Community Social Capital and Board Advising: Evidence from the Structure of Board Committees

We investigate how community social capital, captured by the strength of cooperative norms and social networks within a geographical community, affects the internal structure of corporate boards. We find that firms headquartered in high-social-capital US counties have a more advising-intensive board structure, as they are more likely to set up specialised advisory committees and appoint more advisory directors. These findings are robust to endogeneity concerns and a battery of sensitivity tests. Our mediation analysis shows that the increased board advising intensity, induced by community social capital, reduces investment inefficiency. We further reveal that community social capital reduces board monitoring intensity and directors’ monitoring efforts. Overall, our results are consistent with the argument that community social capital serves as a societal monitoring mechanism to reduce firms’ need for board monitoring and, hence, firms’ boards located in high-socialcapital communities focus more on advising.


Introduction
Community social capital, captured by the confluence effects arising from the cooperative norms and the density of associational networks in a geographical community, is an important construct across various disciplines, including sociology, economics and management. 1 As individuals are susceptible to social influences in the geographical areas in which they reside, community social capital helps to build trust, reciprocity, information sharing and cooperation, thus encouraging honest dealings and discouraging individuals' unethical behaviours , La Porta et al., 1997. 2 A burgeoning literature shows that community social capital matters in the corporate setting, as corporate managers are individuals subject to the influence of social capital in the community where their firms are headquartered (Jha and Chen, 2014). Managers' self-interested behaviours are contrary to the prescribed values of cooperative norms, and dense social networks compel individuals to comply with the codes of conduct associated with cooperative norms . Therefore, community social capital disciplines managers and alleviates agency issues (Gao et al., 2019, Hoi et al., 2019, an important responsibility of the board of directors. However, we still know little about how community social capital affects the functioning of corporate boards. The board of directors, as an integral element, performs both monitoring and advising duties (Adams and Ferreira, 2007). The monitoring function of the board oversees the management and guards against harmful conduct, while the advising function of the board guides the management to apply appropriate strategies and approves major expenditures 1 Prior studies in these disciplines have found that community social capital reduces the crime rate (Paolo Buonanno et al., 2009), enhances local and national governmental performance (Knack, 2000) and facilitates economic growth (Knack and Keefer, 1997). 2 For instance, Hong et al. (2004) and Hong et al. (2005) show that social interactions in local geographical areas affect stock-market participation and fund managers' trading behaviours. Pirinsky and Wang (2006) observe strong co-movement in the stock returns of companies headquartered in the same geographic area due to the trading patterns of local residents. A large number of previous studies, including , Elster (1989), Guiso et al. (2004), and Spagnolo (1999) have found that strong cooperative norms and dense social networks in a community foster an environment that constrains narrow and self-interested behaviours and limits opportunistic behaviours. . Although monitoring the management is essential for firm success, excessive board monitoring can be counterproductive. Increased focus on board monitoring not only comes at a substantial cost to board advising but also weakens the CEO's perception of board support (Adams andFerreira, 2007, Hillman andDalziel, 2003), which leads to managerial myopia and poor performance . While antecedents for effective board monitoring have been well established, the optimum way to structure the board for effective advising remains an essential but understudied issue.  claim that board monitoring becomes less demanding, and even redundant, when external governance mechanisms discipline managers. Adams and Ferreira (2007) theoretically contend that a friendly board that does not monitor too much but focuses on advising is more optimal when other governance mechanisms exist. Therefore, we conjecture that firms headquartered in areas with higher levels of social capital can assemble a more advising-intensive board to avoid the adverse consequences of excessive monitoring and improve board efficiency.
To test our conjecture, we empirically explore the effect of social capital at the county level in the US on board advising intensity for firms headquartered in the county. Using a sample of 12,174 firm-year observations from S&P 1,500 firms over the period 2000 -2018, we find that firms headquartered in counties with higher levels of social capital are more likely to set up specialised advisory committees and appoint more directors devoted to advising, suggesting a positive relationship between community social capital and board advising intensity.
According to , advisory directors offer strategic advice about the firm's investment opportunities. We expect that firms that receive better advice will invest wisely and efficiently (Kim et al., 2014). Our mediation analysis confirms that board advising intensity mediates the relationship between community social capital and firm investment efficiency. In other words, the increased board advising intensity, induced by higher levels of community social capital, results in more efficient firm investments, suggesting that community social capital improves board advising efficiency.
We then conduct several tests to underpin the argument that community social capital reduces board monitoring need and allows for more advising. First, we corroborate previous studies by showing that community social capital reduces discretionary accruals, CEO compensation, and costs of equity. Second, we find that firms in high-social-capital counties appoint fewer monitoring directors and are less likely to form a monitoring-intensive board.
Third, we use board meeting attendance as a proxy for directors' efforts (Masulis and Mobbs, 2014) and show that community social capital significantly increases meeting absence among directors holding monitoring duties, but not among those specialising in advising. Collectively, these results provide strong evidence that board monitoring is less demanding for firms in highsocial-capital counties, and suggest that the board may shift its focus from monitoring to advising.
To further strengthen our premise, we conduct two cross-sectional analyses to examine whether a firm's need for board monitoring plays a role in the positive relationship between community social capital and board advising intensity. As the firm's need for board monitoring is further reduced by the external market monitoring (Guo et al., 2015, Randøy and, we find that the observed relationship is more prominent for firms covered by more analysts and for firms operating in highly competitive industries. The cross-sectional variation evidence further confirms the importance of reduced board monitoring needs in the interplay between community social capital and board advising intensity. We also investigate an alternative interpretation of our results. Specifically, higher levels of community social capital develop trust and dense networks, which are considered valuable resources for strategic advising. As a result, the board can take advantage of these resources and increase its advising intensity. This proposes a direct effect of community social capital on board advising, while our premise and previous results suggest an indirect effect of community social capital on board advising through a reduction in board monitoring. While our premise is consistent with the agency theory perspective of community social capital and board advising (Adams and Ferreira, 2007, Hoi et al., 2019, Masulis and Mobbs, 2014, the alternative explanation is consistent with the resource dependence theory, which argues that the board responds to the external environment and changes its composition directly (Hillman et al., 2009, Pfeffer andSalancik, 2003). We do not consider the two interpretations mutually exclusive, as Hillman and Dalziel (2003) contend that board function can be explained by integrating the two perspectives. We perform a path analysis that supports our view. We find that community social capital is positively related to board advising intensity when board monitoring is held constant, confirming the direct effect. However, community social capital also indirectly improves board advising intensity through its effect on reducing board monitoring after taking into account its direct effect. Thus, both paths contribute to the positive relationship between community social capital and board advising intensity. However, since excessive board monitoring leads to inferior performance , the indirect path is important for firm success.
Our main results hold when addressing the endogeneity concerns with the instrumental variable (IV) approach, propensity score matching (PSM) technique and difference-indifferences (DiD) analysis, suggesting that our results are causal instead of correlational. The results also survive a battery of robustness tests relating to omitted variables and alternative measures of community social capital and advisory directors.
We first contribute to the limited research exploring the dynamics of the optimal internal structure of board committees Wu, 2016, Faleye et al., 2013). As we show that community social capital is a key factor influencing internal board structure regarding committee setup and director assignments, we conform to the notion that optimal board structure depends on a firm's operating environment (Boone et al., 2007. Second, we elaborate on the emerging literature on the dual role of board functions. We confirm that community social capital improves monitoring outcomes (e.g., mitigates earnings management and rent extraction) and show that the increased board advising intensity, induced by community social capital, enhances board advising effectiveness (e.g., reduces investment inefficiency). Third, we highlight the importance of recognising the heterogeneity of independent directors sitting on different board committees , Zalata et al., 2019. By examining independent directors sitting on monitoring and advisory committees, rather than treating all independent directors homogenously, we add to the research that explores the different roles played by independent directors , Faleye et al., 2013. Fourth, we extend the theoretical development of Adams and Ferreira (2007) by answering the call to investigate circumstances in which a friendly board that does not monitor excessively but focuses on advising is optimal. To the best of our knowledge, this study is the first to investigate the interactions between external monitoring mechanisms (e.g., community social capital) and the subordinate board structure regarding advising. Our findings suggest that firms can alleviate the competing tensions between board monitoring and advising when considering the community social capital surrounding their headquarters.

Community social capital and opportunistic behaviours
The theoretical foundation of social capital was first systematically developed by , who defines social capital as a variety of different entities generated by trust, information, norms, obligations and effective sanctions that facilitate individual or group actions.  further perceives the use of social capital as a general theoretical strategy that involves taking rational action but rejects extreme individualistic premises.
Since then, social theorists have developed various operating definitions of social capital and argue that social capital encourages honest dealing and deters opportunistic behaviours (Adler andKwon, 2002, Scrivens andSmith, 2013). Because the environmental secular of social norms and networks are at the core of social capital, two approaches are commonly adopted in previous definitions. In the 'norms' approach,  sees social capital as a tendency of people within a group to collaborate to achieve socially productive outcomes and emphasises the norms of reciprocity and trustworthiness that arise from connections between individuals. Fukuyama (2001) argues that social capital is the existence of the same set of informal values or norms shared among members of a group that allows for cooperation. In the 'network' approach, social capital is modelled as a set of networks from which efficient information sharing and better communication are derived , Lin, 1999, Payne et al., 2011. Given that individuals need to maintain a moral self-concept (Mazar et al., 2008), dense social networks intensify the costs and the punishment of unethical and opportunistic behaviours (Coleman, 1994, Spagnolo, 1999. 3 As a result, repeated interactions within a dense network promote greater trust among its members over time and foster the norms of cooperation and honesty , Fischer and Pollock, 2004, Fukuyama, 2001, Putnam, 2000, Uzzi, 1996. Economists, however, criticise the lack of conceptual and analytical frameworks in social capital, as it is difficult to disentangle the effects of cooperative norms and social networks (Sobel, 2002). Since individual behaviours are influenced by the community, economists characterise social capital as a community-level attribute that collectively reflects individuals' behaviours, beliefs and values (Rupasingha et al., 2006). Therefore, economics studies often do not distinguish between social norms and networks but instead adopt the approach advocated by Knack and Keefer (1997), Woolcock (1998), and Guiso et al. (2004) to define social capital as the environmental element that jointly captures the confluence effects of cooperative norms and dense networks within a geographical community, a definition that we follow in this study.

Board committee reforms, monitoring and advising
Agency theory posits that, due to the separation of ownership and control, managers tend to engage in self-interested, opportunistic behaviours that benefit themselves but at the cost of the shareholders (Jensen and Meckling, 1976). Shareholders thus appoint the board of directors to discipline managers and protect their interests. Given that board committees drive the functioning of the board (Adams et al., 2015, Kesner, 1988, Klein, 1998, regulators in the US have gradually turned their attention to the composition of board committees concerning monitoring. 4 After several major accounting scandals in the early 2000s, the Securities and Exchange Commission (SEC) passed the Sarbanes -Oxley Act (SOX) in 2002 in order to scrutinise board monitoring. This Act mandates firms to set up several monitoring committees namely the audit, compensation, governance and nomination committees, which are composed solely of independent directors. 5 Early research, including Fama and Jensen (1983), acknowledges that the board also has an advising role as they provide counsel to the CEO, set strategy and approve major expenditures. However, it was not until 2004 that the Corporate Director's Guidebook of the American Bar Association explicitly recognised advising as one of the two basic functions of 4 The Securities and Exchange Commission (SEC) began to require firms to establish an audit committee comprised of outside directors in 1940 (Birkett, 1986), and to mandate firms to disclose audit committee composition in the 1970s (Reeb and Upadhyay, 2010). 5 In response to SOX, major US stock exchanges (i.e. the New York Stock Exchange and Nasdaq) also issued requirements regarding board committees. The New York Stock Exchange (NYSE) Listed Company Manual Section 303A.03 requires complete independence of audit, compensation, nominating and governance committees. Nasdaq requires complete independence of these major committees or independent directors to oversee the executive compensation and requires a majority of independent directors to select or recommend director nominees if such committees do not exist. the board (Adams, 2010). The advising role of the board guides the management to formulate strategies and assist with decision-making . However, intensive board monitoring weakens board advising, as it creates an information conflict between the management and the board (Adams and Ferreira, 2007). Since managers are concerned that information disclosed to the board can be used to monitor their behaviours, they become reluctant to share key strategic information with a board that monitors intensively.
Consequently, the lack of valuable information provided to the board weakens strategic advising and reduces shareholder value. Adams and Ferreira (2007)

Empirical literature review and hypotheses development
As previously discussed, the board of directors has monitoring and advising duties. Many prior studies adopt an 'inside-outside' approach to proxy for the strength of board monitoring and advising. It follows that independent directors mainly contribute to the monitoring function, since they are independent of the management, while inside directors primarily perform the advising duty, because they have more firm-specific knowledge (Duchin et al., 2010, Lehn et al., 2009, Linck et al., 2008. However, Baldenius et al. (2014) conclude that the 'inside-outside' approach oversimplifies the role of independent directors and leads to inconclusive empirical evidence. 6 The emerging literature has shifted the focus from the 'inside-outside' approach toward a holistic understanding of board committees when evaluating board monitoring and advising intensity. Since the board sets up committees that are either of an advising or a monitoring nature to address firms' specific needs (Klein, 1998 propose that observing board committees is a better way to proxy for the strength of board monitoring and advising.  show that a monitoring-intensive board, where the majority of independent directors are allocated to monitoring committees, results in significantly weaker strategic advising and greater managerial myopia. Consistent with the theoretical prediction in Adams and Ferreira (2007),  confirm that the costs of weaker advising outweigh the benefits of intensive monitoring, as firm value is significantly lower for those with monitoring-intensive boards.
Due to the information conflict, weak advising also poses a threat to boards whose directors perform both monitoring and advising duties (Adams and Ferreira, 2007). Faleye et al. (2013) argue that it is vital for the board to separate committees specialising in advising from those performing monitoring duties, because the separation alleviates information conflict and serves as a substitute for a commitment to not use the revealed information against the CEO (Laux andLaux, 2009). Zalata et al. (2019) show that directors appointed to monitoring committees mitigate managerial opportunism, but not those appointed to advisory committees, confirming that directors serving on advisory committees are minimally involved in monitoring 6 The 'inside-outside' approach ignores two important facts. First, independent directors can acquire firm-specific information through board meetings and interaction with the management or other directors, to contribute to the advising function (Brickley andZimmerman, 2010, Hillman andDalziel, 2003). Second, prior studies have acknowledged that the independent director is a valuable source of expertise, as independent directors with specific characteristics and backgrounds can help to achieve superior performance (Dalton et al., 1999, Hermalin and Weisbach, 1988, Yermack, 1996. Hence, independent directors can not only monitor managers but can also provide strategic advice (Bhagat andBlack, 1999, Chen et al., 2020). activities. As information conflicts are alleviated by separating advising and monitoring committees, Faleye et al. (2013) show that firms with specialised advisory directors enjoy enhanced advising performance and have higher shareholder value.
'Friendly board theory' acknowledges that excessive board monitoring impedes information exchange between the CEO and the board, while a friendly board that does not monitor too much receives more valuable information and is better at advising (Adams and Ferreira, 2007). However, the 'friendly board theory' also argues that, in order for a management-friendly board to be optimal, other governance mechanisms need to pick up the slack of board monitoring (Adams and Ferreira, 2007) since managers may still engage in opportunistic behaviours that hurt shareholder value without being disciplined by other governance mechanisms. Prior research, including Cremers et al. (2008), Ferreira et al. (2011) and Guo et al. (2015), shows that external governance mechanisms, including stock price informativeness, market competition, and takeover threats, substitute for the internal governance imposed by directors. Therefore, strong external monitoring mechanisms provide a prerequisite for the board to reduce its internal monitoring and improve its advisory capacity.
However, the current empirical literature has not systemically linked the strength of external monitoring mechanisms to the intensity of board advising.
Community social capital is a societal monitoring mechanism identified in a growing body of literature. Prior research shows that community social capital deters opportunistic corporate practices, such as auditing misconduct and litigation risk (Jha and Chen, 2014), tax avoidance  and conflicts between shareholders and debtholders (Hasan et al., 2017b), because corporate decisions are made by executives who are disciplined by social capital surrounding corporate headquarters (Bertrand andSchoar, 2003, Hilary andHui, 2009).
Since agency issues are caused by managerial opportunism and represent a violation of the trust vested by shareholders, social capital can mitigate this principal-agent problem. Indeed,  document a negative relationship between the social capital of the county where the firm resides and the cost of equity, as equity holders require lower returns for firms with less severe agency issues. Gao et al. (2019) find evidence suggesting that community social capital induces managers to use corporate resources more efficiently. Hoi et al. (2019) conclude that high community social capital mitigates the agency issue by restraining managerial rent extraction. These studies demonstrate that community social capital is an external monitoring mechanism that ameliorates agency conflicts.
The monitoring role of community social capital suggests that the firm's need for board monitoring is low when community social capital already serves as an incremental monitoring mechanism that reduces the agency issue (Hoi et al., 2019). Based on the 'friendly board theory' (Adams and Ferreira, 2007), we conjecture that shareholders would increase board advising intensity when community social capital is high to prevent information conflicts and sustain efficient advising. We, therefore, develop our first hypothesis (H1) as follows: H1: Firms headquartered in high-social-capital regions are associated with greater board advising intensity.
Because the majority of independent directors are full-time employees of other firms, the board takes a more hands-off approach when performing the advising duties. Thus, directors rely on the firm-specific information provided by the CEO to make advising decisions (Adams and Ferreira, 2007). Therefore, the advisory performance depends on the completeness of the information that the management provides (Armstrong et al., 2010). According to the 'friendly board theory', a board that does not monitor too much receives more valuable information and is better at advising (Adams and Ferreira, 2007). If community social capital promotes a friendly board that has separated advising committees, managers will be more willing to provide valuable information to advisory directors (Faleye et al., 2013). The more firm-specific knowledge the management shares, the better the board's advisory performance will be. Since advisory directors guide the CEO to set strategy and approve major expenditures , we should expect firms that receive better advice from the board to invest wisely and have lower levels of investment inefficiency (Kim et al., 2014). Therefore, we develop our second hypothesis (H2) as follows: H2: The increased board advising intensity induced by community social capital results in lower investment inefficiency.

Data source
Our sample consists of S&P 1,500 firms for the period 2000 -2018, excluding firms from the financial (SIC 6000 -6999) and utility sectors (SIC 4900 -4999). 7,8 We manually track firm' headquarters' counties during the sample period using the address information stated in firm 10-K filings from the SEC Electronic Data Gathering, Analysis, and Retrieval (EDGAR) database. Firms headquartered outside the US are excluded. We use the Federal Information Processing Standards (FIPS) codes for each firm headquartered in US counties to match county-level data. The social capital index for each county is constructed using data from the Northeast Regional Center for Rural Development (NRCRD). We collect the county-level economic outputs and demographic profiles from the Bureau of Economic Analysis (BEA) and the United States Census Bureau. Firm fundamental variables are retrieved from Compustat, stock market price data are from CRSP, directors' committee assignments are from BoardEx and director meeting attendance is from ISS. Our final sample consists of 1,281 unique firms and 12,174 firm-year observations. 7 We start from 2000 because data prior to 2000 are limited in BoardEx. 8 Prior research provides inconclusive evidence on the long-term presence of social capital in societies (Paxton, 1999. Researchers argue that the inconclusive evidence is mainly due to the lack of reliable data on measuring social capital (Rupasingha et al., 2006). Rupasingha et al. (2006) are the first to develop the most reliable measure of social capital that captures both cross-sectional and time-series variations in social capital based on US county-level data. The method is widely adopted in academic studies, including Hasan et al. (2017b) and Hoi et al. (2019). We, therefore, chose the US context for our study.

Variables used in the study
Dependent variables. Following Reeb and Upadhyay (2010), Faleye et al. (2013), we define finance, investment, strategy, acquisitions, science and technology and executive committees as advisory committees while audit, compensation, nominating and governance committees are considered monitoring committees. 9 Advisory_Committee is a dummy variable that equals one if the firm sets up at least one advisory committee, and zero otherwise. N_Advisory_Committee is the logarithm transformation of the number of advisory committees in a firm plus one. Following Faleye et al. (2013), we define independent directors who sit on at least one of the advisory committees but do not serve on any monitoring committee as advisory directors. Advisory_Director_Ratio is the number of advisory directors to the total number of independent directors. 10

Main independent variable: SC_Index
Following Hoi et al. (2019), we define community social capital as the joint effect of cooperative norms and social networks within a US county.
Following Rupasingha et al. (2006), we measure community social capital (SC_Index) as the first principal component of the voter turnout for the presidential election (Pvote), census mail response rate (Pespn), the aggregate number of social organisations (Assn) and the number of not-for-profit organisations (Nccs) for each county provided by NRCRD. The NRCRD only provides data for 1997, 2005, 2009 and 2014 over our sample period. 11 We, therefore, follow Hoi et al. (2019) to backfill the missing data using the available SC_Index from the most recent 9 To identify committees devoted to the advising function and minimally involved in the monitoring function, committees that share both monitoring and advising responsibilities, such as the audit and finance committee, are considered monitoring committee, as in Faleye et al. (2013). 10 Following Faleye et al. (2013), we focus on independent directors, because inside directors do not typically serve on board committees (Chen and Wu, 2016 Table A1 from the Appendix.
(1) when the dependent variable is N_Advisory_Committee and Advisory_Director_Ratio. The main variable of interest, _ , , , is the estimated social capital index for county j where firm i is headquartered at time t. , , is a vector of the five sets of variables described in the previous section. λ and λ +1 are industry and year dummies, respectively. Industry is defined by the two-digit Standard Industrial Classification (SIC) codes. H1 predicts a positive and statistically significant coefficient on community social capital ( 1 ).
We perform a mediation analysis to test H2. The intuition behind the mediation analysis is illustrated in Figure 1. Path ABC represents the total effect of the treatment (community social capital) on the outcome (investment inefficiency), which can be decomposed into direct and indirect effects. Path A corresponds to the effect of the treatment on the mediator (board advising intensity) and Path B demonstrates the effect of the mediator on the outcome. Paths A and B comprise the indirect effect (mediating effect) of community social capital on investment inefficiency, while Path C shows the direct effect of community social capital on investment inefficiency.
[Insert Figure 1 Around Here] We then follow Baron and Kenny (1986) and Li et al. (2021) to estimate the following structural equation models to test the mediation effect.
As expected, the coefficient on SC_Index is positive (0.091) and highly significant (p<0.000), suggesting that firms residing in high-social-capital counties are more likely to set up committees that specialise in advising. Columns (2) and (3) present the OLS regression results when the dependent variable is N_Advisory_Committee and Advisor_Director_Ratio, respectively. The positive and significant coefficients on SC_Index present clear evidence that higher community social capital is related to more advisory committees within the board and more directors that specialise in advising. The effect of community social capital is also economically significant. For example, an interquartile increase in SC_Index leads to a 9.92% increase in the number of advisory committees and a 19.24% increase in the ratio of advisory directors from their mean. 15 Consistent with H1, our findings support the view that higher 14 The increase from Q1 to Q4 for all three board advising intensity variables is statistically significant at the 1% level. 15 The 25th (75th) percentile of social capital is -1.119 (-0.064). For N_Advisory_Committee in Column (2), an interquartile increase in social capital leads to a 0.033 (= (-0.064 -(-1.119)) × 0.031) increase in the logarithm of levels of community social capital result in a more advising-intensive board. Our results also support the theoretical prediction from Adams and Ferreira (2007), which posits that the board should focus on advising when external governance mechanisms discipline the managers.
[Insert Table 2 Around Here]

The mediating effect of board advising intensity
Since community social capital results in a more advising-intensive internal board structure, we perform a mediation analysis to test whether increased advising intensity can reduce investment inefficiency.
Results from the mediation analysis are presented in Table 3. Panel A presents the results from the structural equations. Specifically, Columns (1), (3) and (5) Baron and Kenny's (1986) approach and shows the mediating effect that operates through Advisory_Committee (N_Advisory_Committee and Advisory_Director_Ratio) is -0.034 (-0.043 and -0.048), accounting for 5.52% (7.08%, and 7.88%) of the total effect of SC_Index on Inefficiency. The z-statistics suggest that these the number of advisory committees. Given that the mean value of number of advisory committees without logarithm is 0.504, an interquartile increase in social capital increases the number of advisory committees to 0.554 (=exp(ln(1+0.504)+0.033)-1), representing a 9.92% increase from its mean. For Advisor_Director_Ratio in Column (3), an interquartile increase in social capital leads to a 1.522 (= (-0.064 -(-1.119)) × 1.441) increase in the ratio of advisory director. With a mean value of 7.907 of Advisory_Director_Ratio, the 1.522 increase represents a 19.24% (=1.522/7.907) increase from its mean.
indirect effects are statistically significant at the 1% level. These results are also illustrated in Figure 4. Overall, the mediation analysis supports H2 and confirms that the increased board advising intensity, driven by community social capital, leads to reduced investment inefficiency.
[Insert Table 3 Around Here] [Insert Figure 4 Around Here] The effect of community social capital on board monitoring needs We develop our hypothesis based on the 'friendly board theory' (Adams and Ferreira, 2007) and the premise that community social capital reduces board monitoring needs, allowing firms in high-social-capital areas to assemble an advising-intensive board. In this section, we strengthen this argument by confirming the negative effect of community social capital on board monitoring needs.
The disciplining effect of community social capital on managers' opportunistic behaviours is extensively documented in the literature , Hoi et al., 2019, Jha, 2019. Notwithstanding the prior evidence, we corroborate the notion that community social capital reduces agency issues by examining its effect on discretionary accruals, CEO compensation and costs of equity in Internet Appendix IB.2. Consistent with prior studies, we find evidence suggesting community social capital reduces agency issues. Since monitoring the management to alleviate agency issues is one of the two primary responsibilities of the board of directors , less severe agency issues suggest lower board monitoring need ).
Next, we directly test our premise by examining the effect of community social capital on board monitoring intensity and director meeting attendance. We follow  to proxy for board monitoring intensity and test the effect of community social capital on board monitoring intensity in Panel A of  (1) and (2), but is negative, albeit statistically insignificant, for Advisor_Attendance_Problem (-0.000) and Advisor_Attendance_Problem_Ratio (-0.032) in Columns (3) and (4) [Insert Table 4 Around Here] Taken together, the results from this section conform with previous literature by showing that higher levels of community social capital reduce the agency issue , Hoi et al., 2019, and that the need for board monitoring is low . The reduced board monitoring need for firms in high-social-capital counties reduces 16 Director meeting attendance data are obtained from ISS. Unfortunately, ISS does not provide director board assignments other than the audit, compensation, nominating and governance committees, which makes it extremely difficult to accurately identify advisory directors from ISS data. Following  and Zalata et al. (2019), we assume that independent directors that are not classified as monitoring directors are advisory directors in this test only. We then examine board attendance using the director-year data from ISS. The results reported in Internet Appendix IB.3 also confirm that monitoring directors from firms in high-social-capital counties are more likely to miss board meetings, but advisory directors are not. 17 Our results do not necessarily suggest that monitoring directors violate professional guidance, which requires them to monitor the management closely. Instead, it indicates that the extent of monitoring depends on the social capital surrounding their firms' headquarters.
monitoring intensity and suggests our findings are due to the board shifting its focus from monitoring to advising.

Cross-sectional analysis
In this section, we conduct several moderating tests on the interplay between community social capital and board advising intensity. If the positive effect of community social capital is more prominent when the firm operates in environments that are already subject to strict monitoring, then this provides further assurance that reduced board monitoring needs are the key to explaining our findings. In Table 5, we explore the interactive effect of external monitoring mechanisms imposed by financial analysts (Healy and Palepu, 2001) and market competition (Nickell et al., 1997) that make board monitoring less demanding .
We create High_Coverage to indicate firms with above-median analyst coverage in each industry each year and High_Competition for firms primarily operating in a competitive industry. We then interact SC_Index with the two dummy variables, respectively. The interaction term coefficients (SC_Index*High_Coverage and SC_Index*High_Competition) are positive and significant, suggesting a more prominent effect of community social capital on board advising intensity when strong external monitoring further reduces the firm's needs for board monitoring. Thus, results from Table 5 further confirm that the reduction in board monitoring needs is the key factor facilitating the positive relationship between community social capital and advising intensity.
[Insert Table 5 Around Here]

Path analysis
So far, our premise and results are consistent with the agency theory perspective on community social capital, which states that community social capital reduces board monitoring intensity through its effect on reducing board monitoring needs and intensity. In this path, board monitoring is the mediator, and community social capital is perceived to have an indirect effect on board advising intensity through a reduction in board monitoring.
However, the resource dependence theory (Pfeffer and Salancik, 2003) proposes that the board changes its composition by responding to the external environment, as the board's capital can bring resources (e.g., advice and counsel, links to other organisations.) to minimize environmental dependence. 18 Given that trust and networks are the core components of community social capital, they may directly influence the board's capital regarding its advising capacity. For example, since the development of trust facilitates better information sharing that is essential for strategic advising (Kor and Sundaramurthy, 2009), managers in high-socialcapital communities are influenced by the norms of trustworthiness and become more willing to share information with the board and more likely to trust and value external advice. In addition, the dense social network in high-social-capital communities may help to build the board's relational capital by increasing their social ties, which becomes a valuable resource for board advising. Therefore, in an environment where trust and networks are important, selecting directors that are trustworthy and can bring external resources will more effectively utilise directors' capital for board advising. As a result, shareholders may assemble a more advisingintensive board. Thus, the resource dependence perspective predicts a direct path from community social capital to board advising intensity.
To investigate the path through which community social capital affects board advising intensity, we adopt a mediating analysis and present the results in Table 6. We use Monitoring_Director_Ratio as the mediator. 19 The coefficients of SC_Index (the treatment) in Column (1) and Monitoring_Director_Ratio (the mediator) across Columns (2) to (4) are all 18 It is worth noting that community social capital is different from the board's capital. Community social capital is the environmental factors arising from social norms and networks that influence an individual's behaviours (Rupasingha et al., 2006), while the board's capital refers to directors' human capital (i.e., experience and expertise) and relational capital (i.e., network of ties to external contingencies) (Hillman and Dalziel, 2003). 19 Faleye et al. (2013) argue that Advisory_Director_Ratio and Monitoring_Director_Ratio capture distinct functions of the board and the two variables do not mirror each other.
significantly related to board advising intensity (the outcome), showing an indirect effect from community social capital to board advising intensity through board monitoring. Panel B of Table 6 tests the significance of the indirect effects with Sobel's (1982) statistics and finds that the indirect effects are all statistically significant at the 1% level. In addition, Columns (2) to (4) in Panel A confirm that there is a direct effect from community social capital on all three measures of board advising intensity, as the coefficient on SC_Index is positive and significant when Monitoring_Director_Ratio is included in the regressions. Thus, firms in high-socialcapital community can also increase board advising intensity independent of the monitoring intensity.
[Insert Table 6 Around Here] Thus, the results support the agency theory perspective that community social capital indirectly affects board advising through its effects on mediating board monitoring, but also the resource dependence explanation that community social capital can directly affect board advising. The two effects are not mutually exclusive. This finding is consistent with Hillman and Dalziel (2003), who argue that integrating the two theories helps to explain findings in the research on boards of directors. 20 Despite the direct effect, we argue that the indirect effect is vital to board functioning and corporate governance reform. As Adams and Ferreira (2007) and  show that excessive board monitoring hurts firm value, it is crucial for the board to reduce its monitoring intensity and increase board advising when external monitoring mechanisms already discipline the managers. 21 20 Hillman et al. (2009) argue that, although the agency theory is the predominant theory used in board of directors studies (Dalton et al., 2007, Johnson et al., 1996, the resource dependence theory has an important construct on board studies that have often been dwarfed by applications of agency theory. 21 In addition, we decompose SC_Index into Cooperative_Norms and Social_Networks and show that the trust captured by Cooperative_Norms can significantly reduce the agency issue, suggesting that trust can also influence board advising through the indirect channel. Furthermore, we additionally control for directors' network size and confirm that greater directors' relational capital does not drive our results. We also test whether risk aversion that is correlated with religion can explain our findings, as prior research recognises that religion can affect agency problems and the role of the board , Miller, 2000. The results are discussed in Internet Appendix IB.4 to IB.6.

Endogeneity
Unlike monitoring committees that are required by regulation, firms have the discretion to set up advisory committees. As a result, the presence of advisory committees may be a function of observable and non-observable characteristics that correlate with community social capital. In addition, if the corporate headquarters location is self-selected and endogenously determined, community social capital could also be endogenous. We allay the potential endogeneity concerns in this section.
We first address the endogeneity concerns with an IV approach using the firm's distance to the US -Canadian border (Border_Distance) as the instrument for community social capital, because  claims that areas closer to the border have higher social capital. The first-stage result, presented in column (1) of Table 7, confirms a negative relationship between Border_Distance and SC_Index. 22 The fitted value of community social capital (Fitted_SC_Index) is predicted, and its coefficient is positive and significant in the second-stage regressions in Columns (2) to (4). 23 [Insert Table 7 Around Here] Next, we control for the observable differences in firm attributes for firms that reside in high-and low-social-capital counties by employing a PSM technique as in Hoi et al. (2019).
We sort counties with SC_Index in the top (bottom) quartile into the high-social-capital (lowsocial-capital) group. High_Social_Capital is a dummy variable set to one for firms in highsocial-capital counties and zero for firms in low-social-capital counties. A propensity score is computed based on all firm-level controls in Eq.(1). We then match, without replacement, each firm located in high-social-capital counties with a unique firm residing in low-social-capital 22 We have three second-stage regressions, and therefore, three corresponding first-stage regressions. The firststage regression results are very similar. For brevity, Column (1) of Table 7 only reports the corresponding firststage results for Column (2). 23 In addition, we follow Hasan et al. (2017b) by adding Racial_Diversity as an additional instrument for SC_Index in Internet Appendix IB.7. We continue to find that the fitted value of community social capital results in a more advisory-intensive board. counties using the closest propensity score within the 1% caliper. The regression results based on the matched samples in Panel A of Table 8 show that firms in high-social-capital counties are more advising-intensive, confirming our baseline results. The balance tests of the matching variables in Panel B reveal no significant difference in any variables across the two groups, suggesting a good match of the PSM sample. 24 [Insert Table 8 Around Here] We further address the endogeneity concern with a DiD analysis on firms that relocate headquarters to a different county. 25 Following Hasan et al. (2017b), we define a social-capitalchanging relocation as a firm that moved its headquarters to another county, each with at least two years of available data before and after the relocation. 26 We create a dummy variable, Increase_Relocation, that equals one for firms relocated to counties with higher levels of social capital, and zero for firms relocated to counties with lower levels of social capital.
Post_Relocation is a variable indicating years after relocation. The positive and statistically significant coefficients on the interaction term (Increase_Relocation*Post_Relocation) across Panel A of Table 9 indicate that firms that relocated to counties with higher levels of social capital increased their board advising intensity. These results give us more confidence in inferring the causal relationship of our findings. Panel B shows that differences in board advising intensity and firm characteristics are insignificant for firms in the two groups in the year prior to the headquarters relocation, suggesting that the parallel trend assumption of DiD analysis is likely to be met in our analysis. 27 24 In unreported tables, we find that the results remain robust when using sample median or tertile as the benchmark to define High_Social_Capital, lifting the no replacement restriction or including county-level variables in the matching process. 25 The DiD analysis can also address the concern that our results only capture the cross-sectional variations in community social capital due to the lack of mobility of firm headquarters' locations. 26 We removed firms with multiple relocations to avoid the confounding effect. We identified 145 relocation events that meet our requirements, of which 65 firms moved to counties with higher social capital and 80 firms relocated to counties with lower social capital. These relocations yielded 1,496 firm-year observations, of which 611 are from the pre-relocation period and 885 are from the post-relocation period. 27 We also obtain qualitatively similar results when addressing the endogeneity concerns with the generalised method of moments (GMM) models. Results are reported in Internet Appendix IB.8.
[Insert Table 9 Around Here] Collectively, results from this sub-section attenuate the endogeneity concern of our study and confirm that high community social capital drives up board advising intensity.

Robustness
We address the concern that omitted variables drive our results in Table 10. In Panels A and B, respectively, we include the State_GDP_per_Capita and Metro in Eq.
(1) to address the concern that state-level and metropolitan factors may plague our findings. 28 Following Hoi et al. (2019), we capture the influence of unknown omitted county-level factors by additionally controlling for the median value of our dependent variables in Panel C. Unknown firm-level variables are captured by the long-window change-on-change analysis in Panel D. 29 In Panel E, we address the related concern that board structure among firms located in the same county might be correlated by replacing the corresponding firm-level variables with county-median values. 30 The positive and significant coefficients on SC_Index (or ΔSC_Index in Panel D) across all panels confirm that our main findings are robust to unobserved state-, metropolitan-, county-and firm-level factors influencing internal board structure concerning advising intensity.
[Insert Table 10 Around Here] We also show that our results are robust to 1) alternative proxies for community social capital and advisory directors; 2) Poisson and negative binomial models for the non-negative 28 The related concern for controlling for the metropolitan setting is that firms headquartered in metropolitan areas enjoy agglomeration benefits such as lower communication and transportation costs and increased efficiency . It might be easier for firms to find suitable advisory directors. 29 Since both the board structure and social capital of each county are relatively stable, the standard county fixedeffect or firm fixed-effect model is inappropriate and not applicable (Griffin et al., 2021). Zhou (2001) and Roberts and Whited (2013)  These results are reported and discussed in Internet Appendix IB.9 to IB.12, respectively.

Summary and conclusion
This study integrates two lines of research. The first one uncovers the role of community social capital in corporate settings and the second one explains the dynamics of board structure concerning committees. We find robust empirical evidence that firms headquartered in communities with higher levels of social capital are more likely to set up specialised advisory committees and appoint more advisory directors. The increased advising intensity of the board leads to improved investment efficiency.
We make several contributions to the literature. First, we add to the limited research that explores the dynamics of optimal internal board structure Wu, 2016, Faleye et al., 2013) by investigating the impact of community social capital on advisory committees and directors. Building on the perspective that community social capital is a societal monitoring mechanism that alleviates the agency issue (Hoi et al., 2019), we show that social capital surrounding the firm's headquarters is a crucial factor influencing board internal structure regarding advising intensity. Our results refute the one-size-fits-all criterion that regulatory actions apply to board composition and conform with the notion that optimal board structure depends on the firm's operating environment (Boone et al., 2007.
Second, we elaborate on the emerging literature on the dual role of board functions and board effectiveness. We confirm that community social capital improves board monitoring effectiveness, as it reduces accounting manipulation and rent extraction. More importantly, we show that board advising intensity can mediate the relationship between community social capital and firm investment inefficiency, suggesting that increased board advising intensity, induced by community social capital, can improve board advisory effectiveness.
Third, we reveal the importance of investigating the heterogeneity of independent directors sitting on different board committees , Zalata et al., 2019. By showing the important role of independent directors on advisory committees, we shed light on research proxying for board monitoring and advising from the internal structure of board committees, rather than adopting the 'inside-outside' approach that treats independent directors on the board as a homogenous element.
Fourth, our investigation advances the theoretical view of Adams and Ferreira (2007) by showing that community social capital is an important factor in optimising a friendly board.
Specifically, we respond to the call from Adams and Ferreira (2007) to investigate circumstances in which external monitoring mechanisms substitute board monitoring for an advisory board to be optimal. We provide direct evidence showing that community social capital is an external monitoring mechanism that reduces board monitoring needs. We also confirm that an advising-intensive board for firms in high-social-capital areas is optimal as it can improve board advising effectiveness without impairing monitoring effectiveness.
Our findings have economic and policy implications regarding the optimal board structure. Shareholders face a trade-off when appointing directors to monitoring and advising duties as the increased focus on one responsibility is often at the cost of the other (Masulis and  (Adams and Ferreira, 2007). To regulators, we provide valuable advice on how best to efficiently push governance reform that improves overall board effectiveness when board monitoring and advising duties are considered. In addition, both the academic and popular press are debating whether social capital is declining in the US (Paxton, 1999 Second, social capital is a broad concept as it is the aggregate effects of multiple entities that are difficult to disentangle , Sobel, 2002 Percentage of people who are 25-years-old or above in the county with a Bachelor's degree or higher.

United States Census Bureau
County_Median_Age Natural logarithm of the population median age in the county.

United States Census Bureau Inefficiency
Industry-adjusted investment inefficiency estimated from  model. Compustat

Monitoring_Director_Ratio
Number of monitoring directors to the total number of independent directors. A monitoring director is an independent director sitting on at least two monitoring committees (audit, compensation and nominating/governance committees).

Monitoring_Intensive_Board
Dummy variable coded to one if the majority of independent directors are monitoring directors. A monitoring director is an independent director sitting on at least two monitoring committees (audit, compensation, and nominating/governance committees).

Monitor_Attendance_Problem
Dummy variable that equals one if at least one of the monitoring directors of the firm attends less than 75% of the board meetings during a year, and zero otherwise. A monitoring director is an independent director sitting on at least two monitoring committees (audit, compensation and nominating/governance committees).

ISS
Monitor_Attendance_Problem_Ratio Number of monitoring directors of the firm who attend less than 75% of the board meetings during a year, scaled by the total number of monitoring directors. A monitoring director is an independent director sitting on at least two monitoring committees (audit, compensation and nominating/governance committees).

Advisor_Attendance_Problem
Dummy variable that equals one if at least one of the advisory directors of the firm attends less than 75% of the board meetings during a year, and zero otherwise. An advisory director in this variable is defined as an independent director who is not a monitoring director. A monitoring director is an independent director sitting on at least two monitoring committees (audit, compensation and nominating/governance committees).

Advisor_Attendance_Problem_Ratio
Number of advisory directors of the firm who attend less than 75% of the board meetings during a year, scaled by the total number of advisory directors. An advisory director in this variable is defined as an independent director who is not a monitoring director. A monitoring director is an independent director sitting on at least two monitoring committees (audit, compensation and nominating/governance committees).

Ave_Director_Age
Natural logarithm of the average age of the board of directors. ISS

Ave_Director_Tenure
Natural logarithm of the tenure of the board of directors plus one.

ISS US_Director_Ratio
The ratio of US directors to the total number of board directors.

ISS Female_Director_Ratio
The ratio of female directors to the total number of board directors. BoardEx Board_Size Natural logarithm of the number of board directors. BoardEx Ave_N_Outside_Directorships Average number of directorships the board of directors hold outside the focal firm. ISS ROA Earnings before interests, taxes, depreciation and amortisation, scaled by total assets. Compustat High_Coverage Dummy variable that equals one if the number of analysts covering the firm is above the median in each industry each year, and zero otherwise.

High_Competition
Dummy variable that equals one if the industry Herfindahl-Hirschman Index is below the sample median, and zero otherwise.

Low_Education
Dummy variable that equals one if the county's Education is below the sample median, and zero otherwise.

High_Religiosity
Dummy variable that equals one if the county's Religiosity is above the sample median, and zero otherwise.

Association of Religion Data Archives Border_Distance
Natural logarithm of the shortest distance between the firm's headquarters county and the US -Canadian border.

High_Social_Capital
Dummy variable that equals one for firms residing in counties in the top quartile of social capital, and zero for firms residing in counties in the bottom quartile of social capital.

Northeast Regional Center for Rural Development (NRCRD) Increase_Relocation
Dummy variable that equals one if the firm relocates to a county with higher social capital, and zero if the firm relocates to a county with lower or equal social capital.

Post_Relocation
Dummy variable that equals one in the year of headquarters relocation and afterwards, and zero for the years preceding the headquarters relocation.

State_GDP_per_Capita
Natural logarithm of per capita GDP of the state.

Bureau of Economic Analysis Metro
Dummy variable that equals one if the firm is located within a 250-kilometre radius of a metropolitan area with more than one million population in the 2010 census.
United States Census Bureau   Table A1.

Tables Table 1. Summary Statistics
This table presents the number of observations (N), the mean (Mean), the standard deviation (Std), the 25th percentile (P25), the median (Median) and the 75th percentile (P75) for the main variables used in this study. The sample consists of 12,174 firm-year observations for the period between 2000 and 2018. SC_Index is the county-level social capital measure based on data from the NRCRD. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise.
N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one. Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. Detailed variable definitions are given in Table A1 (1) through (3) are Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one.
Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. SC_Index is the county-level social capital measure based on data from the NRCRD. Column (1) uses the probit model, and Columns (2) and (3) are OLS models. Detailed variable definitions are given in Table A1 from the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.
(1)   (1), (3) and (5) (2), (4) and (6) estimate Eq. (4) when the mediator is Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Inefficiency is the industry-adjusted investment inefficiency estimated from  model. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one. Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. SC_Index is the county-level social capital measure based on data from the NRCRD. Detailed variable definitions are given in Table A1 from the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes the same set of control variables as in Table 2. Panel B presents the total, direct, and indirect effect. The indirect effect is tested with Sobel (1982) z-statistics. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.   (1) and (2) is Monitoring_Director_Ratio and Monitoring_Intensive_Board, respectively. Monitoring_Director_Ratio is the number of monitoring directors to the total number of independent directors. Monitoring_Intensive_Board is a dummy variable that equals one if the majority of independent directors are monitoring directors. Column (1) uses OLS models, and Column (2) uses the probit model. Panel B presents the results for director meeting attendance. The dependent variables in Columns (1) through (4) are Monitor_Attendance_Problem, Monitor_Attendance_Problem_Ratio, Advisor_Attendance_Problem and Advisor_Attendance_Problem _Ratio, respectively. Monitor_Attendance_Problem is a dummy variable that equals one if at least one of the monitoring directors of the firm attends less than 75% of the board meetings during a year, and zero otherwise.
Monitor_Attendance_Problem_Ratio is the ratio of the number of monitoring directors of the firm that attend less than 75% of the board meeting during a year to the total number of monitoring directors. A monitoring director is an independent director sitting on at least two monitoring committees (audit, compensation, and nominating/governance committees). Advisor_Attendance_Problem is a dummy variable that equals one if at least one of the advisory directors of the firm attends less than 75% of the board meeting during a year, and zero otherwise. Advisor_Attendance_Problem_Ratio is the ratio of the number of advisory directors of the firm that attend less than 75% of the board meeting during a year to the total number of advisory directors. Columns (1) and (3) use the probit models and Columns (2) and (4) (1) and (4) is Advisory_Committee, in Columns (2) and (5) it is N_Advisory_Committee and in Columns (3) and (6) it is Advisory_Director_Ratio. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one. Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. Columns (1) and (4) use the probit models and Columns (2), (3), (4) and (6) use OLS models. SC_Index is the county-level social capital measure based on data from the NRCRD. High_Coverage is a dummy variable that equals one if the number of analysts covering the firm is above the median in each industry each year, and zero otherwise. High_Competition is a dummy variable that equals one if the industry Herfindahl-Hirschman Index is below the sample median, and zero otherwise. Each column includes the same set of control variables as in Table 2, the year and 2-digit SIC dummies. Detailed variable definitions are given in Table A1 from the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.
Monitoring_Director_Ratio is the number of monitoring directors to the total number of independent directors. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialized advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one. Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. SC_Index is the county-level social capital measure based on data from the NRCRD. Each column includes the same set of control variables as in Table 2. Detailed variable definitions are given in Table A1 from the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Panel B presents the total, direct, and indirect effects. The indirect effect is tested with Sobel's (1982) z-statistics. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

Table 7. Instrumental Variable Two-Stage Least Square Analysis
This table presents the regression analysis of the instrumental variable approach. Column (1) presents estimates from the first-stage analysis, where the dependent variable is SC_Index. SC_Index is the county-level social capital measure based on data from the NRCRD. The instrument for SC_Index is Border_Distance, measured as the natural logarithm of the shortest distance between the firm's headquarters county and US -Canadian border. Columns (2) through (4) present the second-stage analysis. The dependent variables in Columns (2) through (4) are Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one. Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. Detailed variable definitions are given in Table A1 from the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.   (3) are Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one. Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. Column (1) uses the probit model, and Columns (2) and (3) use OLS models. High_Social_Capital is a dummy variable that equals one for firms that reside in counties in the top quartile of social capital, and zero for firms that reside in counties in the bottom quartile of social capital. Panel B presents the balance tests between the two groups. Detailed variable definitions are given in Table A1 from the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.   Table 9.

Difference-in-Differences Analysis of Firm Headquarters Relocation
This table presents the results from the difference-in-difference analysis on 145 firms with headquarters relocations, of which 65 firms move to counties with higher social capital and 80 firms relocate to counties with lower social capital. Panel A presents the regression analysis using the sample with headquarters relocations. The dependent variable in Columns (1) through (3) is Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one. Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. Increase_Relocation is a dummy variable that equals one for firms moving to counties with higher social capital, and zero for firms that move to counties with lower social capital. Post_Relocation is a dummy variable that equals one for years after headquarters relocation. Panel B tests firm characteristics prior to the headquarter relocations for firms that experienced social-capital-increasing relocation and firms that experienced social-capital-decreasing relocation. Column (1) uses the probit model and Columns (2) and (3) use OLS models. All columns include the same set of control variables as in Table 2. Detailed variable definitions are given in Table A1 from the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. The industry is defined by the two-digit SIC codes. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.    Advisory_Director_Ratio) for other S&P 1,500 firms residing in the same county in a given year. Panel D presents the results of a long-window change-on-change analysis to remove time-invariant unobserved firm-level variables. The dependent variable is measured as the change from year t to t+5, while all independent variables are measured as the change from year t-6 to t-1. Panel E replaces all firm-level variables with the corresponding median variables for firms located in the same county in a given year. Industry dummies are dropped in Panel E as it is not meaningful to use median industry dummies. Column (1) uses the probit model and Columns (2) and (3) use OLS models. Each column includes the same set of control variables as in Table 2, the year and 2-digit SIC dummies. Detailed variable definitions are given in Table A1 from the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.

Investment Inefficiency -Richardson (2006) Model
Firm investment inefficiency is estimated by following the The total investment, I_total, is then split into two main components: (i) required investment to maintain assets in place, I_maintenance, and (ii) investment in new projects, I_new.
Where I_maintenance is amortisation and depreciation (Compustat item dpc), scaled by previous year's book value of total assets (Compustat item at).
I_new can then be decomposed into expected (or optimal) investments, * , and unexpected (or abnormal) investments, .
As in , we estimate the expected investments using the following dynamic model: Where V/P is a measure of growth opportunities, calculated as the ratio of the value of the firm ( ) to the market value of equity (Compustat item csho × prcc_f). is computed . is set to 12%, and = 0.62 is the abnormal earnings persistence parameter from the Ohlson (1995) framework. BV is the book value of common equity (Compustat item ceq), is the dividends (Compustat item dvc) and is operating income after depreciation (Compustat item oiadp).
Since the abnormal investments can be either positive or negative, the absolute value of the residual ( , ) from Eq. (4) is our estimate for inefficient investment for firm i at year t.

Item IB.1 Correlation Matrix
We have a large set of control variables in our study. If some control variables are highly correlated, multicollinearity could be a potential concern for our empirical setting. We present the Pearson correlation matrix in Table IB.1. The correlation coefficients for each pair of control variables are at the level that does not raise significant multicollinearity concerns. More importantly, SC_Index is positively correlated with all three board advising intensity measures (Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio), consistent with our theoretical prediction that firms which reside in high-social-capital areas focus more on board advising.

Item IB.2 The Effect of Community Social Capital on Agency Issues
Our hypothesis is based on the premise that social capital is a societal monitoring mechanism that mitigates the agency issue and, hence, reduces board monitoring needs. The monitoring role of community social capital has been extensively documented in the literature, including , Hoi et al. (2019), and Jha (2019). Notwithstanding the prior evidence, we corroborate the argument that community social capital reduces the agency issue by examining its effect on discretionary accruals, CEO compensation, and costs of equity in Table IB where TACC is the total accruals, computed as the difference between earnings before extraordinary items and discontinued operations from the cash flow statement (Compustat item ibc) and cash from operation (Compustat item oancf), scaled by the previous year's total assets (Compustat item at). TA is the book value of total assets (Compustat item at). ∆REV is total revenue (Compustat item revt) in year t minus that in year t-1. ∆REC is receivables (Compustat item rect) in year t minus that in year t-1. PPE is gross property, plant, and equipment (Compustat item ppegt). ROA is income before extraordinary items (Compustat item ib) to the average book value of total assets (Compustat item −1 + 2 ). The absolute value of the estimation residual is the measure of discretionary accruals (Discretionary_Accruals). Thus, the higher values of Discretionary_Accruals indicate poorer earnings equality and more severe agency issue. Column (1) of Table IB.2 confirms that community social capital reduces discretionary accounting accruals, thus improving accounting quality (Jha, 2019).
In Column (2) of Table IB.2, we examine the effect of social capital on CEO_Total_Compensation, measured as the natural logarithm of CEO total compensation (Execucomp item tdc1) as in Hoi et al. (2019). The coefficient on SC_Index is negative and statistically significant, confirming that social capital alleviates the rent extraction problem, a form of agency issue.
The dependent variable in Column (3) of Table IB.2 is the firm cost of equity capital, measured from the modified PEG ratio model (Easton, 2004). Cost_of_Equity ( ) is solved from the following equation.
Where * is the closing price in the June following the latest fiscal year-end. +1 is the median forecasted earnings per share from IBES. is the yield on 10-year Treasury bonds.
is the current payout ratio.
Consistent with , community social capital significantly reduces the firm's cost of equity capital, implying that equity holders view firms in high-social-capital areas as having less severe agency issue, hence requiring lower returns.
In sum, results from Table IB.2 are consistent with previous studies showing that community social capital reduces agency issues. Since monitoring the management to alleviate the agency issues is one of the two primary responsibilities of the board of directors (Adams et al., 2010), less severe agency issues suggest lower board monitoring needs . Hence, these results support our premise that community social capital reduces board monitoring needs.

Table IB.2. The Effect of Community Social Capital on Agency Issues
This table presents the regression analysis of the effect of community social capital on agency issues. The dependent variable in Column (1) is Discretionary_Accruals, measured as the absolute residual from the modified Jones model. The dependent variable in Column (2) is CEO_Total_Compensation, measured as the natural logarithm of CEO total compensation reported in Execucomp. The dependent variable in Column (3) is firm Cost_of_Equity, estimated from the modified PEG ratio model (Easton, 2004). SC_Index is the countylevel social capital measure based on data from the NRCRD. Detailed variable definitions are given in Table  A1 in the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively. (1)

Attendance -Director-level Regressions
In Panel B of Table 4 in the main text, we present regression results for the effect of community social capital on director meeting attendance at the firm level using firm-year data. In Table   IB.3, we also present the regression results using director-year data.
Column (1) investigates the monitoring directors' sub-sample and Column (2) investigates the advisory directors' sub-sample. Attendance_Problem is a dummy variable that equals one if the director attends less than 75% of the board meetings during a year, and zero otherwise. Both columns use the probit model. Consistent with Panel B of Table 4, these supplementary results show a positive and significant relationship between community social capital and attendance problems for the monitoring director subsample, but not for the advisory director subsample. These results confirm that monitoring directors reduce their effort to monitor managers if their firms reside in high-social-capital counties, suggesting directors perceive that firms' need for intensive monitoring is low when community social capital is high.

Table IB.3. The Effect of Community Social Capital on Director Meeting Attendance -Director-level Regressions
This table presents the effect of community social capital on director board meeting attendance using director-year data from ISS. Attendance_Problem is a dummy variable that equals one if the director attends less than 75% of board meetings in the year. Monitoring directors are independent directors who serve on at least two monitoring committees (audit, compensation, and nominating/governance committees) within the board. Due to the data limitation of ISS, advisory directors in this table are assumed to be independent directors other than monitoring directors. Both columns use probit models. SC_Index is the county-level social capital measure based on data from the NRCRD. Detailed variable definitions are given in Table A1 in the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.

Agency Issues
Our path analysis shows that community social capital has a direct effect on board advising intensity. A potential explanation for this direct effect is that firms in high-social-capital counties are more likely to trust and value external advice, resulting in an advising-intensive board. However, we also show that the community social capital has an indirect effect on board advising intensity through its effect on board monitoring intensity. In this Internet Appendix Item IB.4, we test if trust can affect agency issues to support the agency theory perspective of our results (the indirect path).
Since voting in presidential elections and participating in census surveys are voluntary activities without direct material incentives, Pvote and Pespn can reflect the strength of cooperative norms (Alesina and La Ferrara, 2000). Assn and Nccs, on the other hand, measure repeated and face-to-face interactions that are likely to strengthen the network-associated norms . Following Hasan et al. (2017b), we measure the strength of Cooperative_Norms (Social_Networks) as the first principal component of Pvote and Respn (Assn and Nccs) in a county. In Table IB.4, we then examine the effect of the social norms and networks on discretionary accruals, CEO compensation, and costs of equity as in   while social networks facilitate efficient information sharing and better communication (Payne et al., 2011, Lin, 1999, the results suggest that both trust and networks can significantly reduce the firm agency problems, supporting our premise and the agency theory perspective of community social capital. Thus, although trust may directly affect board advising, it may also indirectly affect board advising through reducing the firm's monitoring needs.

Table IB.4. The Relative Effects of Cooperative Norms and Social Networks on Agency Issues
This table presents the regression analysis of the relative influence of cooperative norms and social networks on agency issues. The dependent variable in Column (1) is Discretionary_Accruals, measured as the absolute residual from modified Jones model. The dependent variable in Column (2) is CEO_Total_Compensation, measured as the natural logarithm of CEO total compensation reported in Execucomp. The dependent variable in Column (3) is firm Cost_of_Equity, estimated from the modified PEG ratio model (Easton, 2004). Corporative_Norms is from the principal factor analysis based on Pvote and Respn, while Social_Networks is from the principal factor analysis on Assn and Nccs. Detailed variable definitions are given in Table A1. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in the parentheses and are clustered at the county level to control for potential correlation in the error terms. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. (1)

Controlling for Director Network Size
We have shown that community social capital can directly affect board advising intensity as higher levels of community social capital may help directors to build relational capital (i.e. the network of ties to external contingencies). A related concern here is that the positive relationship between community social capital and board advising intensity could be driven by directors' relational capital, because the firm's directors have a greater network size and can find independent directors to serve on advisory committees at lower costs. Although we have controlled for the supply of directors using Local_Director_Pool, we address this concern further by controlling for Director_Network_Size, measured as the total number of overlaps for all directors of the firm through employment, education and other activities. Results reported in Table IB.5 show that a larger director network size is positively associated with board advising intensity. More importantly, the coefficient on SC_Index remains positive (0.091) and highly significant (p<0.000), suggesting that our results are not driven by a dense network related to a large director network size.  (3) are Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one.
Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. SC_Index is the county-level social capital measure based on data from the NRCRD. Director_Network_Size is the natural logarithm of the total number of overlaps for all directors of the firm through employment, education and other activities. Column (1) uses the probit model, and Columns (2) and (3) are OLS models. Detailed variable definitions are given in Table A1 in the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively. (1)

The Role of Religiosity
We also test whether risk aversion correlated with religion can explain our findings. Prior research recognises that religious individuals are more risk-averse (Miller, 2000, and managers of firms in religious areas are less likely to misbehave (McGuire et al., 2011).
Therefore, it might be the case that risk-averse managers in high religious counties engage in fewer opportunistic behaviours, allowing shareholders to assemble an advising-intensive board.
In addition, research on board and religion shows that a specific religious supervisory board imposes strong governance and restrains risk-taking behaviours (Mollah et al., 2021). Our results can then be driven by religious monitoring board members in high social capital areas that mitigate opportunistic risk-taking, hence, allowing more advising.
Although we have controlled for the strength of religiosity of the county with the variable Religiosity (measured as the percentage of residents in a county that adheres to organised religions) in our regressions, we address this concern more explicitly in Internet Appendix IB.6. We use the dummy variable, High_Religiosity, indicating counties with an above-median value of Religiosity. We then interact SC_Index with High_Religiosity in our regressions when examining board advising intensity. However, in Columns (1) to (3) of Table   IB.6, we find that the coefficients on High_Religiosity and SC_Index*High_Religiosity are statistically insignificant, while coefficients on SC_Index remain positive and significant. Thus, our results are unlikely to be explained by the county-level risk aversion that is related to the religion of the county members.  (1) to (3) is Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one.
Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. Column (1) uses the probit models, and Columns (2) and (3) use OLS models. SC_Index is the countylevel social capital measure based on data from the NRCRD. High_Religiosity is a dummy variable that equals one if the county's Religiosity is above the sample median, and zero otherwise. Each column includes the same set of control variables as in Table 2, year and 2-digit SIC dummies. Detailed variable definitions are given in Table A1. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. The standard errors are presented in the parentheses and are clustered at the county level to control for potential correlation in the error terms. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.
(1)  Table 7 because  perceives that the best single predictor of US social capital is the distance to the Canadian border. Here in Table   IB.7, we adopt an additional instrument for community social capital in the IV approach.  and Knack and Keefer (1997) argue that people are less likely to trust each other when they belong to different races. Alesina and La Ferrara (2000) model that more homogenous communities witness higher levels of social interactions, which enhances social capital. Indeed, Alesina and La Ferrara (2000) and Putnam (2007) provide evidence that social capital is lower in more racially and ethnically fragmented communities. These findings suggest that the racial heterogeneity of a county can affect its level of social capital. However, it is unlikely that racial diversity is correlated with the appointment of directors because directors are assigned to monitoring or advising committees based on their skills (Bhagat andBlack, 1999, Faleye et al., 2011), rather than because of their race or ethnicity.
In light of the above discussions, we follow  and Hasan et al. (2017b) to adopt the measure of racial heterogeneity as our second instrument for social capital. We calculate the Racial Herfindahl Index across the ethnic categories reported in the US Census Bureau and adopted by Hasan et al. (2017b): Non-Hispanic White, Non-Hispanic Black or African American, Asian and Hispanic. Race_Diversity is measured as one minus the Racial Herfindahl Index. Therefore, a higher value of Race_Diversity represents a higher level of racial fragmentation. Column (1) of Table IB.7 reports the first-stage result of the IV approach.
As expected, we find a negative association between Race_Diversity and SC_Index. Columns (2) to (4) of Table IB.7 present the second-stage results on the three board advising intensity measures. The results are consistent with our baseline analysis in Table 2 and the IV analysis   in Table 7, as community social capital significantly increases board advising intensity. These results further accentuate the endogeneity concerns of our study.  (2) through (4) are Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one.
Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. Detailed variable definitions are given in Table A1 in the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.

Generalised Method of Moments (GMM) Model
We have addressed the endogeneity concern in our study with IV, PSM, and DiD analysis. We provide further results in Table IB.8 with the generalised method of moments (GMM) model.
Although GMM is a powerful tool to address endogeneity concerns, the GMM estimator for a probit model with an endogenous regressor is not consistent (Dagenais, 1999, Lucchetti, 2002.
We, therefore, only examine the effect of community social capital on the number of advisory committees and the allocation of advisory directors with the GMM model. In the GMM model, we treat all independent variables as potentially endogenous variables, and the first differences of these variables are lagged twice as additional instruments in the level equations of the GMM system. Results in Table IB.8 show that despite the significant coefficient on the lagged dependent variable, the coefficient on SC_Index remains positive and statistically significant.

Table IB.8. Generalised Method of Moments (GMM) Model
The table presents the effect of social capital on board advising intensity using the generalised method of moments (GMM) models. The dependent variables in Columns (1) and (2) are N_Advisory_Committee and Advisory_Director_Ratio, respectively. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one. Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. SC_Index is the county-level social capital measure based on data from the NRCRD. All independent variables are treated as potentially endogenous variables, and the first-differences of these variables are lagged twice as additional instruments in the level equations of the GMM system. Detailed variable definitions are given in Table A1 in the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively. Item IB.9

Directors
We test the sensitivity of our results by using alternative proxies for social capital and advisory director in Table IB.9. In Columns (1), (3) and (5), we follow Jha and Chen (2014) (2), (4) and (6) (1) is Advisory_Director_Ratio_Non-Monitor, which is the number of the alternatively defined advisory directors, scaled by the total number of independent directors. In addition, it is argued that inside directors may have more firm-specific knowledge and can be a valuable source of information for advising activities. Hence, in our second alternative definition of advisory director, we perceive both independent and inside directors (excluding the CEO) that serve on advising committees but do not serve on any monitoring committee as advisory directors.
Advisory_Director_Ratio_All in Column (2) is computed as the number of the inside and independent advisory directors to the total number of board directors, excluding the CEO. In both columns of Panel B, we find that SC_Index remains positively related to our alternative measures of advisory director ratios, suggesting that our findings are not sensitive to the choice of advisory director measure.  1997, 2005, 2009, and 2014. Organ_Donation is the annual data on the number of total donors of all organ types scaled by the total population in the state. The dependent variables in Columns (1) through (3) are Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one. Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. Columns (1) and (2) use the probit model, and Columns (3) through (6) use OLS models. Panel B presents the results using two alternative measures of the advisory director ratio. Advisory_Director_Ratio_Non-Monitor in Column (1) is the number of independent directors who do not sit on monitoring committees, scaled by the total number of independent directors. Advisory_Director_Ratio_All in Column (2) is the number of the inside and independent directors who serve on advisory committees, scaled by the total number of board directors, excluding the CEO. SC_Index is the county-level social capital measure based on data from the NRCRD. Each column includes the same set of control variables as in Table 2, year and 2-digit SIC dummies. Detailed variable definitions are given in Table  A1 in the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.

Poisson and Negative Binomial Model
We use the log transformation of the number of advising committees and adopt the OLS model in Column (2) of Table 2 to test the effect of community social capital on the number of advising committees. Whilst this aids interpretation, the concern is that if the number of advising committees is a non-negative integer value, OLS may not be appropriate. Therefore, in Table IB.10, we estimate the Poisson and negative binomial models when the dependent variable is the number of advising committees without the logarithm. The results in Table IB.10 corroborate our OLS-based results by continuing to show a positive relationship between community social capital and the number of advising committees in the board.  (1) and (2) is the integer number of advisory committees within a firm in a given year. SC_Index is the county-level social capital measure based on data from the NRCRD. Column (1) estimates the Poisson model, and Column (2) estimates the negative binomial model. Each column includes year and 2-digit SIC dummies. Detailed variable definitions are given in Table A1 in the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively. (1)

Principal Factors for Independent Variables
We have a large number of control variables in our models. Concerns of multicollinearity arise if some independent variables are highly correlated. Although our correlation matrix in Table   IB.1 does not reveal significantly high correlations among our variables, we replace our independent variables with their principal factors to increase the power of regression analysis and circumnavigate multicollinearity concerns . Specifically, Complexity is the first principal factor of Firm_Size, Firm_Age, Leverage and N_Segments.
Information_Costs is the first principal factor of Market-to-Book, R&D and Return_Volatility.
CEO_Entrenchment is the first principal factor of CEO_Tenure, CEO_Ownership and  Table IB.11. The coefficients on SC_Index continue to be positive and significant at 5% or better.  (1) through (3) are Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one.
Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. SC_Index is the county-level social capital measure based on data from the NRCRD. Complexity is the first principal factor of Firm_Size, Firm_Age Leverage, and N_Segments. Information_Costs is the first principal factor of Market-to-Book, R&D and Return_Volatility. CEO_Entrenchment is the first principal factor of CEO_Tenure, CEO_Ownership and CEO_Duality. Governance_Structure is the first principal factor of Institutional_Ownership, Blockholder_Ownership and Board_Independence. Geographic_Factor is the first principal factor of Local_Director_Pool, Per_Capita_Income Population_Growth, Population_Density, Education, Religiosity and County_Median_Age. Column (1) uses the probit model, and Columns (2) and (3) are OLS models. Detailed variable definitions are given in Table A1 in the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.

The Relative Effects of Cooperative Norms and Social Networks
In our analysis, we do not distinguish between cooperative norms and social networks in our definition of community social capital. In Table IB.12 we test the possibility that cooperative norms and social networks may have distinct effects on board advising intensity.
We replace SC_Index with the two separate measures and present the results in Table   IB.12. We show that the coefficients on both Cooperative_Norms and Social_Networks are positive and statistically significant. We test the difference between the coefficients on Cooperative_Norms and Social_Networks by using the chi-square test in Column (1) and the F-test in Columns (2) and (3). The results do not reveal significant differences between the two coefficients in all columns, suggesting that both the cooperative norms and dense social networks within a county can lead to increased board advising intensity.

Table IB.12. The Relative Effects of Cooperative Norms and Social Networks
This table presents the regression analysis of the relative influence of cooperative norms and social networks on board advising intensity. The dependent variables in Columns (1) through (3) are Advisory_Committee, N_Advisory_Committee and Advisory_Director_Ratio, respectively. Advisory_Committee is a dummy variable that equals one if the firm sets up at least one specialised advisory committee, and zero otherwise. N_Advisory_Committee is the natural logarithm of the number of advisory committees within the board in a given year plus one. Advisory_Director_Ratio is the number of advisory directors scaled by the total number of independent directors. Corporative_Norms is from the principal factor analysis based on Pvote and Respn, while Social_Networks is from the principal factor analysis on Assn and Nccs. Column (1) uses the probit model, and Columns (2) and (3) are OLS models. Detailed variable definitions are given in Table A1 in the Appendix. All continuous variables are winsorised at the 1st and the 99th percentiles to eliminate the influence of outliers. Each column includes year and 2-digit SIC dummies. The standard errors are presented in parentheses and are clustered at the county level to control for potential correlation in the error terms. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively. (1)