and practical significance for the field of HRD. This CEO trait is known to influence important areas of HRD (e.g., career development, executive coaching, organizational learning, and firm performance) (e.g., Ellinger, Ellinger, Yang, & Howton, Reference Ellinger, Ellinger, Yang and Howton2002; Godkin & Allcorn, Reference Godkin and Allcorn2009; Lundgren, Poell, & Kroon, Reference Lundgren, Poell and Kroon2019). Our research findings may help HRD researchers explore promising future research areas. HRD professionals may also be guided by our research findings to properly coach upper level employees to make effective use of their traits (Berr, Church, & Waclawski, Reference Berr, Church and Waclawski2000; Harland, Reference Harland2003).
The second question we aim to answer is “under what condition(s) might CEO overconfidence influence firm performance?” Chatterjee and Hambrick (Reference Chatterjee and Hambrick2007) and Hiller and Hambrick (Reference Hiller and Hambrick2005) suggest that executive traits are associated with extreme and fluctuating performance – either big wins or big losses. Hambrick (Reference Hambrick2007) notes that there is still much to learn about the conditions under which executives influence firm performance. Engelen, Neumann, and Schmidt (Reference Engelen, Neumann and Schmidt2016) also point out that the executive trait literature is paradoxical, because the existing literature is unclear about whether executive traits have beneficial or detrimental effect on firm performance. To clarify at least some of these paradoxes and to reconcile conflicting findings, we take a contingency approach (Luthans & Stewart, Reference Luthans and Stewart1977) and examine potential CEO overconfidence–firm performance moderators. Such a contingency approach also answers calls from prior studies which indicated research regarding executive traits should go beyond the “good-versus-bad” discussion to a more refined “it depends” perspective (e.g., Fatfouta, Reference Fatfouta2019).
Finally, to answer a call from Hambrick (Reference Hambrick2007), the third question we explore in this meta-analysis is what mechanism translates CEO overconfidence into actions that influence firm performance? Prior research suggests that CEO characteristics are associated with strategic actions which subsequently influence performance (Wang, Holmes, Oh, & Zhu, Reference Wang, Holmes, Oh and Zhu2016). Existing research on CEO overconfidence, however, is dominated by studies examining the direct effect of overconfidence on firm outcomes even though potential mediating mechanisms are routinely discussed (e.g., Chen, Ho, & Ho, Reference Chen, Ho and Ho2014). Building on the resource orchestration literature (Sirmon & Hitt, Reference Sirmon and Hitt2003; Sirmon, Hitt, & Ireland, Reference Sirmon, Hitt and Ireland2007), we expect entrepreneurial orientation (EO) – a multidimensional construct describing an organization’s strategy-making process including dimensions such as product innovation, proactiveness, and risk-taking behavior (Covin & Slevin, Reference Covin and Slevin1991; Rosenbusch, Rauch, & Bausch, Reference Rosenbusch, Rauch and Bausch2013) – to mediate the CEO overconfidence–firm performance relationship. We view EO as the means through which CEOs apply their managerial decision-making style to direct how resources are used in pursuit of competitive advantage and higher performance (Sirmon & Hitt, Reference Sirmon and Hitt2003; Sirmon et al., Reference Sirmon, Hitt and Ireland2007). Thus, EO represents CEOs’ resource utilizing visions as discussed in the resource orchestration literature (Chirico, Sirmon, Sciascia, & Mazzola, Reference Chirico, Sirmon, Sciascia and Mazzola2011; Sirmon & Hitt, Reference Sirmon and Hitt2003; Wales, Patel, Parida, & Kreiser, Reference Wales, Patel, Parida and Kreiser2013). Aside from contributions to the entrepreneurship and strategy literatures, examining EO as a mediator also has far-reaching implications for the field of HRD. How to manage change and innovation is a critical topic in the field of HRD (e.g., Kontoghiorghes, Awbre, & Feurig, Reference Kontoghiorghes, Awbre and Feurig2005). Since EO is a construct closely linked to change and innovation, our research findings may result in a better understanding of how CEO traits influence top executives to manage change and innovation in order to improve firm performance.
Existing knowledge and key questions regarding the relationship between CEO overconfidence and firm performance vis-à-vis the contributions of the present manuscript are presented in Table 1. Our manuscript proceeds as follows: (1) theoretical background, including upper echelons theory and trait activation theory; (2) the CEO overconfidence–firm performance relationship; (3) contextual moderators including national culture and firm industry; (4) EO as a mediator of the CEO overconfidence–firm performance relationship; (5) research methods; (6) results; (7) discussion; (8) and conclusion.
A summary of existing knowledge, key questions, and contributions from the present study
Theoretical background
Upper echelons theory, trait activation theory, CEO personality, and firm performance
Upper echelons theory suggests managerial characteristics may influence firm performance through executives’ strategic decisions (Hambrick, Reference Hambrick2007; Hambrick & Mason, Reference Hambrick and Mason1984). Upper echelons theory builds upon Simon’s (Reference Simon1957) concept of bounded rationality suggesting CEOs are limited in their ability to access, process, and use information. These limitations lead CEOs to rely on personality traits to direct their attention, frame their perceptions, and form interpretations of particular situations or decisions (Hambrick & Mason, Reference Hambrick and Mason1984; Wang et al., Reference Wang, Holmes, Oh and Zhu2016). In particular, CEOs’ psychological traits may guide their decisions and actions in a given competitive environment, thus influencing firm performance (Hambrick, Reference Hambrick2007; Hambrick & Mason, Reference Hambrick and Mason1984; Hiller & Hambrick, Reference Hiller and Hambrick2005). Until recently, the process through which executive psychological characteristics influence firm performance has been a “Black Box.” Sirmon et al. (Reference Sirmon, Hitt and Ireland2007) have opened this “Black Box” by describing how managers structure, bundle, and leverage firm resources to create competitive advantage and increase firm performance. According to Penrose (Reference Penrose1959), how executives choose to use resources is at least as important as the resources possessed by the firm. Zott (Reference Zott2003) provides support for this perspective by demonstrating that firms with similar resources may produce very different performance outcomes depending on how managers choose to use available resources. The key here is that managers make decisions regarding how to structure, bundle, and leverage available resources and that these decisions can impact firm performance. From an upper echelons perspective then, it is important to understand managerial characteristics, such as psychological traits, because these traits influence managers’ decisions regarding how to structure, bundle, and leverage available resources in search of competitive advantage and superior performance.
Historically, upper echelons researchers have focused their attention on executives’ demographic characteristics (e.g., age, experience, and education, etc.) because it is difficult to directly measure executives’ psychological characteristics (Hambrick & Mason, Reference Hambrick and Mason1984; Wang et al., Reference Wang, Holmes, Oh and Zhu2016). More recently, however, scholars have shifted focus from demographic characteristics to psychological characteristics by applying either direct psychological measures or indirect indicators to capture CEO psychological characteristics (Wang et al., Reference Wang, Holmes, Oh and Zhu2016). One CEO psychological characteristic that has received significant attention is CEO overconfidence (Chen et al., Reference Chen, Crossland and Luo2015).
Overconfident executives are generally viewed as harmful to their organizations (e.g., Adams, Reference Adams2012; Sahadi, Reference Sahadi2018). For example, an overconfident CEO such as Steve Jobs may well be a successful innovator, but his success is often tempered by product failures such as the Apple III, which Jobs insisted have no fans or air vents to minimize noise but which also had the effect of overheating the unit to the extent that connections between chips would melt (Isaacson, Reference Isaacson2013). Interestingly, despite these overwhelmingly negative general views, theoretical arguments and empirical evidence regarding the CEO overconfidence–firm performance relationship are inconclusive suggesting a potential disconnect between public perception and reality. For example, one literature stream theorizes a positive relationship between CEO overconfidence and firm performance. This literature suggests that trait overconfidence should influence CEOs to pay attention to positive cues, have favorable views toward opportunities, and identify more opportunities; in addition, overconfident CEOs generally feel confident and optimistic, are more decisive, focus less on their limitations, experience less anxiety, and set higher goals. Additionally, researchers have reported positive relationships between CEO overconfidence and innovation which can lead to first-mover advantages and competitive advantage (Galasso & Simcoe, Reference Galasso and Simcoe2011; Hirshleifer et al., Reference Hirshleifer, Low and Teoh2012; Tang, Li, & Yang, Reference Tang, Li and Yang2015). Thus, because overconfident CEOs are more likely to have favorable views toward opportunities and set higher goals, they may make decisions to utilize organization resources in bold and risky ways, thus potentially leading to greater firm performance (Sirmon et al., Reference Sirmon, Hitt and Ireland2007; Sirmon, Hitt, Ireland, & Gilbert, Reference Sirmon, Hitt, Ireland and Gilbert2011; Wang et al., Reference Wang, Holmes, Oh and Zhu2016).
A second literature stream proposes a negative relationship between CEO overconfidence and firm performance. Overconfident CEOs consistently feel they are “better than average,” underestimate uncertainties, overestimate their acumen, and ignore criticisms and corrective feedback from others (Chen et al., Reference Chen, Crossland and Luo2015). Due to their arrogance, self-admiration, and strong sense of personal ability, they may utilize resources inefficiently such as by pursuing grand projects that are flawed, naïve, and/or foolish which may ultimately harm firm performance (Gerstner, König, Enders, & Hambrick, Reference Gerstner, König, Enders and Hambrick2013; Hiller & Hambrick, Reference Hiller and Hambrick2005; Sirmon et al., Reference Sirmon, Hitt and Ireland2007, Reference Sirmon, Hitt, Ireland and Gilbert2011).
A final literature stream suggests CEO overconfidence may create either “big wins” or “big losses” and there may not be a systematic positive or negative relationship with firm performance (Hiller & Hambrick, Reference Hiller and Hambrick2005). On the positive side, if assumptions and calculations are correct, overconfident CEOs who are prone to make grandiose, bold, and speedy decisions to invest firm resources in activities such as large acquisitions, large-scale new product launches, and aggressive international expansion will lead their firms to first-mover advantage and superior performance (Eisenhardt, Reference Eisenhardt1989; Hiller & Hambrick, Reference Hiller and Hambrick2005). Due to these same grandiose, bold, and speedy decisions, however, resources may be incorrectly or inefficiently utilized leading to below-average performance, substantial losses, and firm failure (Finkelstein, Reference Finkelstein2004; Hiller & Hambrick, Reference Hiller and Hambrick2005; Invernizzi et al., Reference Invernizzi, Menozzi, Passarani, Patton and Viglia2017).
As previously stated, the predicted relationship between CEO overconfidence and firm performance is inconsistent across different theoretical perspectives. Due to this reason, we choose to provide a research question rather than formulate a directional hypothesis to examine the effect of this CEO trait on firm performance.
Research Question 1: (How) Does CEO overconfidence relate to firm performance?
CEO overconfidence–firm performance moderators
Theoretically, the CEO overconfidence–firm performance literatures need to be reconciled; otherwise, researchers and managers may be unable to make useful, actionable decisions regarding this relationship. CEOs have the authority to spend millions or billions of dollars; thus, if our understanding with respect to CEOs’ overconfidence remains unclear, firms may be placed in precarious financial and/or strategic situations. Comprehending why and how CEO overconfidence leads to better or worse performance may also help stakeholders and CEOs mitigate the corrosive effects of trait overconfidence and amplify their beneficial effects. Drawing on contingency theory (Luthans & Stewart, Reference Luthans and Stewart1977), we suggest environmental influences may limit or enhance the effect of CEO overconfidence on decision-making. A contingency approach is designed specifically to consider relationships among environmental, management, and performance relationships (Luthans & Stewart, Reference Luthans and Stewart1977). In this research, we are especially concerned with how environmental factors moderate the relationship between CEO overconfidence and firm performance. More specifically, we examine how environmental contingency variables (e.g., national cultures and industry) influence the relationship between CEO overconfidence and firm performance (Engelen et al., Reference Engelen, Neumann and Schmidt2016; Gerstner et al., Reference Gerstner, König, Enders and Hambrick2013; Maccoby, Reference Maccoby2003). Our choice of cultural and industry moderators is guided by the principle that CEOs decisions are embedded in multiple external contexts including national culture and industry characteristics as reflected by the person–environment perspective.
National culture
Trait activation theory suggests that traits are more predictive of outcomes when a context (i.e., job, social, organizational, and societal situations) contains trait-relevant cues that activate trait expression and prompt one to behave in ways consistent with the cues (Tett & Burnett, Reference Tett and Burnett2003). For example, employees’ conscientiousness is more likely to activated into responsible and diligent behaviors (e.g., getting assignments done on time and meeting performance goals) if their leaders monitor and reward goal achievement. Consistent with trait activation theory, an approach motivation orientation (Elliot & Covington, Reference Elliot and Covington2001; Foster & Brennan, Reference Foster and Brennan2011) also supports a set of cultural contextual cues that may influence individuals to act in accordance with cultural cues. National cultural contexts may contain trait-relevant cues that trigger trait expression because national cultures form social norms that define acceptable behaviors within a society (DiMaggio & Powell, Reference DiMaggio and Powell1983; Scott, Reference Scott2001) and stipulate values that are rewarded (Oh et al., Reference Oh, Guay, Kim, Harold, Lee, Heo and Shin2014). For CEOs, this culture effect may even dictate whether they are able to make decisions consistent with overconfidence such as investing in risky projects (Li, Sun, Taris, Xing, & Peeters, Reference Li, Sun, Taris, Xing and Peeters2021; Rabl, Jayasinghe, Gerhart, & Kühlmann, Reference Rabl, Jayasinghe, Gerhart and Kühlmann2014). Thus, cultural values influence and constrain individual actions and perspectives (Huang & Van De Vliert, Reference Huang and Van De Vliert2003), thereby influencing executive decision-making and behavior (Crossland & Hambrick, Reference Crossland and Hambrick2007; Davis, Schoorman, & Donaldson, Reference Davis, Schoorman and Donaldson1997; Foster & Brennan, Reference Foster and Brennan2011).
For example, in countries scoring high on assertiveness, individualism (low collectivism), and performance orientation (emphasizing agency, achievement, and independence – positive self-concepts/trait-relevant cues for society), CEOs’ positive self-concepts/traits such as overconfidence are more strongly activated into trait-relevant behaviors (reflected in EO behaviors) which, in turn, may positively affect firm performance. These are important research questions for cross-cultural management, as our study attempts to address whether and how country-level cultural values are related to the micro-foundation of firm performance (i.e., the role of CEO traits in firm performance). Since national culture may also condition the association between CEO overconfidence and firm performance by constraining executive decision-making and behavior (Crossland & Hambrick, Reference Crossland and Hambrick2007; Davis et al., Reference Davis, Schoorman and Donaldson1997), it is important to investigate how GLOBE cultural dimensions (i.e., performance orientation, uncertainty avoidance, humane orientation, institutional collectivism, in-group collectivism, assertiveness, gender egalitarianism, future orientation, and power distance) moderate the CEO overconfidence–firm performance relationship. We chose GLOBE cultural dimensions, because the GLOBE cultural project offers the most recent and comprehensive cultural classifications of countries developed by a global team of researchers and has been widely applied in prior studies (e.g., Li et al., Reference Li, Sun, Taris, Xing and Peeters2021; Saeed, Yousafzai, & Engelen, Reference Saeed, Yousafzai and Engelen2014).
Research Question 2: (How) Do national cultural dimensions moderate the relationship between CEO overconfidence and firm performance?
Industry
CEO decision-making studies have frequently noted industry environment as a CEO characteristic–firm performance relationship moderator (Crossland & Hambrick, Reference Crossland and Hambrick2007; Finkelstein & Hambrick, Reference Finkelstein and Hambrick1990; Haleblian & Finkelstein, Reference Haleblian and Finkelstein1993; Hambrick, Geletkanycz, & Fredrickson, Reference Hambrick, Geletkanycz and Fredrickson1993). For example, highly dynamic industries, such as high-technology industries, are characterized by the speed and magnitude of change in consumer preferences and behaviors and by their level of market uncertainty; thus, these conditions complicate a CEO’s job, present them with information processing challenges, and decrease their ability to influence firm performance (Davis, Eisenhardt, & Bingham, Reference Davis, Eisenhardt and Bingham2009; Haleblian & Finkelstein, Reference Haleblian and Finkelstein1993). However, it is also possible that overconfident CEOs may underestimate industry limitations and/or overestimate their ability to overcome these limitations. In this case, overconfident CEOs may be more likely to pursue aggressive growth or innovation strategies, because they are more likely to underestimate the likelihood of failure (Galasso & Simcoe, Reference Galasso and Simcoe2011).
Prior research has also provided support for CEOs having less decision-making latitude in more munificent, low discretion industries, such as those found in low-technology industries, and that CEO characteristics may be less strongly associated with firm performance in these industries (Chen et al., Reference Chen, Crossland and Luo2015). In low-technology industries where uncertainty is low and change is predictable, CEOs have few strategic decisions to make and thus have fewer opportunities to over or under estimate their abilities (Finkelstein & Boyd, Reference Finkelstein and Boyd1998). It is possible then that overconfident and less confident CEOs may be similarly influenced by a low-technology industry environment. Additionally, some notable CEO trait studies (Brown & Sarma, Reference Brown and Sarma2007; Chatterjee & Hambrick, Reference Chatterjee and Hambrick2007; Gerstner et al., Reference Gerstner, König, Enders and Hambrick2013) have not considered industry contexts. We therefore believe it is theoretically reasonable to examine whether industry characteristics (i.e., high vs. low technology) moderate the CEO overconfidence–firm performance relationship. We chose industry technology level as a moderator because industry technology level is widely known as a major contingency factor in affecting top executives’ behaviors, top executives’ decisions, firm structure, firm behaviors, and firm performance (Ang, Reference Ang2008; Makri, Lane & Gomez-Mejia, Reference Makri, Lane and Gomez-Mejia2006).
Research Question 3: (How) Do industry characteristics (high vs. low technology) moderate the relationship between CEO overconfidence and firm performance?
CEO overconfidence–firm performance relationship mediator
Researchers have examined how CEO characteristics influence organization performance (Dalton, Daily, Ellstrand, & Johnson, Reference Dalton, Daily, Ellstrand and Johnson1998; Haleblian & Finkelstein, Reference Haleblian and Finkelstein1993; Simsek, Reference Simsek2007), yet less is known regarding processes through which CEO characteristics influence firm strategies that ultimately influence performance (Sirmon & Hitt, Reference Sirmon and Hitt2003). Resource orchestration (Sirmon et al., Reference Sirmon, Hitt, Ireland and Gilbert2011) suggests firm resources (including CEO characteristics such as overconfidence) are not sufficient to explain firm performance differences. Rather, it is important to understand how and why CEOs might use resources at their disposal to influence performance (Sirmon et al., Reference Sirmon, Hitt and Ireland2007). EO is a key manifestation of how CEOs use firm resources. An excellent example of how firm resources are used to impact firm performance is the work of Wales, Patel, and Lumpkin (Reference Wales, Patel and Lumpkin2013) who reported that EO partially mediates the relationship between CEO narcissism and firm performance. This finding suggests that EO may be a key mechanism through which CEO traits ultimately influence firm performance. EO is a multidimensional construct describing an organization’s strategy-making process including dimensions such as product innovation, proactiveness, and risk-taking (Covin & Slevin, Reference Covin and Slevin1991; Rosenbusch et al., Reference Rosenbusch, Rauch and Bausch2013). We expect EO to mediate the CEO overconfidence–firm performance relationship because EO is the means by which CEOs apply their managerial decision-making style to direct how resources are used. Thus, EO represents CEOs’ resource mobilizing visions as described in the resource orchestration literature (Chirico et al., Reference Chirico, Sirmon, Sciascia and Mazzola2011; Sirmon & Hitt, Reference Sirmon and Hitt2003; Sirmon et al., Reference Sirmon, Hitt and Ireland2007). For example, overconfident CEOs are inclined to take aggressive and bold strategic actions to aggrandize themselves and to attract attention from others which coincide with the key characteristics of EO such as new product innovation and risk-taking (Lumpkin & Dess, Reference Lumpkin and Dess1996; Wales, Gupta, & Mousa, Reference Wales, Gupta and Mousa2013). Empirical evidence supports this view as overconfident CEOs have made risky decisions to invest resources in innovative activities (Hirshleifer et al., Reference Hirshleifer, Low and Teoh2012) and acquisitions (Brown & Sarma, Reference Brown and Sarma2007). Irrespective of resource limitations, risks, and/or feedback from others, overconfident CEOs are likely to lead their organization to experiment with high-risk, high return ideas to seek admiration from their followers and to affirm their uniqueness and invincibility (Wales et al., Reference Wales, Gupta and Mousa2013). Additionally, overconfident CEOs may underestimate the likelihood of failure and therefore make riskier decisions (Galasso & Simcoe, Reference Galasso and Simcoe2011). Taken together, overconfident CEOs are prone to utilize firm resources in support of EO characteristics such as risk-taking and innovation to impact firm performance, thus suggesting EO may mediate the relationship between CEO overconfidence and firm performance.
Research Question 4: (How) Does EO mediate the relationship between CEO overconfidence and firm performance?
Method
Literature search
To capture all relevant literature, we utilized a series of keywords (and several variations of them) for our search which are comprised of chief executive officer(s), CEO(s), executive(s), managerial, overconfidence, overconfident, entrepreneurial orientation, strategic orientation, performance, growth, and profitability/profit(s). We searched electronic databases, and this database list consisted of ABI/INFORM, EBSCO Host, Google Scholar, ProQuest Dissertations and Theses, PsycNET, ScienceDirect, and Web of Science. Google Scholar article search captured both published studies and unpublished studies (e.g., unpublished working papers and conference papers) as long as they are available online, thus further ensuring the comprehensiveness of our literature search. We searched relevant journals in accounting, entrepreneurship, finance, management, psychology, and strategic management, and this journal list consisted of Academy of Management Journal, Administrative Science Quarterly, Entrepreneurship Theory and Practice, Journal of Accounting Research, Journal of Applied Psychology, Journal of Business Venturing, Journal of Finance, Journal of Financial Economics, Journal of Management, Journal of Management Studies, Journal of Organizational Behavior, Journal of Personality and Social Psychology, Leadership Quarterly, Organizational Behavior and Human Decision Processes, Organization Science, Personnel Psychology, Strategic Entrepreneurship Journal, Strategic Management Journal, and The Accounting Review. We also searched pertinent conference papers.
Inclusion criteria and coding procedures
Studies were included in our meta-analysis when they met the following criteria. First, the studies must be empirical and quantitative; all qualitative studies were excluded. Second, studies must measure the focal CEO trait (CEO overconfidence). Third, studies had to report at least a correlation coefficient between CEO overconfidence and firm performance and/or between CEO overconfidence and EO. We used Lipsey and Wilson’s (Reference Lipsey and Wilson2001) method to perform effect size conversions when studies failed to report correlation matrices but reported sufficient statistics that allow effect size conversions. After applying these inclusion criteria to winnow the articles identified in the initial search, we found 50 eligible samples (N = 269,512) for the CEO overconfidence–firm performance relationship. References for the included studies are provided in supplemental materials.
In addition to coding effect sizes for the main effect as stated above, we also coded a set of moderators, including national cultural dimensions, industry characteristics, firm performance measurements, and CEO overconfidence scales. Since House, Hanges, Javidan, Dorfman, and Gupta (Reference House, Hanges, Javidan, Dorfman and Gupta2004) provided scores for national cultural dimensions, we coded national cultural dimensions of each included study according to House et al.’s (Reference House, Hanges, Javidan, Dorfman and Gupta2004) GLOBE study based on the country of each included study. While the way that we coded cultural dimensions is an acceptable practice and consistent with the approaches in previous meta-analyses (e.g., Oh et al., Reference Oh, Guay, Kim, Harold, Lee, Heo and Shin2014) this approach still relies on approximation because it does not take within-culture heterogeneity into consideration by implicitly assuming within-culture homogeneity. We coded nine GLOBE cultural dimensions, including performance orientation, uncertainty avoidance, humane orientation, institutional collectivism, in-group collectivism, assertiveness, gender egalitarianism, future orientation, and power distance. We performed a mean split on the cultural scores and assigned the studies having cultural scores above the mean as the high-level group and the studies having cultural scores below the mean as the low-level group.
Measurement methods for our focal variables (CEO overconfidence and firm performance) vary across studies, suggesting the need to examine measurement methods as a moderator because measurement heterogeneity may influence meta-analytic effect sizes (Miao, Rutherford, & Pollack, Reference Miao, Rutherford and Pollack2017b). For example, many studies used the CEO overconfidence measure developed by Malmendier and Tate (Reference Malmendier and Tate2005). There are studies, however, that adapted Malmendier and Tate’s (Reference Malmendier and Tate2005) measurement methods to their specific study context. These alternative approaches leave researchers without empirical evidence as to whether these measurement differences provide consistent or different results. Similarly, measurement of our outcome of interest, firm performance, varies across studies. The important role firm performance plays for researchers and business leaders warrants close attention to how performance is conceptualized and measured (Venkataraman & Ramanujam, Reference Venkataraman and Ramanujam1986). We examine both market and accounting-based performance measures as some researchers may have a personal or domain-specific preference (Hax & Majluf, Reference Hax and Majluf1984). Consistent with our reasoning for examining different CEO overconfidence measures, alternative approaches to measuring firm performance leave little empirical evidence as to whether measurement differences provide consistent or different results (Junni, Sarala, Taras & Tarba, Reference Junni, Sarala, Taras and Tarba2013). To be consistent with prior research (e.g., Stam, Arzlanian, & Elfring, Reference Stam, Arzlanian and Elfring2014; Wang et al., Reference Wang, Holmes, Oh and Zhu2016), we coded firm performance into the categories of profitability, growth, and market-based performance. Profitability includes studies that used profits as an indicator of performance such as various types of accounting-based measures of profitability. Growth includes performance measures such as sales growth and asset growth. Market-based performance includes performance measures such as Tobin’s Q and shareholder return. We coded CEO overconfidence scales into Malmendier & Tate, adapted Malmendier & Tate, and other measures. Some authors revised the original Malmendier & Tate’s scale by altering, deleting, and/or adding items from the original scale to generate adapted scales. This was done typically with the goal of improving the original scale and/or to tailor the scale to the research contexts. The category of other measures refers to scales that have little or no link with the original Malmendier & Tate’s scale.
We followed approaches recommended by Stam et al. (Reference Stam, Arzlanian and Elfring2014) to classify the included studies into high- and low-technology industries. High-technology industries are characterized by greater technological turbulence and higher dynamism (Stam et al., Reference Stam, Arzlanian and Elfring2014). Examples of high-technology industries include electronics, while examples of low-technology industries include banks, manufacturing, and restaurants. Two coders independently coded each included study. The initial coding agreement was 95%. Coding disagreements were resolved via discussion and 100% consensus was achieved after discussion.
Meta-analytic procedures
We performed random-effects psychometric meta-analysis (Schmidt & Hunter, Reference Schmidt and Hunter2015) to synthesize the literature. We used two-tailed significance tests to examine nondirectional research questions in our study. As discussed in Cho and Abe (Reference Cho and Abe2013), using two-tailed significance testing is at least more conservative than one-tailed significance testing. The input correlations were weighted by sample size when being synthesized via meta-analysis, which means that larger N correlations were given greater weight than small N correlations if all other things are equal. In line with prior research (e.g., Dalton et al., Reference Dalton, Daily, Ellstrand and Johnson1998), we used a conservative 0.8 reliability estimate for the objectively measured variables when correcting for measurement errors. We corrected for measurement errors in both independent and dependent variables for each individual effect size. We calculated 
$ \hat{\bar{\rho}}$ (corrected sample-size-weighted mean correlation) in addition to r̅ (uncorrected sample-size-weighted mean correlation). We calculated corrected 95% confidence interval to assess the statistical significance of effect sizes. An effect size is considered to be statistically significant at 0.05 level when a corrected 95% confidence interval of an effect size does not include zero. Given the use of nondirectional hypotheses (research questions), we used a two-tailed significance level of 0.05 (Type I error), which is closely aligned with the use of 95% confidence interval. That is, if the 95% confidence interval does not include zero, we are at least 95% confident that the expected relationship is unlikely to be zero (either positive or negative), thus statistically significant. However, if the 95% confidence interval includes zero, we are less than 95% confident that the expected relationship is non-zero, thus statistically nonsignificant. Both Varart% statistic and corrected 80% credibility interval were computed to examine heterogeneity in effect sizes and the potential existence of moderators. A meta-analytic distribution may be influenced by moderators if statistical artifacts account for less than 75% of the variance in effect sizes (i.e., Varart% < 75%).
We examined moderator effects by performing subgroup analyses (to evaluate the statistical significance of between-group differences in effect sizes) and conducted meta-analytic path analysis to analyze the mediating role of EO (Bergh et al., Reference Bergh, Aguinis, Heavey, Ketchen, Boyd, Su and Joo2016). The path analyses were performed with maximum likelihood estimation. Relative to bivariate meta-analysis, this technique is preferable as it computes effect sizes that control for other variables in the model and provides the degree of model fit. In addition, it also allows us to test mechanisms in the “Black Box” (i.e., mediating mechanisms) and to compare different nested theoretical models, which are conducive to theory development and theory testing. The harmonic mean sample size was computed because sample sizes differed across the cells in the meta-analytically derived correlation matrices and using harmonic mean sample size is a conservative approach (when sample sizes vary greatly) due to less weight given to large samples (e.g., Colquitt, Scott, & LePine, Reference Colquitt, Scott and LePine2007; Landis, Reference Landis2013; Miao, Humphrey, & Qian, Reference Miao, Humphrey and Qian2018; Viswesvaran & Ones, Reference Viswesvaran and Ones1995).
Results
Main and moderator effects
The main effect for the relationship between CEO overconfidence and firm performance is shown in Table 2. The corrected correlation between CEO overconfidence and firm performance is 0.13, which is positive and statistically significant with a 95% confidence interval ranging from 0.09 to 0.17 (see Table 2). Moderators may exist if statistical artifacts explain less than 75% of the variance in the meta-analytic correlations (Schmidt & Hunter, Reference Schmidt and Hunter2015). The CEO overconfidence–firm performance meta-analytic distribution has a small Varart% value that is far less than 75% (see Varart% column in Table 2), thus indicating the potential existence of moderators and justifying moderator testing.
CEO overconfidence and firm performance meta-analysis results
Note. k = number of independent samples; N = sample size; r̅ = uncorrected sample-size-weighted mean correlation; SDr = sample-size-weighted standard deviation of observed mean correlations; 
$ \hat{\bar{\rho}}$ = corrected sample-size-weighted mean correlation; SDρ = sample-size-weighted standard deviation of corrected mean correlations; CI LL and CI UL = lower and upper bounds of corrected 95% confidence interval; CV LL and CV UL = lower and upper bounds of corrected 80% credibility interval; Varart% = percent of variance in 
$ \hat{\bar{\rho}}$ explained by statistical artifacts; Sig. Dif. = significant difference. Letters in this column correspond to the letters in rows and indicate that effect sizes significantly differ from one another at 0.05 level. The sign “-” shows there is no significant between-group difference. Subgroup analyses are conducted to assess the statistical significance of between-group difference in effect sizes.
Regarding the first set of moderators – national cultural dimensions, we found that eight cultural dimensions are statistically significant moderators. For instance, subgroup analysis demonstrated that the relationship between CEO overconfidence and firm performance is stronger in high assertive cultures (0.15) than in low assertive cultures (0.05). We repeated the same procedure to analyze each of the cultural dimension moderators and tabulated results in Table 3. The last column in Table 2 also shows moderator testing results.
Summary of meta-analytic results for the CEO Overconfidence–firm performance relationship
With regard to the CEO overconfidence–firm performance relationship, this relationship does not vary across low- versus high-technology industries, firm performance measures, overconfidence measures, and low versus high humane orientation cultures. Specific to our industry results, we speculate that overconfident CEOs may self-select into high-technology industries where their overconfidence and subsequent decision-making can maximize “big wins” (Hiller & Hambrick, Reference Hiller and Hambrick2005). Overconfident CEOs are prone to make grandiose, bold, and speedy decisions and these decisions are most likely associated with “big wins” in high-technology industries as opposed to low-technology industries where strategies and performance remain relatively stable over time (Chen et al., Reference Chen, Crossland and Luo2015; Eisenhardt, Reference Eisenhardt1989; Hiller & Hambrick, Reference Hiller and Hambrick2005). Nevertheless, the relationship between CEO overconfidence and firm performance is stronger in high assertive, low institutional collectivistic, low in-group collectivistic, high future-oriented, high gender egalitarian, high performance-oriented, low power distance, and low uncertainty avoidance cultures.
Mediator effect
We performed meta-analytic path analysis to analyze the mediating effect of EO. We analyzed whether EO mediates the relationship between CEO overconfidence and firm performance (see Fig. 1). We found that the standardized path coefficients from CEO overconfidence to EO (β = 0.20, p < 0.001) and from EO to firm performance (β = 0.27. p < 0.001) are statistically significant. In addition, the indirect effect from CEO overconfidence to firm performance is statistically significant (indirect effect = 0.05, p < 0.001). Finally, model fit indices also show good model fit (χ 2[1] = 2.754 (p = 0.097), CFI = 0.964, NFI = 0.947, RMSEA = 0.064, SRMR = 0.031). In sum, we found that EO fully mediates the relationship between CEO overconfidence and firm performance. We chose the full mediation model for two reasons. First, the full mediation model is more parsimonious than the partial mediation model. Second, the full mediation model has good model fit and nearly nothing is lost after the removal of the path from CEO overconfidence to firm performance. Hence, a more parsimonious model – the full mediation model – is chosen (Berry, Lelchook & Clark, Reference Berry, Lelchook and Clark2012).
Meta-analytic path analysis results
Publication bias analyses
Publication bias (also known as the file drawer effect) represents the degree to which studies with lower effect sizes (correlations) and nonsignificant results are suppressed from publication (put in the file drawer and unpublished/unavailable). To examine the degree to which our meta-analytic results are affected by publication bias, we used Begg and Mazumdar’s (Reference Begg and Mazumdar1994) rank correlation test and Egger’s test of the intercept (Egger, Smith, Schneider, & Minder, Reference Egger, Smith, Schneider and Minder1997). Begg and Mazumdar’s rank correlation test indicates the absence of publication bias when the rank order correlation is statistically nonsignificant (Kepes, Banks, & Oh, Reference Kepes, Banks and Oh2014). Egger’s test of the intercept utilized the intercept to examine the presence of publication bias. Publication bias is considered to be absent if the intercept is statistically nonsignificant (Kepes, Banks, & Oh, Reference Kepes, Banks and Oh2014).
Regarding the distribution of the CEO overconfidence–firm performance relationship, Begg and Mazumdar’s rank correlation test yields a correlation of 0.05. This correlation is statistically nonsignificant at the 0.05 level, which shows that publication bias may not exist. Similarly, Egger’s test of the intercept produces an intercept of −3.94, which is also statistically nonsignificant at the 0.05 level. Both tests demonstrated the absence of publication bias for the distribution of the CEO overconfidence–firm performance relationship. In sum, the meta-analytic distribution is robust and is not noticeably influenced by publication bias.
Discussion
Theoretical implications
Researchers studying CEO trait performance effects have reported varied results (e.g., Engelen et al., Reference Engelen, Neumann and Schmidt2016) perhaps because these CEO traits result in extreme firm performance – either big wins or big losses (Hiller & Hambrick, Reference Hiller and Hambrick2005). This study’s purpose then, was to examine the relationship between CEO overconfidence and firm performance. In doing so, our aim was to answer three research questions: (1) is there an overall relationship between CEO overconfidence and firm performance? (2) do cross-cultural and industry conditions moderate the relationship between CEO overconfidence and firm performance? and (3) is EO a mediator of the CEO overconfidence–firm performance relationship? To answer these questions, we meta-analytically assessed the CEO overconfidence literature.
In response to our first research question (is there an overall relationship between CEO overconfidence and firm performance?), our finding demonstrates that CEO overconfidence has a weak, yet statistically significant, positive relationship with firm performance. This result provides some meta-analytic support for upper echelons theory (Hambrick, Reference Hambrick2007; Hambrick & Mason, Reference Hambrick and Mason1984) and addresses a paradox in the executive traits literature (e.g., Engelen et al., Reference Engelen, Neumann and Schmidt2016). Further, our finding extends prior HRD literatures by providing evidence that personality traits (e.g., overconfidence) influence executive behaviors which subsequently affect firm performance (e.g., Berr, Church, & Waclawski, Reference Berr, Church and Waclawski2000; Harland, Reference Harland2003). Despite its small effect on firm performance, we still consider this effect to be practically important. First of all, on a conceptual level, the absolute practical significance of our effect size is reflected by the fact that our criterion of interest is firm (vs. employee or group) performance, which is not only difficult to predict and influence because it is also affected by many extraneous variables which we cannot easily predict (Park & Shaw, Reference Park and Shaw2013), such as economic conditions and pandemic, but also has societal-level implications (for example, when a firm performs well, they tend to invest more, thus creating more jobs and contributing to economic growth and welfare of society in general). Thus, as discussed in Wang et al. (Reference Wang, Holmes, Oh and Zhu2016), “because the outcome is firm financial performance, even “small” effect sizes can have practical significance” (p. 826). Second, typical effect size benchmarks (0.10 = small, 0.30 = medium, 0.50 = strong) have almost no similarity to actual findings in the field (which tend to much lower than the typical rule-of-thumbs). And effect size distributions and benchmarks vary across different research domains or the prediction of different focal outcomes (Bosco, Aguinis, Singh, Field, & Pierce, Reference Bosco, Aguinis, Singh, Field and Pierce2015). Relatedly, these effect size benchmarks and rule of thumbs are often based on employee- or group-level findings, and, thus, they are not directly applicable to firm-level findings. With all these in mind, one way to better demonstrate the absolute effect size of our findings is using an alternative effect size. As discussed in Wang et al. (Reference Wang, Holmes, Oh and Zhu2016, pp. 826-827), according to binomial effect size displays (BESD; Rosenthal & Rubin, Reference Rosenthal and Rubin1982), the uncorrected mean correlation of 0.10 between CEO overconfidence (assuming there are dichotomies like high versus low levels of overconfidence) and firm performance (assuming successful versus unsuccessful dichotomies) means that companies with CEOs having higher levels of overconfidence have 10% higher success rates than those of other companies (55% success rates vs. 45% success rates). Since the focus of our study is how CEO overconfidence influences firm performance, we compare our result with other meta-analyses that examine either CEOs or top management teams to allow for a suitable comparison because effect size distributions and benchmarks vary across different research domains or the prediction of different focal outcomes (Bosco et al., Reference Bosco, Aguinis, Singh, Field and Pierce2015). For example, according to Bergh et al. (Reference Bergh, Aguinis, Heavey, Ketchen, Boyd, Su and Joo2016), corrected correlations between seven predictors of firm performance were: CEO tenure (0.019), top management team tenure (0.036), top management team diversity (−0.015), top management team size (0.049), board size (0.070), board leadership structure (0.010), and board independence (0.023). Our study reported a corrected correlation between CEO overconfidence and firm performance to be 0.13, which was larger than any of the seven correlations reported in Bergh et al. (Reference Bergh, Aguinis, Heavey, Ketchen, Boyd, Su and Joo2016). Hence, compared to other predictors of firm performance, the performance effect of 0.13 at the firm level may be viewed small in absolute terms yet appears to be practically important in relative terms. In addition to this bivariate relation, we identified sizable CEO overconfidence–firm performance relationship effect size distribution variations, suggesting the presence of moderators. The fact that the moderators didn’t alter effect sizes to a noteworthy degree shows further evidence for the consistency with previous research having firm performance as an outcome variable, demonstrating that firm performance is intractable and difficult to influence (e.g., Park & Shaw, Reference Park and Shaw2013). However, it is noted that a majority of meta-analytic distributions across subgroups still have small Varart% values (see Table 2), thus suggesting the presence of further moderators in these meta-analytic distributions. This provides a roadmap for future research to identify more moderators which may influence the effect size distributions for the relationship between CEO overconfidence and firm performance. The findings about moderators are discussed in the following sections.
Results associated with our second research question (do cross-cultural and industry conditions moderate the relationship between CEO overconfidence and firm performance?) revealed several interesting relationships (and statistically nonsignificant relationships). With respect to moderator testing for the CEO overconfidence–firm performance relationship, and consistent with trait activation theory, our results indicate that this relationship is context-dependent on national cross-cultural dimensions. These cross-cultural findings, based on a contingency theory approach, answer the call from prior HRD studies by demonstrating that executive trait research should go beyond the “good-versus-bad” discussion to a more refined “it depends” perspective (e.g., Fatfouta, Reference Fatfouta2019). Our findings also shed light on the growing body of international HRD research (e.g., Bartlett, Lawler, Bae, Chen, & Wan, Reference Bartlett, Lawler, Bae, Chen and Wan2002; Moon, Choi & Jung, Reference Moon, Choi and Jung2012) by demonstrating cross-cultural differences in the CEO overconfidence–firm performance relationship.
With regard to the first cultural dimension – assertiveness – which refers to “the degree to which individuals are assertive, confrontational, and aggressive in their relationships with others” (House et al., Reference House, Hanges, Javidan, Dorfman and Gupta2004, p. 30), we found that the CEO overconfidence–firm performance relationship is stronger in high assertive than low assertive cultures. Overconfident CEOs are inclined to be involved in challenging tasks, to be in control of all activities, and to deviate from proven practices to pursue success (Tang, Qian, Chen, & Shen, Reference Tang, Li and Yang2015); hence, the performance effect of CEO overconfidence is likely to be activated in high assertive cultures where success, progresses, and control over the environment are valued.
In-group collectivism refers to “the degree to which individuals express pride, loyalty, and cohesiveness in their organizations or families” and institutional collectivism refers to “the degree to which organizational and societal institutional practices encourage and reward collective distribution of resources and collective action” (House et al., Reference House, Hanges, Javidan, Dorfman and Gupta2004, p. 30). High in-group collectivistic and high institutional collectivistic cultures value group loyalty, interest, and interdependence and prioritize relatedness with in-groups. We found that the CEO overconfidence–firm performance relationship is stronger in low institutional collectivistic and low in-group collectivistic cultures. Since overconfident CEOs are inclined to dominate other top management team members (Engelen et al., Reference Engelen, Neumann and Schmidt2016), their trait overconfidence is likely to be deactivated in high institutional collectivistic and high in-group collectivistic cultures where domination over others is discouraged, thus weakening trait overconfidence’s performance effect.
Future orientation is “the degree to which a collectivity encourages and rewards future-oriented behaviors such as planning and delaying gratification” (House et al., Reference House, Hanges, Javidan, Dorfman and Gupta2004, p. 282). We found that the CEO overconfidence–firm performance relationship is stronger in high future-oriented cultures. Since overconfident CEOs may underestimate the uncertainty of future states of the world and overestimate their certainty about being correct, they are likely to be flexible in making quick entrepreneurial or “out-of-the-box” decisions which a highly future-oriented culture strongly values and promotes (Chen et al., Reference Chen, Crossland and Luo2015; Engelen et al., Reference Engelen, Neumann and Schmidt2016; Hiller & Hambrick, Reference Hiller and Hambrick2005), thus activating the performance effect of trait overconfidence.
Gender egalitarianism is “the degree to which a collective minimizes gender inequality” (House et al., Reference House, Hanges, Javidan, Dorfman and Gupta2004, p. 30). We found that the CEO overconfidence–firm performance relationship is stronger in high gender egalitarian cultures. In high gender egalitarian cultures where both men and women enjoy equal workplace opportunities and are expected to undertake dual roles rather than dividing labor along traditional lines (Shockley et al., Reference Shockley, Douek, Smith, Peter, Dumani and French2017), this may boost the overall performance effect of overconfidence because both men and women equally benefit from being overconfident in high gender egalitarian cultures.
Performance orientation refers to the degree to which a community supports excellence, innovation, and performance improvement (House et al., Reference House, Hanges, Javidan, Dorfman and Gupta2004; Miao, Coombs, Qian, & Sirmon, Reference Miao, Coombs, Qian and Sirmon2017a). We found that the CEO overconfidence–firm performance relationship is stronger in high performance-oriented cultures. Overconfident CEOs have a propensity to pursue difficult tasks and entrepreneurial opportunities to enhance firm performance because they are obsessed with their personal abilities to produce success and excellent performance (Engelen et al., Reference Engelen, Neumann and Schmidt2016). Thus, the performance potential of trait overconfidence will be unlocked in high performance-oriented cultures because such cultures promote and reward the expression of overconfidence.
Power distance refers to the extent of societal acceptance toward power distributions/differences, status privileges, and authority (House et al., Reference House, Hanges, Javidan, Dorfman and Gupta2004). Our result showed a stronger performance effect of CEO overconfidence in low power distance cultures. Followers in low power distance cultures are more willing to share their constructive ideas/advices with CEOs and there may be less CEOs’ dominance in decision-making (i.e., more participative decision-making), thus helping overconfident CEOs make more shrewd decisions and avoid making costly mistakes (She, Li, London, Yang, & Yang, Reference She, Li, London, Yang and Yang2020). Thus, the positive performance effect of CEO overconfidence may be enhanced due to improved decision-making processes and idea exchanges in low power distance cultures.
Uncertainty avoidance indicates “the extent to which a society, organization, or group relies on social norms, rules, and procedures to alleviate unpredictability of future events” (House et al., Reference House, Hanges, Javidan, Dorfman and Gupta2004, p. 30). We found that the CEO overconfidence–firm performance relationship is stronger in low uncertainty avoidance cultures. High uncertainty avoidance cultures discourage risk taking due to high levels of formalization and bureaucracy (Covin, Green, & Slevin, Reference Covin, Green and Slevin2006; Miao et al., Reference Miao, Coombs, Qian and Sirmon2017a). The hallmark of CEO overconfidence is risk taking; thereby, cultural norms of high uncertainty avoidance suppress the expression of trait overconfidence.
The moderator effect of humane orientation is statistically nonsignificant, suggesting that the positive CEO overconfidence–firm performance relationship is generalizable across this cultural dimension. Moderator effects of industry technology intensity, firm performance measurement, and overconfidence scales are also statistically nonsignificant, suggesting the generalizability of the positive performance effect of CEO overconfidence across some of these moderator subgroups.
In response to our final research question (is EO a mediator of the CEO overconfidence–firm performance relationship?) our result demonstrates that EO fully mediates the CEO overconfidence–firm performance relationship. This finding supports upper echelons theory because it suggests that CEO personality traits may shape EO, which captures firm strategic actions, and in turn influences firm performance (Hiller & Hambrick, Reference Hiller and Hambrick2005; Miao et al., Reference Miao, Coombs, Qian and Sirmon2017a; Wang et al., Reference Wang, Holmes, Oh and Zhu2016). This finding also supports the notion that overconfident CEOs are more prone to pursue bold entrepreneurial opportunities and navigate firms to take “large-stakes initiatives” to enhance firm performance because they perceive these actions and opportunities to be attractive and likely to produce profits (Wang et al., Reference Wang, Holmes, Oh and Zhu2016). Our result also supports previous meta-analytic findings (Miao et al., Reference Miao, Coombs, Qian and Sirmon2017a) regarding EO as a mediator between both human and social capital and firm performance. Taken together, this finding provides evidence that EO is a primary mechanism through which executives inject their personal characteristics into decisions regarding resource utilization and also provides strong support for the resource orchestration literature (Sirmon & Hitt, Reference Sirmon and Hitt2003; Sirmon et al., Reference Sirmon, Hitt and Ireland2007, Reference Sirmon, Hitt, Ireland and Gilbert2011) which aims to open the “Black Box” of executive decision-making by describing how managers structure, bundle, and leverage firm resources to create competitive advantage and increase firm performance.
This mediation finding also has implications for EO research, the resource orchestration literature, and the field of HRD. Previous research suggests that researchers know little about EO’s antecedents (Miao et al., Reference Miao, Coombs, Qian and Sirmon2017a; Rosenbusch et al., Reference Rosenbusch, Rauch and Bausch2013; Wales et al., Reference Wales, Gupta and Mousa2013). Our research contributes to the EO literature by examining a new antecedent – CEO overconfidence. In addition, the resource orchestration literature suggests that EO is a conduit or mobilizing vision via which managers use systems of practices and managerial styles to direct how resources are utilized to impact firm performance (Chirico et al., Reference Chirico, Sirmon, Sciascia and Mazzola2011; Sirmon et al., Reference Sirmon, Hitt, Ireland and Gilbert2011). Managing change and innovation is an important topic in HRD because change and innovation will foster organizational learning and lead to competitive advantage (Kontoghiorghes, Awbre, & Feurig, Reference Kontoghiorghes, Awbre and Feurig2005). Our research findings contribute to the field of HRD by demonstrating that CEO overconfidence may influence CEO decision-making processes and behaviors to manage change and innovation so as to influence firm performance, but through different mechanisms. Practically speaking, excess belief in one’s own abilities (overconfidence) positively influences firm performance through strategic decision-making (EO). In sum, EO plays a critical role in transmitting the performance effect of CEO overconfidence.
Practical implications
Much of the popular business press describes CEO overconfidence in a negative light (Adams, Reference Adams2012; Maccoby, Reference Maccoby2003; Sahadi, Reference Sahadi2018), suggesting that this CEO trait should be avoided. Conflicting research results have added to these concerns. Our meta-analytic findings clarify this relationship, provide useful information, and highlight practical implications for practicing managers, board members, and other stakeholders. First, CEO overconfidence positively influences firm performance because overconfident CEOs cultivate a strategic posture (i.e., EO) that positively influences firm performance. Second, this positive performance effect is stable across firms in industries with different technology intensities and different ways of measuring firm performance (i.e., profitability, growth, and market-based performance). Third, CEO overconfidence has a more pronounced positive performance effect in high assertive, low institutional collectivistic, low in-group collectivistic, high future-oriented, high gender egalitarian, high performance-oriented, low power distance, and low uncertainty avoidance cultures. CEOs in these cultures should capitalize on advantages from these cultural norms in order to unlock the performance potential of their trait overconfidence.
From a practical perspective, HRD professionals, managers, board members, and other stakeholders should clearly understand that corporate leaders such as CEOs are generally more overconfident than the general population (Goel & Thakor, Reference Goel and Thakor2008; Young & Pinsky, Reference Young and Pinsky2006). Recognizing this fact, board of director members (who themselves are likely to be more overconfident), HRD professionals, and policy makers may consider the following ideas to either benefit from positive aspects, or control negative aspects of CEO overconfidence.
For example, HRD professionals may follow the best practices to administer personality tests and provide executive coaching and/or leadership development opportunities to upper level employees, in order to develop and enhance their self-awareness and help them understand their strengths and weaknesses as a result of their overconfidence (Berr, Church, & Waclawski, Reference Berr, Church and Waclawski2000; Harland, Reference Harland2003). Such HRD activities or practices will help senior executives comprehend what dysfunctional behaviors as a result of their overconfidence should be avoided and what functional behaviors due to their overconfidence should be practiced more frequently (Berr, Church, & Waclawski, Reference Berr, Church and Waclawski2000). Changes to CEO compensation structure may be considered such as increasing incentive compensation for overconfident CEOs to best take advantage of their positive bias toward their organization (Humphery-Jenner, Lisic, Nanda, & Silveri, Reference Humphery-Jenner, Lisic, Nanda and Silveri2016). In addition, international HRD professionals may consider following the best practices to develop and implement effective executive training programs that have a special focus on cross-culture and psychological traits (e.g., Ballou, Bowers, Boyatzis, & Kolb, Reference Ballou, Bowers, Boyatzis and Kolb1999; Cherniss, Grimm, & Liautaud, Reference Cherniss, Grimm and Liautaud2010).
Research limitations and directions for future research
First, we only investigated EO as a mediator in the present study. Future research may examine additional mediators such as how CEO overconfidence influences strategic flexibility (i.e., the capacity to adapt quickly to environmental changes). For example, a prior study has examined how Big Five personality traits influence strategic flexibility (Nadkarni & Herrmann, Reference Nadkarni and Herrmann2010). Since CEO overconfidence is related to change and strategic adaptations, it would be theoretically relevant to test how it influences strategic flexibility. Future research may also examine how CEO overconfidence impacts other firm processes related to change and innovation, such as dynamic capabilities and organizational ambidexterity.
Second, most CEO overconfidence studies only examine direct, indirect, or moderated CEO overconfidence effects, while they largely ignore possible curvilinear effects. According to our research findings, the overall performance effect of CEO overconfidence is positive. However, little is known regarding whether there is a “too much of a good thing” effect of CEO overconfidence. Would firm performance decline if a CEO is too overconfident (i.e., an inverted U-shaped relationship)? Behavioral threshold theory of nonlinear effects suggests that low levels of CEO overconfidence may result in adaptive manifestations of overconfidence that can amplify its performance benefits, whereas high levels of CEO overconfidence may lead to maladaptive manifestations of overconfidence that impair CEOs’ decision-making and detract from firm performance (Grijalva et al., Reference Grijalva, Harms, Newman, Gaddis and Fraley2015). We encourage future research to examine such curvilinear effects based on Grijalva et al. (Reference Grijalva, Harms, Newman, Gaddis and Fraley2015) and Palich, Cardinal, and Miller (Reference Palich, Cardinal and Miller2000) to explore the complexity that lies in the relationship between CEO overconfidence and firm performance.
Third, although we tested a series of moderator and mediator variables in our study, our examination is still exploratory due to the lack of relevant studies and theoretical rationales. For these reasons, we call for future studies to develop theory regarding boundary conditions and underlying mechanisms for CEO traits–firm performance relationships.
Fourth, we could have used meta-analytic regression analysis (MARA) for moderator testing as this analysis takes less variance from the analysis as it analyzes continuous, rather than, categorical variables. Although MARA has its strengths, we chose subgroup analyses to test moderator effects in order to be statistically conservative. Since the size of our meta-analytic database is not large enough to be suitable for MARA, performing MARA may subject our results to second-order sampling error. Schmidt and Hunter (Reference Schmidt and Hunter2015) warned that the number of studies is the sample size in meta-regression; hence, regression estimates are prone to be influenced by second-order sampling error if the sample size in a meta-regression (i.e., number of studies) is not large enough. Thus, MARA results can be biased and statistically nonsignificant if second-order sampling errors exert noticeable impacts on meta-analytic estimates. In order to address this limitation, we performed subgroup analyses because such analyses have higher statistical power than other meta-analytic moderator analyses (Wang, Oh, Courtright, & Colbert, Reference Wang, Oh, Courtright and Colbert2011). In addition, the number of samples in our meta-analyses is also consistent with the sample requirement of subgroup analyses (i.e., a minimum of 10 samples for the sum of number of samples across all subgroups for a moderator) (Kirca, Hult, Deligonul, Perryy, & Cavusgil, Reference Kirca, Hult, Deligonul, Perryy and Cavusgil2012).
Conclusion
Using meta-analytic techniques, we examined whether there is a relationship between CEO overconfidence and firm performance based on 50 samples with subjects drawn from multiple countries. Our study confirmed a positive performance effect for CEO overconfidence. A set of possible moderators including industry characteristics, CEO overconfidence measures, and cultural dimensions were examined. The relationship between CEO overconfidence and firm performance varied noticeably across a set of cultural dimensions. Further, our results demonstrate that EO fully mediates the positive performance effect of CEO overconfidence. Our study provides multiple avenues for future research and important practical advice for executive decision-makers.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/jmo.2023.58.
Competing interests
The authors declare none.
