The CEO−CFO pay gap: executive power and celebrity at play

Vertical pay gaps between top management team (TMT) members and rank-and-file employees are well-documented in upper echelons and compensation literature (Boone et al., 2024; Mueller et al., 2017; Song et al., 2019). Interestingly, however, consulting firms and executive search firms have recently identified widening horizontal pay gaps between TMT members, particularly between CEOs and CFOs (BDO, 2020; Hodgson, 2022; Schneider, 2024). Such widening executive pay gaps may have negative consequences for firms, with existing studies stressing impaired CEO−CFO collaboration, increased greenwashing, more earnings management, and lower firm value as outcomes (Kubick and Masli, 2016; Lee et al., 2026; Li and Chen, 2024; Park, 2017; Siegel and Hambrick, 2005). While there is empirical evidence about the implications CEO−CFO pay gaps have at the firm level, the causes of widening CEO−CFO pay gaps remain unclear. Investigating the latter, however, is essential to avoid negative firm-level consequences. In this paper, we thus follow prior calls in top management literature to examine horizontal pay gaps, with a particular focus on their antecedents (Fredrickson et al., 2010; Pissaris et al., 2017; Uhde et al., 2017).

By studying antecedents of horizontal pay gaps, we close an important research gap in upper echelons and compensation literature. Hereby, we focus on differences in the power and celebrity of CEOs and CFOs, both of which are considered key characteristics of top managers (Finkelstein, 1992; Heavey et al., 2020; Lovelace et al., 2018; Ozgen et al., 2025; van Essen et al., 2015). Up to now, their interplay has been insufficiently theorized and empirically tested within academic literature. We draw on managerial power theory and impression management literature to explain how executives exploit relative power and engage in impression management to boost their celebrity with the aim of improving their pay compared to others (i.e. to widen pay gaps). Precisely, we theoretically develop the following novel mediation framework: (i) CEO−CFO power gaps affect CEO−CFO pay gaps, (ii) CEO−CFO power gaps affect CEO−CFO celebrity gaps, and (iii) CEO−CFO celebrity gaps mediate between CEO−CFO power gaps and CEO−CFO pay gaps.

To test our hypotheses, we use a sample of US S&P 500 CEOs and CFOs (for years 2015–2018). The United States (US) is a particularly interesting context to study executive power, celebrity, and pay. Executive pay in the US is market-driven, top executives have a high level of discretion and influence, and individuals at the apex of a firm matter also in public perception—giving power and celebrity room to unfold. We employ panel regression analysis to test our hypotheses. With regard to (i), our analyses widely confirm the positive association between CEO−CFO power gaps and CEO−CFO pay gaps. With respect to (ii), we show that CEO−CFO power gaps mostly explain CEO−CFO celebrity gaps. Finally, for (iii), our analyses demonstrate that the CEO−CFO celebrity gap mediates the relationship between CEO−CFO power gaps and pay gaps, thus empirically confirming our proposed framework.

Differentiating between different dimensions of power, i.e. ownership power, prestige power, expert power, and structural power (Blagoeva et al., 2020; Finkelstein, 1992; Kerai and George, 2025; Ozgen et al., 2025) enables us to gain fine-grained insights. Precisely, prestige power, expert power, and structural power gaps explain CEO−CFO pay gaps, whereas ownership power is not a significant explanatory factor. Ownership, prestige, and structural power determine CEO−CFO power gaps' influence on CEO−CFO celebrity gaps, while expert power gaps do not.

Responding to calls for research on horizontal pay dispersion in the TMT (Uhde et al., 2017, p. 128), we study a widely disregarded phenomenon. Compared to the study of vertical pay dispersion, literature on horizontal pay dispersion is underdeveloped (Boone et al., 2024; Cabezon, 2025; Lee et al., 2019; Mueller et al., 2017). Our study thus addresses a relevant gap in the literature, i.e. why pay gaps between CEOs and CFOs emerge. We herein make several contributions. First, we contribute to upper echelons literature on pay, power, and celebrity, uncovering novel pathways through which executive power and celebrity interact to explain pay gaps. Particularly, we introduce a novel array, according to which power gaps shape celebrity gaps, which, in turn, mediate the impact on pay gaps. This adds a new perspective to existing literature, which has previously concentrated on the effects of celebrity on both power and pay, either in isolation (as direct effects) or in a different sequence (Edwin and Benjamin, 2017; Hayward et al., 2004; Ozgen et al., 2025; Park et al., 2014). Second, by combining managerial power theory and impression management literature, we develop novel theoretical lines of argument about how power and celebrity interact to shape pay gaps between CEOs and CFOs. By further differentiating between distinct power dimensions—ownership, prestige, expert, and structural power—we provide nuanced arguments into which sources of power matter most for explaining pay disparities. Our empirical analyses confirm the applicability of managerial power theory and impression management arguments in the context of explaining executive pay gaps. Third, through our focus on executive power and celebrity as relative constructs, we complement executive compensation literature, which has thus far concentrated on absolute pay (Cabezon, 2025; Fredrickson et al., 2010; Long et al., 2026; Pissaris et al., 2017; Uhde et al., 2017). In this connection, our findings underscore the importance of gradients in executive characteristics to explain pay gaps. By focusing on the CEO−CFO dyad, i.e. the main leadership duo of the firm, we follow recent calls articulated by Firk et al. (2024) and Harrison and Malhotra (2024) to go beyond investigations of single C-suite members and their characteristics.

Examining key antecedents of pay gaps, our research also provides value for practitioners and policymakers. Like academic literature, practitioner-oriented literature and press releases typically focus on CEOs' absolute pay levels or on vertical pay differences, mostly in the form of differences between CEO pay on the one hand and pay for rank-and-file employees on the other hand (Lichtenberg, 2024). However, horizontal pay differentials between top executives—particularly between CEOs and CFOs—are largely overlooked. Our insights are especially relevant to boards, compensation committees, compensation consultants, and other stakeholders, as they enable more informed decisions or recommendations surrounding relative executive compensation. Based on our results, we advocate for better consideration of relative executive characteristics, such as power and celebrity, not only in pay setting but also in executive selection.

Our paper is structured as follows. In Chapter 2, we provide an overview of the literature on TMT pay gaps and develop our hypotheses based on managerial power theory and impression management literature. Afterwards, we describe our method (Chapter 3) and present the results of the main analysis and additional analyses (Chapter 4). In Chapter 5, we discuss our empirical findings; this discussion informs our contributions to literature and theory, as well as our practical implications elaborated on in Chapter 6. We conclude with limitations and avenues for future research (Chapter 7), outlining the potential of alternative data sources (e.g. social media data) or investigating other study contexts (like China or Europe).

While literature on absolute executive compensation is abundant (Antoons and Vandebeek, 2024; Bebchuk and Fried, 2004; Cabezon, 2025; Sanchez-Marin and Samuel Baixauli-Soler, 2014), studies on relative compensation between TMT members are rare. Within relative executive compensation literature, pay gaps in the TMT have received less attention (Fredrickson et al., 2010; Park, 2017; Sahib et al., 2018). Among the few existing studies, scholars have investigated the individual- and team-level consequences of pay gaps, such as perceived inequality and unfairness, reduced job satisfaction (Fong et al., 2009), motivation (Connelly et al., 2016), productivity, and team performance (Lee et al., 2008). Pay gaps may increase executive turnover rates (Pissaris et al., 2017) or decrease commitment to the firm as executives feel undervalued (Trevor and Wazeter, 2006).

Other studies link pay gaps to firm-level outcomes. Cheng et al. (2024) establish a connection between pay gaps and forecasting quality. Bebchuk et al. (2011) show that pay gaps between CEO and TMT pay positively relate to lower firm value and stock returns. Park (2017) demonstrates that larger CEO−TMT pay gaps reflect an increase in earnings management activities. Siegel and Hambrick (2005) reveal a negative relationship between the CEO−TMT pay gap and firm performance, while Kale et al. (2009) and Lee et al. (2008) identify a positive relationship. Fredrickson et al. (2010) partially solve this dilemma by revealing that pay gaps between the CEO and the TMT are negatively related to firm performance when they exceed a certain level.

While existing literature has studied the consequences of pay gaps, their determinants remain poorly understood (Uhde et al., 2017). Addressing this gap, we focus on the two most important positions in the C-suite, i.e. the CEO and the CFO (Schmid and Altfeld, 2018). Nowadays, the CEO−CFO dyad is often referred to as the strategic leadership duo in a firm making key decisions (Han et al., 2015; Ren and Zhong, 2025). While historically, in most firms, CFOs were primarily responsible for financial functions, their general strategic responsibilities have grown in recent times (Ginesti et al., 2021; Zhang et al., 2024; Zorn, 2004). The CFO is the closest working partner of the CEO (Wang et al., 2020) and “second in command” (Six et al., 2013; Talukdar et al., 2025), also often referred to as the CEO's “co-pilot” or “strategic partner” (Uhde et al., 2017). The CFO's importance has also risen due to the growing effect of financial strategy on overall firm performance (Karaevli and Ozcan, 2022).

Recent reports by consulting and executive search firms document a widening pay gap between the CEO and the CFO (BDO, 2020; Hodgson, 2022; Schneider, 2024). Focusing on the CEO−CFO pay gap thus appears highly relevant, rather than analyzing CEO pay and CFO pay in isolation—as in prior studies (see, for example, Barbi et al., 2024; Firk et al., 2024; Lewellyn and Muller-Kahle, 2022). The few studies that have focused on CEO−CFO pay gaps have concentrated on the consequences of such pay gaps. Kubick and Masli (2016) demonstrate that a larger CEO−CFO pay gap leads to tax avoidance. Jia (2017) finds a positive link with audit fees. Han et al. (2022) report a positive relation with aggressive investment policies.

We depart from this outcome-centric focus by concentrating on antecedents. In addition, instead of vertical pay gaps, we place emphasis on horizontal pay gaps. Vertical pay gaps, such as those between TMT members and rank-and-file employees, have already been studied thoroughly (Connelly et al., 2016; Kiatpongsan and Norton, 2014; Weng and Yang, 2024), including their antecedents (Mueller et al., 2017; Song et al., 2019). Horizontal pay gaps refer to differences in pay between members at the same hierarchical level, for instance, between different TMT members (Park, 2017; Pissaris et al., 2017). For these horizontal pay gaps, antecedent-centric research is missing despite having been called for (e.g. Uhde et al., 2017). Understanding horizontal pay gaps' antecedents is, however, crucial if we want to better understand managerial influence on firms.

Literature on the CEO−CFO power gap has thus far focused on its consequences. Baker et al. (2019) suggest that a larger CEO−CFO power gap is positively related to an increase in earnings management, while Wu (2019) offers evidence that a lower CEO−CFO power gap leads to less earnings management. Prior research also establishes that the ability of the CEO to pressure the CFO to engage in earnings management is dependent on the respective power constellation in the C-suite (Alkebsee et al., 2025; Florackis and Sainani, 2021). Dikolli et al. (2021) investigate the link between CEO and CFO power and remuneration, suggesting that CEOs with high power urge CFOs with less power to overstate results with the aim of boosting their own remuneration. Other studies contend that powerful executives receive higher remuneration compared to their less powerful counterparts (see, for example, Choe et al., 2014; Collins et al., 2018).

The latter findings resonate with literature on executive power within TMTs as a relative concept (Ocasio, 1994; Ozgen et al., 2025; van Essen et al., 2015). Following this logic, strategic outcomes are not only determined by absolute levels of power (Brahma and Economou, 2024; Burkert, 2026) but also by relative levels of power. Consequently, gaps in power might explain extractive practices of CEOs to increase their own pay (as in Dikolli et al., 2021) and thus differences in pay between themselves and CFOs.

Based on managerial power theory, we reason that executives who are more powerful can increase their pay, as they have greater access to—or control over—a variety of tools and mechanisms to affect compensation, compared to their less powerful counterparts (Abernethy et al., 2015; Bebchuk et al., 2002). We suggest that these differences between more and less powerful executives materialize via various power dimensions, i.e. ownership, prestige, expert, and structural power (Finkelstein, 1992; Ozgen et al., 2025).

Ownership power. Ownership power results from executives' equity. Holding significant amounts of shares in a firm improves an executive's position within the organization (Daily and Johnson, 1997; Jensen and Meckling, 1976; Nel et al., 2024). Ownership enables executives to exercise greater influence over the firm's strategic direction (Chen et al., 2011). Executives who are shareholders have a strong position vis-à-vis the firm's board and can thus partially avoid internal corporate control (McClelland et al., 2012). Specifically, ownership allows executives to exercise influence over boards, their agendas, and decisions (Bebchuk et al., 2001; Combs et al., 2007; Finkelstein, 1992).

Drawing on these channels, both CEOs and CFOs can use ownership power to influence the firm and their own compensation, respectively (Duong and Evans, 2015). Nevertheless, CEOs and CFOs do not necessarily have the same ownership power. Often, the ownership power of the CEO is considered greater than that of the CFO. Yet, some CFOs may also hold significant shares, enhancing their ownership power (Xu et al., 2022). If the ownership power of CEOs is high and the ownership power of CFOs is low, CEOs' respective ability to influence key actors and important stakeholders in remuneration-setting processes to their advantage is heightened (or vice versa). A larger CEO−CFO ownership gap thus increases pay differences, while a smaller gap reduces them. We hypothesize:

Prestige power. Another important source of managerial power is prestige, built up via a strong and large personal network of executives (Boivie et al., 2012; Diga and Kelleher, 2009; Engelberg et al., 2013). Prestige power grants executives access to rare resources, high-quality information, guidance from renowned peers, or political favors (Acharya and Pollock, 2013; Blagoeva et al., 2020). Consequently, it facilitates governance via social network effects, increasing the overall influence of executives (Westphal and Zajac, 2013).

Consistent with this logic, Belliveau et al. (1996) find a positive relationship between prestige power and executive compensation. Prestige power can be interpreted as a key resource that causes those responsible for remuneration-setting to offer higher executive compensation. Wade et al. (1990) illustrate that prestige power leads to securing CEO compensation with a special severance payment, called “golden parachute.” In terms of absolute pay, Engelberg et al. (2013) suggest that CEOs with more prestige power earn more; Datta and Iskandar-Datta (2014) reach similar findings for CFOs.

Although the CEO usually holds the most prestigious and publicly visible role in a firm, CFOs are also able to accumulate substantial prestige power. By serving on boards of other firms or by engaging in additional networking activities, both CEOs and CFOs can boost their prestige relative to each other (Fogel et al., 2018). With prestige power affecting their opportunities to leverage networks and stakeholder relationships, executives' prestige power puts them in a better position in pay negotiations. We hypothesize:

Expert power. Expert power results from knowledge, experience, and skills gained throughout a career, which make executives capable of handling job-related challenges (Haynes and Hillman, 2010; Lammers and Pape, 2025). Expert power enables executives to better carry out strategic analyses to the benefit of their firms (Kunisch et al., 2019) and to take more informed decisions (Park and Tzabbar, 2016). A track record of strong performance in the past increases an executive's expert power, which is why tenure has often been linked to expert power (Blagoeva et al., 2020; Combs et al., 2007). Overall, expert power can be interpreted as a means to improve internal processes and operational efficiency, making it a valuable resource in itself that also leverages the employment of other resources (Kor and Mahoney, 2005).

While any executive who serves the organization can bring in expert power, there may be different degrees of expert power. Literature reveals that expert power differences impact executive compensation, as specific knowledge, experience, and skills can be considered leverage in pay negotiations (Faulkender and Yang, 2013; van Essen et al., 2015). More expert power, accordingly, would enable a CEO or a CFO to convince compensation-setters more effectively of the value their expert resource would provide to advance the organization (compared to their counterparts with lower expert power). Expert power thus serves as a signal affecting pay gaps. We hypothesize:

Structural power. Structural power originates from the hierarchical setup in the firm, i.e. it results from the formal governance position held by an executive (Crossland and Hambrick, 2011; Harrison and Malhotra, 2024). Holding a key position in the organizational hierarchy reinforces an executive's structural power (Finkelstein and D'Aveni, 1994; Haynes and Hillman, 2010; Krause et al., 2022). This, in turn, enables them to exercise more control in general (Acharya and Pollock, 2013; Harrison and Malhotra, 2024), e.g. by guiding decision-making processes via organizing and determining the agenda for board meetings (Feng et al., 2021).

Structural power, for example, reflected by CEO (or CFO) duality, highlights the relative position of an executive compared to other executives (Florackis and Sainani, 2021). CEOs or CFOs with relatively higher structural power have a more relevant say in important decisions, such as nominating and reappointing directors, as well as director compensation in general (Bedard et al., 2014). Stronger structural power enables a CEO and a CFO to forge social ties, creating deeper interdependencies between themselves and board directors and, consequently, allowing them to influence compensation-setting and improve their relative position (Bedard et al., 2014; Hoitash, 2011). In case of a structural power gap between the CEO and the CFO, the more powerful party can exploit their standing more effectively vis-à-vis key stakeholders, increasing the pay gap. We hypothesize:

While managerial power is built through an executive's personal progress and acquired authority, executive celebrity is regularly caused by enhanced media attention (Yoon et al., 2025). Media attention causes celebrity executives to be recognized and triggers emotional responses of a broader public audience (Cabano et al., 2024; Cho and Choi, 2024; Hambrick and Wowak, 2021; Lovelace et al., 2018; Pollock et al., 2019). Accordingly, managerial power and celebrity are distinct constructs: while managerial power reflects top managers' channels of direct influence on strategy and decision-making in the firm, celebrity rather illustrates the radiance an executive exhibits to the general public and stakeholders not directly involved in the firm's core business (Cao et al., 2025; Finkelstein, 1992; Hayward et al., 2004; Wade et al., 2006).

Following impression management literature, executives boost celebrity by managing communication as well as delaying, manipulating, or reframing information (Friedman, 2014). To elevate their celebrity strategically, executives employ impression management techniques (Park and Berger, 2004; Westphal and Deephouse, 2011). They use a variety of approaches when purposely generating, carefully selecting, and planning information about their abilities (Bolino et al., 2016; Liu and Atinc, 2021; Moreno and Jones, 2022; Treadway et al., 2009).

We reason that impression management, as a means to enhance celebrity, has a foundation in executive power (Davidson et al., 2004). The more powerful executives are, the more influence they exert over a firm's internal and external communication in general and how information is conveyed (Friedman, 2014). Furthermore, the power of executives to allocate resources to marketing campaigns or communication projects can influence executive celebrity (Haynes and Hillman, 2010). Thus, we propose that powerful executives are more likely to attain a higher level of celebrity compared to their less powerful counterparts. How this translation of managerial power into celebrity happens is determined by power, with its different dimensions.

Ownership power. Ownership power endows executives with a greater degree of independence from corporate control and the board of directors (Combs et al., 2007; McClelland et al., 2012). This allows them to allocate financial resources more freely and strengthen impression management (Blagoeva et al., 2020; Haynes and Hillman, 2010). Executives with substantial ownership power can exploit their power for their personal benefit, fostering impression management to enhance not only the firm's but also their own celebrity. We argue that differences in CEO−CFO ownership power may also lead to differences in the extent to which executives can leverage their celebrity. We hypothesize:

Prestige power. Informal power derived from prestige, the willingness to use prestige power, and the ability to demonstrate this power to others are key elements in successful impression management (Pollach and Kerbler, 2011). An executive's prestige power may be reflected through high-profile affiliations, such as membership on prestigious boards or leadership positions in influential nonprofit organizations highly visible to stakeholders (Malhotra et al., 2021). High visibility, combined with being well-connected, can enhance the effectiveness of an executive's impression management (Bolino et al., 2016) and thus further increase their celebrity. We thus posit that differences in prestige power between the CEO and CFO may lead to differences in celebrity, since the party with more prestige power can engage in more effective impression management. We hypothesize:

Expert power. Expert power matters since media prefer to engage with experts due to their credibility and reliability in providing insights on strategic issues (Westphal and Deephouse, 2011). Executives with high levels of expert power can therefore leverage the media's demand for experts, making positive impression management easier to achieve. Additionally, executives with significant expert power often have superior communication skills (Berger, 2005). We therefore argue that executives with greater expert power will be more successful in engaging in impression management, resulting in increased levels of celebrity. Differences in expert power between the CEO and CFO may thus lead to differences in celebrity. We hypothesize:

Structural power. Previous research reasons that structural power facilitates the allocation of organizational resources (Finkelstein, 1992; Lewellyn and Fainshmidt, 2017). Executives with high structural power can exercise strong control over firm communication, which enhances opportunities for successful impression management (Davidson et al., 2004). Executives may wield their structural power in ways that serve their own interests rather than those of the firm (Carty and Weiss, 2012; Finkelstein and D’Aveni, 1994). Accordingly, we suggest that executives with more structural power are more successful in leveraging control over communication and, therefore, achieve higher levels of celebrity using impression management. Gaps in structural power between the CEO and the CFO are thus expected to broaden celebrity gaps. We hypothesize:

Beyond the aforementioned relationships, we suggest that CEO−CFO celebrity gaps mediate the effects of CEO−CFO power gaps on their pay gaps. When it comes to the link between celebrity and pay in general, celebrity executives' media presence contributes to the notion of them being responsible for the firm's success (Hayward et al., 2004; Pollock et al., 2024). If media outlets spread positive attributions, the firm and its stakeholders may develop an improved (potentially exaggerated) perception of the capabilities of celebrity executives, which enhances their reputation (Hambrick and Wowak, 2021; Rindova et al., 2006; Wade et al., 2006). Research also suggests that executives utilize their celebrity to their advantage during compensation negotiations (Long et al., 2026; Wade et al., 2006; Yoon et al., 2025). Furthermore, embracing their ascribed status allows them to claim firm successes for themselves (Hayward et al., 2004; Malmendier and Tate, 2009) and causes them to try to extract higher rents (Park et al., 2014; van Essen et al., 2015). Firms giving in to these demands may justify the higher compensation granted to celebrity executives by pointing to positive media portrayals (Graffin et al., 2008).

Relative to each other, executives can differ significantly in their celebrity, e.g. due to variations in the degree of media exposure (Ahmed et al., 2025; Lovelace et al., 2022). We transfer this logic to CEOs and CFOs, suggesting that they differ in their celebrity. We posit that CEO−CFO celebrity gaps, shaped by power differences, influence pay gaps; we herein interpret celebrity gaps as hinges between power gaps and pay gaps, or mediators of the relationships between power gaps and pay gaps. Executives use power to expand their celebrity, with the ultimate goal of increasing their compensation, causing widening gaps between themselves and other C-suite members.

We propose that said mediation effect happens via the transformation of power differences into symbolic and reputational value (i.e. celebrity differences). Through strategic control and resource access, powerful executives, compared to their less powerful counterparts, shape media narratives and dominate external communication. Accordingly, power allows them to develop relatively more celebrity capital (Rindova et al., 2006). This immaterial asset permits them to signal competence and indispensability, which, in turn, they can draw on to justify compensation differences when compensation setting takes place. Particularly, if their counterparts are less powerful and thus possess less celebrity capital, this will expand the CEO−CFO pay gap. We hypothesize as follows:

Figure 1 summarizes our conceptual framework and hypotheses.

To test our hypotheses, we use a sample based on a panel of all S&P 500 CEOs and CFOs, covering the years 2015–2018. The choice of this sample period is motivated by several factors. First, we start our sample period in 2015, as this was the first year in which CFO information was broadly available for US firms. The Dodd-Frank Act, introduced in 2010, imposed additional reporting duties on firms regarding their CFOs, including their compensation (Huntington et al., 2010). Hence, although regulations have mandated the disclosure of CFO compensation since the Dodd-Frank Act, comprehensive and consistent data has only been available for 2015 and the following years due to standardized data collection and improved reporting practices. While partial CFO data exists for earlier years, reporting and database coverage were still incomplete and less standardized prior to 2015. Second, in our endeavor to develop a baseline understanding of the nexus between CEO−CFO power gaps, celebrity gaps, and pay gaps, we aimed for a sample period unaffected by major macro developments. We thus opted against the inclusion of years in which the consequences of Covid-19 were felt most heavily (2020–2022); this approach is well in line with other research in the field (Gong et al., 2024; Harrison and Malhotra, 2024). As we work with lagged variables (to account for the time lag between power and celebrity, on the one hand, and pay, on the other hand), for some variables, our database covers the years 2016–2019 (instead of 2015–2018). Third, while we could have opted for a very recent sample excluding Covid-19 years, this would have left us with a two-year sample at best (2023–2024). Therefore, we decided to choose a larger sample period, albeit with the consequence of using slightly older data. This is in line with literature advocating longer sample periods, and again aligns with approaches of other scholars in the field who make similar sample choices (Byrka-Kita and Bulasiński, 2024; Lee and Landers, 2022). Our approach is thus in line with Ketchen et al. (2023), who suggest that using older data is justified when newer data intended for the purpose of the research is not available (here, a longer sample period relatively shielded from major macro developments). Fourth, our data is not merely extracted from databases but also partially results from cumbersome manual data collection and treatment, which is another reason for drawing on older data (Ketchen et al., 2023). Altogether, we thus consider 2015–2018 a suitable sample period.

To construct our final sample, we first excluded firms for which CEO and CFO data was incomplete. This left us with 414 CEO−CFO pairs (i.e. 414 firms) and 1,656 firm-year observations. Afterwards, we removed those firm years from the reduced sample during which the CEO−CFO pair was not jointly in office (since either the CEO or the CFO was not serving their role in a specific year). The resulting final baseline sample comprised 1,243 firm-year observations for the 414 CEO−CFO pairs (firms).

Our data originates from various databases. Data on CEO and CFO compensation comes from ExecuComp, which was also used to retrieve other individual-level data (such as gender or education). Data on celebrity was collected via LexisNexis. To extract media coverage data, we developed an algorithm to analyze 18,116 news articles about the respective CEOs and CFOs. Data on executive power and board characteristics (such as board size or CEO duality) was sourced from BoardEx and ExecuComp. Firm-level data was retrieved from Compustat. Missing data was collected manually; particularly, information on CFO characteristics (such as tenure, educational background) unavailable in primary databases was extracted by hand from annual reports, company websites, and other publicly accessible sources.

Compensation is measured as a CEO's or CFO's total annual compensation (Bebchuk et al., 2011; Kalogeraki and Georgakakis, 2022). Total compensation comprises fixed and variable components, i.e. salary, bonuses, restricted stock, stock options, and long-term incentives. We build on previous research to calculate the CEO−CFO pay gap as the natural logarithm of their compensation difference (Han et al., 2022; Kubick and Masli, 2016). Following prior research, and to circumvent heteroscedasticity issues, we apply a logarithmic transformation to the CEO−CFO pay gap variable (Datta and Iskandar-Datta, 2014). To prevent reverse causality issues, compensation is measured in t = 1, while power and celebrity are measured for t = 0.

We use four measures to reflect our power dimensions (Finkelstein, 1992, 2009; van Essen et al., 2015). We measure the ownership power, prestige power, expert power, and structural power of each CEO and CFO. For each of the four dimensions and each CEO−CFO pair, we calculate the difference between CEO power and CFO power to arrive at the CEO−CFO power gap.

In line with previous literature, we measure ownership power as the percentage of shares outstanding held by the executive (Blagoeva et al., 2020; Lammers and Pape, 2025; Xu et al., 2022). To calculate the ownership power gap, we create an ownership power measure for each executive (shown in $EQ$₁), where $OP$ illustrates ownership power and $Sp$ stands for shares outstanding held by the executive $i$ as a percentage. The index takes a value between 0 and 1, with higher values characterizing higher ownership power. The ownership power gap equals CEO ownership minus CFO ownership.

Based on authoritative literature, prestige power is gauged by current board mandates and an individual's network size (Malhotra et al., 2021). We employ a composite index (⁠$EQ$₂), where $PP$ represents the prestige power of the executive $i$⁠, $NBM$ represents the number of board mandates, and $NS$ represents the network size of the executive. The normalized index ranges from 0 to 1, with higher values indicating higher prestige power. The prestige power gap equals CEO prestige power minus CFO prestige power.

Expert power is measured by an executive's tenure in the firm and time in office (see also Combs et al., 2007; Florackis and Sainani, 2021). We employ a composite index (⁠$EQ$₃), where $EP$ represents expert power of the executive $i$⁠, $YO$ represents the number of years the executive has been in office, and $YF$ represents the number of years the executive has worked for the respective firm. The normalized index ranges from 0 to 1, with higher values indicating higher expert power. To determine the expert power gap, we subtract the CFO expert power from the CEO expert power.

We link structural power to CEO or CFO duality within the corresponding firm (Broye et al., 2017). CEO duality is determined by simultaneously serving as board chair, granting formal authority over the TMT and the board. This is measured as a dummy variable (1 if the CEO is board chair, 0 otherwise). Following Baker et al. (2019), we define CFO structural power as board membership, enabling influence over the board and strategic decisions. Similar to Baker et al. (2019), we calculate the structural power gap by subtracting CFO power from CEO power. This results in a categorical variable for the structural power gap, which is 1 for CEO−CFO pairs where the CEO is board chair, but the CFO is not part of the board. The structural power gaps are zero on two occasions: first, if the CEO is board chair and the CFO is also part of the board; second, if the CEO is not board chair and the CFO is not a board member. If the CEO is not the board chair, but the CFO is a board member, the variable is coded as −1.

In line with prior literature and our conceptual approach, we reason that the celebrity of a CEO or a CFO is reflected by media exposure (Hayward et al., 2004; Lovelace et al., 2022; Park et al., 2014; Pollock et al., 2024). Accordingly, we apply an established approach of operationalizing celebrity based on the number of news articles published about CEOs and CFOs (Park et al., 2014; Pfarrer et al., 2010). We retrieved CEO/CFO news articles from LexisNexis and then identified the relevant news articles by searching for articles that included the name of the CEO/CFO and referred to their respective company. Our search involved a set of renowned newspapers and business magazines, i.e. New York Times, USA Today, Los Angeles Times, Tampa Bay Times, and Star Tribune, as well as Forbes, Wired, Fast Company, and Inc. The publications have significant nationwide coverage in the United States (Park and Berger, 2004). We opted for measuring celebrity in a more sophisticated way compared to prior research (Park et al., 2014; Pfarrer et al., 2010) by including business magazines (Forbes, Wired, Inc.) and online articles, hence broadening the basis for our celebrity assessment. For our search, we looked for the names of the respective CEOs and CFOs included in our S&P 500 sample in the LexisNexis database. For our sample period, we gathered 18,116 news articles referencing the CEOs and CFOs in our sample.

Following Pfarrer et al. (2010) and Pollock et al. (2019), we consider celebrity measured by positive news coverage rather than overall news coverage. To arrive at our positive news coverage measure, i.e. the number of positive news articles, we exclude negative news coverage from overall news coverage. To identify negative news coverage, we use the linguistic dictionary provided by LexisNexis (2025), which helps assess whether an article contains negative expressions. The dictionary specifies stems of negative terms such as “corrupt!”, “illegal!”, or “violat!”. If an article contained any of these negative expressions, it was flagged as “negative” and excluded before constructing our celebrity measure. In other words, only articles without negative connotations enter our count variables for CEO and CFO celebrity—and thus also our celebrity gap measure. This approach ensures that our celebrity measure captures positive media attention and is not affected by negative press coverage (for a similar approach, see Pfarrer et al., 2010; Pollock et al., 2019). To decrease the potential influence of outliers, we winsorize our celebrity variable at the 5% level (Reifman and Keyton, 2010; Wan, 2014). The CEO−CFO celebrity gap is calculated by subtracting the CFO's article count from the CEO's article count.

We control for a variety of individual-, board-, firm-, and industry-level factors that may also influence pay gaps between CEOs and CFOs. At the individual level, we control for gender differences, since gender influences compensation (Chen et al., 2022). Given education's link to executive pay (Datta and Iskandar-Datta, 2014), we take into account MBA degrees (Graham et al., 2012).

Prior studies have demonstrated that board characteristics such as diversity can be significant predictors of executive compensation (Antoons and Vandebeek, 2024). Hence, we consider the proportion of female supervisory directors as well as board nationality diversity in our analysis. Furthermore, research indicates that larger boards struggle with coordination and with executive oversight, which may in turn affect executive compensation (van Essen et al., 2015). Thus, we control for board size. At the firm level, we consider firm size, as a vast body of literature confirms its influence on executive compensation (Finkelstein, 2009). It is for the same reason that we include firm age as a control (McKnight and Tomkins, 2004). Since executive compensation often includes performance-based components, we control for return on assets as a measure of firm performance (Lovelace et al., 2022). We also control for leverage (total debt to assets), as firm risk affects executive remuneration (Datta and Iskandar-Datta, 2014). To consider industry effects, we include dummy variables based on the one-digit Standard Industrial Classification (SIC) codes of businesses. Lastly, we add year dummies to control for time-specific effects.

We draw on authoritative literature to set up our empirical models for mediation analysis (Baron and Kenny, 1986; Lee and Park, 2006). In line with this literature, to translate our conceptual model with its direct and indirect effects into empirical models, the following three equations are required:

• (i; H1a-d) CEO−CFO pay gap = CEO−CFO power gap + controls

• (ii; H2a-d) CEO−CFO celebrity gap = CEO−CFO power gap + controls

• (iii; H3) CEO−CFO pay gap = CEO−CFO power gap + CEO−CFO celebrity gap + controls

To investigate our hypothesized relationships adhering to these equations, we conduct panel regression analyses. Using larger sample sizes, panels enhance statistical inference accuracy and model parameter estimates, increasing the likelihood of identifying underlying relationships and dynamics between variables. Furthermore, panel regression analysis increases statistical power by correcting for unobserved heterogeneity and incorporating individual and sample-wide effects (Greene, 2018). To address issues such as within-unit serial correlation, we use generalized least squares (GLS) (Greene, 2018; Halaby, 2004).

To determine whether a fixed effects model or a random effects model should be used, we conduct a Hausman specification test (HST) (Wooldridge, 2010). The results of the HST (see  Appendix 1) indicate that using a random effects model is appropriate for our analysis, as it helps reduce the impact of collinearity as well as unobserved heterogeneity (Bell et al., 2019). However, it should be noted that the HST assumes that the random effects estimator is efficient, which is not always the case. Therefore, we perform a Breusch-Pagan Lagrange multiplier test (BPLMT), as suggested by Greene (2018). The BPLMT indicates significant panel-level variance, meaning that a random effects specification is preferred (see  Appendix 1).

The correlations and descriptive statistics for the dependent, independent, and control variables are presented in Table 1. Mean values of our main variables of interest show that there are indeed meaningful differences between CEOs and CFOs in terms of their power and celebrity on average. Standard deviations indicate heterogeneity within our sample, which is not surprising, given that our sample covers, for instance, a variety of different industries. In terms of correlations, pay gaps are significantly correlated with power gaps for several dimensions. It becomes apparent that, for the most part, power and celebrity are not correlated. This finding confirms our conceptual distinction—even for seemingly similar constructs like prestige power and celebrity, which are uncorrelated—and justifies our reasoning behind using both power and celebrity as direct and indirect explanatory factors for CEO−CFO pay gaps.

Relationships between CEO−CFO power gaps and pay gaps. Hypotheses 1a–1d postulate a positive relationship between the different dimensions of the CEO−CFO power gap and the CEO−CFO pay gap. As demonstrated in Table 2, the hypothesized positive associations are empirically supported for three out of the four power dimensions. The associations between prestige power, expert power, and structural power gaps and the CEO−CFO pay gap are all significant at a level of p < 0.01. Therefore, we can confirm hypotheses H1b, H1c, and H1d. However, for hypothesis H1a, our analysis yields a non-significant association between the ownership power gap and pay gap.

Relationship between CEO−CFO power gaps and celebrity gaps. Hypotheses 2a–2d suggest a positive association between the four different dimensions of the CEO−CFO power gap and the CEO−CFO celebrity gap. As illustrated in Table 3, the positive association between the CEO−CFO ownership power gap and the CEO−CFO celebrity gap (H2a) is significant (p < 0.1). Furthermore, the CEO−CFO prestige power gap reveals a positive association with the CEO−CFO celebrity gap (H2b) at a significance level of p < 0.1. We also find support for our hypothesis on the positive association of the CEO−CFO structural power gap and the CEO−CFO celebrity gap (H2d) (p < 0.1). However, we do not find support for H2c, i.e. we cannot confirm a positive association between the CEO−CFO expert power gap and the CEO−CFO celebrity gap. Overall, the results imply that, for the most part, power differences between the CEO and the CFO are positively linked to the CEO−CFO celebrity gap.

Relationship between CEO−CFO power gaps, celebrity gaps, and pay gaps. Hypothesis 3 contends that the CEO−CFO celebrity gap functions as a mediator through which power gaps translate into pay gaps. As Table 4 shows, at p < 0.01, the CEO−CFO celebrity gap is consistently and positively related to the CEO−CFO pay gap across all model specifications (Models 1–6). In Model 2, the inclusion of the ownership power gap does not affect the significance of the celebrity gap, which remains positive and significant at p < 0.01. In Model 3, when adding the prestige power gap, both the prestige power gap (p < 0.01) and the celebrity gap (p < 0.01) are significantly related to the pay gap. Model 4 displays a significant effect for the expert power gap (p < 0.01) alongside a stable and highly significant association between the celebrity gap and the pay gap (p < 0.01). In Model 5, the structural power gap is also significant (p < 0.01), while the celebrity gap remains significant at p < 0.01. Finally, in the full Model 6, the celebrity gap continues to be positively and significantly associated with the pay gap (p < 0.01), while the prestige power gap (p < 0.01) and the structural power gap (p < 0.05) also retain their significance. Therefore, our findings support our proposed mediation in that the CEO−CFO celebrity gap consistently and significantly explains variations in the CEO–CFO pay gap across models. Figure 2 summarizes our hypotheses and our expected relationships as well as the results.

We perform several additional analyses to address endogeneity concerns. To deal with identification concerns (Manski, 1995), we introduce an alternative measure for our dependent variable. While our main measure, i.e. the CEO−CFO pay gap, is based on authoritative literature (Han et al., 2022; Kubick and Masli, 2016), some studies use the CEO pay slice (CPS) to assess executive pay gaps (Bebchuk et al., 2011; Chen et al., 2013). Hence, we calculate the CPS, i.e. the ratio of CEO to CFO total compensation, and test our hypotheses with this instead of the CEO−CFO pay gap. As highlighted in Table 5, the results overall align with the findings of our main analysis. All four dimensions of the CEO−CFO power gap—ownership, prestige, expert, and structural power—mostly show significant positive associations with the CPS. In addition, the CEO−CFO celebrity gap remains positively significant, as also seen in the fully specified Model 6 (p < 0.01). Our results thus further substantiate our argument that the CEO−CFO celebrity gap mediates the relationship between power gaps and compensation gaps.

In a second additional analysis, we address potential endogeneity concerns related to measurement errors (Roberts and Whited, 2013). While our main analysis relies on positive media attention (see also Pollock et al., 2019; Rindova et al., 2006), other studies conceptualize executive celebrity more broadly, irrespective of whether media coverage is positive or negative (Denner et al., 2018; Fralich and Papadopoulos, 2020). To account for this, we choose an alternative conceptualization and operationalization. We examine whether our results hold when celebrity is measured by overall media coverage, including positive and negative articles—to ensure that our findings are not driven by our initial measurement choice. We opt for a celebrity measure that includes all media articles mentioning the CEO or CFO, regardless of tone. Using this adjusted celebrity measure, we re-run our models. The results are largely consistent with our main analysis. As displayed in Table 6, we find that the CEO−CFO prestige power gap (p < 0.1), expert power gap (p < 0.05), and structural power gap (p < 0.05) remain positively and significantly associated with the CEO−CFO celebrity gap, while the ownership power gap reveals a positive but statistically insignificant relationship. These findings confirm that our results are not solely driven by the exclusion of negative media coverage and that power differences between the CEO and CFO continue to explain variations in the celebrity gap, even when overall (positive and negative) media exposure is considered. As detailed in Table 7, our additional analysis also reveals a strong and consistent association between the CEO−CFO celebrity gap and the CEO−CFO pay gap. Across all model specifications, the CEO–CFO celebrity gap is positively and significantly associated with the CEO–CFO pay gap (p < 0.05 or p < 0.01), broadly confirming a mediation effect.

In a third additional analysis, to address concerns related to reverse causality, we employ an instrumental variable approach to investigate the relationships between CEO−CFO power gaps, celebrity gaps, and pay gaps. Following authoritative literature (Semadeni et al., 2014; Staiger and Stock, 1997), we utilize a leave-one-out industry-year measure of the CEO−CFO celebrity gap as an external instrument. This instrument captures exogenous variations in executives' visibility arising from broader industry-level celebrity (Breuer, 2022) while excluding the focal firm to prevent interference by firm-specific factors. To reflect the actual distribution of media attention within each industry year, the industry-year averages are weighted by the total number of news articles published about each remaining firm. As seen in  Appendix 2, the first-stage estimates confirm the relevance of the instrument, as the leave-one-out celebrity gap strongly predicts firm-level CEO−CFO celebrity gaps (β = 0.031, p < 0.01; Kleibergen–Paap F = 93.58). In the second stage, the instrumented CEO−CFO celebrity gap remains positively and significantly associated with the pay gap (β = 0.03365, p < 0.05; Anderson–Rubin (AR) F = 22.23, Anderson-Rubin (AR) p = 0.0181), indicating that exogenous variations in celebrity causally affect pay gaps. Additional instrumental variable models using the four CEO−CFO power gaps as dependent variables yield mostly insignificant effects, suggesting that celebrity generally does not endogenously reinforce executive power ^[1].

Our first set of hypotheses contends that larger CEO−CFO power gaps result in greater CEO−CFO pay gaps. Drawing on different dimensions of power, as advocated by prior literature (Blagoeva et al., 2020; Finkelstein, 1992; Kerai and George, 2025), allows a better conceptual and empirical grasp of this relationship. The empirical results of our study support the theoretical notion of a connection between power gaps and pay gaps for three out of four power dimensions. Specifically, prestige power, expert power, and structural power gaps have a significant positive relation with the CEO−CFO pay gap. This largely confirms our arguments based on managerial power theory and underscores the relevance of power as a key executive characteristic (Ozgen et al., 2025; van Essen et al., 2015). As proposed in our hypothesis development, we empirically show that networks and personal connections (prestige power), knowledge, skills, and competencies (expert power), as well as formal ties and hierarchical positions related to other directors (structural power), can be leveraged to affect pay and thus enhance pay gaps. Contrary to our theoretical reasoning, the ownership power gap does not have a consistent direct effect on the CEO−CFO pay gap. This empirical finding contradicts the theoretical argument that personal equity in a firm necessarily enhances opportunities for further personal gain by influencing pay. In terms of predicting the effect of ownership power on pay gaps, managerial power theory may thus have limits, as it does not properly account for monitoring mechanisms. The exploitation of ownership power that we focus on in our theoretical development may be restricted by such monitoring mechanisms. Monitoring may potentially prevent executives from using ownership power for exploitative behavior (see also Conyon and He, 2011; Davila and Penalva, 2006; Sapp, 2008). Moreover, incentive alignment may be another counteractive measure. Existing literature suggests that the main reason why executives are granted ownership in firms is for the convergence of shareholder and executive interests (Hoskisson et al., 2002; Ozgen et al., 2025). Although executives with higher relative ownership power have the potential to use power to increase their compensation, they may also have reduced—or no—motivation to seek personal benefits, which could explain this particular result.

Our second set of hypotheses builds on the premise that power gaps influence celebrity gaps. Both power and celebrity have already been referred to as key executive characteristics (Heavey et al., 2020; Lovelace et al., 2018). Investigating their relationship empirically, we close a research gap: the interplay of executives' power and celebrity has been insufficiently researched, particularly in relative terms. Our empirical findings confirm most of our hypotheses and expand scholarly conversations surrounding managerial power theory and impression management. We confirm that relative shares held (ownership power), networks and personal connections (prestige power), as well as duality (structural power), directly improve top managers' capacity to engage in impression management, enhance public visibility in their favor, and boost celebrity. These effects have so far neither been conceptualized nor empirically investigated by prior literature (see also Davidson et al., 2004; Long et al., 2026; Wade et al., 2006). Interestingly, according to our analyses, expert power exerts no significant effect. Our reasoning, rooted in managerial power theory, builds on the idea that CEOs or CFOs may be the preferred and immediate choice if experts on specific topics are sought after (Jin, 2025; Westphal and Deephouse, 2011). A possible explanation for the insignificant findings of our study may lie in common media practices of choosing experts. When media search for industry-specific knowledge, other experts like academics or representatives of non-governmental organizations may be perceived as more “trustworthy” sources (Reif et al., 2020; Waisbord, 2011). Top executives may not be the media's primary choice in this case, even if they have a long tenure and thus have been exposed to a specific industry environment for a long time. In other words, there is no opportunity to leverage expert power gaps in the media since other target groups are favored by the media. Our additional analyses show that power gaps can affect both positive and negative celebrity. Power may thus serve executives to facilitate positive news coverage in their favor, but it may also entail negative reports that may not have been intended (e.g. negative coverage resulting from an abuse of excessive power). Our findings on negative celebrity thus complement literature centered around positive celebrity, which continues to dominate the academic debate (Hayward et al., 2004; Long et al., 2026; Lovelace et al., 2022; Pfarrer et al., 2010; Pollock et al., 2019).

Our third hypothesis connects power gaps, celebrity gaps, and CEO−CFO pay gaps—investigating the interplay of power and celebrity as antecedents of horizontal pay differences. Our empirical tests of this hypothesis thus target our core research gap. Both our main and additional analyses provide empirical evidence that executive celebrity gaps serve as mediators between power gaps and pay gaps, broadly confirming our third hypothesis. Precisely, as hypothesized, we show that celebrity is not just a standalone determinant of pay (see also Ahmed et al., 2025; Rindova et al., 2006), as implied by impression management literature, but that celebrity gaps may also serve as mediators between power gaps and pay gaps. By showing that power gaps may have an indirect effect (via mediation) beyond a direct effect on pay gaps, our study connects managerial power theory and impression management literature in a novel way. Beyond that, empirically confirming our third hypothesis, we also complement existing studies which have previously focused on the interplay between absolute executive celebrity and executive compensation (e.g. Graffin et al., 2008; Malmendier and Tate, 2009; Wade et al., 2006). We hereby add a novel perspective to the existing literature (Choe et al., 2014; Collins et al., 2018; Dikolli et al., 2021). Altogether, our findings advance the understanding of managerial power and celebrity as key characteristics of executives, which is meaningful for both managerial power theory and impression management literature.

Aside from the results for our main variables of interest, the results for several control variables uncover additional explanatory factors in CEO−CFO pay gaps. Interestingly, firm size negatively affects CEO−CFO pay gaps across model specifications, which suggests that larger firms tend to exhibit narrower pay differentials between CEOs and CFOs. A possible explanation in this case is that larger and more visible organizations are subject to stronger monitoring through governance structures and external scrutiny, which constrain excessive CEO compensation and thereby reduce the relative gap (Core et al., 1999; Elias et al., 2025). Furthermore, there is an indication of a negative effect of firm performance (ROA) on the CEO−CFO pay gap, which adds nuance to existing literature on the performance-pay link. An explanation could be that good firm performance is also credited to CFOs who are tasked with overseeing the financials of the firm (Caglio et al., 2018), allowing them to narrow the pay gap between them and CEOs. Board nationality diversity appears to impact CEO−CFO pay gaps negatively, which might indicate that a more diverse board will more cautiously monitor not only pay levels, but also excessive pay gaps (Usman et al., 2018). Differences in holding an MBA degree show a weak yet consistently negative effect on CEO−CFO pay gaps across several models. This confirms prior arguments about the importance of education in executive compensation (Datta and Iskandar-Datta, 2014; Graham et al., 2012). The pay gap thus tends to narrow when the CFO (rather than the CEO) holds an MBA or when both have similar educational backgrounds. This finding is notable given the well-documented importance of education for career progression and executive success (Bühlmann et al., 2022; Judge et al., 1995). Advanced education may reduce compensation disparities, possibly because higher educational attainment strengthens professional legitimacy and credibility, and thus bargaining positions in pay negotiations. In addition, executives with an MBA degree might signal analytical capability, financial expertise, and professionalism—attributes that are particularly valued in financial leadership positions—thereby enhancing executives' ability to negotiate compensation more effectively, especially in the US context (Datta and Iskandar-Datta, 2014; Salimi and Danesh, 2024). In line with our theoretical reasoning, education may thus serve as a subtle equalizer for the compensation of the CEO−CFO dyad.

We acknowledge that all (un)observed effects could be specific to our US study context. Studies carried out in other contexts, e.g. in Europe or Asia, might reveal different relationships. We also elaborate on this topic in the limitations section.

Compared to studies on vertical pay gaps (e.g. pay gaps between CEOs and rank-and-file employees), literature on horizontal pay gaps within the C-suite is scarce. Accordingly, Pissaris et al. (2017, p. 323) argue that compensation research “may benefit from a more focused examination of pay disparity,” while Fredrickson et al. (2010, p. 1050) state that “the antecedents of … [pay] dispersion have received very little attention.” More precisely, the few existing studies on horizontal pay gaps develop and test theory with a focus on outcomes, such as earnings management or managerial decision-making (Henderson and Fredrickson, 2001; Liu and Pan, 2022; Park, 2017; Vieito, 2012). What is missing is theorization about and empirical testing of the antecedents of pay gaps. Through our study, we seek to close this gap and advance theory and literature by not investigating the antecedents of absolute executive pay levels (as done in the past; see, for example, Bouteska and Mefteh-Wali, 2021; Cabezon, 2025; Gabaix and Landier, 2008; Tosi et al., 2000), but executive pay in relation to peers. Specifically, we answer Uhde et al.’s (2017, p. 128) call for “research on horizontal pay dispersion” with an emphasis on “pay gaps between the CEO and the CFO.” Studying how power and celebrity gaps interact when explaining pay gaps allows us to make relevant contributions not only to different literature streams and theory but also to practice.

First, we add to upper echelons literature, particularly on executive power, celebrity, and pay, by providing a novel perspective on how they interact. Both power and celebrity have already been considered as executive pay determinants (Abernethy et al., 2015; Collins et al., 2018; Long et al., 2026; Malmendier and Tate, 2009; Ozgen et al., 2025). However, they have rarely been considered jointly (Wade et al., 2006). Proposing a novel mediation effect based on arguments from managerial power theory and impression management literature, we contend that power gaps affect celebrity gaps, which then subsequently influence compensation gaps between CEOs and CFOs. This array of effects has been hinted at in the past (Bolino et al., 2016; Davidson et al., 2004) but has not been studied empirically. While the predominant theoretical notion in literature is that executive celebrity first affects executive power (which subsequently affects organizational or governance-related outcomes; Malmendier and Tate, 2009; Park et al., 2014; Wade et al., 2006), we mirror this perspective and provide empirical evidence for the effect power gaps have on celebrity gaps. Given the importance and popularity of managerial power and celebrity as theoretical topics and constructs in contemporary literature (Ahmed et al., 2025; Hambrick and Wowak, 2021; Ozgen et al., 2025), these novel insights constitute a meaningful addition to the literature, allowing for a more complete picture of the effects they have.

Second, beyond our general contribution to upper echelons literature, we contribute to both managerial power theory and impression management literature. Precisely, we combine managerial power theory and impression management literature in a novel way to theorize how power gaps and celebrity gaps explain CEO−CFO pay gaps. This also enables us to address our research gap surrounding the investigation of the antecedents of pay gaps. Managerial power theory builds on the premise that CEOs and CFOs have some degree of power to shape their own compensation packages, departing from models that assume perfect contracting between the company and CEOs/CFOs (Bebchuk et al., 2002; van Essen et al., 2015). A theoretical novelty of our study lies in theorizing that power is utilized as a means to engage in impression management. Building on impression management literature, we further theorize that CEOs' and CFOs' celebrity helps them to increase their compensation (Graffin et al., 2011; Heavey et al., 2020; Lee et al., 2020; Long et al., 2026). By considering different dimensions of power, we add further nuance to managerial power theory in the context of our research question, arguing that different power bases matter for both celebrity and pay (gaps). Our significant empirical findings confirm that managerial power theory and impression management literature are applicable to explaining pay gaps. What is more, we confirm the notion that executives operate within a dynamic social system, i.e. the TMT, where comparisons with peers are useful in explaining differences in their pay (Yang and Wang, 2014).

Third, we also contribute to executive compensation literature by theorizing about the relative pay of the two key executives, i.e. the CEO and CFO. With our study, we close gaps in existing literature and study horizontal pay gaps. The focus on the CEO−CFO duo has so far been a blank spot in executive compensation research (Ferramosca and Allegrini, 2021; Schmid and Altfeld, 2018; Six et al., 2013). However, the CEO−CFO duo is of fundamental strategic importance for firms, with the CEO considered as the head of the firm and the CFO usually being considered “second in command” (Caglio et al., 2018, p. 266; see also Ren and Zhong, 2025; Talukdar et al., 2025). Prior research has already shed light on CEOs' relative positions compared to the board (Fiss, 2006; Pearce and Zahra, 1991) or the entire TMT (Georgakakis et al., 2022; Junge et al., 2025). We provide empirical evidence on the implications of the power and celebrity wielded by arguably the two most central individuals in top management on compensation, more specifically on the pay gap. This emphasizes that compensation differences are influenced by the dynamics between different individuals, thereby adding a novel perspective to compensation literature.

Just like the academic debate, the broader public debate has long revolved around CEOs' absolute pay levels (Bloomberg, 2023) or vertical pay differences, e.g. pay gaps between CEOs and rank-and-file employees (Boone et al., 2024; Cabezon, 2025; Lee et al., 2019; Mueller et al., 2017). In this connection, the discussion often centered on the justification or the perceived fairness of CEO pay. We direct attention to horizontal pay disparities within the TMT, which are often overlooked in practice, despite potentially carrying organizational consequences. Our study informs a range of stakeholders by facilitating a better understanding of the antecedents of CEO−CFO pay gaps, in particular managerial power and celebrity. Such an understanding serves as a basis for appropriately handling horizontal pay differences in practice.

Specifically, our study bears relevance for compensation consultants, compensation committees, or boards involved in decisions about executive compensation (Murphy and Sandino, 2010; Nanda et al., 2025; Sirkin et al., 2024). This is particularly important, as previous studies have shown that high pay discrepancies in the TMT often have negative firm-level consequences, such as higher earnings management or lower firm value (Kubick and Masli, 2016; Li and Chen, 2024; Park, 2017; Siegel and Hambrick, 2005). If firms wish to prevent these negative consequences, they may take respective action informed by our study. Precisely, firms could strive for better balancing CEO and CFO pay, which may be achieved by balancing CEO and CFO power and celebrity. In this connection, compensation committees and boards (as well as consultants providing advice to them) may seek to ensure that pay structures are calibrated such that they facilitate a balanced consideration of power and celebrity. Concrete measures may vary between firms. For instance, fixed pay components might become more balanced, taking into account the relative network access that CEOs and CFOs have and use to the benefit of the focal firm (instead of the sheer size of networks; see also Engelberg et al., 2013). Variable pay components might turn out more balanced if structural power gaps are reduced, e.g. by avoiding CEO or CFO duality (as CEO duality, for instance, could allow for extractive, self-serving behavior with regard to variable pay; see also Krause et al., 2014).

Beyond that, to decrease the odds of potentially negative consequences of CEO−CFO pay discrepancies, our findings might be relevant in the context of executive selection. That is, when selecting a new CEO or CFO, considering the relative power and celebrity in comparison to the CEO or CFO in office might be incorporated into selection procedures (e.g. by gathering information on their networks or their public appearances). This allows firms to anticipate power and celebrity differences between CEOs and CFOs and the potential consequences for pay early on. This could help prevent negative outcomes at the CEO−CFO dyad/team level (such as cooperation or conflict) or at the firm level (such as decision-making or performance). Aside from the findings related to our main constructs, our study also offers additional practical insights. For instance, we show that boards with high nationality diversity can reduce pay gaps between CEOs and CFOs. Likewise, individual executives' formal education, here at the MBA level, can lower compensation gaps. These findings show that board diversity and MBA education of C-suite members are relevant when firms wish to avoid high CEO−CFO pay gaps.

Our study has several limitations, which provide promising avenues for future research. Regarding our key constructs, our operationalization of power is established and widely recognized in top management research (Collins et al., 2018; Finkelstein, 1992; van Essen et al., 2015). Nevertheless, we believe that future research could also apply non-conventional approaches to gather data on executive power, e.g. with approaches inspired by linguistics (Brunzel, 2023; Matthews et al., 2024). Analyzing documents, such as speeches, reports, or conference calls, may reveal additional insights into power-related aspects like CEO and CFO masculinity (Nair et al., 2022), (over-)confidence (Effah and Su, 2026; Heavey et al., 2022), charisma (Tosi and Greckhamer, 2004), or narcissism (Junge et al., 2025; O'Reilly et al., 2014; Rovelli and Curnis, 2021). Beyond that, for the operationalization of celebrity, social media data may be considered. While social media was less important for executives in the past, its influence is growing and will likely become even more important in the future (Brunner-Kirchmair and Dick, 2025; Heavey et al., 2020; Lovelace et al., 2022). Hence, future studies could consider social media data in the assessment of executives' impression management, whilst also considering factors like audience engagement (Acharya et al., 2025; Matthews et al., 2022; Zhou et al., 2024).

Another limitation is context. Our study focuses on US firms, which operate in a specific institutional environment characterized by distinctive governance and compensation practices (Conyon and Murphy, 2000; Fernandes et al., 2013). These institutional characteristics shape the importance of how executive power, celebrity, and pay interact. Hence, our results need to be interpreted by considering the constraints of the US context. Other institutional settings could yield different results. For example, government influence, ownership structures, or career horizons, as observed in China (Li et al., 2024; Long et al., 2026; Maqsood et al., 2025; Ren and Zhong, 2025; Zahid et al., 2025), or the stronger employee representation in boards of European two-tier systems (Bruce et al., 2005; Conyon and Schwalbach, 2000), may cause different results. Building on this notion, future research could examine CEO−CFO power, celebrity, and pay gaps across alternative institutional settings. Governance-related theories (e.g. institutional theory or stakeholder theory) could help to develop a deeper, contextualized understanding. Studies in Asian contexts may uncover alternative mechanisms linking executive power and celebrity to compensation outcomes (Ouyang et al., 2021). In addition, comparative analyses between shareholder-oriented economies, such as the US, and stakeholder-oriented economies, such as those in Europe, could reveal whether governance systems and levels of managerial discretion (Bottenberg et al., 2017; Crossland and Hambrick, 2011) systematically alter the observed relationships.