Emphasizing the “I” in “Innovation”: the influence of CEO narcissism on the firm’s relative exploration orientation, and the moderating role of CEO tenure
by
George Lijzenga S3189538
MSc BA Strategic Innovation Management Supervisor: dr. P.J. Steinberg
Co-assessor: M.K. Weck Word count: 13.184
June 22nd, 2020
Abstract
Research following the upper echelon theory has mainly focused on executives’ demographic characteristics, which is why scholars have pleaded for more insights into executives’ psychological processes underlying organizational decisions, such as in innovation. Accordingly, this research uncovers how CEO narcissism influences the firm’s relative exploration orientation, and how this effect changes with CEO tenure. Building on the self-verification theory, I predict that narcissistic CEOs establish a greater relative exploration orientation in order to obtain feedback that verifies their self-concept of uniqueness, while this effect attenuates with increasing CEO tenure as the effectiveness of this ‘tool’ decreases. I test my hypotheses using panel data from 128 firms in the S&P 100 between 2008-2018. Contrary to my expectations, I find that CEO narcissism significantly decreases the relative exploration orientation, and that this negative effect amplifies with increasing CEO tenure. However, my robustness checks raise concerns about the results’ robustness. These insights contribute to the upper echelon theory by showcasing that CEO narcissism influences the innovative orientation of a firm’s activities towards exploitation. Moreover, these insights could aid the board of directors in appointing CEOs, as narcissism can be a trait to either seek or avoid, depending on the firm’s desired innovative orientation.
Key words:
1
Table of contents
1. Introduction ... 2
2. Literature review ... 6
2.1 Upper echelon theory ... 6
2.2 CEO Narcissism ... 7
2.3 Exploration, exploitation and the relative exploration orientation ... 9
2.4 Hypothesis development ... 11
2.4.1 CEO narcissism and relative exploration orientation ... 11
2.4.2 The moderating role of CEO tenure ... 13
3. Methodology ... 15
3.1 Sample and data collection ... 15
3.2 Measurements ... 16 3.2.1 Dependent variable ... 16 3.2.2 Independent variable ... 17 3.2.3 Moderator ... 18 3.2.4 Control variables ... 18 3.3 Analytical method ... 20 4. Results ... 21
4.1 Descriptive statistics and correlations ... 21
4.2 Regression results ... 23
4.3 Robustness checks ... 24
5. Discussion and conclusion ... 27
5.1 Theoretical implications ... 29
5.2 Managerial implications ... 30
5.3 Limitations and future research ... 31
References ... 34
Appendices ... 42
Appendix 1: Validity check of narcissism construct (correlation) ... 42
Appendix 2: Variance Inflation Factors (VIF) test ... 42
Appendix 3: Robustness check with CEO picture as independent variable ... 42
Appendix 4: Exploration and exploitation orientation dictionaries from Moss, Payne & Moore (2014)... 43
Appendix 5: Robustness check with relative exploration orientation measured following Moss, Payne & Moore (2014) ... 44
2
1. Introduction
Over the last decades, there has been a growing interest inresearch following the upper echelon theory, which examines how organizational strategy and performance are influenced by the CEOs’ and other top executives’ personal characteristics, including their demographic characteristics, values, psychological traits and knowledge bases (Hambrick & Mason, 1984; Carpenter, Geletkanycz & Sanders, 2004; Hambrick, 2007). The lion’s share of research following the upper echelon theory has focussed on the executives’ demographic characteristics such as age, functional background and education (e.g., Bantel & Jackson, 1989; Barker & Mueller, 2002; Bertrand & Schoar, 2003). However, as Lawrence (1997), Carpenter, Geletkanycz & Sanders (2004) and Hambrick (2007) have explained, this focus on demographics has led to a ‘black box’ problem, meaning there is little concrete understanding of the psychological and social processes influencing executives’ strategic decision making. In response to this, a stream of literature has aimed to ‘open up the black box’ by comprehensively uncovering how the psychological traits of CEOs (and other executives) influence organizational behaviour and performance (e.g., Peterson et al., 2009; Wales, Patel, & Lumpkin, 2013; Ham, Seybert, & Wang, 2018).
One of the psychological traits that has been deemed particularly salient to analyse in this context is narcissism, as narcissists often are intensely attracted to leadership positions, resulting in the widespread prevalence of narcissism under leaders and CEOs (Kets de Vries & Miller, 1985; Resick, Whitman, Weingarden & Hiller, 2009). Narcissism refers to “the degree to which an individual has an inflated sense of self and is preoccupied with having that self-view continually reinforced” (Chatterjee & Hambrick, 2007: p. 353). Research has examined the effect of CEO and top executive narcissism on some aspects of organizational behaviour, strategy and performance, such as Buyl, Boone & Wade (2019) who found that narcissistic CEOs engage in more corporate risk-taking, Oesterle, Elosge & Elosge (2016) who showed that narcissistic CEOs are more likely to engage in internationalization, Rijsenbilt & Commandeur (2013) who found that CEO narcissism is positively related to fraudulent behaviour, and Chatterjee & Hambrick (2007) who indicated that narcissistic CEOs achieve more extreme and fluctuating firm performances.
Surprisingly, however, one potentially interesting area of organizational strategy remains underexplored in relation to CEO narcissism: innovation. Studies linking CEO narcissism to certain aspects of
innovation have been few and far between, with scarce examples being Gerstner et al. (2013) who showed that CEO narcissism causes firms to more aggressively adopt discontinuous technologies from the
3
incomplete and fragmented, with some potentially salient linkages remaining unexamined (Barker & Mueller, 2002).
One such unexamined linkage is the relationship between CEO narcissism and the firm’s innovative orientation towards exploitation, that is, the incremental exploitation of current capabilities, or
exploration, that is, the exploration of fundamentally new capabilities (March, 1991; Raisch et al., 2009). Research has illustrated the vital importance of these innovative orientations for firm performance and survival, with exploitation being crucial for short-term viability and exploration being crucial for long-term viability (Levinthal & March, 1993; He & Wong, 2004; Greve, 2007). However, these orientations have principally different requirements in terms of strategy, processes, routines and structures (March, 1991; He & Wong, 2004; Uotila et al., 2009). Therefore, research has emphasized that there is a trade-off between the two orientations and that firms resultantly establish a position in the continuum between the two orientations, indicated as the relative exploration orientation or the extent to which firms emphasize exploration activities over exploitation activities (Uotila et al., 2009; Heyden et al., 2015). A firm’s relative exploration orientation can have important strategic consequences, as it determines the focus towards either short-term or long-term viability, which is why a stream of research has aimed to discover which factors could influence, adjust and bias a firm’s innovative orientation towards exploitation or exploration (e.g., Sidhu, Volberda & Commandeur, 2004; Heyden et al., 2015; Wang & Dass, 2017).
As the firm’s primary strategic decision maker, the CEO can strongly influence the innovative orientation of organizational activities (Mackey, 2008; Busenbark, 2016). Therefore, following the upper echelon theory, the CEO’s psychological traits, such as narcissism, could affect the firm’s relative exploration orientation. The strategic salience of this potential effect is that, depending on the firm-specific
contingencies, it might hinder or facilitate firms in adjusting their innovative orientation to focus more on short-term or long-term viability, or in maintaining a desired status quo. Given this strategic relevance, it is then also important to understand when this effect becomes stronger or weaker through the analysis of boundary conditions, such as how the effect differs over time, as this develops a more complete
comprehension of the effect’s implications and deepens the understanding of the underlying mechanism.
4
What is the effect of the degree of CEO narcissism on the firm’s relative exploration orientation, and how is this effect moderated by CEO tenure?
Following the mechanism of the self-verification theory, I argue that CEO narcissism has a positive effect on the firm’s relative exploration orientation. Narcissistic CEOs believe that they are unique and
differentiated from other CEOs, and they crave to ‘verify’ this self-concept by engaging in explorative innovations which differentiate themselves from the firm’s previous CEO, thereby increasing the chances of obtaining feedback from others that reinforces their self-concept of uniqueness (Swann, Pelham & Krull, 1989; He & Wong, 2004; Reve, 2007). Instead, engaging in exploitative innovations would be less fruitful for narcissistic CEOs in their endeavour to be perceived as unique, as this emphasizes the
continuation of the capabilities and routines that were established under the previous CEO (He & Wong, 2004; Reve, 2007).
I furthermore argue that CEO tenure acts as a boundary condition of this mechanism, due to the changes over time in the relative aspect of the mechanism. In the early stages of CEO tenure, exploration and exploitation determine how differentiated narcissistic CEOs are perceived relative to the prior CEO, but when CEO tenure increases, exploration and exploitation determine how differentiated the narcissistic CEOs are perceived compared to their prior selves (Hambrick & Fukotimi, 1991; Miller, 1991; Campbell, Goodie & Foster, 2004). Therefore, narcissistic CEOs’ ‘tool’ of emphasizing exploration over
exploitation to differentiate themselves from others becomes less effective over time, which is why I predict that an increased CEO tenure will attenuate the pattern of narcissistic CEOs focusing on exploration over exploitation.
I test my hypotheses using codified transcripts of quarterly earning calls from the Standard & Poor (S&P) 100 firms from 2008-2018. Through the application of a programming script, I distilled the CEO’s contributions from 5083 earning calls and analysed them through Computer-Aided Text Analysis (CATA). Following Chatterjee & Hambrick (2007), I measured CEO narcissism through the number of first-person singular pronouns the CEO used (I, me, myself, mine, my), divided by the sum of these pronouns plus all first-person plural pronouns used (we, us, our, ours, ourselves). I measured the firm’s relative exploration orientation through CATA of the firm’s financial reports, according to the measures of Uotila et al. (2008). The data on CEO tenure was obtained from ExecuComp. Contrary to my
5
This research is of theoretical interest, as it contributes to addressing the ‘black box’ problem of the upper echelon theory as stated by Lawrence (1997), Carpenter, Geletkanycz & Sanders (2004) and Hambrick (2007), as they pleaded for more insights into how the CEOs’ psychological processes influence strategic decision making. Specifically, this thesis aims to do so by furthering the currently limited understanding of the relationship between CEO narcissism and innovation, through analysing the previously
unexamined effect of CEO narcissism on the relative exploration orientation. As the results suggest that narcissistic CEOs establish a lower relative exploration orientation, scholars should be aware that CEO narcissism not only affects the adoption of external technologies as prior research discovered (Gerstner et al., 2013), but also affects the relative exploration orientation of the firm’s own activities.
Moreover, this research is of managerial interest, as practitioners will understand what the consequences of appointing a narcissistic CEO will be for the firm’s relative exploration orientation, and how this effect changes across the CEO’s tenure. This is a strategically salient insight for practitioners, since they will know whether and when the appointment of a narcissistic CEO can either bolster or distort the firm’s efforts to focus more on short-term or long-term viability, or to remain in the desired current position, depending on what the firm deems prudent. Thus, this insight can aid the board of directors in appointing CEOs and it can furthermore provide practical indications for when the board should aim to ‘compensate’ for a CEO’s potential bias towards exploitation or exploration when this would be incongruent with the firm’s desired orientation.
6
2. Literature review
As this thesis aims to extend the upper echelon theory, through filling in the gap of how CEO narcissism would influence the firm’s relative exploration orientation, I first discuss the theoretical background of the upper echelon theory and its status quo in Section 2.1. Afterwards, I more concretely review the literature surrounding CEO narcissism (Section 2.2) and the relative exploration orientation (Section 2.3), after which I develop my hypotheses (Section 2.4).
2.1 Upper echelon theory
The fundamental principle of the upper echelon theory is that organizational outcomes, both in terms of strategic decisions and consequential performance, are affected by the personal characteristics of top executives in the organization (Hambrick & Mason, 1984). The original model of Hambrick & Mason (1984) states that these top executives make strategic choices under conditions of bounded rationality, as they cannot fully process all environmental and organizational stimuli. To cope with this bounded rationality, top executives rely on their knowledge base and personal values to firstly focus their attention on particular stimuli and secondly subjectively interpret these stimuli, thereby creating the managerial perceptions that - in tandem with their knowledge base and personal values - form the basis of strategic decision making (Hambrick & Mason, 1984).
This core line of reasoning of the upper echelon theory focuses on the knowledge base and personal values of executives, and thereby emphasizes the importance of the psychological components of executives’ personal characteristics. However, as Hambrick & Mason (1984) and Hambrick (2007) explained, such psychological components are difficult or sometimes even impossible to measure, as they claimed that top executives rarely want to directly participate in psychological tests. Therefore, they proposed that the more easily observable demographic variables of executives - such as their age,
functional background, education and financial position - could be used as proxies for these psychological variables in order to advance the theory (Hambrick & Mason, 1984; Hambrick, 2007).
As a result, the vast majority of studies following the upper echelon theory has analysed such
demographic variables, both on the CEO/single executive level (e.g., Barker & Mueller, 2002; Slater & Dixon-Fowler, 2009; Kish-Gephart & Campbell, 2015) and on the top management team (TMT) level (e.g., Bantel & Jackson, 1989; Wiersema & Bantel, 1992; Tihanyi et al., 2000). However, scholars in the field have criticized this dominant focus on demographic variables (Lawrence, 1997; Carpenter,
7
used as valid proxies for the psychological variables, they also indicated that such demographics are incomplete and imprecise in capturing the psychological components. Lawrence (1997) further reasoned that since demographics at most provide a noisy, muddied indicator of psychological variables, empirical investigations of these demographics do not truly test the core of the upper echelon theory. This criticism on the use of demographics to test the upper echelon theory has been synopsized as the ‘black box’ problem: the analysis of demographics does not provide insights into the psychological and social processes influencing executives’ decision making (Lawrence, 1997; Carpenter, Geletkanycz & Sanders, 2004; Hambrick, 2007).
In response to this problem, a stream of research has aimed to open up the black box through analysing executives’ personalities and psychological traits to understand why executives make the decisions that they do. Examples of this stream of research are Peterson et al. (2009) who showed how the
psychological traits hope, optimism and resilience cause CEOs to adopt transformational leadership styles which in turn improve firm performance, and Nadkarni & Herrmann (2017) who linked the ‘big five’ personality traits amongst CEOs with strategic flexibility and found that conscientiousness, openness and extraversion are positively related to it, while emotional stability decreases it and agreeableness shows an inverted-U shape with it. These studies provide insights that are more directly related to the original model of Hambrick & Mason (1984), as they analyse executives’ psychological traits and personalities which determine their personal values and managerial perceptions, that in turn influence their strategic decision making. Therefore, this thesis aims to join this stream of research that opens up the black box, as I aim to further the current understanding of the effects of the psychological trait narcissism amongst CEOs, by linking this trait to the firm’s relative exploration orientation. The next sections further elaborate on these concepts.
2.2 CEO Narcissism
8
Campbell, Goodie & Foster, 2004; Resick et al., 2009). Chatterjee & Hambrick (2007) built on this idea and stated that narcissism has a cognitive dimension of an overconfident belief in the excellence and uniqueness of one’s qualities, as well as a motivational dimension of an intense need to have one’s excellence reaffirmed by others.
In research following the upper echelon theory, narcissism has been deemed a salient psychological trait to analyse amongst CEOs, especially considering its widespread prevalence amongst CEOs and leaders in general (Kets de Vries & Miller, 1985; Maccoby, 2000; Resick, et al., 2009). This is because narcissism often is the driving force for obtaining leadership positions, as these positions aid narcissists in building their legacy of achievements, which in turn can lead to their desired attention and praise (Kets de Vries & Miller, 1985; Wales, Patel, & Lumpkin, 2013). Moreover, certain aspects of narcissism, such as
boastfulness, aggressiveness and an intense desire to compete, in fact can help individuals to achieve these high leadership positions through promotions (Maccoby, 2000; Resick et al., 2009). Due to this interlocking relationship between narcissism and the CEO position, and the resulting pervasiveness of narcissism amongst CEOs, a stream of research has analysed how CEO narcissism influences strategic decision making and organizational performance (e.g., Rijsenbilt & Commandeur, 2013; Wales, Patel & Lumpkin, 2013; Oesterle, Elosge & Elosge, 2016).
9
In line with these findings on the variance and extremity of performance, Buyl, Boone & Wayde (2019) found that narcissistic CEOs engage in more corporate risk-taking. Oesterle, Elosge & Elosge (2016) also elaborated on the risk-taking aspect by finding that narcissistic CEOs are more likely to engage in
internationalization. Other studies, however, demonstrated some clear drawbacks of this overconfidence in risk-taking: Rijsenbilt & Commandeur (2013) showed that narcissistic CEOs are more likely to engage in fraud, and O'Reilly, Doerr & Chatman (2018) found that narcissistic CEOs are more likely to subject the organization to lawsuits.
This prior discussion highlighted some areas of strategic decision making and organizational performance that have been examined in relation to CEO narcissism. Still, one of the areas of organizational decision making that has received relatively little attention in this regard is innovation, as only a few studies have linked CEO narcissism to some dimension of innovation. The priorly mentioned study of Wales, Patel & Lumpkin (2013) fits this category, as innovativeness is one of the dimensions of the entrepreneurial orientation that this study examined in relation to CEO narcissism. Next to this, Ham, Seybert & Wang (2018) analysed the impact of CEO narcissism on R&D spending and found that firms led by narcissistic CEOs are more likely to overinvest in R&D, due to the potential recognition that successful R&D can provide for the CEO. Additionally, Gerstner et al. (2013) found that CEO narcissism causes the firm to more aggressively adopt discontinuous technologies from the environment, as these bold technologies can help CEOs gain attention. Still, as this body of research linking CEO narcissism to innovation is rather limited, some potentially interesting areas of innovation have not yet been examined in this context. The firm’s relative exploration orientation is one of those areas, and is discussed in the next section.
2.3 Exploration, exploitation and the relative exploration orientation
The concepts of exploration and exploitation were introduced by March (1991) in the context of organizational learning. As March (1991, p. 71) stated: “Exploration includes things captured by terms such as search, variation, risk taking, experimentation, play, flexibility, discovery, innovation.
10
to short-term viability through realising returns from current capabilities and refining these capabilities, while exploration is key for long-term viability through adapting to new opportunities and avoiding obsolescence (Levinthal & March, 1993; He & Wong, 2004).
Still, research has emphasized that there is a trade-off in establishing an orientation towards exploration and exploitation, as both draw on the organization’s limited resources and attention, yet require
substantially different strategies, processes, routines, and structures (March, 1991; He & Wong, 2004; Uotila et al., 2009). As exploitation focuses on leveraging current capabilities, it may introduce organizational inertia and an inability to adapt to future changes, thereby limiting the organization’s potential for exploration (March, 1991; He & Wong, 2004). The other way around, exploration requires substantial resources and attention in order to experiment with uncertain and new capabilities, thereby reducing the speed at which the organization’s current capabilities can be refined and leveraged (March, 1991; He & Wong, 2004). As a result of these incompatibilities, firms have to divide their resources between these orientations and establish a position in the continuum between exploration and exploitation (March, 1991; Andriopoulos & Lewis, 2009; Uotila et al., 2009). This position is referred to as the relative exploration orientation, which is “extent to which the firm emphasizes exploration activities over exploitation activities” (Uotila et al., 2009: p. 223). The ‘optimal’ relative exploration orientation is dependent on the specific contextual contingencies that a firm faces, such as its environmental demands (Levinthal & March, 1993; Uotila et al., 2009).
Importantly, it can be difficult for firms to adjust their orientation from an emphasis on exploitation to more exploration or vice versa when they perceive this to be necessary, because the tensions between exploitation and exploration may cause firms to become ‘locked-in’ to either of them (Levinthal & March, 1993; He & Wong, 2004). Firms emphasizing an exploitation orientation may suffer from the ‘success trap’ as exploitation has more certain short-term success and positive returns than exploration, with past exploitation furthermore increasing the efficiency of future exploitation through improved competence (Levinthal & March, 1993). As a result, these firms become locked-in to the exploitation side of the continuum since the opportunity costs of switching to more exploration become increasingly higher (Levinthal & March, 1993). At the other end, firms emphasizing an exploration orientation may suffer from the ‘failure trap’ as these firms become locked-in to vicious cycles of failed exploration and
11
These lock-ins can make it difficult for firms to re-adjust their position in the continuum when they perceive this to be necessary, such as when they aim to improve their short-term viability through more exploitation or their long-term viability through more exploration (Levinthal & March, 1993; He & Wong, 2004). Therefore, a stream of research has aimed to identify which factors may influence, bias and shift a firm’s relative exploration orientation. For example, Sidhu, Volberda & Commandeur (2004) showed that a stronger organizational mission and more slack resources increase the relative exploration orientation, and more in line with the upper echelon theory, Heyden et al. (2015) found that heterogeneity of functional backgrounds in the board of directors is positively related to the relative exploration
orientation, and Wang & Dass (2017) discovered that TMT innovativeness increases the relative exploration orientation. This thesis aims to join this stream of research in identifying factors that can influence or bias a firm’s relative exploration orientation, as I expect that CEO narcissism will positively affect it. I elaborate on this in the next section.
2.4 Hypothesis development
In this section, I develop my hypotheses as presented in the conceptual model of Figure 1:
Figure 1: Conceptual model
2.4.1 CEO narcissism and relative exploration orientation. In order to fill the gap in the upper echelon
12
Following this theory, a narcissistic CEO has an inflated self-concept as they believe that they are
extraordinary, differentiated and special compared to others (Campbell, Goodie & Foster, 2004; Campbell et al., 2011; Galvin, Lange & Ashforth, 2015). They perceive themselves to stand out compared to their peers as they would possess unique capabilities that others would not possess (Campbell, Goodie & Foster, 2004; Chatterjee & Hambrick, 2007). Importantly, this identity of uniqueness is a relative
construct, which is determined through social comparison: in order to be regarded as unique, other actors must perceive that the CEO is special and different compared to others, such as the firm’s previous CEO (Campbell, Goodie & Foster, 2004; Chatterjee & Pollock, 2017).
Then, in accordance with the self-verification theory, a narcissistic CEO has an intense need to obtain external reaffirmation of this relative uniqueness compared to others, such as the firm’s previous CEO (Campbell, Goodie & Foster, 2004; Chatterjee & Pollock, 2017; Wales, Patel & Lumpkin, 2013). As a result, a narcissistic CEO engages in strategies and activities that they believe will lead to feedback from other actors inside and outside the organization that would verify their self-concept of uniqueness (Campbell, Goodie & Foster, 2004; Chatterjee & Hambrick, 2007). Moreover, they will use the firm as a tool in achieving this (Chatterjee & Hambrick, 2007; Resick et al., 2009). This is because a narcissistic CEO believes that their needs should take precedence over the firm’s needs (Kets de Vries & Miller, 1985). The resulting ‘narcissistic organizational identification’ that Galvin, Lange & Ashforth (2015) describe, entails that the CEO will shape the organization based on their own identity and needs, effectively treating the organization as a means for their own goals instead of an end in itself.
13
Instead, if the narcissistic CEO were to emphasize exploitation, the organization would focus on exploiting its existing capabilities (Raisch et al., 2009; Uotila et al., 2009). This would prompt variation-reducing processes: the organization would continue to leverage the capabilities and technologies that have been established under the reign of the previous CEO, and aims to do so more efficiently through small and incremental refinements (He & Wong, 2004; Reve, 2007). Consequently, other actors perceive that the current CEO adheres to the present circumstances that were instituted by the previous CEO, meaning that they do not regard the current CEO as unique or special but rather as commonplace and similar to the previous CEO. This means that an exploitation orientation would be an ineffective tool for narcissistic CEOs, as the feedback they would obtain from others would debunk their self-concept of uniqueness, causing them to experience psychological dissonance (Swann, Pelham & Krull, 1989). Therefore, I hypothesize that narcissistic CEOs will prefer exploration over exploitation in the firm’s innovative orientation:
H1: A higher degree of CEO narcissism leads to a greater relative exploration orientation.
2.4.2 The moderating role of CEO tenure. Next, I aim to develop more insights into the boundary
conditions of the previously explained mechanism of the self-verification theory that would cause narcissistic CEOs to emphasize exploration over exploitation. In order to do so, I analyse the role of the length of the CEO tenure in moderating this effect. Even though I focus on its moderating effect, CEO tenure could also have a potential direct effect on the relative exploration orientation. In this regard, CEOs tend to make fewer major changes as their tenure increases, as longer tenured CEOs typically settle on their priorly established directions and strategies that turned out to be successful, because the continuation of successful firm performance can ensure that CEOs continue their mandate (Hambrick & Fukotimi, 1991; Musteen, Barker & Baeten, 2006). Thus, increased levels of CEO tenure could negatively affect the firm’s relative exploration orientation, as the CEO would prefer exploitation over exploration.
Still, the main emphasis of this research is on the moderating role of CEO tenure on the effect of CEO narcissism on the relative exploration orientation. As mentioned in Section 2.4.1, an important
14
However, the ‘unit’ of social comparison of this tool changes over the course of the CEO’s tenure. When the CEO is in the early stages of their tenure, they are close in temporal space to the organization’s previous CEO (Hambrick & Fukotimi, 1991; Miller, 1991). In such instances, the routines and
capabilities currently in use in the organization were established and achieved under the previous CEO (Miller, 1991). Thus, when a narcissistic CEO engages in exploration to discover new capabilities and routines, they will distinguish themselves from the prior CEO and their establishments (Miller, 1991; Campbell, Goodie & Foster, 2004; Greve, 2007). Instead, if a narcissistic CEO were to emphasize exploitation, they would continue to leverage the establishments of the prior CEO, meaning they would be perceived as similar compared to the prior CEO (Miller, 1991; Campbell, Goodie & Foster, 2004; Greve, 2007). In other words, when the narcissistic CEO’s tenure is still short, emphasizing exploration over exploitation is an effective tool for them to be perceived as unique because the unit of social comparison is the prior CEO.
In contrast, when the narcissistic CEO’s tenure increases, emphasizing exploration or exploitation would determine the degree to which they are perceived as differentiated compared to their own prior selves (Hambrick & Fukotimi, 1991; Miller, 1991; Campbell, Goodie & Foster, 2004). This is because exploration would entail that the CEO deviates from their own prior establishments such as capabilities and routines, while exploitation would mean that the CEO continues to leverage their own prior establishments (Miller, 1991; Musteen, Barker & Baeten, 2006). As this only determines whether the CEO is perceived as unique or similar compared to their past selves, it does not aid the CEO in obtaining feedback that would verify their desired self-concept of relative uniqueness (Campbell, Goodie & Foster, 2004; Greve, 2007). Thus, when the narcissistic CEO’s tenure increases, emphasizing exploration over exploitation becomes an ineffective tool since the unit of social comparison shifts to the CEO’s past self.
Concludingly, these arguments lead me to expect that as the CEO tenure increases, the pattern of narcissistic CEOs focusing on exploration over exploitation will be attenuated compared to when the CEO tenure is still short. I therefore hypothesize:
15
3. Methodology
In this section, I describe the methodology that I used to conduct the research. I firstly discuss the sample of the study and the data collection, after which I explain the measurements that I used for the variables, and I finally clarify the analytical method that I used to test the hypotheses.
3.1 Sample and data collection
The study’s sample consists of 128 firms from the Standard & Poor (S&P) 100 Index from 2008-2018. The S&P 100 is a stock market index composed of the largest publicly traded firms in the US. This makes for a suitable sample for the study, as these publicly traded firms disclose documents - such as annual reports and transcripts of earnings calls - which can be used to gain rich insights into both the CEO and the firm. The analysis of these documents provides an indirect approach to circumvent the earlier mentioned concerns of Hambrick & Mason (1984) and Hambrick (2007), who stated that top executives often would be unwilling to participate in direct research on psychology, for example through surveys. Therefore, this thesis makes use of Computer-Aided Text Analysis (CATA) of documents disclosed by the S&P 100 firms, including annual reports and transcripts of earnings calls.
16
3.2 Measurements
3.2.1 Dependent variable. Both the dependent and independent variables were measured through the use
of Computer-Aided Text Analysis (CATA). CATA is a software-enabled approach to measure theoretical constructs by counting how frequently certain words representing a construct occur within a text (Short et al., 2010). The specific software I used was Linguistic Inquiry and Word Counter (LIWC), which counted specific dictionaries of words that operationalized my dependent and independent variables.
For my dependent variable, the relative exploration orientation, I adopted the dictionaries developed by Uotila et al. (2008) for exploration and exploitation orientation. They based their dictionaries on the definitions of March (1991) that were mentioned in Section 2.3. As such, organizational exploration orientation was measured as LIWC counted the following word roots: explor∗ , search∗ , variation∗ , risk∗ , experiment∗ , play∗ , flexib∗ , discover∗ , innovat*. Similarly, organizational exploitation orientation was measured as LIWC counted these word roots: exploit∗ , refine∗ , choice∗ , production∗ , efficien∗ , select∗ , implement∗ , execut*. The * signs indicate that any combination of characters could follow after the word root (Uotila et al., 2008). The relative exploration orientation was calculated through dividing the sum of words counted for exploration orientation by the sum of total words counted for exploration and exploitation orientation.
I chose these dictionaries to measure the exploration and exploitation orientation through CATA, because they allow for a sophisticated distinction between exploration and exploitation orientation regardless of the operational context of a firm (Uotila et al., 2008). This is because the word roots in the dictionaries cover a wide array of organizational activities, therefore making them applicable across multiple industries, which is important for my sample given the diversity of industries in the S&P 100 (Uotila et al., 2008). Moreover, as these dictionaries are directly based on the original definitions of March (1991), they capture the essence of the meaning of these constructs as March (1991) had initially introduced them.
17
3.2.2 Independent variable. I also relied on CATA to measure my independent variable, CEO
narcissism, as this allows me to circumvent the concerns of Hambrick & Mason (1984) and Hambrick (2007) that CEOs generally are unwilling to participate in psychology research. I followed the CATA measurement of Chatterjee & Hambrick (2007), which consists of the proportion of first-person pronouns used by the CEO that is singular. This means that CEO narcissism was measured by LIWC counting the number of first-person singular pronouns the CEO used (I, me, myself, mine, my), and dividing this by the sum of these pronouns plus all first-person plural pronouns used (we, us, our, ours, ourselves). Chatterjee & Hambrick (2007) based this measure on the research of Raskin & Shaw (1988), who indicated that the trait narcissism is positively correlated with the use of first-person singular pronouns, but negatively correlated with the use of first-person plural pronouns. This is due to the egocentrism of narcissists, who are primarily concerned with their own needs over the needs of others (Kets de Vries & Miller, 1985; Raskin & Shaw, 1988).
The strength of analysing the CEO’s dialect to measure their narcissism lies in the fact that speech is a form of behavioural expression that is strongly influenced by - and thus representative of - a person’s psychological and personal traits (Ramsay, 1968; Raskin & Shaw, 1988). However, for this to be the case, the CEO’s speech has to be voluntary and spontaneous, rather than pre-instructed and controlled (Raskin & Shaw, 1988; Chatterjee & Hambrick, 2007). Therefore, the textual source that I used for the CATA measures of CEO narcissism are transcripts of quarterly earnings calls from the S&P 100 during 2008-2018, and specifically the CEO’s contributions during the question and answer sessions of the calls. The quarterly earnings calls typically consist of two components: first, the CEO and other executives present several pre-written statements on the firm’s financials from the last quarter, and second, the executives answer questions from analysts, partners and investors in a question and answer session. I specifically focused on the CEO’s contributions during the question and answer session because the CEO is more spontaneous in their responses during this part, meaning it would provide the most accurate reflection of the CEO’s psychological and personal traits.
18
calls with only the CEO text during the question and answer session remaining, and 610 earnings calls which did not have any CEO text as there was no CEO participating in them.
Some imperfections of the script should be noted, due to both time constraints and the capability range from the programmer. First, the program identified the function of CEO in the calls through the inclusion of ‘CEO’ in the title of the speaker. This however means that the program was not able to discern between the CEO of the focal firm in the call, and the CEO of subdivisions of that firm or the CEO of a partner firm in the call. Therefore, the text of these ‘additional’ CEOs also had to be included, but these additional CEOs appeared relatively infrequently in the calls. Second, some lines of text from the CEO were excluded from the calls, as they included the word ‘operator’. Since the program had to cut out the text from the operator in the calls, the lines of text from the CEO including the word ‘operator’ also had to be removed as a by-product.
To test whether these imperfections would pose a problem, and to thus check the validity of the script, my fellow students and I manually parsed the CEO’s statements in the question and answer sessions from 216 quarterly earnings calls in 2018 (5% validity check). From these calls, I again constructed the narcissism measure as explained before in this section, and checked its correlation with the narcissism measure that I constructed from the earnings calls that were parsed by the script. Appendix 1 shows a correlation of r=0.95 between the narcissism measures, indicating that the script functioned correctly. The lacking 5% could be the result from the earlier indicated limitations of the script.
3.2.3 Moderator. CEO tenure was measured by the number of years that the CEO has been in charge of
the company (Wu, Levitas & Priem, 2005). This data was obtained from ExecuComp, as was explained in Section 3.1.
3.2.4 Control variables. My analysis included both CEO-level control variables as well as firm-level
control variables to control for confounding effects.
CEO gender. Female CEOs tend to be less overconfident and more risk-averse compared to male CEOs, which is why firms led by female CEOs engage in less risky activities and investments (Martin,
19
female CEOs tend to be less narcissistic compared to male CEOs, meaning that the inclusion of the CEO gender control variable also helps to account for this potential effect (Ingersoll et al., 2019).
CEO age. Older CEOs are more risk-averse and conservative compared to younger CEOs, who prefer actions that are unprecedented, novel, and risky (Hambrick & Mason, 1984; Serfling, 2014). Thus, older CEOs might have a bias towards an exploitation orientation while younger CEOs might be biased towards an exploration orientation.
CEO duality. CEO duality is measured as a dichotomous variable that indicates whether the CEO also is the chairman of the board (Kim et al., 2009). Organizations with the CEO duality model typically have a power structure that is more densely concentrated in the hands of the CEO, meaning that the CEO experiences more freedom to establish their personal agenda, including their preferred degree of organizational exploration and exploitation (Kim et al., 2009; Krause, Semadeni, & Cannella Jr., 2014).
Firm age. Compared to younger firms, older firms typically have more established institutional routines and practices that might lead to inertia, therefore potentially causing a bias for exploitation over
exploration (Sørensen & Stuart, 2000; Koryak et al., 2018). Following Yasuda (2005) and Koryak et al. (2018), I measured firm age as the natural logarithm of the number of years since a venture was founded.
Firm size. Similarly, larger firms typically are more associated with established routines and inertia, while smaller firms typically are more agile, flexible and fast to change (Rothwell, 1989; Koryak et al., 2018). Therefore, larger firms could be biased towards exploitation compared to smaller firms. In accordance with prior research, I measured firm size as the natural logarithm of the number of employees in the firm (Uotila et al., 2009; Koryak et al., 2018).
Year dummy and industry dummy. Lastly, I controlled for potential year effects by including a year dummy, and similarly I controlled for industry effects with an industry dummy consisting of the first digit of the firms’ SIC codes.
20
3.3 Analytical method
To test the hypotheses of this thesis, I performed hierarchical regression analyses, as this allowed me to observe the step-by-step effects that adding my control variables, independent variables and moderator would have on my dependent variable. For panel data, which are used for this thesis, two commonly used models for analyses are fixed effects models and random effects models (Cameron & Trivedi, 2005). A fixed effects model assumes that the unobserved individual effects of each entity (firm in this case) are correlated with the regressors, causing biases (Cameron & Trivedi, 2005). This is why a fixed effects model analyses the within-firm effect of a predicting variable on an outcome variable over time while controlling for potentially biasing time-invariant effects. Instead, a random effect model assumes that the unobserved individual effects of each entity or firm are uncorrelated with the regressors, therefore incorporating time-invariant effects as well through a between-firm analysis (Cameron & Trivedi, 2005).
From a theoretical standpoint, a random effects model is more appropriate than a fixed effects model for this thesis since the main predictor, narcissism, is a stable psychological trait for each CEO (Campbell et al., 2011). This means that the degree of CEO narcissism is a relatively time-invariant construct,
especially considering that the average CEO tenure of the firms in my sample was 6.24 years, thus indicating that firms did not change their CEOs - and their degree of narcissism - relatively frequently. I performed a Hausman test to examine my theoretical assumption that a random effects model would be more appropriate. The null hypothesis of this test was that the unobserved individual effects were uncorrelated with other regressors. Since the result of the test was non-significant, it confirmed that a random effects model was the appropriate means of analysis. As a side note, the Hausman test was performed using GLS regression, as it was incompatible with the type of regression used for the main analyses of this thesis, the tobit regression, as will be explained next.
21
4. Results
In this section, I firstly present the descriptive statistics and correlations, after which I show the results of the regressions to test the hypotheses, and lastly I show the results of my robustness checks.
4.1 Descriptive statistics and correlations
Table 1 shows the descriptive statistics of the variables in the study, including the means, standard deviations and minimum and maximum values. All variables have 1026 observations, except for the CEO picture variable, used for the robustness checks, for which only 742 observations were collected. The mean score on relative exploration orientation is 0.49 and the average level of narcissism amongst CEOs on a scale from 0 to 1 is 0.25. CEOs moreover are mostly male, with an average age of 57 years and an average tenure of 6.24 years. 65% of the CEOs also are the chairman of the board.
22
Table 1: Descriptive Statistics
Variable Obs Mean Std.Dev. Min Max
Relative Exploration 1026 .49 .17 0 .89 CEO Narcissism 1026 .25 .09 0 .69 CEO Picture 742 2.69 .86 1 4 CEO Tenure 1026 6.24 5.11 1 36 CEO Gender 1026 .07 .26 0 1 CEO Age 1026 57.33 5.71 28 77 CEO Duality 1026 .65 .48 0 1 Firm Age (ln) 1026 4 .91 .69 5.42 Firm Size (ln) 1026 3.94 1.27 .62 7.74
Table 2: Correlation matrix
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (1) Relative Exploration 1.00 (2) CEO Narcissism 0.10*** 1.00 (3) CEO Picture 0.12*** -0.05 1.00 (4) CEO Tenure -0.10*** 0.18*** -0.03 1.00 (5) CEO Gender -0.12*** -0.14*** 0.04 -0.07** 1.00 (6) CEO Age -0.06* -0.04 0.04 0.26*** -0.04 1.00 (7) CEO Duality -0.07** 0.13*** 0.07* 0.14*** -0.03 0.10*** 1.00 (8) Firm Age (ln) -0.06** 0.07** 0.20*** -0.09*** -0.01 0.08*** 0.12*** 1.00 (9) Firm Size (ln) 0.05* -0.07** 0.07* -0.06* 0.03 0.11*** 0.13*** 0.16*** 1.00
Notes: *** p<0.01, ** p<0.05, * p<0.1; ln = natural logarithm
Table 3: Random effects tobit regression results (for H1 and H2)
Relative exploration orientation
Variables Model 1 Model 2 Model 3
Independent variable CEO Narcissism -0.087** (0.037) -0.007 (0.053) Interaction effect Narcissism x Tenure -0.012** (0.006) Moderating variable CEO Tenure -0.002*** (0.001) -0.002*** (0.001) 0.001 (0.002) Control variables CEO Gender -0.002 (0.017) -0.006 (0.017) -0.004 (0.017) CEO Age -0.000 (0.001) 0.000 (0.001) 0.000 (0.001) CEO Duality 0.002 (0.008) 0.001 (0.008) 0.001 (0.008) Firm Age (ln) 0.018** (0.009) 0.019** (0.009) 0.019** (0.009) Firm Size (ln) -0.001 (0.007) -0.002 (0.007) -0.002 (0.007) Observations 1,026 1,026 1,026 Number of firms 128 128 128 Constant 0.311*** (0.062) 0.328*** (0.063) 0.306*** (0.063) Log likelihood 1080 1082 1085 Sigma_u 0.098*** (0.007) 0.099*** (0.007) 0.099*** (0.007) Sigma_e 0.071*** (0.002) 0.071*** (0.002) 0.071*** (0.002)
23
4.2 Regression results
Table 3 showcases the results from my random effects tobit model regressions on the relative exploration orientation. Model 1 includes the control variables and the direct effect of the moderating variable CEO tenure, thus merely as a control variable (i.e., no interaction effect yet). Model 2 adds the independent variable, CEO narcissism, in order to test H1. Model 3 adds the interaction effect between CEO tenure and CEO narcissism, in order to test H2.
H1 predicted that CEO narcissism would be positively related to the firm’s relative exploration orientation. Contrary to my expectations, Model 2 demonstrates that CEO narcissism actually has a significantly negative impact on the relative exploration orientation at the 5% level (β = -0.087, p = 0.018). This rejects H1, as narcissistic CEOs in fact establish a lower relative exploration orientation. Specifically, every increase of one standard deviation in CEO narcissism accounts for a decrease of 0.087 standard deviation in the relative exploration orientation.
Before continuing with the interaction effect of CEO tenure as hypothesized in H2, Model 1 and Model 2 first show a significantly negative direct effect of CEO tenure on the relative exploration orientation at the 1% level (β = -0.002, p = 0.003 in Model 1; β = -0.002, p = 0.006 in Model 2). Next, H2 predicted that CEO tenure would attenuate the positive effect of CEO narcissism on the relative exploration orientation, and thus predicted a negative interaction effect. Model 3 shows that there indeed is a significantly
24
Figure 2: Visual representation of the interaction effect between CEO narcissism and CEO tenure
4.3 Robustness checks
In order to test the robustness of these findings, I performed three additional robustness checks. First, my initial analysis provided a limited representation of the trait narcissism, by only considering its
manifestation in the CEO’s speech. However, narcissism is a complex trait that can manifest itself in multiple ways in the CEO’s behaviour, such as through their interactions, their relative pay, or through their prominence and presence in press releases (Petrenko et al., 2006; Chatterjee & Hambrick, 2007). Therefore, to cover a different manifestation of CEO narcissism, I used a different measure of the independent variable for my first robustness check: instead of a CATA-based measure, I used a measure that reflects CEO narcissism through the prominence of the CEO’s photograph in annual reports
25
by themselves that was smaller than half a page, and 4 if there was a picture of the CEO by themselves that was larger than half a page.
Appendix 3 contains Model 4 and Model 5 which showcase the robustness checks with CEO picture as the independent variable. Model 4 presents the main effect of CEO picture on the relative exploration
orientation, which now is non-significant and positive (β = 0.004, p = 0.369). This therefore does not confirm the robustness of the results on H1. Model 5 showcases the interaction effect of CEO picture with CEO tenure, which is non-significant and positive (β < 0.000, p = 0.608). Thus, the CEO picture analysis also does not confirm the robustness of the results on H2.
Second, the measurement choice of the dependent variable, relative exploration orientation, also might have affected the results as I only considered one potential dictionary to measure the construct, namely that of Uotila et al. (2009). Therefore, I conducted a robustness check using an alternative CATA measure for relative exploration orientation. Specifically, I used the dictionaries from Moss, Payne & Moore (2014) for exploration and exploitation orientation, which can be found in Appendix 4. The starting point of their dictionaries also were the definitions of March (1991), but they expanded the dictionaries to cover more words such as pioneer, prospect and patent for exploration orientation, and automate, maintain and salespeople for exploitation orientation (Moss, Payne & Moore, 2014). Even though their dictionaries were developed for their research on high-tech industries, most terms should be applicable across multiple contexts (Moss, Payne & Moore, 2014). Relative exploration orientation was again calculated by dividing the sum of words for exploration by the sum of total words for exploration and exploitation.
Appendix 5 contains Model 6-8, showcasing the results of this robustness check on the alternative
dependent variable. As can be seen from Model 7, CEO narcissism has a non-significant positive effect on the relative exploration orientation (β = 0.007, p = 0.788), thus not confirming the robustness of the results for H1. Also, CEO tenure loses its significant direct effect on relative exploration orientation following this measure (β = 0.001, p = 0.296). Model 8 shows the interaction effect of CEO tenure and CEO narcissism, which is non-significant and positive (β = 0.002, p = 0.586). Therefore, this robustness check also fails to confirm the robustness of H2.
Lastly, the choice of the regression type might have had an influence on the results. Therefore, I
26
regressions. First, the direct effect of CEO tenure on the relative exploration orientation is negative and significant at the 10% level (β = -0.002, p = 0.055 in Model 9, β = -0.002, p = 0.059 in Model 10), which is in line with the results from Model 1 and Model 2. Also, Model 10 shows that CEO narcissism has a significantly negative effect on the relative exploration orientation at the 10% level (β = -0.087, p = 0.081), which is in line with the results from the main analysis from Model 2.
However, an interesting note is that the R-squared of Model 9, which includes the control variables and the direct effect of CEO tenure, is 0.470, while the R-squared of Model 10 which also includes the direct effect of CEO narcissism is lower at 0.466. This is surprising, given that CEO narcissism has a significant effect on the relative exploration orientation, and thus should explain a part of the variance of the
dependent variable, meaning the R-squared should go up when it is included. It cannot be the result of methodological issues, as the number of firms and observations is the same and the regressions also are the exact same except for the addition of CEO narcissism in Model 10. Therefore, it is puzzling why the R-squared has actually decreased after the addition of CEO narcissism.
Next, Model 11 shows a negative interaction effect between CEO tenure and CEO narcissism that is insignificant (β = -0.012, p = 0.137). Thus, the random effects GLS regressions confirm the robustness of H1, but do not confirm the robustness of H2.
27
5. Discussion and conclusion
This final section first contains an overview of the research and the concluding results, accompanied by theoretical and methodological explanations. Next, I discuss some cautious theoretical and managerial implications. Finally, I review the limitations of the thesis and provide suggestions for future research.
In this thesis, I aimed to answer the research question: What is the effect of the degree of CEO narcissism on the firm’s relative exploration orientation, and how is this effect moderated by CEO tenure? I
expected that narcissistic CEOs would emphasize the “I” in “Innovation” by establishing an orientation that could aid them in achieving their personal goals. Building on the mechanism of the self-verification theory, I predicted that CEO narcissism would be positively related to the relative exploration orientation (H1), since an exploration orientation would be an excellent tool for narcissistic CEOs in self-verifying their concept of relative uniqueness while an exploitation orientation would make other actors perceive them as more commonplace. Moreover, I expected that an increased CEO tenure would attenuate this effect, indicating a negative interaction effect (H2), since the unit of social comparison changes over time from the previous CEO to the narcissistic CEO’s past self.
Based on random effects tobit regressions on panel data from my sample of 128 firms from the S&P 100 between 2008 and 2018, I found that CEO narcissism has a significantly negative effect on the relative exploration orientation, contrary to my expectation of H1. The expected negative interaction effect between CEO tenure and CEO narcissism of H2 was confirmed by the main model, now meaning that the negative effect of CEO narcissism on the relative exploration orientation amplifies as the tenure increases.
The result of narcissistic CEOs establishing a lower relative exploration orientation is surprising, as this would cause them be perceived as more commonplace while literature has emphasized that they aim to self-verify their identity of uniqueness (Campbell, Goodie & Foster, 2004; Campbell et al., 2011; Galvin, Lange & Ashforth, 2015). Therefore, the result does not fit with other studies that show that narcissistic CEOs engage in bold and daring activities that help them stand out more, such as risk-taking (Buyl, Boone & Wayde, 2019), dynamic strategies (Chatterjee & Hambrick, 2007) and adopting discontinuous technologies (Gerstner, 2013).
28
their concept of uniqueness, it could be that narcissistic CEOs use it as a tool to verify their self-concept of successfulness and being the right person for the job (Morf, Weir & Davidov, 2000; Morf & Rhodewalt, 2001; Petrenko et al., 2016). Even though narcissistic CEOs are overconfident, they also are fragile as they have the need to obtain feedback from other actors reinforcing that they are performing well and that they are successful at their job (Morf, Weir & Davidov, 2000; Petrenko et al., 2016). Importantly, this craving for verifying feedback is continuous: narcissistic CEOs need to obtain the positive feedback frequently and consistently, rather than occasionally (Morf & Rhodewalt, 2001; Chatterjee & Hambrick, 2007).
Following this reasoning, an exploitation orientation is a particularly useful tool for narcissistic CEOs, as its continuous short-term results provide CEOs with many frequent opportunities to obtain
self-reinforcing feedback from others (Levinthal & March, 1993; He & Wong, 2004). Instead, an exploration orientation would be less suitable, as its many uncertainties and failures often mean that exploratory innovations can take a tremendously long time to be realized (March, 1991; Levinthal & March, 1993). Therefore, it can take considerably longer for narcissistic CEOs to obtain their reinforcing feedback, they could receive it infrequently, and they might not even receive feedback for the success of an exploratory innovation when it is not finished before the end of their tenure.
Building on this theoretical reasoning, the moderating role of CEO tenure in amplifying the negative effect of CEO narcissism on the relative exploration orientation can also be explained. As the tenure increases, the negative effect of CEO narcissism on the relative exploration orientation becomes stronger, because CEOs will have increasingly more time to ‘re-shape’ the organization, its strategy and its
employees to their personal preferences through reinforcing decisions (Hambrick & Fukutomi, 1991; Miller, 1991). That is why Hambrick & Fukutomi (1991) stated that an organization and its characteristics will increasingly reflect the CEO’s personality as the CEO tenure becomes longer.
29
However, most of the robustness checks I conducted failed to confirm the results from the main models, therefore raising concerns about the results’ robustness. From a theoretical perspective, these inconsistent results may be the result of the counteracting forces of narcissistic CEOs on the one hand aiming to verify their self-concept of uniqueness, thus requiring more exploration, and on the other hand aiming to
frequently verify their self-concept of successfulness, requiring more exploitation.
Nonetheless, it is also likely that the robustness checks have failed to confirm the results due to
methodological reasons. First, the robustness check using the size of the CEO picture in annual reports as an alternative measure of narcissism had the issue that it may not have been as ‘pure’ of a measure of narcissism compared to the CATA measures. This is because firms typically use their annual reports as tools for impression management, meaning that analysts in the firm tend to double-check the reports to ensure that they contain no elements that could lead to unfavourable suggestions, such as an overly large picture of the CEO that could cause others to perceive the CEO as selfish (Stanton, Stanton & Pires, 2004).
Second, the robustness check using the alternative CATA dictionaries to measure the relative exploration orientation following Moss, Payne & Moore (2014) might have had the concern that they were developed for high-tech industries, meaning some words might not have been applicable to every firm in the S&P 100. Also, compared to the dictionaries of Uotila et al. (2009), the dictionaries of Moss, Payne & Moore (2014) included considerably more words that were not directly based on the original definitions of March (1991), meaning they might not have captured the essence of the exploration and exploitation constructs.
Lastly, my robustness check using random effects GLS regressions confirmed my first hypothesis, but failed to confirm the second hypothesis. This could be since GLS regressions are more suited to continuous dependent variables that are unlimited in range (Cameron & Trivedi, 2005), while my dependent variable was limited in range between 0 and 1.
5.1 Theoretical implications
30
underlying strategic decision making, and this thesis aimed to do so through furthering the currently limited understanding of how CEO narcissism influences innovation. While prior studies have shown that narcissistic CEOs overinvest in R&D (Ham, Seybert & Wang, 2018) and more aggressively adopt
discontinuous technologies from the environment (Gerstner et al., 2013), the prior state of research offered no understanding of whether and how CEO narcissism would influence the innovative orientation of the firm’s activities.
Through the integration of the self-verification theory with the upper echelon theory, this thesis aimed to fill this research gap. Specifically, the results of the main model suggest that narcissistic CEOs establish a lower relative exploration orientation, as this aids them in obtaining frequent feedback that continuously verifies their self-concept of successfulness and being the right person for the job. Moreover, this thesis deepened the knowledge on this underlying mechanism by showing that the negative effect of CEO narcissism on the relative exploration orientation amplifies as the CEO tenure increases, because CEOs will have had more time to re-establish the firm’s innovative orientation to their preferences.
Thus, while Gerstner et al. (2013) found that narcissistic CEOs more aggressively adopt discontinuous technologies from the environment, this thesis showed that narcissistic CEOs shift the innovative orientation of the firm’s own activities towards the exploitation of existing competencies and technologies, such as externally sourced discontinuous technologies. As such, this thesis shows that scholars should take into account that CEO narcissism not only influences the adoption of external technologies, but also influences the innovative orientation of the firm’s own activities.
Lastly, this thesis contributes to further research on CEO personality and psychology through CATA, through the development of a script that isolates the CEO’s contributions in question and answer sessions from textual transcripts of corporate earning calls.
5.2 Managerial implications
31
find CEO narcissism to be counterproductive. The same implications hold for firms that aim to maintain their desired status quo of their innovative orientation: when this status quo is at the exploitation end of the spectrum, appointing a narcissistic CEO can be advantageous, but when the status quo is towards the exploration end of the spectrum, CEO narcissism could be unfavourable.
Moreover, following this same reasoning, these insights can aid the board in when they should ‘compensate’ for and oppose to the narcissistic CEO’s bias towards exploitation. If a shift towards exploitation (or maintenance of a low relative exploration orientation) is incongruent with the firm’s desired innovative orientation, the board should try to counterbalance the narcissistic CEO’s tendency to emphasize exploitation. This becomes especially salient with higher levels of CEO tenure, since then the negative effect of CEO narcissism on the relative exploration orientation amplifies.
5.3 Limitations and future research
In this thesis, several limitations arose that limited the robustness and generalizability of the findings. These are issues that could be addressed in future research.
First, there were limitations regarding the sample and context of the study. The sample consisted of firms from the S&P 100, which are the largest publicly traded firms in the US. This limits the generalizability of findings, as the results are purely confined to large public firms, while no SMEs were included in the study. Therefore, it could be interesting for future research to replicate this study with a sample of SMEs, for example because CEOs of SMEs typically have more influence, which could mean that an SME’s characteristics - such as its innovative orientation - could be an even stronger reflection of the CEO’s personality (Hsu, Chen & Cheng, 2013). Next to this, the study’s geographical context was limited to firms from the US, and it did not account for differences in the nationality of the CEO. Therefore, future research could consider whether the study’s results would be different between firms from diverse countries and between CEOs from diverse countries, since countries with individualistic cultures, such as the US, display greater narcissism (Foster, Campbell & Twenge, 2003).
32
r=0.95 with the CEO narcissism measure from the manually parsed CEO contributions, future research could still aim to improve the script to increase its accuracy.
Next to this, it could be that the CATA-based measurement of narcissism, constructed as the relative number of first-person singular pronouns used, only provided a limited representation of how the complex trait narcissism can manifest itself (Chatterjee & Hambrick, 2007). To gain a more complete overview of the degree of CEO narcissism, future research could measure the trait from multiple angles: the CATA-based measure could be complemented with surveys, such as the comprehensive Narcissism Personality Inventory (NPI) survey from Raskin & Hall (1979), as well as with measures of relative CEO
compensation and CEO prominence in press releases by the firm (Chatterjee & Hambrick, 2007;
Rijsenbilt & Commandeur, 2013), the size of the CEO signature (Ham, Seybert & Wang, 2018) and third-party ratings of video samples from the CEO (Petrenko et al., 2016).
Another limitation could have arisen from the measurement of the dependent variable, relative
exploration orientation. This measure was constructed through CATA of annual reports, but basing the measure on publicly available documents has the possible concern that firms might obscure or disguise that they are pursuing potential innovations that could provide them with a competitive edge (Arvidsson, 2011). This in turn might have decreased the accuracy of the measure of the relative exploration
orientation. Therefore, future research might consider also using surveys to measure the relative exploration orientation, following Sidhu, Volberda & Commandeur (2004), as firms might be more willing to fully discuss their innovation orientation when it remains confidential.
Moreover, this thesis only considered the effects of CEO narcissism on the firm’s innovative orientation towards exploration or exploitation, meaning it could be interesting for future research to analyse whether the effects of CEO narcissism are similar on more concrete outputs of exploration and exploitation, such as exploratory and exploitative patents (Guan & Liu, 2016).
33
The results of this thesis might furthermore have been susceptible to endogeneity, as narcissistic CEOs might be drawn to certain aspects of the firm which were not included in the model such as a low return on assets, since this facilitates narcissistic CEOs in obtaining feedback for achieving improvement (Chatterjee & Hambrick, 2007). Also, there might have been reverse causality as narcissistic CEOs might be drawn towards firms with a low relative exploration orientation, because this would aid them in achieving frequent positive feedback. Future research could aim to more elaborately deal with these problems through, for example, using the Generalized Methods of Moments (GMM), which is a tool that addresses endogeneity through transforming the variables by subtracting their past value from their present value (Ullah, Akhtar & Zaefarian, 2018).
Additionally, due to time constraints, this study was limited in the sense that it only considered the influence of narcissism of the CEO on the relative exploration orientation, as CEOs are the primary strategic decision makers in the firm (Busenbark, 2016). Still, following the TMT stream of the upper echelon theory, future research could examine whether narcissism of other top executives, such as the CFO, and narcissism across the entire TMT would influence the relative exploration orientation.
34
References
Andriopoulos, C., & Lewis, M. W. (2009). Exploitation-exploration tensions and organizational ambidexterity: Managing paradoxes of innovation. Organization science, 20(4), 696-717.
Arvidsson, S. (2011). Disclosure of non‐financial information in the annual report. Journal of intellectual capital, 12(2), 277-300.
Bantel, K. A., & Jackson, S. E. (1989). Top management and innovations in banking: Does the composition of the top team make a difference? Strategic Management Journal, 10(1), 107-124.
Barker, V. L., & Mueller, G. C. (2002). CEO characteristics and firm R&D spending. Management Science, 48(6), 782–801.
Belsley, D. A., Kuh, E. & Welsch, R. E. (1980). Regression Diagnostics: Identifying Influential Data and Sources of Collinearity. John Wiley & Sons.
Bertrand, M., & Schoar, A. (2003). Managing with style: The effect of managers on firm policies. Quarterly Journal of Economics, 118(4), 1169–1208.
Busenbark, J. R., Krause, R., Boivie, S., & Graffin, S. D. (2016). Toward a configurational perspective on the CEO: A review and synthesis of the management literature. Journal of Management, 42(1), 234-268.
Buyl, T., Boone, C., & Wade, J. B. (2019). CEO narcissism, risk-taking, and resilience: An empirical analysis in US commercial banks. Journal of Management, 45(4), 1372-1400.
Cameron, A. C., & Trivedi, P. K. (2005). Microeconometrics: methods and applications. Cambridge University Press.
Campbell, W. K., Goodie, A. S., & Foster, J. D. (2004). Narcissism, confidence, and risk attitude. Journal of behavioral decision making, 17(4), 297-311.
