Board Connectedness and Board Effectiveness

Board Connectedness and Board Effectiveness∗ Vincent J. Intintoli Clemson University Tel: 864-656-2263 Email: [email protected] Kathleen M. Kahle University of Arizona Tel: 520-621-7489 Email: [email protected] Wanli Zhao Southern Illinois University Tel: 618-453-7109 Email: [email protected]

Abstract We examine the effect of the social connectedness of independent, non-co-opted directors on their ability to monitor and advise the firm. We begin by providing evidence that well-connected directors have greater protection from career concerns. We next examine the channels by which director connectedness may improve monitoring and find that audit committee connectedness has a positive effect on the quality of financial reporting. Further, better connected compensation committees are less likely to overpay the CEO. Finally, we examine the effect of well-connected directors on the firm’s information environment. We show that firms with highly connected boards have lower financing costs and higher payout ratios. We also find that better connected boards are better able to shield the firm from the negative impact of a competition shock. Our results are robust to multiple approaches to mitigate endogeneity. Overall, well-connected boards appear to be effective in promoting shareholders’ interests.

JEL Classifications: G30; J33; M52; Keywords: Centrality; Board connectedness; Board effectiveness; Fraud; Compensation; SEOs

∗

We would like to thank Luke DeVault, Daniel Greene, Jayanthi Krishnan, Brandon Lockhart, Angie Low, Angela Morgan, David Reeb, Matthew Serfling, Johan Sulaeman, Mike Weisbach, Jack Wolf, Ryan Williams, Fei Xie, Bernard Yeung, and seminar participants at Clemson University, Nanyang Technological University, and National University of Singapore for useful comments.

1. Introduction Early studies of corporate boards focus on the roles of inside vs. outside directors. Inside directors are presumed to specialize in knowledge of the firm while outside directors are considered to be independent of the CEO’s influence and thus are better positioned to monitor management. Consequently, conventional wisdom suggests that having more outside directors increases firm value. Yet empirical studies provide mixed evidence on the effect of board independence on firm performance (Hermalin and Weisbach, 1998; Bhagat and Black, 1999). Against this backdrop, recent evidence suggests that directors who are conventionally independent may not be truly independent from CEO’s influence, which can result in lower pay-performance sensitivity, worse firm performance, and reduced likelihood of CEO dismissal (Hwang and Kim, 2009; Fracassi and Tate, 2012; Coles, Daniel, and Naveen, 2014). Other studies on boards go beyond simply focusing on director independence, instead borrowing from the social networks literature to explore the social connections of directors, which is often referred to as director “connectedness” or “centrality.” Well-connected board members possess critical positioning in the social network and therefore have greater influence and access to better information sets, offering multiple benefits to the firm. For example, well-connected directors may foresee industry trends, market conditions, and regulatory changes sooner (Mizruchi, 1996; Mol, 2001). These directors may also provide the firm with better business and political contacts (Mol, 2001; Nicholson et al., 2004). Finally, director connections reduce information asymmetry between the firm and the external market (Schoorman et al., 1981) and facilitate information transmission between firms on innovations and value-enhancing practices (Haunschild and Beckman, 1998). These benefits can improve both the monitoring of managers and strategic advising within the firm. Consistent with the benefits of well-connected directors, Larcker, So, and Wang (2013) find that firms with well-connected boards earn higher abnormal returns. We examine the social connectedness of independent directors to gain insights into how board connections aid in the effectiveness of monitoring and advising. Following Coles et al. (2014), we focus on independent directors who are not co-opted by the incumbent CEO, since the connectedness of directors with close ties to the CEO could be used to help entrench management. However, in order for boards to 1

effectively monitor managers, directors must be both independent of management and feel insulated from potential repercussions. We suggest that directors who are well-connected should feel shielded from career concerns and thus be in a better position to monitor. To test this notion, we examine the career consequences of directors after the detection of fraud. We find that in the aftermath of fraud, well-connected directors are less likely to experience turnover and more likely to obtain a board seat at another firm than directors who are not well-connected. 1 After establishing that connectedness aids in shielding directors from career concerns, we next examine the channels by which director connectedness may improve monitoring. In particular, we begin by focusing on the effect of audit committee connectedness on earnings quality. We find that audit committee connectedness has a positive effect on the quality of financial reporting. Our results complement Omer, Shelley, and Tice (2014), who find that firms with well-connected directors are less likely to restate earnings, and other studies that find that specific expertise of the committee members results in reduced earnings management and financial fraud (Wang, Xie, and Zhu, 2013; Krishnan, Wen, and Zhao, 2011; DeFond, Hann, and Hu, 2005). We next turn to the effect of director connectedness on CEO compensation. Hwang and Kim (2009) show that the “informal social networks” of directors (e.g., alma mater, military service, or birthplace) impact their monitoring ability, including disciplinary actions such as CEO dismissal and pay-performance sensitivity. Their results suggest that a considerable percentage of directors currently classified as independent are substantively not as a result of these social ties. We find that compensation committees with greater connectedness grant CEOs lower total pay, consistent with the notion that better connected directors have an informational advantage about CEO ability and are less likely to overpay top executives. While the above results indicate that better-connected directors are more effective monitors, we also examine the effect of well-connected directors on the strategic direction and information environment

1

In a related paper, Cashman, Gillan, and Whitby (2010) find that directors with better professional connections, through both common board appointments and overlapping work experience, are more likely to receive additional board seats and less likely to suffer if they are on the board of a firm that restates its financials.

2

of the firm. If well-connected directors use their knowledge, experience, and connections to better communicate information to the market, then it should be reflected in the firm’s costs of financing. Consistent with this channel, we show that when firms with highly connected boards access the seasoned equity market, the market reaction to the issue announcement is less negative, underpricing is lower, and the offer is more likely to be accelerated. Additionally, we find evidence that the cost of debt decreases when board connectedness increases. We further show that the payout ratio of better-connected boards is higher, although this relation is mitigated in firms with high growth potential. Finally, we use a natural experiment to show that firms that experience an industry competition shock are better able to “buffer” against the negative impact of increasing competition if they have a well-connected board. Overall, our results suggest that board connectedness results in greater firm value especially when firms have greater needs for advising. We address potential self-selection concerns, i.e., better firms are able to attract directors with greater connectedness, using multiple approaches. 2 First, we control for the cross-sectional differences between firms with higher and lower connectedness with propensity-score matching, in which we control for a variety of firm and director characteristics. In particular, we control for CEO tenure, CEO power, CEO connectedness, and CEO-director friendliness. Further, we control for firm fixed effects whenever feasible. Second, even though a matched sample does not completely rule out an omitted variable bias, we use an event-study approach to facilitate the causal interpretation. Third, we explore changes specifications based on instances of director death to establish causality. Finally, we explore an exogenous shock to industry competition to establish identification for the effect of board connectedness on firm strategic reactions. Overall, our results continue to suggest that the net benefit of board connectedness is positive, consistent with the information advantage argument and the career concern proposition for the directors.

2

Untabulated results suggest that self-selection is not likely to drive our results, however. First, we find a stronger effect of board connectedness among smaller firms than in bigger firms. As directors are more likely to self-select into large, prestigious firms, the findings suggest that self-selection is not the primary driver in our results. Second, we find stronger results among firms in which the majority of directors are not “busy”. Such directors do not have many options and thus are less likely to self-select into desired firms, suggesting that self-selection is not the driving force behind our findings.

3

Our paper is related to several studies in the director connectedness literature. Larcker et al. (2013) find that firms with central boards of directors earn superior risk-adjusted stock returns. We complement their findings by providing evidence on the specific channels through which board connectedness aids in monitoring and advising. Our paper fills a gap in the literature by explicitly examining the relation between board connectedness and a full spectrum of board decision-making scenarios. Renneboog and Zhao (2011) analyze the relation between CEO compensation and networks of executive and nonexecutive directors for all listed UK companies. They find that companies with many director interlocks, which they identify as busy boards, have reduced monitoring effectiveness, which leads to CEO compensation that is both higher and less sensitive to firm performance. The governance setting in the U.K. allows the CEO (as well as other insiders) to be on the compensation committee, however, which potentially results in conflicts of interest. Our paper extends their results to the U.S. and shows more specifically that even after controlling for busy boards, compensation committees whose independent, nonco-opted directors are better connected grant lower CEO compensation. These findings are consistent with well-connected directors being better monitors and advisors. Our paper also complements the findings in Coles, Wang, and Zhu (2015) who find that CEO turnover is more likely in firms whose directors are well connected. They conclude that well-connected directors are better monitors due to their informational advantages. Rather than focusing on the effect of connectedness on CEO dismissal, we explore other potential channels by which director connectedness may improve both monitoring and the information environment within the firm, including the effect of connected directors on earnings quality, the costs of debt, payout policy, and the firm’s ability to deal with competition shocks. Finally, Denis et al. (2014) find that directors with more external connections, including mutual board service, employment, education, and military service, seek the quiet life by joining boards that are less likely to threaten their time and reputations. Our findings suggest that well-connected directors are more willing to monitor and do so by improving earnings quality.

4

The remainder of this paper is organized as follows. Section 2 provides a literature review and section 3 discusses our measures of social connectedness and other variables of interest. Section 4 presents our empirical findings and section 5 concludes.

2. Literature Review 2.1 The Role of Boards Fama (1980) and Fama and Jensen (1983) posit that a critical function of the board, and specifically outside directors, is to monitor managers in order to ensure that they are acting in the best interest of shareholders. The effectiveness of board monitoring depends largely on the independence of the members, however. For example, Hermalin and Weisbach (1998) suggest that a director’s willingness to monitor (and potentially replace) the CEO depends on the director’s independence. Other studies examine director monitoring effectiveness by focusing on the conflicts of interest between managers and independent directors. Coles et al. (2014) find that monitoring effectiveness is compromised when supposedly independent directors are appointed by the same CEO that they are expected to evaluate. Shivdasani and Yermack (1999) document that investors react less positively to the appointment of independent directors selected by the CEO. Hwang and Kim (2009) find that directors who have social connections to the CEO grant higher levels of CEO pay that is unrelated to performance. Cohen et al. (2012) provide evidence that firms appoint independent directors who are overly sympathetic to management; following these appointments, firms significantly increase their earnings management activities and CEO compensation. In sum, these studies suggest that director affiliation with management may compromise monitoring diligence. Although outside directors are mainly perceived to play the role of monitor, director experience, knowledge, and external connections also assist in their ability to advise top managers on strategic direction and corporate policies. Adams and Ferreira (2007) argue that it is imperative for directors to have sufficient information and knowledge about the firm’s operations, while Dass et al. (2014) document that directors’ industry experience promotes monitoring effectiveness. Further, regulatory reforms enacted as part of the Sarbanes-Oxley Act call for financial expertise on the audit committee to ensure that members have 5

sufficient skills to detect financial reporting misconduct (DeFond et al., 2005). Given the complexity of information disclosed by managers to the board, the ability to access external information about market trends, industry development, and other relevant information is vital in enabling directors to analyze and understand the information disclosed. Therefore, access to external information is a valuable resource that assists the directors’ effectiveness. 2.2 Board Connectedness While early studies of boards focus on the role of inside vs. outside directors, recent studies in both management and financial economics explore the external connections of directors. Better connections offer a number of advantages. For example, directors with better connections are perceived to have elevated levels of social power that promotes their career prospects (Liu, 2014). While board members generally value their positions, compensation committee chairs with greater connectedness may feel less dependent upon any given board seat and less obligated to any particular CEO (Belliveau et al., 1996). In addition, better connected directors should be highly respected by other members of the board, creating a welldefined informational hierarchy that can improve the efficiency of board interactions (He and Huang, 2011). For these reasons, directors with greater connectedness have superior bargaining power and hence greater influence over managers. Directors with better connectedness also play a pivotal role in information transfer within the social network, which can assist in the spread of information from the firm to the external market. Better access to information offers multiple benefits. Well-connected directors may be better able to (1) foresee industry trends and market conditions (Mizruchi, 1996; Mol, 2001), (2) reduce information asymmetry between the firm and the external market (Schoorman et al., 1981), (3) provide the firm with better contacts, including clients, suppliers, politicians and government officials (Mol, 2001; Nicholson et al., 2004), and (4) facilitate information transmission between firms on innovations and value-enhancing practices (Haunschild and Beckman, 1998). Overall, firms with well-connected directors should be better able to learn from other firms’ successes and avoid their mistakes.

6

However, connectedness can also compromise board monitoring efficacy. Prior work suggests that information spillover from connections may aid in introducing value-decreasing management practices (Bizjak et al., 2009; Armstrong and Larcker, 2009) and that misleading or incorrect information may spread through the board network (Larcker et al., 2013). Well-connected directors may also shift their efforts from firm matters to maintaining their social status. Boardroom networks have also been shown to result in valuedecreasing practices such as option backdating (Core et al., 1999; Loderer and Peyer, 2002; Fich and Shivdasani, 2006). Therefore, the question of whether better board connections aid in monitoring and evaluation is an empirical one. 3 2.3 CEO-Director Friendliness CEO-director connections may enhance the communications and information sharing between the CEO and the board, suggesting that the board may be better able to monitor managers (Adams and Ferreira, 2007; Engelberg, Gao, and Parsons, 2012). On the other hand, recent evidence suggests that a CEO’s connectedness to their own directors is problematic and results in higher compensation, lower payperformance sensitivity, worse firm performance, and reduced likelihood of CEO dismissal (Hwang and Kim, 2009; Fracassi and Tate, 2012; Coles et al., 2014). Further, Chidambaran, Kedia and Prabhala (2012) find that CEO-director connections have a positive and significant effect on the probability the firm commits fraud. Khanna et al. (2015) find that CEO connections developed with top executives and directors through the CEO’s appointment decisions heighten the risk of corporate fraud, while decreasing the likelihood of both fraud detection and CEO turnover after the detection of fraud. CEO-director connections based on prior education, employment or social network ties have an insignificant effect on fraud. Consequently, we control for CEO-director affiliation (i.e., friendliness) in all of our tests.

3

Studies have shown that a director’s “busyness,” defined as the number of board seats a director holds, is negatively associated with monitoring and shareholder wealth (Core et al., 1999; Loderer and Peyer, 2002; Fich and Shivdasani, 2006; Masulis and Mobbs, 2013). While a busy director will tend to have more connections, our connectedness measure encompasses more than just the number of connections, so a well-connected director is not necessarily busy. In fact we find that the correlation between connectedness and busyness is 0.37. We control for busyness in our empirical tests, nevertheless.

7

3. Data In order to measure board connectedness, we use the social network database from BoardEx of Management Diagnostics Ltd., which provides detailed demographic information about the directors and top five disclosed earners of publicly traded U.S. companies. 4,5 BoardEx covers more than 323,485 unique individuals from over 76,000 unique entities, including private and public companies, universities, and other non-profit organizations, with common employment histories going back as far as 1926. BoardEx covers relational links among directors and other corporate officials through cross-referencing their employment history, educational experience, and professional qualifications/experience. Our sample spans from 2001 to 2010. 6 We extract data on each executive/director and their links with all other executives and directors in the database. 7 We repeat this process each year as the cross-reference may change for two reasons: First, newly added individuals may have connections with existing individuals already in the database. Second, new connections will be established or severed as a result of executives/directors

4

BoardEx is commonly used to gauge the social network or connectedness of corporate executives and members of the board of directors (e.g., Fracassi and Tate, 2012; Chidambaran, Kedia, and Prabhala, 2012; Cohen, Frazzini, and Malloy, 2010). Fogel, Ma, and Morck (2014) find that sudden deaths of centrally located independent directors are met with decreases in shareholder value, suggesting that such directors are valuable to firms. However, Omer, Shelley, and Tice (2014) find that firms whose directors are more centrally located with networks have lower performance. Other studies assume that geographic distance is a measure of how “connected” the director is to the firm. For example, Alam et al. (2014) show that geographic distance between directors and corporate headquarters is related to information acquisition and board decisions. 5 BoardEx has two potential data biases. First, BoardEx expands its coverage in 2006 but does not back-fill the missing information for the new firms (Larcker, So, and Wang, 2013). Second, even though BoardEx covers both professional affiliations and other social interactions between executives and directors, it is conceivable that alternative channels through which individuals can establish social connections exist and are not included in BoardEx (e.g., marital connections). We address these concerns in robustness tests. For instance, we find similar results when restricting our sample to firms that exist in the database in each year. In addition, to the extent that the bias is related to certain firm characteristics, we utilize a propensity-score matching sample and find similar results. More specifically, we match firms with high connectedness with firms with low connectedness, based on firm size, volatility, leverage, institutional ownership, ROA, CEO power, board size, board independence, industry and year dummy, using a caliper of 0.1% with no replacement. 6 We ignore 2000 as BoardEx covers significantly more firms in 2001, indicating data incompleteness in 2000. 7 Our definitions of social connections are similar to Liu (2014) and Fracassi (2014). A connection between two individuals is established if they have worked in the same company historically (both as executives, both as directors, or one executive one director). We also include their connections via other mechanisms such as civil services in social clubs, charity organizations, and non-profit entities. We only count connections where the date information is not missing. Further, we count their connections if they obtain their undergraduate or graduate degrees from the same school and graduate within 2 years of each other.

8

changing firm affiliations. We focus on independent directors who serve on the board before the incumbent CEO is appointed, i.e., non-co-opted directors. We find that 48% of independent directors are “truly” independent, which is comparable to the 47% reported in Coles et al. (2014). 3.1 Measures of Social Connectedness Newman (2003) notes that it is important to address the impact of an individual in the entire network (i.e., identify which individuals are most connected to others or have the most influence). Consequently, we utilize four centrality measures developed in network theory that capture not only social ties but, more importantly, the quality of those connections: degree, betweenness, closeness, and eigenvector centrality, as introduced in Proctor and Loomis (1951), Freeman (1977), Sabidussi (1966), and Bonacich (1972), respectively and used in recent studies (Renneboog and Zhao, 2011; Larcker et al., 2013). 8 While there is no theory as to which measure is superior, each measure captures distinct aspects of the relative importance of every individual in the entire network. We discuss each measure below and provide mathematical definitions of the measures in Panel A of Appendix A. 3.1.1 Degree Conceptually, the simplest measure of centrality is degree, which captures the number of direct links an individual has with other individuals in the network. The more connections the individual has, the more important she is in the network. We take the average number of connections of the independent, nonco-opted directors on the board/committee as the board/committee level measure of degree.

8

Our measures assume that the “nodes”, or individuals in the network, are connected via the shortest connections or paths between them. In reality, people often contact those with who they have more prominent shared experience, rather than simple acquaintances. For example, a board member may contact a colleague that he has known for a number of years, rather than someone with whom he recently shared a charity board experience. It is difficult to agree upon a parsimonious measure to rank such connections. For instance, ties to company colleagues may be more important than common board service acquaintances, while corporate board connections may be ranked higher than social boards (e.g., charity, foundations, non-profit hospitals, academic institutions). In robustness tests, we examine the time that two persons have the shared experience. We assume that the longer two persons share a common experience, (a) the stronger the connection and (b) the higher the likelihood that they contact each other in order to transfer information. We then use the common time between two persons as the weight to rank the connection. Results are robust to using the weighted version of each measure.

9

3.1.2 Betweenness Betweenness represents the importance of an individual serving as the shortest information bridge or intermediary for other members (Freeman, 1977). Individuals with a higher betweenness measure have access to richer and more differentiated information. Assuming that individuals connect with each other by the shortest path between them, betweenness is the sum of the shortest paths between all pairs of other individuals that pass through a person, scaled by the total number of shortest paths between the same pair of individuals. We standardize the measure by (n-1)(n-2)/2, where n represents the number of members. 3.1.3 Closeness Closeness is defined as the inverse of the sum of an individual’s distances to all other members (Sabidussi, 1966). The “closer” an individual is, the lower her total distance to all other members. Intuitively, having closer connections to more people makes information exchange quicker and more readily available, thus resulting in greater influence on others and higher social power. 3.1.4 Eigenvector Centrality Eigenvector centrality focuses on the influence of an individual in the network. It is based on the notion that not all individuals connected to a given person are equally important. Google’s PageRank is a variant of this measure, which takes into account the relative importance or popularity of connected webpages. A person who is connected with more important individuals is herself more important in the information dissemination channel and will have higher Eigenvector centrality, all else equal. We use the Perron-Frobenius theorem to ensure that all Eigenvectors are positive and that only the greatest Eigenvalue results in the desired centrality measure. 3.1.5 An Example of Centrality Measures We report formal definitions of the centrality measures in Panel A of Appendix A and present a simple director network in Panel B in order to illustrate our calculations of the centrality measures. This exemplary network has ten directors and every director is connected with each other either directly or indirectly. We briefly describe the calculations of the four centrality metrics for some of the directors in this example, as summarized in Panel C. First, Mark has the highest degree centrality because he is directly 10

connected to 6 other directors. His degree metric is calculated as 6/(10-1) = 0.667. However, his closeness metric is not the highest because it takes four steps to reach Laura, three steps to James, two steps to Nancy and one step to the rest directors. As such, his closeness centrality is (10-1) * (1/(4+3+2+1*6)) = 0.6. Jack and Karen on the other hand have the highest closeness measure as they can quickly reach other members of the network, considering all direct and indirect links. Betweenness involves the shortest paths between all other pairs of the network. For instance, for David, there are 36 pairs of other directors in the network and he is one of the two shortest paths between Tom and John. Also, David lies on one of the three shortest paths between Tom and Jack. No other shortest path passes through him. Therefore, we calculate his betweenness as (1/2 + 1/3)/36 = 0.023. By this logic, we find that Nancy actually has the highest betweenness metric. In other words, she serves as a crucial information flow broker of the network. Eigenvector centrality is a weighted sum of the degree measure. John and Nancy have the same degree metric (i.e., 0.333). However, we observe that John connects to David, Mark, and Jack, while Nancy connects to Jack, Karen, and James. Considering the degree measures of the six connected directors, we can see that (1) David’s degree is higher than James’ and (2) Mark’s degree is higher than Karen’s. As such, John connects to directors with higher degree than Nancy does. The difference is reflected in the Eigenvector measure: John’s Eigenvector centrality is 0.594 and Nancy’s is 0.407. 3.1.6 Aggregate Social Connectedness We present summary statistics of the director-level centrality measures in Panel A of Appendix B. 9 We also take the average of each centrality metric for all independent, non-co-opted directors within each firm-year to obtain the board-level centrality measures. We report the averages of each of the board-level centrality statistics across all firm-years in Panel B. Each measure of social connectedness focuses on a distinct aspect of the individual’s position or importance in the network. The correlation matrix between the board-level centrality measures, shown in Panel C, suggests that they are positively correlated to each

9

Our measures of centrality exhibit similar summary statistics as in Table 3, Panel A of Liu (2010).

11

other. Finally, we use factor analysis to extract the common latent factor that explains the variations across these measures, and use the factor score as our social connectedness measure. For tests at the firm level, the factor score is calculated for all independent, non-co-opted board members, which we identify as Board Connectedness. 10 Similarly, we calculate connectedness separately for the audit (AC Connectedness) and compensation (CC Connectedness) committees. For some tests, we also calculate the factor scores at the director level to determine the connectedness of each individual director. 11 3.2 CEO Power One of the most important responsibilities of the board is to monitor the CEO and curb opportunistic behavior. However, the ability to monitor depends on the relative power of the CEO compared to that of the board (Hermalin and Weisbach, 1998; Morse, Nanda, and Seru, 2011). We use three proxies for CEO power: CEO tenure, CEO equity ownership, and CEO/chairman duality. We apply factor analysis to extract the common underlying latent variable, which we denote as CEO Power. 12 Further, we include the level of CEO connectedness as additional control for alternative manifest of CEO power. 3.3 Measuring CEO-Director Friendliness and Board/Committee Director Busyness We also account for the effect of CEO connections to independent, non-co-opted directors via historical overlapping experience. We first identify overlapping mutual work experience (as directors

10

In Panel D of Appendix B, we show that the factor analysis loads on one factor that has an Eigen value bigger than 1, which explains 63% of the variation. Alternatively, we follow Larcker et al. (2013) and rank each firm into quintiles by degree, closeness, betweenness, and eigenvector centrality, and then use the average ranking to measure overall centrality. The correlation between the factor score and the “N-score” of Larcker et al. is 0.66. In untabulated results, our main findings are robust to the use of the N-score measurement approach. 11 Based on our aggregate connectedness measure, the most connected directors include Bruce L. Claflin (Chairman of the Board at AMD and CEO of 3Com), James E. Daley (director at Adobe and former CFO of EDS), Douglas Alexander Warner III (Chair of the audit committee at GE and former CEO of J.P. Morgan Chase), Kevin W. Sharer (director at 3M, former CEO of Amgen, lecturer at Harvard, US Naval Academy graduate), John Fellows Akers (director at New York Times, former CEO of IBM), Tim D. Cook (director at Nike, COO of Apple), and Walter (Jim) James McNerney Jr. (Chairman of the Board at P&G, CEO of Boeing, CEO of 3Com). The least connected directors include Angelo R. Mozilo (former director at Home Depot, former CEO of Countywide Financial), Jane Margaret O'Brien (director at Norfolk Southern, President at St. Mary’s College), Kent B. Foster (director at JC Penny, CEO of Ingram Micro Inc.), Kathy J. Victor (director at Best Buy, President of Centera), Ann Marie Fudge (director at GE, former CEO of Young & Rubicam), Barry K. Allen (director at Harley Davidson, Executive V.P. of Operations at Qwest Communications), Arthur C. Martinez (director at Pepsico, former CEO at Sears), and Hugh Jesse Arnelle (director at Textron, lawyer). 12 The three proxies exhibit correlations of 0.32, 0.18, 0.33, and the factor analysis yields one common factor with an Eigen value greater than 1 (1.55), which explains 52% of the common variance among the three proxies.

12

and/or executives) between the CEO and each independent, non-co-opted director. We then take the sum of the independent, non-co-opted directors who have a connection to the CEO and divide it by the total number of independent, non-co-opted directors, which gives us the overlap ratio (CEO/Board Overlap Ratio). We calculate overlap ratios for the audit (CEO/AC Overlap Ratio) and compensation (CEO/CC Overlap Ratio) committees when applicable. Further, we control for the effect of “busy” directors on board or committee monitoring. Specifically, busy directors are independent, non-co-opted directors who serve on 4 or more public firms’ boards (Fich and Shivdasani, 2006). We denote Busy Board, Busy AC, and Busy CC by the proportion of busy directors on the overall board, the audit committee, and compensation committees, respectively. 3.4 Financial Reporting Quality Measures We use three accruals-based measures that are commonly used in accounting studies to proxy for firm’s financial reporting quality. Our first accruals based measure (Hribar) is defined as the absolute value of abnormal accruals based on Hribar and Nichols (2007). The second measure (DGLS) is the error term from the estimation of accruals calculated as the industry-adjusted absolute value of the Dechow and Dichev (2002) residual, based on the cross-sectional adaptation of the model in Dechow et al. (2011). Our third measure (AQ) is the Kothari et al. (2005) performance-matched signed discretionary accruals estimate as used by Ashbaugh et al. (2003). Detailed descriptions of each measure are in Appendix C. 3.5 Summary Statistics Table 1 presents summary statistics on the main dependent variables. The first section focuses on the dependent variables in the tests detailed in section 4. We find that the average (median) reporting quality as measured by Hribar is 0.142 (0.110). The means of the other two measures for reporting quality (i.e., DGLS and AQ) is 0.005 and -0.006, respectively. The average annual total compensation for a CEO is $4.76 million and the median is $2.86 million. Dividend payout is 3.9% of total assets on average with the median of 1.6%. Mean (median) SEO underpricing is 3.36% (2.46%) and the average SEO announcement CAR is -0.56%.

13

We next focus on the board characteristics. Board Connectedness has a mean of 0.017 and standard deviation of 0.99. We observe relatively similar statistics for the audit committee and compensation committee connectedness measures. CEO Power has a mean value of 0.04 and standard deviation of 0.914. Roughly 16% of the independent directors have social connections with the CEO, while less than 10% of audit/compensation committee members share social connections with the CEO. We find that 28% of the independent directors are “busy,” i.e., have 4 or more board seats; similar statistics are observed for the audit and compensation committees, which are untabulated for brevity. The median board has 9 members while the mean board independence (based on non-co-opted directors) is approximately 48%. As an important factor for audit committee monitoring, we see that on average 18% of the audit committee members are accounting experts. The last section of Table 1 examines other firm characteristics. Our average firm has total assets of $6.4 billion, with the top (bottom) quartile at $4,500 ($605) million, suggesting a wide coverage of firms in our sample. The mean (median) ROA is 9.2% (9.3%). On average, firms have leverage of 17.6% and a market-to-book ratio of 3.03. Finally, roughly 70% of the common equity of our sample firms is owned by institutional investors and there are on average 12 analysts following each firm.

4. Multivariate Tests 4.1 Director Connectedness and Career Concerns The primary roles of the board is to monitor and advise management. Following Liu (2014), who suggests that a CEO’s outside options influence turnover, we posit that directors who are well-connected are better shielded from career concerns and thus are in a better position to monitor. We use cases of detected fraud as platform to test such a notion. We focus on the independent, non-co-opted audit committee members and examine both director turnover following fraud and the procurement of subsequent board seats. Audit committee members are specifically charged with the oversight of financial reporting and thus are particularly concerned about the consequences of misleading financial reporting, which can result in

14

lawsuits and SEC action. 13 Such litigation is a source of career concerns, in that inefficient monitoring may lead to director turnover and reputation damage (Farber, 2005; Srinivasan, 2005). We rely on Accounting and Auditing Enforcement Releases (AAER), legal actions from the Securities and Exchange Commission and Department of Justice (SEC/DOJ), as well as class action lawsuits to indicate the quality of financial reporting of the firm. We use the Audit Analytics legal case database to proxy for misconduct within our sample of firms and to identify the period in which the firms are alleged to be involved in financial misconduct. Specifically, we create a dummy variable (Fraud) to identify legal actions related to accounting malpractice, mergers & acquisitions, securities law, financial reporting, fraud, AAER, class action, stockholder suits, and initial public offerings (IPOs). Our fraud variable identifies the fraudulent activity as of the time that the public becomes aware of the misconduct, not the time that the alleged misconduct occurred. 14 We begin by examining the likelihood of director turnover in the year following the discovery of fraudulent activity and suggest that higher levels of connectedness shield directors from the negative effects of fraud on their careers. If connectedness helps shield the audit committee members from career concern, then relative to other directors, they should be less penalized by the labor market when a monitoring failure occurs. In Panel A of Table 2, we examine director-level regressions in which the dependent variable is equal to 1 if director turnover occurs. In Column 1, we use our entire sample of non-co-opted audit committee director observations and include annual controls for fraudulent activity (Fraud), an indicator set to one if director connectedness is in the top quartile of the sample (High AC Connectedness), and an

13

There is evidence that lawsuits are positively associated with deficient financial reporting. DuCharme et al. (2004) find that earnings management around stock offers is positively associated with the probability of subsequent litigation. Palmrose et al. (2004, p. 145) document that ‘‘core and pervasive restatements increase the likelihood and severity of lawsuits, incremental to other litigation factors including fraud, the impact of the restatement on net income, and stock prices.” Further, a PricewaterhouseCoopers (2000) survey suggests that accounting issues related to revenue recognition are positively associated with class action lawsuits. 14 As with any work examining legal action, our litigation sample does not capture all possible cases. For instance, the Public Company Accounting Oversight Board (PCAOB), which was created by the Sarbanes-Oxley Act, can prosecute firms for misconduct but only has to disclose winning cases. On the other hand, SEC/DOJ cases are exposed no matter the outcome of the litigation. Nonetheless, by using actual legal actions, we are unable to identify firms that commit undiscovered financial misconduct or cases where firms are innocent and misclassified as offenders.

15

interaction term between highly connected audit committee director and fraud. We also control for director age and an indicator variable if the director is female, as prior studies suggest that older directors suffer more turnover and female directors experience less turnover than their male counterparts (Yermack, 2004; Srinivasan, 2005). Further, we control for industry, firm, and year fixed effects. Consistent with our predictions, the coefficient on Fraud is positive and significant, indicating that fraudulent activity is associated with a higher likelihood of audit committee member turnover. Further, the negative and significant coefficient on High AC Connectedness suggests that, overall, highly connected directors are less likely to be dismissed from the board. The negative and significant coefficient on Fraud * High AC Connectedness indicates that in the case of fraud, audit committee members with greater connectedness face less turnover than less connected members. The economic significance of our findings is also strong; highly connected audit committee members have an 11.8% lower chance of turnover. Additionally, when fraud occurs, these directors have a 47% lower likelihood of turnover. 15 In order to alleviate the concern that certain audit committee members choose to serve on the boards of specific types of firms (self-selection), we utilize a propensity score matched sample and report the results in Column 2. We match firms with detected fraud to firms without based on the following variables: firm size, ROA, CEO tenure, CEO power, CEO connectedness, ROA, volatility, institutional ownership, analyst following, audit committee size, audit committee independence, average audit committee director tenure, average audit committee director age, audit committee female ratio, fraud score (Fscore), industry and year controls. We use one-to-one matching with a caliper of 0.1% without replacement based on the year prior to the fraud being detected. We match on Fscore to mitigate the possibility that better connected directors choose to sit on the boards of firms that have a lower likelihood of engaging in fraudulent

15

An alternative interpretation of our findings is that AC members from firms that have at least one well-connected AC member are less likely to be punished following fraud. In other words, our results could be driven by the firmlevel effect instead of the director effect. To investigate such possibility, we split our sample firms by the median level of AC connectedness and we repeat our test for each subsample. We find that the effect of AC director connectedness on their turnover is more pronounced in the firms that have lower AC connectedness, suggesting that it is the individual director’s connectedness that drives our findings.

16

activity. 16 Alternatively, we use litigation risk to proxy for such reputational concerns of the directors and we find very similar results. Similar to our Column 1 findings, the results suggest that (1) audit committee members with a high level of connectedness are 18% less prone to turnover, and (2) in the case of fraud, audit committee members with greater connectedness face 28% less chance of turnover than other directors. In Column 3 we limit the sample to firms who experience fraudulent activity; the results are consistent with those in Columns 1 and 2. In Panel B of Table 2, we further examine the career concerns of directors by studying the likelihood of audit committee members obtaining board seats at other firms following the detection of fraud. Our dependent variable is the total number of additional board seats the director obtains in the three years following the detection of fraudulent activity. As in Panel A, we report results on the full, matched, and fraud only samples in Columns 1-3, respectively. Examination of the full sample indicates that on average audit committee members receive fewer future board seats following fraud. The positive and significant coefficient on High AC Connectedness suggests that better connected audit committee members obtain more future board seats, all else equal. Interestingly, after the discovery of fraud, these directors obtain relatively more board seats than audit committee members with low connectedness, as shown by the positive and significant coefficient on the Fraud * High AC Connectedness interaction. Results in Columns 2-3 provide similar evidence. These findings are consistent with the idea that better connected directors have more external opportunities and are less likely to be punished by the labor market after a monitoring failure is detected. Alternatively, these directors may have discovered the fraudulent activity, which constitutes a positive signal to the labor market. 17

16

We construct a fraud score (FScore) using the coefficients from Dechow et al. (2011), who use data on legal actions from the SEC’s Accounting and Auditing Enforcement Releases to develop a prediction model for accounting fraud. See Appendix C for the model and variable definitions. Dechow et al. (2011, p. 61) suggest that the ratio of the computed predicted probabilities of misstatements from their model to the unconditional probability of misstatements can be used as a measure of fraud likelihood relative to a random firm taken from the population. However the unconditional probability is specific to the time period of their study (1982-2005). Because our time period does not correspond to theirs, we use the predicted probabilities and use its sample distribution to demarcate situations of highand low-likelihood of potential misstatements. 17 We also explore the notion that better connected directors may be less prone to firm performance to obtain future board seats. Untabulated results suggest that the sensitivity of future board seats on performance is much lower among directors with high level of connectedness than directors with low connectedness.

17

4.2 Audit Committee Connectedness and Financial Reporting Quality The previous section provides evidence that well-connected directors are shielded from career concerns and thus are in a stronger position to monitor and advise. In this section we begin to explore the channels by which director connectedness may promote shareholders’ interests by examining the effect of audit committee connectedness on earnings quality; we focus on the audit committee since it is specifically charged with the oversight of financial reporting and disclosure. Financial reporting quality is of crucial importance to the firm’s information environment and the level of investor risk. In the wake of recent accounting misconduct, the Sarbanes-Oxley Act and Regulation Fair Disclosure focus on strengthening the audit committee and increasing the committee’s responsibilities and authority. Recent studies document that audit committee financial expertise is negatively associated with earnings management (Bédard et al., 2004; Carcello et al., 2011; Dhaliwal et al., 2010) and internal control problems (Krishnan 2005; Hoitash and Hoitash, 2009), and positively associated with accounting conservatism (Krishnan and Visvanathan, 2008). We examine whether earnings quality is higher in firms with better connected audit committees. Given that managers often withhold information from the board in an attempt to mitigate diligent monitoring (Adams and Ferreira, 2007), audit committee connectedness may substitute for such incomplete information by incorporating outside information regarding the firm’s operations and industry practice obtained through the committee’s connections. Such information enables committee members to better understand financial statements and improve corporate governance. Therefore, we propose that better connectedness is associated with better monitoring and higher quality financial reporting. In Table 3, we present the results examining the effect of audit committee connectedness on firm’s financial reporting quality, using the accruals measures discussed in section 3.4. We control for audit committee accounting expertise (Acct Expertise), as prior studies show that it is the SEC’s narrowly-defined accounting expertise, rather than the broader financial expertise of members, that drives the committee’s monitoring effectiveness (DeFond et al., 2005; Krishnan and Visvanathan, 2008). To do so, we examine the prior work experience and qualifications of each independent, non-co-opted audit committee member 18

using BoardEx. We classify as accounting experts those individuals who currently hold or have held the position of CFO, CPA/CFA, controller, comptroller, treasurer, or any other position that is financial reporting related. We also control for audit committee size and independence to capture the quality of the committee. Finally, we control for industry, firm, and year fixed effects. Table 3 presents our results. In Columns 1-3, we use OLS to examine the relation between audit committee connectedness and the three accruals measures. All three regressions show significant coefficient estimates on Audit Committee (AC) connectedness, suggesting that AC members’ connectedness is associated with lower earnings management (i.e., higher financial reporting quality). Economically, a one standard deviation increase in AC Connectedness is associated with 2% - 20% increase in financial reporting quality, depending on the accruals measure used. These results suggest that director connectedness facilitates their understanding of the firm’s operations and improves monitoring of the CEO. Alternatively, our findings are consistent with the argument that well-connected directors have better career opportunities, which reduces the influence that the CEO has over these directors. We also find that accounting expertise (Acct Expertise) is positively associated with financial reporting quality, albeit marginally. Audit committee size is positively associated with audit quality, consistent with the notion that larger audit committees have more or better “eyes” with which to monitor managers. Audit committee independence (AC Independence) is unrelated to audit quality, which is not surprising given that independence is less of a concern issue postSOX due to the 100% independence requirement for audit committee membership. In Columns 4-6 of Table 3, we use an instrumental variable approach to mitigate endogeneity concerns that our results may be driven by unobserved omitted variables. Studies such as Useem and Karabel (1986) and Domhoff (2002) find that individuals with degrees from elite institutions are more likely to ascend to the upper level of social hierarchies. Consequently, we use the proportion of audit committee members with an undergraduate degree from an elite college or a prominent MBA degree as our

19

instrumental variable.18 For all three financial reporting quality measures, audit committee connectedness exhibits negative and significant coefficients. Economically, a one standard deviation increase in audit committee connectedness is associated with 2.8% lower earnings management, as measured by Hribar. In order to further alleviate self-selection concerns, we restrict the sample to include only exogenous shocks to audit committee memberships, which are in the form of member deaths. Board member deaths are identified by Audit Analytics Director and Officer Changes database, which indicates the reason for director committee membership and their turnover details. In order to quantify the effect of this shock, we calculate the change (year t+1 minus year t-1) in each variable relative to the year of the director death (t = 0). Results are provided in Columns 7-9 of Table 3. For the sample of director deaths, the coefficient on the change in audit committee connectedness (ΔAC Connectedness) is negative and significant, indicating that increases (decreases) in AC connectedness are associated with improvements (declines) in audit quality. 4.3 Compensation Committee Connectedness and CEO Pay We next explore the effect of director connectedness on CEO compensation, focusing on total compensation (Total Pay). Better connected committee members should have an enhanced understanding of CEO ability due to their access to external information and be in a better position to monitor managers. More effective monitoring could substitute for CEO compensation and result in less need for high pay. 19 We control for compensation committee size and independence as well as CEO Power, CEO connectedness, firm performance (ROA), leverage, capital expenditures (CAPX), market-to-book, and institutional ownership. In addition, we control for the proportion of compensation committee members

18

Following Useem and Karabel’s (1986) classification of top-ranked colleges and prominent MBA programs, we collect the number of directors with degrees from such schools for each firm. Since the rankings may not be unanimous, we also used U.S. News & World Report for college and MBA rankings and Business Week for MBA rankings. Not surprisingly, the rankings from different sources generate a very high correlation and yield virtually identical results regardless of the source of ranking. In the first-stage of our 2SLS regression, we find that the proportion of audit committee directors that have prominent degrees is positively and significantly related to the audit committee connectedness. The Cragg-Donald Wald statistic for weak-instrument test yields a value of 64.03, which exceeds the threshold as suggested by Stock and Yogo (2005). 19 In untabulated results, we confirm the findings in Coles et al. (2015) that CEO turnover is higher in firms with better connected boards.

20

that have historical overlapping experience with the incumbent CEO (CEO/CC Overlap Ratio). Finally, we control for industry, firm, and year fixed effects. We convert Total Pay to log form, although we find similar results using untransformed measures. Results are reported in Table 4. In Column 1, we examine the relation between total compensation and CC Connectedness, which measures the connectedness for all independent non-co-opted compensation committee (CC) members. Previous studies suggest that CEO compensation is positively related to CEO power (Morse, Nanda, and Seru, 2011) and the strength of CEO-board connections (Liu, 2010), and negatively related to the degree to which the board is socially independent of the CEO (Coles et al., 2014; Hwang and Kim, 2009). Our results suggest that a better connected compensation committee is associated with lower level of total compensation, even after controlling for factors such as committee size, independence, member busyness, CEO power, and historical CEO-member mutual experience. Economically, a one standard deviation increase in CC Connectedness is associated with 1% decrease in total pay. These results suggest that compensation committee connectedness enables the board to better monitor the CEO. We use two strategies to address endogeneity concerns. In column 2, we use the fraction of compensation committee members that obtain degree from a prominent undergraduate or MBA program to instrument for CC Connectedness. The results show that the predicted CC Connectedness is associated with lower total pay for the CEO, consistent with the OLS results in Column 1. In Column 3 we use a change specification to facilitate causal interpretation. We identify 72 cases of unexpected compensation committee member deaths. After controlling for other determinants of pay, in Column 3 we find that increases in compensation committee connectedness due to this exogenous shock is associated with decreases in CEO total compensation. We interpret the finding as consistent with the notion that better connected compensation committee members have better knowledge about the CEO ability and a greater concern of overpaying their executives. In the end, we find causal evidence that compensation committee member connectedness is negatively associated with the probability of overpaying their CEO, suggesting that director connectedness is associated with stronger corporate governance. Economically for instance, column 2 results indicates that a one standard deviation increase in change of CC Connectedness is 21

associated with a 2.9% decrease in total pay. Overall, our results are consistent with the notion that a compensation committee with the ability to access greater external information sources is better able to assess the project selections of the CEO since they are better able to judge on CEO ability. 20 It is also consistent with the notion that better connected compensation committee members are more concerned about their reputation and thus have a lower tendency to overpay the CEO. 4.4 Board Connectedness and Seasoned Equity Offers (SEOs) To further explore the effect of well-connected directors on the strategic direction and information environment of the firm, we examine the effect of better connected boards on the cost of equity financing, as measured using a sample of seasoned equity offers (SEOs) occurring over the period January 1, 2001 to December 31, 2010. SEO characteristics are obtained from Securities Data Company’s (SDC) Global New Issues database. Similar to previous work, we exclude unit and rights offerings, non-U.S. issuers, closedend funds, REITs, and utilities. Offer firms must have available CRSP returns for at least 30 days prior to the issue date and offer prices greater than $3.00. We further exclude offers with stock splits in the 11-day window surrounding the offer date and outliers with underpricing that exceeds the absolute value of 60%. Lastly, we delete all SEOs that occur within a year of the firm’s last seasoned offer. The SEO offer discount, or SEO Underpricing, is defined as negative one times the return from the previous day’s closing price to the offer price. 21 The measure is structured such that underpricing is positive when the offer price is less than the previous day’s closing price. Further, we examine SEO announcement returns (SEO CAR), which are defined as the cumulative abnormal return during the (-1, +1) window around the announcement date.

20

An interesting angle on CEO compensation is whether the board differentiates between luck and skill. Garvey and Milbourn (2006) document that CEO pay is positively related to both luck and skill. In untabulated results, we also find that both luck and skill are positively related to CEO total compensation. Further, we find that better compensation committee connectedness is associated with larger (smaller) delta when the skill (luck) is high, consistent with the notion that well-connected committees are better able to differentiate CEO ability. 21 Due to the fact that many SEOs take place following the close of trading (Safieddine and Wilhelm, 1996), we apply a volume-based offer date correction to our sample. Specifically, we adjust the offer date to the following trading day if trading volume on the day following the SEO is more than twice that on the offer date and more than twice that of the average previous 250 day trading volume.

22

We control for a number of known determinants of underpricing as in Corwin (2003) and Intintoli and Kahle (2010). More specifically, we include firm size, stock return volatility, number of managing underwriters, the maximum underwriter ranking, the number of analysts, shares offered to float, an indicator variable (Increase Dummy) set to one if the offer price is higher than the original filing price, offer price, price clustering, an indicator variable set to one for stocks that trade on the NYSE, the ratio of secondary shares offered to total shares offered (Percent Secondary Shares), and cumulative abnormal returns over the period t = -5 to t = -1 relative to the offer date (CAR). Column 1 of Table 5 reports our results. After controlling for the standard determinants of underpricing, we find a negative and significant coefficient on Board Connectedness, supporting the idea that better connected boards improve the information environment and reduce underpricing. Our results are economically significant. Average underpricing for our sample is 3.363%, so the negative coefficient on our connectedness measure (-0.402) implies that a one standard deviation increase in board connectedness results in an 11.9% decrease in underpricing. To address endogeneity concerns, we employ the 2SLS-IV methodology used in earlier tables. Specifically, in the first stage we regress board connectedness on the proportion of independent non-co-opted directors that have degrees from prominent undergraduate universities or MBA programs. We find that the instrument variable is positive and highly significant; a weak-instrument test reveals that the instrument has very strong power on the endogenous variable. Second stage results, shown in Column 2, are consistent with our OLS findings, i.e., better connected boards are associated with lower levels of underpricing. In Column 3 of Table 5, we examine the effect of board connectedness on SEO announcement abnormal returns. If better connected boards aid in the information environment of the firm, then uncertainty regarding the rationale behind the seasoned offer (e.g., overvaluation, market timing) should be alleviated and market reactions to the announcement of the SEO should be less severe. Consistent with this prediction, the positive and significant coefficient on Board Connectedness provides evidence that the market reaction

23

is less negative when firms with better connected boards announce a new equity offering. 22 In Column 4 we report results from our 2SLS-IV specification, which are consistent with our OLS estimates. To further examine the relation between board connectedness and the cost of raising equity, we next examine the decision to use an accelerated SEO, in which traditional marketing efforts are forgone, over the fully marketed offer. Marketing efforts and underwriter certification help to mitigate informational asymmetries between market participants and managers (Huang and Zhang, 2011; Smith, 1986; Booth and Smith, 1986). For this reason, firms with a better informational environment may choose to skip the costly marketing process in favor of an accelerated offer. We follow Bortolotti, Megginson, and Smart (2008) and identify accelerated offers (SEO Acceleration) as any offer that contains a SDC designation of accelerated trade (AT), accelerated bookbuilt (ABO), bought deal (BD) or block trade (BT). Further, any offer with less than three trading days between filing and issuance date is identified as accelerated. 23 Overall, 26% of our sample consists of accelerated offers. In Column 5 of Table 5, we report logit regressions using the accelerated offer indicator as the dependent variable. Independent variables include our determinants of underpricing. Results suggest that firms with better connected boards are more likely to implement accelerated offers. Economically, a one standard deviation increase in board connectedness is associated with 17.7% greater odds of choosing an accelerated SEO over the fully marketed offer. We use an instrumental variable approach similar to that in our underpricing regression to address endogeneity concerns and find similar inferences. 4.5 Board Connectedness and Cost of Debt We next explore Dealscan data to examine the effect of board connectedness on the cost of debt. Specifically, we examine whether a change in board/AC connectedness influences the firm’s future cost of borrowing from banks. We calculate the cost of bank loans as the log of the loan spread for each loan

22

We exclude accelerated offers from the SEO announcement regressions because the period relative to the SEO announcement often overlaps with the issuance period. However, we find similar results when including the accelerated sample, which is untabluated for brevity. 23 This method is similar to what is used in Gustafson (2014). The three day cutoff also relates to Gao and Ritter’s (2010) observation that fully marketed offers take anywhere from 3 to 150 calendar days from filing to offering.

24

facility. We use all-in-spread drawn as the measure of the loan spread, which is the interest spread over LIBOR plus associated loan origination fees (Huston et al., 2014). We then average the loan spread at firm level. We adopt a change regression specification where all variables are year t+1 value minus year t-1 value, with year t being the year of director death incident. We present the results in Table 6. The results in Column 1 show that an increase in board connectedness is associated with lower spreads on bank loans, whereas audit committee connectedness is negative but statistically unrelated to spreads, shown in Column 3. 24 In order to address endogeneity concerns, we rely on an exogenous shock to the board based on the deaths of non-co-opted directors, identifying 322 cases at the board level and 108 cases at the audit committee level. Results are consistent with our entire sample findings, with the exception that an increase in AC connectedness is also significantly associated with lower spreads on bank loans, shown in Columns 2 and 4. 4.6 Board Connectedness and Payout Policy Our next test centers on the notion that if better connected directors are effective monitors and strive to alleviate agency problems between managers and outside investors, then better connected boards should choose higher payouts, ceteris paribus. This idea follows from the free cash flow hypothesis, which suggests that since managers cannot credibly convince investors that they will not invest excess cash in negative NPV projects, dividend payout can mitigate this problem (Easterbrook, 1984; Jensen, 1986; Lang and Litzenberger, 1989). 25 Following Grinstein and Michaely (2005), we use the sum of annual dividends and shares repurchased scaled by the book value of assets as the dependent variable for Payout. We control for other firm characteristics that are related to the payout decision, i.e., firm size, market beta, cash holding

24

Our findings are robust to utilizing a connectedness measure consisting of all board members, which is consistent with the conjecture that social connections of insiders and co-opted outsiders may also facilitate information transparency similar to that of external, non-co-opted members. 25 The alternative signaling hypothesis suggests that managers use dividends to signal their inside information about the firm, implying that firms with better connected boards do not need to use dividends to signal since information asymmetry is already low in such firms. Empirical evidence indicates that increasing (decreasing) dividends is associated with a positive (negative) stock price reaction, consistent with both signaling and excess-cash hypotheses.

25

(Cash), cashflow, performance (ROA), in addition to our common controls such as board size, board independence, CEO power, CEO Connectedness, CEO Board Overlap Ratio, volatility and institutional ownership. In addition, we control for industry, firm, and year fixed effects. Table 7 presents our results using a Tobit model due to the censored structure of our dependent variable. Column 1 shows that Board Connectedness is positively associated with payout, consistent with well-connected boards having a better ability to prevent the CEO from engaging in actions that are related to empire building and free cash flow concerns (Jensen, 1986). Economically, a one standard deviation increase in Board Connectedness is related to a 10.2% increase in dividend payout. In Column 2, we examine whether better connected boards are more sensitive to the firm’s growth opportunities in their payout decision. We use market-to-book as the proxy for growth potential and a dummy variable indicating if the firm’s market-to-book is within the top quartile of the sample (High Market-to-book). Board connectedness is again positively associated with payout. However, the interaction term is negative and significant, suggesting that better connected boards are sensitive to the firm’s growth opportunity. That is, when growth potential is high, boards are more likely to encourage a lower payout policy, consistent with the notion that social connectedness assists directors in their information gathering and consequently helps to match the firm’s payout policy with its growth needs. 4.7 Industry Competition Shock, Board Connectedness, and Firm Strategy Our final test centers on an exogenous shock to industry competition to establish identification for the effect of board connectedness on firm strategic reactions. We suggest that a sudden change in industry competition has significant implications for the firms’ strategic reactions. Heightened competition is expected to reduce a firm’s market power and profitability (Tirole, 2010), reduce growth opportunities and induce managers’ to modify their investment decisions (Grenadier, 2002) and/or innovation efforts (Aghion et al., 2005). In addition, greater competition increases the riskiness of cash flows, suggesting that managers may also consider cutting investments or decreasing leverage (Raith, 2003; Gaspar and Massa, 2006; Irvine and Pontiff, 2009). Hou and Robinson (2006) and Valta (2012) suggest that increased competition also lowers lender’s willingness to provide financing. If a firm lacks the means to obtain external resources to 26

defend against competition, it may be forced to cut investment below the optimal level, thereby compromising the long-term prospects of the firm. However, board connectedness may mitigate these issues by aiding in the financing effort and providing strategic advice to the managers during periods of escalating competition. In sum, we posit that a better connected board will be better able to shield the firm from the negative impact of a competition shock. In order to capture shocks to industry competition, we rely on sudden decreases in import tariff rates. Following Fresard (2010), we use the Census Bureau imports database to identify industries that experience drastic import tariff drops, which causes the intensification of competition among peer firms (Feenstra, 1996; Feenstra and Romalis, 2014). 26 We classify 4-digit SIC (2011 to 3999) industries as experiencing increased competition if year-to-year tariff decreases are over 2.5 times that of the average of tariff changes. During our period, 14.6% of industries (among SIC 2011-3999) experience competition shocks due to a sudden drop in imports tariff. Our treatment group consists of firms in industries that experience shocks. We next identify a matched sample that consists of firms that operate in industries that do not face a competition shock. More specifically, we create our control sample using propensity-score matching based on the following variables: firm size, CEO tenure, CEO connectedness, CEO power, growth opportunities (market-to-book ratio), ROA, institutional ownership, cash flow, cash holdings, and leverage in the year prior to the shock. We require that the matched firms be in a different industry than the “treated” firm. The matched sample yields 211 treated firms and 211 matched firms that do not experience a competition shock. We use the following specification to test our hypothesis:

∆Investment = α + β1 * Competition Shock + β2 * Board Connectedness + β3 * Board

26

Products imported to the U.S. are coded based on the Harmonized System (HS) established by the World Customs Organization (WCO) with each product assigned a ten-digit HS code. Feenstra (1996) and Pierce and Schott (2009) develop concordance tables that map each HS product code into four-digit SIC codes. Because HS codes are only based on product characteristics, and SIC codes incorporate the method of production, HS codes cannot be directly matched to SIC codes. Consequently, it is possible that a given HS category is matched to several four-digit SIC codes. In practice however, we find no cases in which a specific product (HS code) is assigned to multiple (four-digit) SIC codes in the industries in our sample.

27

Connectedness * Competition Shock + βX * Controlsi,t + ξi,t where ∆Investment is the change in capital expenditures from the year prior to the shock to the year after the shock. Competition Shock is a dummy variable equal to 1 if the firm experiences a competition shock. The interaction term Board Connectedness*Competition Shock examines the incremental effect of board connectedness on the firm’s investment changes, over the effect of board connectedness on firms that do not experience an increase in competition. We expect that the interaction term is positive, suggesting that board connections facilitate better handling of the competition shock. Table 8 shows the results. In Column 1, we find that the effect of Competition Shock is negative and significant while that of Board Connectedness on investment is positive and insignificant. 27 Moreover, we find that the interaction term is positive and significant, indicating that the effect of board connectedness on firms experiencing increased competition is larger than that for firms that do not experience a competition shock. These results are consistent with the notion that a better connected board is better able to shield the firm from the negative impact of a competition shock. In Column 2 and 3 we focus on the effect of board connectedness on future performance after the firm receives a competition shock. More specifically, we explore the notion that a well-connected board is better able to advise managers in their handling of the industry shock, which should result in better long-term performance. We rely on the sample of firms with competition shock and adopt change specification where the dependent variable is the change of ROA and Tobin’s Q from the year before the shock to the year after the shock. The independent variables are differenced similarly. Our focus is on the interaction term of Board Connectedness and change of investment of the firm (Investment Change). The notion is that if a better connected board provides improved firm advising, then investment decisions should be more closely linked to firm performance. Results in both columns show that, conditional on receiving an industry shock, firms with better connected boards are associated with higher firm value and future performance, indicating that board connectedness play an important advising role especially when the need for such advisement is high.

27

In robustness, we measure investment as (1) capital expenditures plus acquisitions and (2) R&D, and obtain similar results.

28

5. Conclusion We examine the effect of social connectedness on the ability of independent, non-co-opted directors to monitor and advise the firm. In order for boards to effectively monitor, directors must be both independent of management and feel insulated from potential repercussions and career concerns. Consequently, we begin by examining the effect of corporate fraud on directors’ career paths. We find that independent, well-connected audit committee members are less likely to experience turnover and more likely to be appointed on subsequent board seats following instances of fraud. We next examine the channels by which director connectedness may improve monitoring in the firm. We find that audit committee connectedness has a positive effect on the quality of financial reporting within the firm. Further, better connected compensation committees are less likely to overpay the CEO. Finally, we examine the effect of well-connected directors on the strategic direction and information environment of the firm. If well-connected directors can use their knowledge and connections to reduce information asymmetry, then it should be reflected in the firm’s corporate policies. Consistent with this channel, we show that when firms with highly connected boards issue equity, they are able to offer equity at a lower discount and are more likely to use accelerated offers. Increases in connectedness also decrease the spread on bank loans. We further show that the payout ratio of better-connected boards is higher, consistent with better connected boards mitigating free cash flow problems. Finally, we find that better connected boards are able to shield the firm from the negative impact of a competition shock. Overall, well-connected boards appear to be more effective in both monitoring the firm and promoting shareholders’ interests.

29

References Adams, R. B., and D. Ferreira, 2007. A theory of friendly boards. Journal of Finance 62, 217-250. Aghion, P., R. Burgess, S. Redding, and F. Zilibotti, 2005. Entry liberalization and inequality in industrial performance. Journal of the European Economic Association 3, 291-302. Alam, Z. S., M. A. Chen, C. S. Ciccotello, and H. E. Ryan, Jr., 2014. Does the location of directors matter? Information acquisition and board decisions. Journal of Financial and Quantitative Analysis 49, 131-164. Armstrong, C., and D. Larcker, 2009. Discussion of the impact of the options backdating scandal on shareholders and taxes and the backdating of stock option exercise date. Journal of Accounting and Economics 47, 50-58. Ashbaugh, H. A., R. LaFond, and B. W. Mayhew, 2003. Do nonaudit services compromise auditor independence? Further evidence. The Accounting Review 78, 611-639. Bédard, J., S. M. Chtourou, and L. Courteau, 2004. The effect of audit committee expertise, independence, and activity on aggressive earnings management. Auditing: A Journal of Practice & Theory 23, 15-37. Belliveau, M. A., C. A. O’Reilly, and J. B. Wade, 1996. Social capital at the top: Effects of social similarity and status on CEO compensation. Academy of Management Journal 39, 1568-1593. Bhagat, S., and B. S. Black, 1999. The uncertain relationship between board composition and firm performance. Business Lawyer 54, 921-963. Bizjak, J. M., M. L. Lemmon, and R. J. Whitby, 2009. Option backdating and board interlocks. Review of Financial Studies 22, 4821-4847. Bonacich, P., 1972. Factoring and weighting approaches to status scores and clique identification. Journal of Mathematical Sociology 2, 113-120. Booth, J. R., and R. L. Smith, II, 1986. Capital raising, underwriting and the certification hypothesis. Journal of Financial Economics 15, 261-281. Bortolotti, B., W. Megginson, and S. B. Smart, 2008. The rise of accelerated seasoned equity underwritings. Journal of Applied Corporate Finance 20, 35-57. Carcello, J. V., T. L. Neal, Z.-V. Palmrose, and S. Scholz, 2011. CEO involvement in selecting board members, audit committee effectiveness, and restatements. Contemporary Accounting Research 28, 396430. Cashman, G. D., S. L. Gillan, and R. J. Whitby, 2010. Expertise, connections, and the labor market for corporate directors. Working paper. Chidambaran, N. K., S. Kedia, and N. R. Prabhala, 2012. CEO-director connections and corporate fraud. Working paper. Cohen, L., A. Frazzini, and C. Malloy, 2012. Hiring cheerleaders: Board appointments of “independent” directors. Management Science 58, 1039-1058. Coles, J. L., N. D. Daniel, and L. Naveen, 2014. Co-opted boards. Review of Financial Studies 27, 17511796. Coles, J. L., A. Y. Wang, and C. Zhu, 2015, Director Network and CEO Turnover, University of Utah working paper. Core, J. E., R. W. Holthausen, and D. F. Larcker, 1999. Corporate governance, chief executive officer compensation, and firm performance. Journal of Financial Economics 51, 371-406.

30

Corwin, S. A., 2003. The determinants of underpricing for seasoned equity offers. Journal of Finance 58, 2249-2279. Dass, N., O. Kini, V. Nanda, B. Onal, and J. Wang, 2014. Board expertise: Do directors from related industries help bridge the information gap? Review of Financial Studies 27, 1533-1592. Dechow, P. M., and I. D. Dichev, 2002. The quality of accruals and earnings: The role of accrual estimation errors. The Accounting Review 77, 35-59. Dechow, P. M., W. Ge, C. R. Larson, and R. G. Sloan, 2011. Predicting material accounting misstatements. Contemporary Accounting Research 28, 17-82. DeFond, M. L., R. N. Hann, and X. Hu, 2005. Does the market value financial expertise on audit committees of boards of directors? Journal of Accounting Research 43, 153-193. Denis, D. K., J. Lee, and K. J. Lee, 2014. Director connections, board appointments, and director reputation. Working paper. Dhaliwal, D., V. Naiker, and F. Navissi, 2010. The association between accruals quality and the characteristics of accounting experts and mix of expertise on audit committees. Contemporary Accounting Research 27, 787-827. Domhoff, G. W., 2002. The power elite, public policy, and public opinion. Navigating public opinion: Polls, policy, and the Future of American democracy. Oxford University Press, 124-137. Easterbrook, F. H., 1984. Two agency-cost explanations of dividends. American Economic Review 74, 650659. Engelberg, J., P. Gao, and C. A. Parsons, 2012. Friends with money. Journal of Financial Economics 103, 169-188. Farber, D. B., 2005. Restoring trust after fraud: Does corporate governance matter? The Accounting Review 80, 539-561. Fama, E. F., and M. C. Jensen, 1983. Separation of ownership and control. Journal of Law and Economics 26, 301-325. Fama, E. F., 1980. Agency problems and the theory of the firm. Journal of Political Economy 88, 288-307. Feenstra, R. C., 1996. Trade and uneven growth. Journal of Development Economics 49, 229-256. Feenstra, R. C., and J. Romalis, 2014. International prices and endogenous quality. Quarterly Journal of Economics 129, 477-527. Fich, E. M., and A. Shivdasani, 2006. Are busy boards effective monitors? Journal of Finance 61, 689-724. Fracassi, C., and G. Tate, 2012. External networking and internal firm governance. Journal of Finance 67, 153-194. Freeman, L. C., 1977. A set of measures of centrality based on betweenness. Sociometry 40, 35-41. Fresard, L., 2010. Financial strength and product market behavior: The real effects of corporate cash holdings. Journal of Finance 65, 1097-1122. Gao, X., and J. R. Ritter, 2010. The marketing of seasoned equity offerings. Journal of Financial Economics 97, 33-52. Garvey, G. T., and T. T. Milbourn, 2006. Asymmetric benchmarking in compensation: Executives are rewarded for good luck but not penalized for bad. Journal of Financial Economics 82, 197-225. Gaspar, J.-M., and M. Massa, 2006. Idiosyncratic volatility and product market competition. Journal of Business 79, 3125-3152. 31

Grenadier, S. R., 2002. Option exercise games: An application to the equilibrium investment strategies of firms. Review of Financial Studies 15, 691-721. Grinstein, Y., and R. Michaely, 2005. Institutional holdings and payout policy. Journal of Finance 60, 1389-1426. Gustafson, M., 2014. The consequences of accelerated seasoned equity offerings. Working paper. Haunschild, P. R., and C. M. Beckman, 1998. When do interlocks matter? Alternate sources of information and interlock influence. Administrative Science Quarterly 43, 815-844. He, J., and Z. Huang, 2011. Board informal hierarchy and firm financial performance: Exploring a tacit structure guiding boardroom interactions. Academy of Management Journal 54, 1119-1139. Hermalin, B. E., and M. S. Weisbach, 1998. Endogenously chosen boards of directors and their monitoring of the CEO. American Economic Review 88, 96-118. Hoitash, R., and U. Hoitash, 2009. The role of audit committees in managing relationships with external auditors after SOX: Evidence from the USA. Managerial Auditing Journal 24, 368-397. Hou, K., and D. T. Robinson, 2006. Industry concentration and average stock returns. Journal of Finance 61, 1927-1956. Hribar, P., and D. C. Nichols, 2007. The use of unsigned earnings quality measures in tests of earnings management. Journal of Accounting Research 45, 1017-1053. Huang, R., and D. Zhang, 2011. Managing underwriters and the marketing of seasoned equity offerings. Journal of Financial and Quantitative Analysis 46, 141-170. Hwang, B.-H., and S. Kim, 2009. It pays to have friends. Journal of Financial Economics 93, 138-158. Intintoli, V. J., and K. M. Kahle, 2010. Seasoned equity offers: The effect of insider ownership and float. Financial Management 39, 1575-1599. Irvine, P. J., and J. Pontiff, 2009. Idiosyncratic return volatility, cash flows, and product market competition. Review of Financial Studies 22, 1149-1177. Jensen, M. C., 1986. Agency cost of free cash flow, corporate finance, and takeovers. American Economic Review 76, 323-329. Khanna, V., E. H. Kim, and Y. Lu, 2015. CEO connectedness and corporate frauds. Forthcoming, Journal of Finance. Krishnan, J., 2005. Audit committee quality and internal control: An empirical analysis. The Accounting Review 80, 649-675. Krishnan, J., Y. Wen, and W. Zhao, 2011. Legal expertise on corporate audit committees and financial reporting quality. The Accounting Review 86, 2099-2130. Krishnan, G. V., and G. Visvanathan, 2008. Does the SOX definition of an accounting expert matter? The association between audit committee directors' accounting expertise and accounting conservatism. Contemporary Accounting Research 25, 827-858. Kothari, S. P., A. J. Leone, and C. E. Wasley, 2005. Performance matched discretionary accrual measures. Journal of Accounting and Economics 39, 163-197. Lang, L. H.P., and R. H. Litzenberger, 1989. Dividend announcements: Cash flow signaling vs. free cash flow hypothesis? Journal of Financial Economics 24, 181-191. Larcker, D. F., E. C. So, and C. C.Y. Wang, 2013. Boardroom centrality and firm performance. Journal of Accounting and Economics 55, 225-250. 32

Liu, Y., 2010. The impact of networks on CEO turnover, appointment, and compensation. working paper. Liu, Y., 2014. Outside options and CEO turnover: The network effect. Journal of Corporate Finance 28, 201-217. Loderer, C., and U. Peyer, 2002. Board overlap, seat accumulation and share prices. European Financial Management 8, 165-192. Mizruchi, M. S., 1996. What do interlocks do? An analysis, critique, and assessment of research on interlocking directorates. Annual Review of Sociology 22, 271-298. Mol, M. J., 2001. Creating wealth through working with others: Interorganizational relationships. Academy of Management Executive 15, 150-152. Morse, A., V. Nanda, and A. Seru, 2011. Are incentive contracts rigged by powerful CEOs? Journal of Finance 66, 1779-1821. Newman, M. E. J., 2003. The structure and function of complex networks. SIAM Review 45, 167-256. Nicholson, G. J., M. Alexander, and G. C. Kiel, 2004. Defining the social capital of the board of directors: An exploratory study. Journal of Management and Organization 10, 54-72. Omer, T. C., M. K. Shelley, and F. M. Tice, 2014. Do director networks matter for financial reporting quality? Evidence from restatements. Working paper. Proctor, C. H., and C. P. Loomis, 1951. Analysis of sociometric data. In Research Methods in Social Relations, With Especial Reference to Prejudice, by Marie Jahoda, Morton Deutsch, and Stuart W. Cook. New York: Dryden 2, 561-585. Raith, M., 2003. Competition, risk, and managerial incentives. American Economic Review 93, 1425-1436. Renneboog, L., and Y. Zhao, 2011. Us knows us in the UK: On director networks and CEO compensation. Journal of Corporate Finance 17, 1132-1157. Sabidussi, G., 1966. The centrality index of a graph. Psychometrika 31, 581-603. Schoorman, F. D., M. H. Bazerman, and R. S. Atkin, 1981. Interlocking directorates: A strategy for reducing environmental uncertainty. Academy of Management Review 6, 243-251. Shivdasani, A., and D. Yermack, 1999. CEO involvement in the selection of new board members: An empirical analysis. Journal of Finance 54, 1829-1853. Smith, Jr., C. W., 1986. Investment banking and the capital acquisition process. Journal of Financial Economics 15, 3-29. Srinivasan, S., 2005. Consequences of financial reporting failure for outside directors: Evidence from accounting restatements and audit committee members. Journal of Accounting Research 43, 291-334. Tirole, J., 2010. The theory of corporate finance. Princeton University Press. Useem, M., and J. Karabel, 1986. Pathways to top corporate management. American Sociological Review 51, 184-200. Valta, P., 2012. Competition and the cost of debt. Journal of Financial Economics 105, 661-682. Wang, C., F. Xie, and M. Zhu, 2013. Industry expertise of independent directors and board monitoring. Journal of Financial and Quantitative Analysis, Forthcoming. Yermack, D., 2004. Remuneration, retention, and reputation incentives for outside directors. Journal of Finance 59, 2281-2308.

33

Panel C: Centrality Measures of the Directors in Panel B Director Degree Betweenness Closeness David 0.444 0.023 0.529 Tom 0.444 0.023 0.529 John 0.333 0.000 0.500 Mark 0.667 0.102 0.600 Kathy 0.333 0.000 0.500 Jack 0.556 0.231 0.643 Karen 0.556 0.231 0.643 Nancy 0.333 0.389 0.600 James 0.222 0.222 0.429 Laura 0.111 0.000 0.310

35

Eigenvector 0.732 0.732 0.594 1.000 0.594 0.827 0.827 0.407 0.100 0.023

Appendix B Director Centrality Summary Statistics This appendix presents the summary statistics of director level and firm level centrality measures. Degree and Betweenness are multiplied by 104 and Closeness and Eigenvector are multiplied by 102. Board-Level centrality measures are average of non-co-opted directors on the board. Panel A: Director-Level Centrality Measures Mean Median Std. Dev. Degree 3.031 2.446 3.665 Closeness 13.252 15.011 9.256 Betweenness 0.530 0.000 2.147 Eigenvector 1.256 0.117 4.753

Lower Quartile 0.529 2.785 0.000 0.020

Upper Quartile 5.215 21.662 0.190 0.585

Skew 3.228 -0.772 10.320 8.073

Panel B: Board-Level Centrality Measures Mean Median Std. Dev. Degree 2.537 2.122 4.295 Closeness 12.012 13.710 9.219 Betweenness 0.482 0.000 1.989 Eigenvector 1.105 0.106 4.886

Lower Quartile 0.422 2.211 0.000 0.015

Upper Quartile 4.661 19.337 0.167 0.522

Skew 3.833 -0.625 8.902 7.211

Panel C: Correlations between Board-Level Centrality Measures Degree Closeness Betweenness Degree 1.000 Closeness 0.882 1.000 Betweenness 0.812 0.775 1.000 Eigenvector 0.820 0.922 0.721

Eigenvector

1.000

Panel D: Factor Analysis Degree Closeness Betweenness Eigenvector Eigenvalue % variance explained

Factor 1 0.622 0.317 0.559 0.592

Factor 2 0.021 0.678 -0.211 -0.149

Factor 3 -0.172 0.020 -0.310 0.693

Factor 4 -0.602 0.119 0.239 0.211

2.410 63.1

0.805 21.1

0.465 12.2

0.138 3.6

36

Appendix C: Earnings Quality Measures Hribar: Unsigned Abnormal Accruals (Hribar and Nichols, 2007) We first estimate the following regression for each year and Fama-French industry: TACC = α + β1ΔREV + β2PPE + ξ where TACC is total accruals, defined as income before extraordinary items minus cash from operations divided by lagged total assets. ΔREV is the change in sales adjusted for the change in receivables, divided by lagged total assets. PPE is gross property, plant, and equipment, scaled by lagged total assets. We then calculate the abnormal accruals as the residual term in the regression, i.e., TACC – (α + β1ΔREV + β2PPE), and Hribar is the absolute value of the residual (abnormal accruals). DGLS: Industry-Adjusted Absolute Value of DD Residual (Dechow et al., 2011) We first regress working capital accruals (WC_ACC) on operating cash flows in the current year (OCFt), the preceding year (OCFt-1), and the following year (OCFt+1): WC_ACCi,t = α0,i + β1,i OCFi,t-1 + β2,i OCFi,t + β3,I OCFi,t+1 + νi,t where WC_ACC = ∆CA - ∆CL - ∆CASH + ∆STDEBT + ∆TAXES, where ∆CA = change in current assets between year t-1 and t, ∆CL = change in current liabilities between year t-1 and t, ∆CASH = change in cash and Short-Term Investments between year t-1 and t, ∆STDEBT = change in short-term debt between year t-1 and t, and ∆TAXES = change in taxes payable between year t-1 and t. All variables are scaled by average total assets and winsorized at the 1 percent and 99 percent levels. We estimate the equation by year for each of the two-digit SIC industry groups. DGLS is the absolute value of each firm’s residual less the average absolute value for the corresponding industry. AQ: Performance-Matched Discretional Accruals (Kothari et al., 2005) We estimate abnormal accruals for each firm-year and subtract the value from the discretionary-accruals of the performance-matched firm. The modified Jones model of abnormal accruals model is estimated crosssectionally each year using all firm-year observations in the same Fama-French industry. TAi,t = β0 + β1(1/ASSETSi,t-1) + β2(ΔSALESi,t – ΔARi,t) + β3PPEi,t + ξi,t where TA (total accruals) is the change in non-cash current assets minus the change in current liabilities excluding the current portion of long-term debt, minus depreciation and amortization, scaled by lagged total assets; ΔSALESi,t is change in sales; ΔARi,t is change in account receivable; and PPEi,t is gross property, plant and equipment, all scaled using lagged total assets, ASSETSi,t-1. We use total assets as the deflator to mitigate heteroscedasticity in the residuals. Residuals from the annual cross-sectional industry regression model in the modified Jones model are used to measure estimated abnormal accruals. We then match each firm-year observation with another firm from the same Fama-French industry and year with the closest return on assets in the current year, ROAi,t (net income divided by total assets). We define the AQ for firm i in year t as the abnormal accrual in year t minus the performance-matched abnormal accrual for year t. FScore: (Dechow et al. 2011) Based on the model in Dechow et al. (2011, 61). Fraudscore = -7.893 x rsst_acc + 2.518 x ch_rec + 1.191 x ch_inv + 1.979 x soft_assets + 0.171 x ch_cs - 0.932 x ch_roa + 1.029 x issue. Where rsst_acc = (ΔWC + ΔNCO + ΔFIN)⁄average total assets, where WC = [Current Assets – Cash and Shortterm Investments]–[Current Liabilities – Debt in Current Liabilities]; NCO = [Total Assets – Current Assets Investments and Advances] – [Total Liabilities – Current Liabilities – Long-term Debt]; FIN = [Short-term Investments + Long-term Investments] – [Long-term Debt + Debt in Current Liabilities + Preferred Stock]; ch_rec =change in accounts receivable scaled by average total assets; ch_inv = change in inventory scaled by average total assets; soft_assets = (total assets - net property, plant and equipment – cash and cash equivalents)/total assets; ch_cs = percentage change in cash sales, i.e., sales minus change in accounts receivable; ch_roa = current year ROA minus last year ROA; ROA is earnings before extraordinary items scaled by the average of this year and last year total assets; issue = 1 if the firm issued securities during the current year.

37

Appendix D: Additional Variables of Interest Variable Name AC Expertise

Accelerated Analyst Following Beta Board (AC/CC) Connectedness Board (AC/CC) Independence Board (AC/CC) Size

Definition Percentage of audit committee members that are considered "experts", where experts are those who currently hold or have held the position of CFO, CPA/CFA, controller, comptroller, treasurer or any other position that are financial reporting related; Dummy equal to one if SDC designation is accelerated trade (AT), accelerated bookbuilt (ABO), bought deal (BD) or block trade (BT). Any additional offer with less than three trading days between filing and issuance date are identified as accelerated; Log number of analysts following the firm during the year;

CAPX

Market beta from Compustat; Factor score based on independent, non-co-opted board (AC/ CC committee) members' degree, betweenness, closeness, and eigenvector centrality; The number of independent board (AC/ CC committee) members scaled by total number of board (AC/ CC committee) members; Log of the total number of board (AC/ CC committee) members; Percentage of independent, non-co-opted board (AC/ CC committee) members that serve on 4 or more public firms' boards; Capital expenditure divided by total assets;

CAR

The cumulative abnormal returns from day -5 to -1 relative to the offer date;

Cash

Cash scaled by book value of total assets;

Cashflow

Operating cash flow divided by book value of total assets;

CEO Connectedness

Director Age

Factor score based on CEOs' degree, betweenness, closeness, and eigenvector centrality; Calculated using factor analysis to extract the common underlying latent variable, using CEO tenure, CEO equity ownership, and CEO/Chairman duality; Percentage of independent, non-co-opted audit committee members that have overlapping experience with the CEO; Percentage of independent, non-co-opted board members that have overlapping experience with the CEO; Percentage of independent, non-co-opted compensation committee members that have overlapping experience with the CEO; Dummy equal to one if the decimal portion of the offer price falls on a $0.25 increment and is equal to zero otherwise; Industry competition intensification, identified using year-to-year tariff decreases over 2.5 times that of the average of tariff changes; Age of the director;

Director Turnover

A dummy variable equals 1 if the director is dismissed from the board;

Female Director

Dummy equal to one if the director is female;

Firm Size

Log of total assets; A dummy variable equals to 1 when the public becomes aware of any fraudulent activity committed by the firm, as evidenced by SEC/DOJ legal actions related to accounting malpractice, mergers & acquisitions, securities law, financial reporting, fraud, AAER, class action, stockholder suits, and initial public offerings (IPOs); Predicted probability of firm committing fraud based on Dechow et al. (2011). See Appendix C; A dummy equal to 1 if the firm's market-to-book ratio is in the top quartile of the sample; A dummy equal to 1 if the difference between the offer price and the original file price is positive; ROA minus industry-median ROA, where industry is identified using Fama-French 49 industry groups; The proportion of common equity owned by institutional investors;

Busy Board/AC/CC

CEO Power CEO/AC Overlap Ratio CEO/Board Overlap Ratio CEO/CC Overlap Ratio Cluster Dummy Competition Shock

Fraud

FScore High Market-to-book Increase Dummy Industry-adjusted Profit Institutional Ownership

38

Leverage

Loan Maturity

Book value of long-term debt divided by total assets; Estimated probability of litigation based on Model (3) in Kim and Skinner (2012, 302), calculated as eSUE/(1+eSUE), where SUEt = -7.883 + 0.566 x FPSt + 0.518 x Assetst-1 + 0.982 x Sales Growtht-1 + 0.379 x Returnt-1 - 0.108 x Returnskewnesst-1 + 25.635 x Returnstddevt-1 + 0.00007 x Turnovert-1. FPS = 1 if the firm is in the biotech (SIC codes 2833-2836 and 8731-8734), computer (3570-3577 and 7370-7374), electronics (36003674), or retail (5200-5961) industry, and 0 otherwise; Assets = log of total assets; Return = Market-adjusted 12-month stock return; Returnskewness = skewness of the firm’s 12-month return; Returnstddev = standard deviation of the firm’s 12-month returns. Sales Growth is current year sales less last year sales scaled by beginning of current year total assets; Turnover = daily trading volume accumulated over the fiscal year scaled by beginning of the year’s shares outstanding (in thousands); log of the loan maturity measured in days;

Loan Size

log of the loan amount measured in millions of dollars;

Loan Spread

log of interest spread over LIBOR plus associated loan origination fees; Log of market capitalization, where market capitalization is defined as price multiplied by the shares outstanding the day before the offer date;

Litigation Risk

Log(Market Cap.) Log(Number of Underwriters) Log(Price) Market-to-book Max Underwriter Rank MBA Number of Analyst Estimates NYSE Dummy

Log of the number of SEO underwriters; Log of the stock price on day prior to the offer; Market value of common equity plus book value of long-term debt divided by total assets; The maximum underwriter ranking of the lead underwriter(s) in the SEO;

Operating Cycle

Market to book assets; The number of analyst estimates in the period just prior to the SEO, from IBES Summary file; Dummy variable equal to one if the stock traded on the NYSE at the time of the SEO; The ratio of shares offered to float, defined as (total shares issued)/(shares outstanding the day prior to the offer)*(1 - (insider ownership)); Log of (days in account receivables + days sales in inventory);

Payout

Annual dividends plus shares repurchased scaled by book value of assets;

% Secondary Shares

Percentage of secondary shares offered;

Recent IPO Dummy

Dummy is set equal to 1 if the IPO took place in past year;

ROA

Earnings before extraordinary items divided by total assets;

SEO CAR

Underpricing

cumulative abnormal return for the window (-1, +1) around SEO announcement date; The standard deviation of daily stock returns over the 30 trading days ending 11 days prior to the issue; Tobin’s Q is market value of equity plus book value of long-term liabilities divided by book value of total assets; ∆Tobin’s Q is change in Tobin’s Q from the year prior to the industry competition shock to the post-shock year; Log of total CEO compensation. We use Execucomp’s TDC1 to measure total compensation and adjust this measure for the post-2006 period as detailed in Coles et al. (2014); Negative one times the return from the previous day's closing price to the offer price;

Volatility

The standard deviation of stock return calculated over prior 60 months;

σ(OCF)

Standard deviation of operating cash flow, measured over the previous 10 years;

σ(Sales)

Standard deviation of sales, measured over the previous 10 years; Capital expenditure change from the year prior to the industry competition shock to the post-shock year, scaled by assets;

Offer to Float

Stdev Returns ∆Tobin’s Q

Total Pay

∆Investment

39

Table 1 Descriptive Statistics This table shows the summary statistics of our variables of interest. Connectedness measures are reported based on their factor scores. Appendix B presents the summary statistics of director level and firm level centrality measures. See Appendix C and D for all variable definitions.

Dependent variables: Hribar DGLS AQ FScore Total Pay ($million) Payout SEO CAR Underpricing Board/Committee Board Connectedness AC Connectedness CC Connectedness CEO Power CEO Connectedness CEO/Board Overlap Busy Board Board Size Board Independence Acct Expertise Other Controls: Firm Size ($million) ROA Leverage Market-to-book Institutional Ownership Analyst Following

Mean

Median

Std. Dev.

Bottom Quartile

Top Quartile

0.142 0.005 -0.006 1.060 4.760 0.039 -0.56% 3.363

0.110 0.004 -0.006 0.972 2.857 0.016 -0.58% 2.458

0.123 0.051 0.103 0.571 5.612 0.060 2.31% 3.554

0.050 -0.021 -0.052 0.645 1.417 0.002 -1.72% 0.901

0.197 0.030 0.038 1.349 5.666 0.049 0.47% 4.765

0.017 -0.011 -0.015 0.040 -0.000 0.159 0.280 9.122 0.479 0.180

-0.283 -0.309 -0.297 0.114 -0.300 0.091 0.167 9.000 0.467 0.200

0.993 0.941 0.903 0.914 0.993 0.218 0.337 2.312 0.252 0.122

-0.612 -0.670 -0.618 -0.811 -0.458 0.000 0.000 7.000 0.250 0.000

0.286 0.324 0.223 0.359 0.073 0.250 0.500 11.000 0.700 0.250

605.468 0.050 0.009 1.478 0.603 6.000

4,500.755 0.143 0.279 3.681 0.900 16.000

6,409.460 1,502.673 18,795.351 0.092 0.093 0.099 0.176 0.154 0.164 3.033 2.290 3.139 0.697 0.783 0.287 11.939 10.000 8.039

40

Table 2 Fraud and Audit Committee Director Career Consequence This table examines the effect of director connectedness on director career consequences. For the propensity score matched sample, we match firms with detected fraud to firms without based on the following variables: firm size, CEO power, CEO tenure, ROA, volatility, institutional ownership, analyst following, audit committee size, audit committee independence, average audit committee director age, average director tenure, audit committee female ratio, Fscore, industry and year dummy. Panel A examines the likelihood of director turnover. Panel B shows results on the directors’ future board seats. In Panels A and B, High AC Connectedness is an indicator equal to one if director connectedness is in the top quartile of the sample. See Appendix C for the detailed definition of FScore and Appendix D for all other variable definitions. The tstatistics (in parentheses) are adjusted for heteroskedasticity using the Huber-White Sandwich estimator and are corrected for clustering of firm effects. ***, **, and * indicates significance at the 1%, 5%, and 10% level using two-tailed tests, respectively. Panel A: Logistic Regressions of Director Turnover (1) Full Sample Dependent Variable: Constant -10.122*** (-20.55) 1.111*** Fraud (5.09) -0.125** High AC Connectedness (-2.09) -0.629** Fraud * High AC Connectedness (-2.46) 0.079*** Director Age (18.82) -0.055 Female Director (-0.59) 0.104** Director Tenure (2.30) 0.029 CEO Power (0.78) 0.037 CEO Connectedness (1.22) 0.144*** Firm Size (5.49) 0.002* AC Size (1.83) 0.326*** AC Independence (3.09) 0.001 Analyst Following (0.55) 0.312 Volatility (0.98) 0.457*** Institutional Ownership (2.69) Industry and Year Dummy Yes Observations 44,260 Pseudo R2 0.203 41

(2) Matched Sample Director Turnover 4.506*** (2.69) 1.303*** (3.19) -0.197* (-1.89) -0.335** (-2.22) 0.075** (2.08) -0.511 (-1.48) 0.222 (1.23) 0.152 (1.25) 0.087 (0.22) 0.080* (1.77) 0.001 (1.42) 0.655 (1.23) 0.011 (1.03) 1.278 (0.50) 0.529 (1.44) Yes 592 0.230

(3) Fraud-Only Sample 2.725 (0.88) -0.237** (-2.41) 0.081*** (3.02) -0.182 (-1.50) 0.050 (0.32) 0.467 (1.30) 0.185 (1.07) 0.477*** (2.71) 0.001 (1.33) 0.882 (1.01) 0.018 (0.70) 3.121* (1.90) 2.199 (1.33) Yes 338 0.222

Panel B: OLS Regressions of Future Board Seats (1) Full Sample Dependent Variable: Constant 0.390*** (3.22) -0.024** Fraud (-2.46) 0.212*** High AC Connectedness (11.96) 0.071** Fraud * High AC Connectedness (2.00) 0.006*** Director Age (5.50) 0.041 Female Director (1.33) -0.018 Director Tenure (-1.46) -0.020** CEO Power (-2.50) 0.017* CEO Connectedness (1.79) 0.061*** Firm Size (9.11) 0.022 AC Size (0.88) 0.055* AC Independence (1.90) 0.001 Analyst Following (0.67) -0.136 Volatility (-1.30) 0.081** Institutional Ownership (2.22) Industry, Firm, and Year Dummy Yes 20,545 Observations 0.138 Adjusted R2

42

(2) Matched Sample Future Boards -0.233 (-0.78) -0.022** (-2.05) 0.277*** (2.70) 0.085** (2.01) 0.022** (2.00) 0.068 (1.22) 0.024 (0.40) -0.033 (-0.95) 0.001 (0.21) 0.023 (0.56) 0.001 (0.77) 0.029 (1.35) 0.002 (0.60) -0.557 (-1.29) 0.032 (0.59) Yes 288 0.396

(3) Fraud-Only Sample -0.491 (-0.77) 0.222** (2.55) 0.017*** (2.91) 0.015 (1.20) -0.019 (-0.25) -0.003 (-1.03) 0.042 (1.38) 0.061 (0.95) 0.002 (1.09) 0.379 (1.42) 0.004 (0.44) -1.833 (-1.60) 0.187 (0.56) Yes 189 0.421

Table 3 Audit Committee Connectedness and Audit Quality This table examines the effect of audit committee connectedness on earnings quality. See Appendix C and D for variable definitions. Columns 1-3 provide full sample OLS results and Columns 4-6 report full sample 2SLS results, where the proportion of audit committee members with an undergraduate or MBA degree from an elite university is used to instrument for AC connectedness. Columns 7-9 results are based on firms with audit committee member deaths. All the variables are year t+1 value minus year t-1 value, with year t being the death year. The t-statistics (in parentheses) are adjusted for heteroskedasticity using the Huber-White Sandwich estimator and are corrected for clustering of firm effects. ***, **, and * indicates significance at the 1%, 5%, and 10% level using two-tailed tests, respectively. (1) Dependent Variable: Constant AC Connectedness ∆AC Connectedness CEO Power CEO Connectedness CEO/AC Overlap Ratio Acct Expertise Firm size ROA AC Size AC Independence Busy AC Leverage Institutional Ownership Market-to-book Analyst Following

(3)

(4)

Hribar 0.150** (2.37) -0.003* (-1.90) -

(2) OLS DGLS -0.040* (-1.88) -0.001** (-2.06) -

Hribar

(5) 2SLS DGLS

AQ -0.134** (-2.55) -0.001** (-2.30) -

-0.003 (-0.77) 0.002 (1.40) 0.006 (0.48) -0.054** (-2.22) 0.006 (1.05) 0.066** (2.20) -0.004** (-2.10) -0.007 (-0.99) 0.006 (0.52) 0.021 (1.27) -0.034* (-1.81) -0.001 (-1.28) -0.001

-0.002 (-1.54) 0.001* (1.90) 0.005* (1.69) -0.011* (-1.82) 0.002 (0.89) 0.065** (2.15) -0.001 (-1.57) -0.005 (-1.23) 0.002 (0.56) 0.015*** (2.70) -0.002 (-0.27) -0.001 (-1.38) -0.001

-0.004 (-1.22) 0.001 (0.56) -0.002 (-0.25) -0.007 (-1.54) 0.004 (1.05) 0.070** (2.48) -0.002 (-1.36) -0.055 (-1.10) 0.006 (0.68) 0.029** (2.16) -0.031** (-2.26) -0.001 (-0.78) -0.001

AQ

0.233** (2.05) -0.013** (-2.29)

-0.062** (-2.15) -0.011** (-2.49)

-0.127** (-2.05) -0.018** (-2.10)

-

-

-

0.001 (0.52)

0.001 (1.25)

0.002 (0.89)

0.003 (1.02)

0.002 (1.29)

0.001 (0.99)

0.003 (0.61) -0.010 (-1.26) 0.004 (1.21) 0.061* (1.90) -0.001* (-1.83) -0.003 (-0.77) 0.003 (0.62) 0.014 (1.02) -0.025* (-1.80) -0.001 (-1.02) -0.001**

0.003 (0.89) -0.010* (-1.71) 0.003 (1.05) 0.047*** (2.66) -0.001 (-1.20) -0.002 (-0.69) 0.003 (0.59) 0.017** (2.42) -0.004 (-0.53) -0.001 (-1.21) -0.001

0.006 (0.78) -0.005 (-1.14) 0.005 (1.25) 0.050* (1.85) -0.002 (-1.46) -0.012 (-1.05) 0.004 (0.72) 0.025* (1.88) -0.031** (-2.03) -0.001 (-0.72) -0.001

43

(6)

(7) (8) (9) AC Member Deaths (OLS) ∆Hribar ∆DGLS ∆AQ 0.005 -0.025 -0.154 (0.04) (-0.46) (-0.75) -0.136** (-2.08) -0.028 (-0.19) 0.108 (1.21) 0.584 (1.11) -0.707 (-0.98) 0.234* (1.87) 0.772** (0.84) -0.196 (-1.44) -0.022 (-0.70) 0.379* (2.02) 1.071 (1.13) -0.578 (-0.62) -0.013 (-0.23) -0.024

-0.017* (-1.86) -0.005 (-0.06) 0.022 (0.38) 0.027 (0.20) -0.017 (-1.06) 0.050 (0.90) 0.172 (1.17) -0.006 (-0.10) -0.015 (-0.46) 0.020 (0.28) 0.164 (0.87) -0.214 (-0.42) -0.007 (-0.52) -0.005

-0.044** (-1.97) -0.052 (-0.16) 0.060 (0.36) 0.418 (1.02) -1.205 (-0.63) 0.127 (0.68) 0.370 (0.54) 0.003 (0.01) -0.037 (-0.50) 0.444 (1.53) 0.477 (0.84) -0.544 (-0.32) -0.042 (-0.61) -0.014

Litigation risk σ(Sales) σ(OCF) Operating Cycle Industry, Firm, Year dummy Observations Adjusted/Pseudo R2

(-1.62) -0.003 (-0.81) 0.053*** (2.93) 0.076* (1.66) 0.016*** (2.90) Yes 6,688 0.210

(-1.11) -0.001 (-0.39) 0.007* (1.75) 0.002 (1.15) 0.006*** (2.64) Yes 6,687 0.211

(-1.23) -0.003 (-1.22) 0.025 (1.46) 0.070* (1.81) 0.003 (1.32) Yes 6,300 0.220

(-2.02) -0.005 (-1.11) 0.015 (0.99) 0.062* (1.75) 0.003 (0.79) Yes 6,676 0.255

44

(-1.08) -0.001 (-0.50) 0.001 (1.03) 0.004 (1.33) 0.004** (2.43) Yes 6,675 0.102

(-0.82) 0.001 (0.44) 0.020 (1.10) 0.016 (1.30) 0.010** (2.14) Yes 6,288 0.088

(-1.34) -0.027 (-0.20) 0.236 (1.25) 0.760 (1.64) 0.149 (1.18) No 122 0.285

(-0.67) -0.087 (-1.19) 0.083 (1.24) 0.494 (1.33) 0.071* (1.94) No 122 0.283

(-0.30) -0.129 (-0.80) 0.377 (1.19) 0.116 (1.08) 0.048 (1.29) No 122 0.188

Table 4 Compensation Committee Connectedness and Compensation Incentives This table examines the effect of compensation committee connectedness on the log of CEO total pay. Column 1 provides full sample OLS results and Column 2 reports full sample 2SLS results, where the proportion of compensation committee members with an undergraduate or MBA degree from an elite university is used to instrument for CC connectedness. Column 3 is based on sample firms experiencing compensation committee member deaths. All variables are year t+1 value minus year t-1 value, where t is the year of committee member turnover. See Appendix D for variable definitions. The t-statistics (in parentheses) are adjusted for heteroskedasticity using the Huber-White Sandwich estimator and are corrected for clustering of firm effects. ***, **, and * indicates significance at the 1%, 5%, and 10% level using two-tailed tests, respectively.

Dependent Variable: Constant CC Connectedness ΔCC Connectedness CEO Power CEO Connectedness CEO/CC Overlap Ratio Firm Size CC Size CC Independence Busy CC ROA Leverage Market-to-book CAPX Institutional Ownership Industry, Firm, Year dummy Observations Adjusted/Pseudo R2

(1) (2) OLS 2SLS Total Pay 4.606*** 4.225*** (23.23) (19.88) -0.046** -0.109* (-2.30) (-1.93) 0.023 (1.41) 0.033*** (2.67) 0.034 (1.07) 0.408*** (10.91) 0.001 (0.83) -0.111*** (-2.71) 0.105*** (3.18) 0.264 (1.51) 0.071 (0.88) 0.002 (1.29) 0.058 (0.20) 0.202** (2.17) Yes 9,074 0.470

45

0.033 (1.13) 0.052** (2.25) 0.005 (1.33) 0.426*** (20.15) -0.001 (-0.22) -0.062 (-1.20) 0.267*** (3.52) 0.425** (2.01) -0.250** (-2.50) 0.002 (0.50) 0.652 (1.55) 0.240*** (2.82) Yes 9,074 0.657

(3) CC Member Deaths (OLS) ∆Total Pay 0.013 (1.01) -0.058** (-2.11) 0.204 (0.79) 0.355* (1.84) 0.576 (0.85) 0.967** (2.27) 0.053* (1.95) -0.177 (-1.32) 0.193*** (2.97) 0.442 (0.89) 0.821 (0.35) 0.041 (0.21) 0.713 (0.98) 0.209 (1.28) No 72 0.366

Table 5 Board Connectedness, SEO Underpricing, and Likelihood of Accelerated Offering This table presents results of the effect of board connectives on SEO underpricing and acceleration likelihood. See Appendix D for variable definitions. Columns 2 and 4 use the proportion of independent, non-co-opted board members with an undergraduate degree from an elite college or a prominent MBA degree to instrument for Board Connectedness. The t-statistics (in parentheses) are adjusted for heteroskedasticity using the Huber-White Sandwich estimator and are corrected for clustering of firm effects. ***, **, and * indicates significance at the 1%, 5%, and 10% level using two-tailed tests, respectively. (1) (2) (3) (4) (5) (6) OLS 2SLS OLS 2SLS Logit 2SLS Dependent Variable: SEO Underpricing SEO CAR SEO Acceleration Constant -0.985 -5.141 -0.025 -0.039 -5.867 1.126 (-0.22) (-0.53) (-0.62) (-0.40) (-1.46) (0.34) Board Connectedness -0.402** -1.222** 0.001** 0.002** 0.163** 0.741** (-2.46) (-2.26) (2.11) (2.44) (2.24) (2.14) CEO Power -0.248 -0.302 -0.001 -0.001 -0.229 -0.058 (-1.03) (-1.18) (-0.40) (-0.44) (-1.10) (-0.52) CEO Connectedness 0.059 0.327 -0.004** -0.003 0.070 0.216 (0.21) (0.52) (-2.31) (-0.67) (0.27) (1.11) CEO/Board Overlap Ratio -0.653 -0.819 -0.001 -0.002 1.026 0.623 (-0.51) (-0.64) (-0.12) (-0.16) (0.97) (1.14) Busy Board 0.834 1.290 0.010 0.012 -0.468 -0.655* (0.92) (1.09) (1.14) (1.20) (-0.70) (-1.76) Stdev Returns 44.443*** 41.883** 0.043 0.037 -9.356 -0.672 (3.16) (3.02) (0.37) (0.31) (-0.85) (-0.12) CAR 9.255*** 9.145*** 1.835 0.962 (2.85) (2.89) (0.73) (0.69) NYSE Dummy -0.989* -0.829 0.000 0.001 0.500 0.090 (-1.68) (-1.24) (0.10) (0.18) (1.12) (0.34) Log(Mkt_Cap) 0.822*** 1.245 -0.002 -0.001 0.059 -0.370 (2.99) (1.43) (-0.71) (-0.09) (0.23) (-1.58) Log(Price) -1.563*** -1.704*** 0.004 0.004 0.030 0.174 (-3.96) (-3.69) (1.14) (1.01) (0.09) (1.08) Percent Secondary Shares -0.537 -0.416 -0.003 -0.003 -0.424 -0.269 (-1.00) (-0.75) (-0.82) (-0.80) (-0.84) (-1.01) Cluster Dummy 1.081** 1.234** -0.001 -0.000 -0.267 -0.301 (2.28) (2.11) (-0.15) (-0.06) (-0.68) (-1.50) Offer to Float 6.448*** 7.390*** -0.017 -0.014 -2.728 -1.874** (3.08) (2.69) (-1.06) (-0.61) (-0.69) (-2.05) Increase Dummy -1.661*** -1.735*** 0.005 0.005 -0.571 -0.152 (-2.96) (-3.09) (1.21) (1.18) (-1.02) (-0.50) Max Underwriter Ranking -0.417 -0.539 0.003 0.003 0.443* 0.329** (-1.06) (-1.25) (1.59) (1.12) (1.85) (2.54) Recent IPO Dummy -0.932 -0.864 0.013 0.016 (-0.52) (-0.62) (0.86) (0.89) (-2.97) (-3.64) Number of Analyst Estimates -0.027 -0.052 0.001** 0.001 -0.004 0.021 (-0.67) (-0.78) (2.13) (0.89) (-0.14) (1.08) MBA -0.107 -0.205 -0.002 -0.002 0.877*** 0.518*** (-0.39) (-0.58) (-0.65) (-0.78) (3.06) (3.27) Log(Number of Underwriters) -0.565 -0.613 -0.002 -0.002 -0.375 -0.122 (-1.35) (-1.38) (-0.64) (-0.67) (-1.21) (-0.71) Industry and Year Dummy Yes Yes Yes Yes Yes Yes Observations 464 464 234 234 464 464 Adjusted/Pseudo R2/Log 0.296 0.277 0.143 0.054 0.127 -404.2

46

Table 6 Board Connectedness and Cost of Debt This table presents regressions of the effect of board connectedness on cost of debt. See Appendix D for variable definitions. Columns 1 and 3 report full sample results for the entire board and audit committee, respectively. Columns 2 and 4 are based on a sample of board and audit committee member deaths. All variables are year t+1 value minus year t-1 value, where t is the year of director death. The t-statistics (in parentheses) are adjusted for heteroskedasticity using the Huber-White Sandwich estimator and are corrected for clustering of firm effects. ***, **, and * indicates significance at the 1%, 5%, and 10% level using two-tailed tests, respectively. (1) Dependent Variable: Constant Board/AC Connectedness CEO Power CEO Connectedness CEO/Board Overlap Ratio Loan Size Loan Maturity Board Size Board Independence Busy Board Firm Size ROA Leverage Market-to-book Industry Dummy Observations Adjusted R2

(3) (4) Audit Committee ∆Loan Spread 0.370 0.042** -0.119 (1.04) (2.00) (-0.45) -0.551 -1.230** -1.372* (-1.52) (-2.09) (-2.38) 0.059* -0.267 -0.529 (1.80) (-0.88) (-0.78) -0.028 -1.032 -1.113 (-0.62) (-0.69) (-0.92) 0.055 1.290 0.783 (0.60) (1.03) (1.39) -0.050** 0.565 0.329 (-2.46) (0.49) (0.55) 0.001** 0.022 0.020 (2.31) (0.73) (0.75) 0.001 -0.040 -0.052 (0.90) (-0.31) (-0.89) 0.036 -1.580 -2.392 (0.50) (-1.23) (-1.55) -0.059 -2.213 -1.392 (-0.78) (-1.34) (-1.21) -0.188*** -4.240 -3.277* (-3.75) (-1.60) (-1.72) -1.660*** -1.239* -1.559* (-3.30) (-1.73) (-1.89) 1.057*** 3.455* 2.749* (5.00) (1.90) (1.78) -0.005 1.838* 2.336* (-0.47) (1.74) (1.70) Yes Yes Yes 3,925 322 108 0.100 0.118 0.125 (2)

Board 0.120*** (3.46) -0.250* (-1.76) 0.056* (1.92) -0.020 (-0.51) 0.047 (0.43) -0.053** (-2.50) 0.001** (2.29) 0.001 (0.67) 0.031 (0.42) -0.081 (-1.05) -0.183*** (-3.81) -1.536*** (-3.37) 1.062*** (5.03) -0.003 (-0.42) Yes 3,925 0.098

47

Table 7 Payout Policy and Board Connectedness This table presents Tobit regressions of the effect of board connectedness on corporate payout policy. See Appendix D for variable definitions. The t-statistics (in parentheses) are adjusted for heteroskedasticity using the Huber-White Sandwich estimator and are corrected for clustering of firm effects. ***, **, and * indicates significance at the 1%, 5%, and 10% level using two-tailed tests, respectively. (1) Dependent Variable: Constant

(2) Payout

-0.061*** (-5.93) 0.004*** (2.59) -

Board Connectedness Board Connectedness*High Market-to-book High Market-to-book

-

Market-to-book

0.003*** (4.91) 0.001 (0.35) 0.001 (1.32) 0.002 (0.19) 0.001 (1.50) 0.009** (2.02) -0.001 (-0.30) 0.002** (2.09) -0.002 (-0.47) 0.015 (0.96) 0.185*** (3.57) -0.013 (-1.48) 0.228*** (4.05) -0.010** (-2.15) -0.118*** (-2.83) Yes 7,693 0.252

CEO Power CEO Connectedness CEO/Board Overlap Ratio Board Size Board Independence Busy board Firm Size Beta Cash Cashflow Leverage ROA Institutional Ownership Volatility Industry, Firm, Year Dummy Observations Adjusted R2

48

-0.056*** (-5.38) 0.005*** (3.38) -0.002*** (-2.33) 0.018** (2.39) 0.001 (0.42) 0.001 (1.27) 0.002 (0.14) 0.001 (1.49) 0.009** (2.07) -0.001 (-0.25) 0.002* (1.85) -0.002 (-0.50) 0.014 (0.88) 0.184*** (3.52) -0.013 (-1.50) 0.229*** (4.41) -0.010** (-2.07) -0.125*** (-2.96) Yes 7,693 0.252

Table 8 Competition Shock, Board Connectedness, and Investment This table presents results of the effect of industry competition shock on change of investment and the sensitivity of performance change to investment change among firms with different levels of board connectedness. Competition Shock is defined as a dummy variable equals to 1 if the industry experiences unexpected import tariffs decrease. See Appendix D for variable definitions. The t-statistics (in parentheses) are adjusted for heteroskedasticity using the Huber-White Sandwich estimator and are corrected for clustering of firm effects. ***, **, and * indicates significance at the 1%, 5%, and 10% level using twotailed tests, respectively. Dependent Variable: Constant Competition Shock Board Connectedness Board Connectedness * Competition Shock Investment Change

(1) ΔInvestment -0.055 (-1.42) -0.016* (-1.79) 0.002 (1.60) 0.035** (2.11) -

Board Connectedness * Investment Change

-

CEO Power

0.001 (1.22) -0.014 (-0.91) -0.019* (-1.85) 0.020 (1.38) -0.002 (-0.55) -0.002 (-1.23) -0.001 (-1.32) -0.015 (-1.09) -0.002 (-0.26) -0.056 (-1.35) 0.025* (1.87) 0.025* (1.89) Yes 422 0.166

CEO Power * Competition Shock CEO Connectedness CEO Connectedness * Competition Shock CEO/Board Overlap Ratio Firm Size Board Size Board Independence Busy Board ROA Leverage Institutional Ownership Industry Dummy Observations Adjusted R2

49

(2) ΔROA -0.054 (-1.09) -

(3) ∆Tobin’s Q 1.555 (1.60) -

0.043* (1.75) -

0.631* (1.88) -

0.092 (1.62) 0.398* (1.83) 0.018 (0.31) -

2.749* (1.89) 6.692** (2.11) -0.553 (-0.84) -

-0.008 (-1.56) -

-0.310 (-1.59) -

0.074 (0.59) 0.098* (1.75) -0.001 (-0.94) 0.038 (0.23) -0.029 (-0.39) -

-1.449* (-1.79) -0.207 (-0.81) 0.001 (0.09) 1.306 (0.87) -0.854 (-0.44) 4.001** (1.99) 1.557 (0.56) 0.571* (0.84) Yes 211 0.258

0.053 (0.20) -0.067 (-0.64) Yes 211 0.228