Top management team (TMT) members have been shown to influence tax avoidance; however, prior literature has not identified whether the intrapersonal diversity of TMT functional backgrounds leads to higher levels of tax avoidance. To study this relationship, we utilize TMT intrapersonal functional diversity, which captures the average heterogeneity of the TMT members' work experience. The skills associated with intrapersonal functional diversity may allow managers to better understand and communicate with various parties related to firm tax policies, thereby facilitating tax avoidance. Overall, we find that TMTs with higher levels of intrapersonal functional diversity achieve lower cash effective tax rates and that these TMTs do not rely on tax strategies that pose high risk.
Data Availability: Data are available from the public sources cited in the text.
Prior research has identified various determinants of tax avoidance including firm characteristics and constraints (e.g., Law and Mills 2015; Zimmerman 1983), tax avoidance opportunities (e.g., Rego 2003), and corporate governance (e.g., Armstrong, Blouin, Jagolinzer, and Larcker 2015). Recently, studies have begun to investigate whether top managers affect tax avoidance. Notably, influential research conducted by Dyreng, Hanlon, and Maydew (2010) documents that top executives play a significant role in determining the level of tax avoidance undertaken by their firms. Subsequent studies (e.g., Law and Mills 2017; Francis, Hasan, Wu, and Yan 2014) have begun to explore whether specific managerial characteristics, such as military experience or narcissism, are associated with tax avoidance. However, significant issues remain unresolved. First, although anecdotes suggest that various forms of functional expertise such as personnel and technology are helpful for managing firm tax planning (e.g., Deloitte 2017a; Deloitte 2017b), research has not explored whether these and other managerial functional backgrounds are associated with tax avoidance. Second, although Dyreng et al. (2010) document that the CEO, CFO, and other top executives affect tax avoidance, subsequent studies (e.g., Law and Mills 2017; Olsen and Stekelberg 2016; Francis et al. 2014) focus on individual managers (i.e., CEO/CFO), rather than the broader top management team (TMT). This paper fills these voids by studying the impact of TMT intrapersonal functional diversity on tax avoidance.
TMT intrapersonal functional diversity refers to the average heterogeneity of functional experiences possessed by TMT members, that is, generalist work experience. TMT members higher in intrapersonal functional diversity generate broader perspectives and are more likely to share common functional backgrounds with the various individuals and groups they interact with. This leads to TMTs that have improved communication with and understanding of various groups, which facilitates information sharing and supports the running of complex cross-functional firm processes (e.g., Bunderson and Sutcliffe 2002; Cannella, Park, and Lee 2008).
As prior tax research argues that the TMT does not directly engage in tax avoidance (Christensen, Dhaliwal, Boivie, and Graffin 2015; Dyreng et al. 2010), specialist-focused TMTs may be unable to apply their specialties in the tax setting. In contrast, TMTs characterized by high intrapersonal functional diversity (i.e., TMTs comprised of generalists) may be beneficial. The modern tax function regularly communicates with the C-suite and must be involved in all firm functions to extract relevant information and enact tax strategies (e.g., PwC 2017, 12). Therefore, to assist, understand, and communicate with or on behalf of the tax function, top executives may be required to possess high levels of intrapersonal functional diversity.
While it may be apparent how the ability to work with certain divisions such as accounting or law could aid the TMT in facilitating tax avoidance, as scholars argue that tax planning involves virtually all aspects of firm operations (Erickson, Hanlon, Maydew, and Shevlin 2019), we propose that the understanding of virtually any function can benefit firm tax avoidance. For example, the operations function handles inventory and supply chains, and is thereby critical to supporting the enactment of transfer pricing (PwC 2013). As another example, the personnel function is required to monitor and manage employee locations in order to support the pursuit of regional labor tax incentives (Deloitte 2017a). While TMT members are unlikely to be directly involved in the above forms of functionally-centric tax avoidance, their knowledge of these and other functions allows them to use their authority to lead the tax-planning process by sharing information with and extracting information from various firm functions. Overall, as taxation is intertwined with all firm functions, intrapersonal functional diversity may provide the TMT with the ability to work across functions in support of tax avoidance.
We note that the ability provided by intrapersonal functional diversity differs substantially from the broad construct of managerial ability (Demerjian, Lev, and McVay 2012) found to be relevant to the tax setting (Koester, Shevlin, and Wangerin 2017). Specifically, managerial ability captures how effectively both the TMT and all other managers utilize firm resources—including those managers directly involved in tax avoidance such as tax managers. In contrast, intrapersonal functional diversity focuses on one directly observable trait of the TMT, allowing for a clear analysis of the TMT's influence on the tax function. Further, while managerial ability is mechanically dependent on firm efficiency and performance, intrapersonal functional diversity is independent of performance, having costs and benefits contingent on the situation. That is, while some settings require in-depth, specialist experience (e.g., Gounopoulos and Pham 2018), broader experience is beneficial in other settings (e.g., Cannella et al. 2008).
Overall, we hypothesize that TMT intrapersonal functional diversity will lead to higher levels of tax avoidance, and test our hypothesis on a sample of 12,431 firm years spanning the period of 2000 through 2016. We define tax avoidance broadly to encompass anything that reduces taxes paid relative to pretax income. Therefore, we focus on the cash effective tax rate (herein Cash ETR).1 We find that the generalism captured by TMT intrapersonal functional diversity, rather than the concentration of a certain specialty, is significantly negatively associated with Cash ETR. Expanding on this finding, we explore whether alternative measures of generalism (e.g., Custódio, Ferreira, and Matos 2013) better explain our results. We find that intrapersonal functional diversity more strongly relates to tax avoidance than these alternatives. Next, we explore a spectrum of somewhat benign to more aggressive forms of tax avoidance, and find that TMTs with higher intrapersonal functional diversity do not rely on risky forms of tax avoidance. Instead, our results imply that these TMTs have the ability to reduce tax liabilities using somewhat less risky forms of tax incentive seeking.
Our results are economically significant. Holding other factors constant, replacing one TMT member having low intrapersonal functional diversity with an individual high in the trait results in a 2.09 percent decrease in mean Cash ETR. This implies a tax savings of around $3.25 million.2 Our findings are decision useful as intrapersonal functional diversity is easily observable. While broad measures of managerial ability are difficult to operationalize as they capture all firm managers and all managerial traits, firms can implement our findings and increase tax avoidance by selecting top managers with high intrapersonal functional diversity.
Our study contributes to the literature in several ways. First, it contributes to the strand of accounting literature examining the influence of managers on tax avoidance and identifies TMT intrapersonal functional diversity as a determinant. This finding is in contrast to prior literature in accounting, which is generally concerned with the impact of theoretically applicable specialties and rarely discusses the value of generalists. Second, our study documents that tax avoidance is driven by a group of top managers rather than a single individual. This is opposed to much of prior literature that focuses on individual executives (e.g., CEO/CFO). Finally, prior studies in management have focused on firm performance (e.g., Bunderson and Sutcliffe 2002; Cannella et al. 2008) and innovation (e.g., Park, Lim, and Birnbaum-More 2009) as outcomes of intrapersonal functional diversity. Our study finds that TMT intrapersonal functional diversity impacts tax avoidance—a financial outcome indirectly managed by the TMT.
The remainder of this paper is organized as follows. Section II discusses prior research and develops our hypothesis; Section III describes our sample selection and research design; Section IV presents our descriptive statistics and main empirical results; Section V explores the elements of functional diversity that drive our results and studies various additional measures of tax avoidance; Section VI discusses additional analyses; and finally, Section VII concludes.
II. PRIOR RESEARCH AND HYPOTHESIS DEVELOPMENT
Prior Research—Management Accounting and Tax Avoidance
Despite a significant increase in tax avoidance research (Hanlon and Heitzman 2010) and public interest in tax avoidance (e.g., Saul and Cohen 2019), empirical tax research has largely focused on firm and governance characteristics as determinants of tax avoidance (Hanlon and Heitzman 2010; Graham 2003; Shackelford and Shevlin 2001). While only a small portion of the tax avoidance literature has studied the impact of managers on tax avoidance, research in the area is increasing. Desai and Dharmapala (2006) indirectly study how managers affect tax avoidance by analyzing the relationship between incentive compensation and tax sheltering. They argue that shareholders are averse to sheltering and that incentive compensation reduces this form of tax avoidance. Dyreng et al. (2010) are the first to study the impact of specific top managers on tax avoidance, tracking the movement of 908 executives to quantify the TMT's impact on firm tax avoidance. They find that the economic impact of executives on tax avoidance is large, however, they do not find strong associations between specific executive skills/backgrounds (e.g., education, experience as CFO) and tax avoidance.
Building on Dyreng et al.'s (2010) finding that the TMT influences tax avoidance, subsequent research has worked to isolate specific managerial characteristics associated with tax avoidance. Chyz (2013) finds that executives suspected of personal tax evasion are more likely to direct their corporations to engage in tax sheltering. Francis et al. (2014) document that female CFOs engage in lower levels of tax aggressiveness. Christensen et al. (2015) find that conservative executives engage in less tax avoidance. Olsen and Stekelberg (2016) find that narcissistic CEOs engage in more tax sheltering. Law and Mills (2017) document that CEOs with military experience pay an estimated $1–2 million more in corporate taxes per firm-year. Next, using Demerjian et al.'s (2012) measure of managerial ability, Koester et al. (2017) find that higher ability managers engage in more tax avoidance. Finally, Hsieh, Wang, and Demirkan (2018) document that firms with overconfident CEOs and CFOs engage in more tax avoidance.
While the above studies illustrate an emerging literature focused on managerial characteristics and tax avoidance, the connection between tax avoidance and management functions/skills remains largely unexplored. Given that intrapersonal functional diversity has not been explored in the above-discussed accounting literature but that anecdotes suggest various functions are involved in tax planning, we next discuss the relevant literature that may support a relationship between intrapersonal functional diversity and firm tax avoidance.
Prior Research—Intrapersonal Functional Diversity
Prior management research argues that intrapersonal functional diversity results in superior communication and understanding skills that facilitate information sharing with various groups (e.g., Cannella et al. 2008; Bunderson and Sutcliffe 2002). Being embedded in various firm functions (e.g., accounting, legal, engineering) implies both an understanding of those functions and an understanding of groups that interact with those functions. For example, experience in accounting may indirectly lead to experience dealing with firm functions that regularly work with accounting (e.g., the legal function on merger issues). Accounting experience may also lead to an understanding of board oversight related to tax avoidance. As such, beyond the direct experiences obtained from having high intrapersonal functional diversity, numerous beneficial indirect experiences are also provided (Burke and Steensma 1998). These various direct and indirect experiences lead to a reduced semantic gap, enhanced understanding and communication skills, and overall improved information sharing between the TMT and the various individuals and groups with which they interact (Cannella et al. 2008; Chattopadhyay, Glick, Miller, and Huber 1999). Overall, these benefits result in managers who “have a good perception of where … knowledge is and how to tap into it” (Bunderson 2003, 460).
Managers with high intrapersonal functional diversity are beneficial to firms in various settings. Bunderson and Sutcliffe (2002) study the impact of intrapersonal functional diversity on 44 business unit management teams of a Fortune 100 company. They find that intrapersonal functional diversity improves team communication and information sharing, which in turn improves unit performance. Cannella et al. (2008) study 207 U.S. firms and find a positive relationship between TMT intrapersonal functional diversity and firm performance. Intrapersonal functional diversity has also been tied to firm innovation outcomes. Specifically, Park et al. (2009) find that intrapersonal functional diversity is beneficial to product innovativeness and new product development time efficiency.
Rather than focusing on various functional specialties, we argue that intrapersonal functional diversity may facilitate tax avoidance. Intrapersonal functional diversity provides managers and teams a greater ability to understand various parties (e.g., Cannella et al. 2008; Bunderson and Sutcliffe 2002) and communicate in the language of various groups (e.g., Bunderson and Sutcliffe 2002). TMTs with high intrapersonal functional diversity have various direct and indirect experiences, which result in TMTs that are more capable of sharing information and handling complex interdepartmental issues (Burke and Steensma 1998).
Tax avoidance is an example of one such complex interdepartmental organizational issue, and the complex nature of tax avoidance increases the need for broad organizational understanding. Modern tax departments must be integrated with all firm departments. TMTs with broad cross-functional understandings have knowledge of and potential connections with the various departments that the tax function must be integrated with, and thus, may be more capable of leading tax departments. For example, when a firm is involved with tax avoidance related to depreciation/depletion, tax departments require information on various items (e.g., equipment, intangibles) housed in various functions. Top executives who understand these functions (e.g., operations and R&D) are aware of the information housed within these functions, can effectively communicate with technically minded leaders in these functions, have the authority to request relevant information to support the tax function, and may be able to identify circumstances in which requested information is incomplete based on their knowledge of these functions.3
We argue that an understanding of and the ability to communicate with nearly every major firm function can support tax avoidance due to the systemic involvement of taxation in virtually all aspects of business strategy and operation (e.g., Erickson et al. 2019; PwC 2017). While a TMT member's functional experience may not be directly used to enact tax strategies, it can be used to work with the many functions in which those tax strategies can be enacted. The more functions that a TMT member has experience in, the more capable they are of using their authority to facilitate tax avoidance across various functions.
Overall, we argue that the superior communication and understanding abilities provided by intrapersonal functional diversity facilitate information sharing between the TMT and the wider firm (Bunderson 2003). Therefore, TMTs characterized by high levels of intrapersonal functional diversity may have the skills and authority to support and facilitate the cross functional process of tax avoidance. Based on this discussion, we state our hypothesis as follows:
Firms with higher TMT intrapersonal functional diversity avoid more tax.
III. RESEARCH DESIGN
Data and Sample
In line with prior accounting and finance literature, we define the TMT as the CEO, CFO, and the other three highest compensated managers (e.g., Bertrand and Schoar 2003; Dyreng et al. 2010). These top five managers have been identified as a group that influences corporate tax avoidance (Dyreng et al. 2010) as well as various other firm outcomes such as earnings management (Cheng, Lee, and Shevlin 2016) and voluntary disclosure (Ke, Li, Ling, and Zhang 2019). We identify the CEO and CFO by their titles reported in Execucomp or BoardEx and require that Execucomp reports total compensation for other firm executives so that we can identify the other members of the top five. We obtain TMT functional backgrounds from BoardEx to construct our measure of TMT Intrapersonal Functional Diversity.4 As BoardEx began reporting the work experience of top executives in 2000 and the most current measure of managerial ability is available up to 2016,5 our initial sample is drawn from the intersection of BoardEx and Execucomp for the period of 2000 to 2016. Compustat provides the financial information used to calculate our measure of tax avoidance (Cash ETR) as well as the majority of the study's control variables. Execucomp provides the data used to calculate the control variable for managerial incentives. Overall, after following the above selection procedures and dropping all observations with missing data, the resulting sample contains 12,431 firm-year observations spanning the years of 2000 to 2016.
Measurement of Tax Avoidance
The dependent variable in our regressions is Cash ETR, which captures the actual current year cash disbursements made for all income tax expenses. Cash ETR is relevant to our setting as managers often treat tax avoidance as a method of decreasing cash taxes paid in order to bolster firm operations (Hanlon and Heitzman 2010). Cash ETR is calculated by dividing cash taxes paid by pretax book income less special items. Our study excludes observations with negative denominators as tax avoidance is a lower priority for loss firms (Dyreng et al. 2008; McGuire, Omer, and Wang 2012). Further, negative pretax incomes result in Cash ETRs that are difficult to interpret (Henry and Sansing 2018).6 We winsorize the remaining non-missing Cash ETRs at [0,1] to reduce the influence of outliers. A lower value for Cash ETR indicates more tax avoidance.
Measurement of Top Management Team Intrapersonal Functional Diversity (IFD)
Following the majority of prior studies (e.g., Bunderson and Sutcliffe 2002; Cannella et al. 2008), we measure TMT intrapersonal functional diversity (IFD) using the Blau Index.7 To construct this index, we obtain the work experience of each executive in our sample, and identify how long the executive has worked in the following functions: accounting/finance, marketing/sales, R&D/engineering, management, production/operations, law, personnel/labor relations, and other (e.g., Cannella et al. 2008). We use these eight functional classifications as our sample incorporates firms across various industries (manufacturing, agriculture, etc.), and prior literature documents that these eight categories are common across various industries (Zhang 2019; Cooper, Patel, and Thatcher 2014; Cannella et al. 2008; Chattopadhyay et al. 1999).8
In Appendix A we display three examples with varying levels of IFD as well as the associated Blau index calculations. Panel A displays the TMT of JCPenney Co., which demonstrates the minimum IFD (i.e., Blau index = 0.00) as each member of the TMT has worked in only one function. In Panel B, Beam Incorporated's TMT has some functional breadth, but is still below median. Finally, the TMT of Intersil Corp. in Panel C displays an above-median level of IFD.
As opposed to the TMT in Panel A, the higher levels of IFD in Panels B and C should better support the tax avoidance process for two reasons. First, high levels of IFD indicate more evenly distributed experiences and imply that superficial (short duration) experiences do not comprise the majority of the TMT's direct and indirect experiences. This indicates a substantive understanding of a function and those groups that work with it, enabling management to understand and work with that function to facilitate tax avoidance. Second, TMTs benefit from having multiple members each possessing a breadth of functional experiences because most managers at large firms have a somewhat defined sphere of influence that often does not overlap with other TMT members on a regular basis (e.g., Carpenter and Sanders 2004).
For example, the VP of power management products at Intersil benefits from having broad functional experiences when working to understand and communicate with the various functions that create and distribute the firm's power management line (e.g., research, production, sales, etc.). Similarly, the VP of consumer products benefits from broad functional experiences when managing the functions that support the consumer line. This argument also applies to TMTs segmented by geography. For example, Beam's President of Europe/Middle East/Africa operations and their President of North American operations each require broad functional expertise to communicate with and understand the various functions in their respective regions. Finally, similar arguments apply to TMTs with functionally defined roles. While the CMO and the COO regularly interact with most firm functions (e.g., Deloitte 2018; EY 2013), relating to the R&D function, the COO may be focused on the execution side (EY 2014) while the CMO may be focused on inspiring product innovations based on the customer experience (Deloitte 2016). Each top manager interacts with different aspects of each function, and each portion of a function contains expenses and investments that may qualify for tax incentives. As such, each manager may benefit from a diverse functional background that allows them to understand and communicate with the various employees across the functions that they regularly interact with.
Overall, in modern, complex firms, we make the intuitive argument that TMT members have certain groups with which they regularly interact. Thus, TMTs benefit from having multiple members with high intrapersonal functional diversity.
Empirical Model for Main Analysis
We first discuss our TMT Controls, beginning with dominant functional diversity (Dominant Functional Diversity). While we do not hypothesize a relationship between this form of diversity and tax avoidance, studies in management generally consider both dominant and intrapersonal functional diversity together (e.g., Buyl, Boone, Hendriks, and Matthyssens 2011; Cannella et al. 2008).10 We measure Dominant Functional Diversity using the same eight functional categories we use for IFD, and define the function an executive spent the most time in as their dominant function (e.g., Cannella et al. 2008; Carpenter and Fredrickson 2001). We then use the Blau Index to capture the diversity of the dominant functional experiences held by a TMT. Next, to ensure that IFD is not capturing the influence of various other forms of diversity, we control for the diversity of tenure (Tenure Diversity), age (Age Diversity), education (Education Diversity), and gender (Gender Diversity) in line with Cannella et al. (2008).
We also control for other TMT characteristics that are expected to affect tax avoidance. We control for TMT age (Age) as prior research argues that younger managers are more likely to pursue risky strategies (Hambrick and Mason 1984). Managers with higher tenure may be more experienced and effective communicators when dealing with various firm groups, so we control for the average tenure (Tenure) of TMT members (Plöckinger, Aschauer, Hiebl, and Rohatschek 2016). Next, we control for TMT members' military experience (Military) as CEOs with military experience engage is less tax avoidance (Law and Mills 2017). Finally, we control for managerial ability (Managerial Ability) as higher ability managers engage in more tax avoidance (Koester et al. 2017).
We then include a range of variables capturing firm characteristics (i.e., Firm Controlsi,t) common in tax avoidance research (e.g., Dyreng et al. 2010; Frank, Lynch, and Rego 2009). Specifically, we control for firm size (Size), growth opportunities (Market to Book), foreign operations (Foreign), profitability (ROA), decreases in net operating losses (NOL Decrease), operating risk (StdROA), managerial incentives (Option Value), leverage (Leverage), as well as other firm characteristics (i.e., R&D, Intangible, Advertising, Free Cash Flow, Cash, Equity Income, PPE, Capital Expenditures) that may affect tax avoidance. We include performance-matched abnormal accruals (Abnormal Accruals) to control for earnings quality. Finally, we include firm and year fixed effects to control for firm- and time-invariant factors that could affect tax avoidance. Therefore, the IFD coefficient captures the association between Cash ETR and IFD within each firm over time. Including firm fixed effects addresses the concern that IFD might capture stationary firm characteristics and removes the cross-firm variation in each variable.
Table 1 presents the summary statistics for the variables in our study. IFD has a mean (median) of 0.384 (0.389). Cash ETR has a mean (median) of 0.239 (0.227) in line with prior literature studying the impact of management characteristics on tax avoidance (e.g., Law and Mills 2017; Dyreng et al. 2010).
Table 2 presents the Pearson correlation coefficients for Cash ETR, IFD, all TMT Controls, and ROA. We find preliminary support for a relationship between IFD and tax avoidance. That is, IFD is negatively correlated with Cash ETR (p < 0.01).11 Next, we find that IFD and Dominant Functional Diversity are positively correlated (p < 0.01).12 Finally, while IFD and Managerial Ability are positively correlated (p < 0.01), the correlation coefficient is only 0.11, indicating that the two measures capture different information.
Table 3 presents the results for our hypothesis, which expects that firms with higher IFD avoid more tax (Equation 1). Column 1 presents the baseline model excluding our variable of interest. The results for our control variables are consistent with prior literature (e.g., Koester et al. 2017; Rego 2003; Mills 1998), and we discuss these variables in greater detail later in this section. Moving to Column 2 and the direct testing of our hypothesis, we find that TMTs characterized by high levels of IFD avoid more taxes. Specifically, in Column 2, IFD has a coefficient of −0.055 (p < 0.05). This finding indicates that the broad functional understanding of a generalist, coupled with the authority held by a top manager, is beneficial to assisting various firm functions in collaborating with the tax department. The beneficial influence of certain types of TMT functional knowledge are intuitive. For example, TMT knowledge of the R&D function may allow the TMT to effectively work with the employees responsible for identifying and classifying various activities into categories that may be eligible for R&D tax credits. However, in line with the argument that taxation can and should be involved with all aspects of a business (Erickson et al. 2019), functions that initially appear separate from taxation can also serve the tax function. For example, the personnel function can place operations in different cities, states, and nations to seek tax incentives (Deloitte 2017a).
While our findings in Column 2 of Table 3 provide support for our hypothesis, one may argue that IFD simply captures the effect of certain potentially influential functions. Therefore, in Column 3 of Table 3, we test whether teams focused on certain functions drive tax avoidance. We do not find that functionally specialized teams are associated with tax avoidance, in line with the education-based skills analysis conducted in Dyreng et al. (2010).13 While each function may provide some benefit, we argue that each specific functional domain captures only one portion of the functional background required to support the interdepartmental tax avoidance process.
As noted above, our control variables are consistent with prior literature. We first examine our control variables that relate to firm characteristics. Larger firms (Size) avoid less taxes as they face greater reputational costs (e.g., Rego 2003). Net operating loss utilization (NOL Decrease) results in a lower Cash ETR due to statutory tax incentives (e.g., Koester et al. 2017). Consistent with the notion that growing firms may make more investments in tax-favored assets (Chen et al. 2010), we find a negative coefficient on Market to Book. Firms with strong cash flows (Free Cash Flow) avoid less taxes as they have a less immediate need for tax avoidance (e.g., Koester et al. 2017). Option Value is negatively related to Cash ETR as management incentives can encourage tax avoidance (e.g., Rego and Wilson 2012). Firms with higher pretax income (ROA) have both a greater need and proclivity for tax avoidance (McGuire et al. 2012). Firms with large cash holdings (Cash) avoid more taxes, potentially due to unrepatriated cash reserves generated from international tax avoidance strategies (e.g., Foley, Hartzell, Titman, and Twite 2007). As firm complexity creates various incentives and opportunities for tax avoidance (McGuire et al. 2012), firms with greater Equity Income avoid more taxes. Finally, consistent with Koester et al. (2017), we also find the within-firm variation of Capital Expenditures is positively associated with Cash ETR. We also find that Abnormal Accruals are positively associated with Cash ETR consistent with McGuire et al. (2012).
Focusing on our TMT managerial characteristic variables, we find a negative coefficient on Age Diversity, suggesting that the age heterogeneity among TMT members increases tax avoidance. Notably, the conservatism (Military) associated with military experience leads to less tax avoidance in line with Law and Mills (2017). Finally, Managerial Ability is non-significant, and we further explore this finding in Section VI of the paper.14
V. EXPLORING DIVERSITY AND TAX AVOIDANCE
Exploring Relevant Aspects of Diversity
While we argue that IFD allows the TMT to interface with different firm functions leading to a greater ability to carry out tax avoidance, various alternative explanations exist. We identify four different explanations for how certain forms of diversity may drive our results, and we test four empirical proxies in order to determine which explanation is most appropriate.15
First, Custódio, Ferreira, and Matos (2013) create a CEO Generalism Index comprised of: number of positions held, number of firms worked at, number of industries worked in, former CEO experience, and experience in conglomerate firms. Custódio et al. (2013) argue that CEOs are better paid when they have high levels of general managerial skills as measured by their Generalism Index. We self-construct an adapted version of Custódio et al.'s (2013),Generalism Index that includes all members of the TMT (see Appendix B). It is possible this broad set of skills drives our findings and that IFD significantly overlaps with this measure (e.g., individuals with experience from various industries may have more functional experiences). Second, while we argue that IFD facilitates tax avoidance by allowing top management to interface with firm functions rather than each other, it is possible that tax avoidance is facilitated by within-TMT communication, which may be assisted by the overlap in functional experiences between TMT members (e.g., Richard, Wu, Markoczy, and Chung 2019). Prior literature credits IFD with various benefits arising from within-TMT communication (e.g., Bunderson and Sutcliffe 2002); therefore, the functional overlap (Overlap) between TMT members (which generally increases with high IFD) may be the driving force behind our main findings. Next, we test both the percentage of functionally broad TMT members (% High IFD) and the aggregate intrapersonal functional diversity of a TMT (Aggregate IFD). These variables allow us to determine whether a TMT with multiple managers each possessing high levels of IFD (% High IFD) is more effective than a TMT that, in aggregate, exhibits a broad range of skills (Aggregate IFD). While we argue that having multiple TMT members with broad experiences allows the TMT to effectively interact with disparate firm functions (as no manager can consistently interact with all firm units), we acknowledge the possibility that Aggregate IFD drives our results.
Table 4 presents the four measures discussed above and their association with tax avoidance. In Column 1, using the Generalism Index, we find non-significant results.16 In Column 2, we find that Overlap is non-significant, implying that IFD is not beneficial due to increased communication effectiveness within the TMT. In Columns 3 and 4, we find that Aggregate IFD is non-significant, but that % High IFD is significant. That is, a TMT is more effective in implementing tax avoidance when more generalists are present, not when a TMT contains a number of different functional specialties spread across the group. This supports our argument that various functionally broad TMT members are helpful in enacting tax avoidance within the groups they regularly interact with. Overall, by considering these various alternative measures and counterarguments, we increase our confidence that the form of generalism most relevant in the tax setting is IFD.
Exploring Tax Strategies Employed
We next explore the tax avoidance approaches employed by firms with high IFD. Much of prior research (e.g., Armstrong, Blouin, and Larcker 2012; Robinson, Sikes, and Weaver 2010; Phillips 2003) demonstrates that, ceteris paribus, tax avoidance leads to increased cash flow, share appreciation, dividend increases (Mills 1998; Mills 1996), and the reduction of a dead-weight cost (Feldstein 1999). Due to these clear benefits, increased tax avoidance is an overarching goal for firm management. Therefore, it is possible that TMTs characterized by high levels of IFD have both the capability and the goal of pushing tax avoidance to its often-risky limits. On the other hand, TMTs characterized by high levels of IFD may wish to avoid the risks associated with certain forms of tax avoidance and rely on less risky (but often effective) forms of tax avoidance.17 That is, C-suite leadership teams may “participate in [tax] decision making to minimize tax risks and to invest in making their tax functions more efficient and robust” (EY 2017, 18). In line with this anecdote, prior literature argues that the broad exposure obtained by generalists results in risk-conscious management teams that are more highly compensated and beneficial to their firms (Hughes-Morgan, Ferrier, and Labianca 2011).
In order to determine the level of tax risk high IFD firms engage in, we consider a broad spectrum of tax avoidance measures. Specifically, we study five additional dependent variables: unrecognized tax benefits (UTB), Wilson's (2009) shelter score (Shelter), Frank et al.'s (2009) discretionary permanent book-tax differences (DTax), permanent book-tax differences (PermBTD), and total book-tax differences (BTD). Prior literature argues that Cash ETR is the broadest and least aggressive metric of tax avoidance, while these additional five measures capture varying levels of aggressiveness from most to least aggressive (e.g., Goh, Lee, Lim, and Shevlin 2016; Lisowsky, Robinson, and Schmidt 2013; Hanlon and Heitzman 2010).18 We regress these additional measures of tax avoidance on IFD while controlling for the same set of variables used in Equation (1). We present our findings in Table 5.
We first note that our measures that capture tax aggressiveness and risk (UTB in Column 1, Shelter in Column 2, and DTax in Column 3) are all non-significant.19 This suggests that these risky methods are not the primary mechanisms used by firms with high IFD. Rather, we find that PermBTD is significant at the 0.05 level (Column 4), and that BTD is significant at the 0.1 level (Column 5). Our finding for PermBTD indicates that IFD drives tax avoidance through various tax incentive seeking activities. We argue that this finding supports our previous arguments, as our measurement of PermBTD captures a wide range of complex but “legitimate”20 tax-planning processes that require working with various firm functions. This includes previously discussed items such as tax credits and the pursuit of location-specific tax incentives, as well as additional tax incentives that require complex, interdepartmental collaboration—e.g., the domestic production activities deduction (KPMG 2016).
As noted above, BTD (Column 5) is significant at the 0.1 level. This finding may indicate that firms with high IFD utilize tax incentives relating to permanent differences as well as tax strategies that result in temporary differences that defer taxation.21 However, additional analyses indicate that the temporary component of BTD is non-significant, indicating that this finding is driven by the underlying permanent differences captured by BTD.22
VI. ADDITIONAL ANALYSES
In this section, we conduct several additional analyses. First, we reconcile our findings to those in Koester et al. (2017) to better differentiate IFD from Managerial Ability. Next, we explore whether IFD allows the TMT to manage the barriers to tax avoidance presented by institutional investors, audit committees, and boards. Finally, we address endogeneity concerns.
Top Management Team Intrapersonal Functional Diversity and Managerial Ability
While we control for Managerial Ability, it may be that firms with high IFD have high levels of Managerial Ability, and that Managerial Ability drives our results. To alleviate this concern, we investigate whether or not IFD is the cause of the non-significant results for Managerial Ability reported in Tables 3 and 4. If IFD is the cause of this result, it may constitute a component of Managerial Ability rather than a distinct construct.
In Table 6, Column 1, we study the impact of Managerial Ability on tax avoidance using the model from Koester et al. (2017). We adopt the most updated version of Managerial Ability, restrict our sample to observations with non-missing IFD in order to present a clear comparison across columns, and calculate all control variables as in Koester et al. (2017) to ensure that our set of control variables is not the cause of the non-significant finding for Managerial Ability reported in Table 3. Consistent with Koester et al. (2017), in Column 1 we find that Managerial Ability is significantly negatively associated with Cash ETR.23 In Column 2, we add our variable of interest, IFD, into the model. Adding IFD results in a somewhat weaker finding for Managerial Ability, that is, Managerial Ability is significant at the 0.10 level while IFD is significant at the 0.05 level (although the magnitude of the coefficient on Managerial Ability is similar). This result implies that, while Managerial Ability and IFD overlap somewhat as evidenced by both this test and the correlation coefficient in Table 2, the two measures are largely different. While Managerial Ability is intrinsically related to firm efficiency and performance (e.g., ROA),24IFD increases performance in certain settings (e.g., dealing with environmental uncertainty; Cannella et al. 2008), while decreasing it in others (e.g., bringing a private firm to its IPO; Gounopoulos and Pham 2018). In line with the argument that Managerial Ability relates to performance and is similar to ROA, we find that adding ROA rather than the other control variables eliminates the significance of Managerial Ability (Table 6, Column 3).25
Overcoming Resistance to Tax Avoidance
In this section, we explore whether IFD is beneficial outside of functional-unit interactions. Various groups both within and outside of the firm are averse to tax avoidance, and even tax incentive seeking can cause risks and be perceived negatively.26 This may lead certain groups to push management to focus on general business operations rather than tax avoidance (e.g., Khurana and Moser 2012; Hoopes, Mescall, and Pittman 2012).
Anecdotes argue that members of the C-suite should understand tax avoidance so that they can work with investors and other groups that are resistant to the practice (Berkman 2017). TMTs characterized by high IFD have the ability to support the tax function and are therefore likely have some understanding of firm tax strategies. These TMTs also have IFD-provided communication skills and access to resistant parties (e.g., investors) due to their high firm status.
We first study whether IFD allows the TMT to overcome institutional investor resistance to tax avoidance. As our purpose is to study institutional investors that are risk averse and less likely to be supportive of investee firm tax avoidance, we study large blockholders. Large blockholders have the influence to monitor firm tax policies and also bear more risk if an investee improperly utilizes their tax function (Khurana and Moser 2012; Hoopes et al. 2012).
We display our results for how institutional investors interact with TMTs characterized by high levels of IFD in Table 7, Columns 1 and 2. We measure large blockholders using a Herfindahl Index of institutional investor ownership concentration (e.g., Ajinkya, Bhojraj, and Sengupta 2005; Hartzell and Starks 2003) in Column 1 (Institutional Concentration), and by a dummy variable denoting the existence of institutional blockholders (5 percent ownership threshold) in Column 2 (Institutional Blockholder). We find that the coefficient on Institutional Concentration is positive and significant (p < 0.10), indicating that institutional investors with large ownership interests reduce investee tax avoidance. However, the interaction between IFD and Institutional Concentration is negative and significant (p < 0.05). This may indicate that IFD helps firms to effectively manage relationships with institutions in order to reduce barriers to tax avoidance. We find similar but somewhat weaker results in Column 2 using Institutional Blockholder.
Similarly, we also explore whether IFD can overcome audit committee and board resistance to tax avoidance (e.g., Richardson, Taylor, and Lanis 2013). In Table 7, Column 3, we find that the coefficient on Audit Committee Independence27 is positive (p < 0.05). This indicates that independent audit committees may be less willing to invest firm resources in tax avoidance, consistent with Richardson et al. (2013). However, the interaction between IFD and Audit Committee Independence is negative and significant (p < 0.05), indicating that high IFD firms are able to confront the resistance of audit committees. We do not find significant results for either a standalone Board Independence variable or its interaction with IFD (Table 7, Column 4), implying that the audit committee is a bigger obstacle to tax avoidance than the wider board.
Addressing Endogeneity Concerns
While firm fixed effects help alleviate time invariant omitted variable bias, it is still possible that IFD is endogenously determined by the firm, and that the same set of factors may jointly affect both IFD and tax avoidance. We adopt an instrumental approach to address this concern. Our first instrumental variable is the industry-year median of IFD (IFD—Industry Median), and we classify industries using two-digit SIC codes. Prior research argues that industry-specific TMT traits are likely to be exogenous as they are not under firms' control (e.g., Kale, Reis, and Venkateswaran 2009). Our second instrument is the occupational diversity of the state in which a firm's headquarters is located (State Occupational Diversity). Even when considering the highest level TMT position (CEO), firms and job candidates have strong regional biases and firms generally hire locally (e.g., Yonker 2016). Therefore, we argue that the functional diversity of a job candidate is contingent on the functional diversity in the region of a firm's headquarters.
Table 8, Column 1, reports our first-stage regression, and our first instrument (IFD—Industry Median) is positively significant in explaining IFD. The second instrument (State Occupational Diversity) has the expected sign, but is not significant. Overall, these results imply that firms are more likely to have high IFD when industry-level functional diversity is high. Diagnostic tests provide evidence that the equations are well-specified.28Table 8, Column 2, reports the second stage results. We find that the coefficient on predicted IFD is significantly negative, indicating that the main findings hold after controlling for endogeneity concerns.
While we argue that TMT hiring is influenced by industry and regional executive availability, we acknowledge that hiring decisions may be influenced by executive self-selection and board selection. Most relevant to our setting, boards aware of the tax benefits of IFD may choose to hire TMT members with high IFD in order to improve firm tax outcomes. If boards choose to hire for IFD, managerial incentives may be a useful method to ensure that TMTs characterized by high levels of IFD leverage their broad skills to support the tax function.29 Without incentives, some TMTs may consider tax to be the domain of specialty divisions and focus their efforts on general business operations.
Further, we acknowledge that both of our instruments are subject to caveats. First, State Occupational Diversity captures the diversity of the general labor force, but may not fully represent the diversity in the highly specialized TMT job market. Second, while industry-level variables are frequently used as instruments, they may not be fully exogenous as firm i is a component of its industry. This both directly (i.e., through inclusion of firm i itself) and indirectly (e.g., through influencing its peer firms) affects relevant industry-level variables.
The modern tax function is increasingly integrated with all facets of a firm, meaning that proper tax function management may be best supported by TMT members with experience in various functions. In line with this argument, we find that TMT intrapersonal functional diversity leads to higher levels of tax avoidance, and that TMTs characterized by intrapersonal functional diversity achieve these tax avoidance outcomes without relying on risky tax avoidance.
Our findings inform the literature in a number of ways. First, we provide a partial answer to the questions raised in Dyreng et al. (2010). Dyreng et al. (2010) find that executives have a large influence on tax avoidance; however, they do not find that the specific skills and background characteristics explored explain this influence. We find that the various functions comprising intrapersonal functional diversity allow TMTs to increase firm tax avoidance. Next, prior management literature has demonstrated a positive relationship between intrapersonal functional diversity and team (Bunderson and Sutcliffe 2002) or firm (Cannella et al. 2008) performance. However, this is the first study that argues intrapersonal functional diversity supports the TMT in indirectly facilitating a financial function outcome. Tax avoidance is an ideal setting to provide support for this argument, as it is an accounting outcome that the TMT is not likely to be directly involved in (e.g., Christensen et al. 2015; Dyreng et al. 2010).
Our findings raise questions for future research. While the tax avoidance setting is largely separate from the direct skill-based influence of the TMT, other financial outcomes are not. Future research can consider whether intrapersonal functional diversity is influential in settings where certain legal/accounting skills may be directly useful (e.g., financial reporting quality).
We find similar results when studying GAAP effective tax rates, but do not tabulate these results for brevity.
We calculate these estimates using % High IFD (Table 4, Column 4), which captures the percentage of generalists on a TMT. For the five-member TMTs studied, replacing one non-generalist TMT member with a generalist increases % High IFD by 20 percent and decreases Cash ETR by 0.005 (% High IFD coefficient of −0.025 × 20%). This is equivalent to 2.09 percent of mean Cash ETR (i.e., 0.005 ÷ 0.239 mean Cash ETR from Table 1), or a tax savings of $3.25 million based on untabulated mean pretax income before special items of $650 million (i.e., 0.005 × $650 million).
Depreciation/depletion example excerpted from interview with a former Fortune 500 top manager.
BoardEx provides biographic information for managers and directors and contains both the past employment history and the current employment status of each TMT member. The database contains the job title(s), provides a role description, and reports the starting and ending dates of the various positions an individual manager has held in his/her professional experience. This data enables us to extract the functional experiences of each TMT member.
All tables in our study use the version of managerial ability recently updated by the original authors of Demerjian et al. (2012), for which data are available through the year 2016. See Appendix B for details. For the previous version of managerial ability, the authors estimated both the first and second stages by Fama-French 48 industry. They now estimate not by industry but by year for both stages. All other aspects (inputs/outputs, control variables) remain the same. Measuring by industry group reduces variation in firms' operating, investing, and financing activities that relate to firm efficiency and generates measures that are more comparable across industries. In contrast, measuring by year reduces temporal variation in these firm activities that relate to firm efficiency.
For example, a firm with positive taxes paid of $20 but a pretax accounting loss of $100 would have the same cash ETR as a firm with a tax refund of $20 and positive pretax accounting income of $100 (Henry and Sansing 2018, 1043).
This classification system differs somewhat from Bunderson and Sutcliffe's (2002). Most importantly, Bunderson and Sutcliffe (2002) provide greater detail regarding the manufacturing function (separating Production/Operations into Manufacturing, Distribution/Warehouse, and Equipment Management). Our classification also uses the categories Law and Others in place of Bunderson and Sutcliffe (2002)'s administrative support function.
While BoardEx provides a robust source of background data for calculating the Blau index, BoardEx has certain weaknesses (e.g., McWilliams, Rupp, Siegel, Stahl, and Waldman 2019). For example, certain employment start/end dates are sometimes omitted or incomplete. To alleviate this problem, we repeat all tests using a version of the Blau index that weighs each functional experience equally regardless of the time served in the given function (e.g., Cannella et al. 2008), thereby making start/end dates irrelevant. This alternative Blau index also prevents the index from changing based on the yearly accumulation of work experience (i.e., a manager moving from four years of experience in management to five) as opposed to management turnover. We find similar results across all tests using this alternative specification.
That is, our theory indicates that the TMT impacts tax avoidance by interfacing with firm functional units. Dominant Functional Diversity's impact, often in opposition to IFD, is attributed to within-team communication breakdowns caused by disparate primary specialties (e.g., Bunderson and Sutcliffe 2002). Due to the within-team focus of Dominant Functional Diversity, we do not expect a relationship with tax avoidance. However, it is possible that different functions are associated with different levels of tax conservatism or aggression, which may cause disagreements between members of a TMT with high Dominant Functional Diversity. We therefore control for Dominant Functional Diversity. To the extent that the within-TMT communication/disagreement affects tax avoidance, we may find that tax avoidance is affected by Dominant Functional Diversity.
While not at the 0.05 level or better, Dominant Functional Diversity is correlated with Cash ETR at the 0.1 level.
For robustness, we also use the decile rank of Demerjian et al.'s Managerial Ability to make the score more comparable across time and industries and to mitigate the influence of extreme observations (Demerjian, Lev, Lewis, and McVay 2013; Demerjian et al. 2012). The decile rank (untabulated) is also non-significant.
We thank an anonymous referee for suggesting this line of inquiry.
We also decompose the Generalism Index and find that each of its five components is non-significant. We argue that these results are intuitive. Consider, for example, the industry generalism component of the Generalism Index. To indirectly support the tax function, the transferable knowledge of and language to communicate with various fields is needed—e.g., finance. This finance knowledge can be obtained from experience in one industry, and the direct and detailed knowledge of the finance practices across industries is not likely to be highly relevant.
Aggressive (risky) tax avoidance is associated with stock price crash risk (Kim, Li, and Zhang 2011; Hanlon and Slemrod 2009), reputational concerns (Austin and Wilson 2017; Dyreng et al. 2016), and increased regulatory enforcement (Kubick, Lynch, Mayberry, and Omer 2016; Hoopes, Mescall, and Pittman 2012). This attention may lead to tax audits and tax liability revisions (Hanlon and Slemrod 2009).
While Cash ETR, BTD, and PermBTD capture some amount of aggressive tax behavior, these measures are much less reflective of tax aggressiveness than alternatives such as DTax, Shelter, and UTB (Goh et al. 2016).
That is, PermBTD captures the normal pursuit of various tax incentives, while “DTax measures permanent BTD (i.e., PermBTD) unexplained by legitimate tax positions” (Lisowsky et al. 2013, 591).
For example, temporary differences commonly result from depreciation, installment sales, or like-kind exchanges.
We thank an anonymous referee for various suggestions that improved this section.
The empirical approach in Table 6 differs from Koester et al. (2017) in two ways. First, we use the updated version of Managerial Ability while Koester et al. (2017) employs the now discontinued 2009 version. Second, we use a sample comprised of observations with non-missing IFD. Despite these differences, we find similar results in Column 1 of Table 6, that is, Managerial Ability is negatively associated with Cash ETR. However, we note that Koester et al. (2017) find significance at the 0.01 level, while we find significance at the 0.05 level and a somewhat lower magnitude (i.e., the coefficient equals −0.047 in our study, while it is −0.1884 in Koester et al. 2017).
In a robustness test Koester et al. (2017) find that controlling for ROA lowers, but does not eliminate, the significance on the 2009 version of Managerial Ability (coefficient reduced from −0.1884 to −0.0708; t-stat. reduced from −11.84 to −4.28). However, using the updated Managerial Ability score we find that controlling for ROA causes Managerial Ability to be non-significant whether we adopt Koester et al.'s (2017) control variables (Table 6) or ours (Table 3), and remains non-significant whether or not we restrict our sample to observations with non-missing IFD.
For example, oil industry incentives (https://www.wsj.com/articles/does-the-oil-and-gas-industry-still-need-tax-breaks-1479092522) or R&D credits (https://www.gao.gov/assets/160/154446.pdf) can cause controversy.
We set Audit Committee Independence to 1 if the committee has independent directors only, and 0 otherwise. We find similar results when using the percent of independent audit committee members.
The Sanderson-Windmeijer F-Statistic for weak identification is 91.54 (p-value < 0.01) for the first-stage regression on IFD, meaning that the model is adequately identified by the instruments. The Hansen J-statistic (p-value = 0.39) is not significant at conventional levels, indicating the null hypothesis that the instruments are uncorrelated with the errors in the second-stage regression cannot be rejected.
Overall, prior literature argues that managerial incentives improve firm tax avoidance outcomes (e.g., Seidman and Stomberg 2017; Gaertner 2014). However, some prior studies argue that certain forms of tax avoidance serve as opportunities for managerial rent extraction (e.g., Desai and Dharmapala 2009; Desai and Dharmapala 2006).