Family Monitoring the Family

Transcription

1 Family Monitoring the Family Joseph P.H. Fan Department of Finance & School of Accountancy The Chinese University of Hong Kong Xin Yu Business School The University of Queensland St Lucia, Brisbane, QLD, Australia, January 2018 * We thank conference participants at the 2016 MIT Asia Conference in Accounting, the Centre for Asian Business and Economics (The University of Melbourne) 3 rd Annual Conference and seminar participants at Massy University, Nanyang Technological University, Shanghai University of Finance and Economics, and University of Queensland. Xin Yu acknowledges financial support from the National Natural Science Foundation of China (Grant No ).

2 Family Monitoring the Family Abstract This study examines how founding family participation in firm ownership and management shapes related-party transactions that concern public investors in more than 1,200 Chinese publicly traded private sector firms. Surprisingly, we find that firms with more family member participation engage in fewer abnormal related-party transactions that are suspicious with respect to expropriation, suggesting a potential monitoring role by firm founding family members. Such effects are stronger in stocks that are thinly traded and followed by few analysts, suggesting a substitutional effect of family governance for weak market governance. Moreover, seniority and closeness of family relationships matter to the strength of family monitoring. The family monitoring effects are stronger when more senior or distantly related family members participate in the firms and weaker when more children of the founders participate. The family members role as owners and/or managers also matters. Shareholding family managers are associated with fewer suspicious related-party transactions than are family managers without shares and family owners who do not act as managers. Overall, this study s evidence supports the view that the checks and balances among founding family members benefit public investors, particularly when market governance is weak in enforcing investor rights. Keywords: Family firms, Family governance, Family monitoring, Related-party transactions, China JEL Classification: G32, G34 1

3 1. Introduction Founding family members long-term participation in the ownership and management of firms is common around the world. To sustain this model, founding families provide unique contributions to their firms, making the firms competitive (Bennedsen et al., 2015). However, founding families decisions are often criticized as self-serving and made at the expense of public investors (e.g., Anderson et al., 2012; Faccio et al., 2001; Faccio et al., 2010; Lins et al., 2013; Liu et al., 2015). This paper examines whether founding family involvement in firms hurts or enhances corporate governance from public investors point of view. Economics and finance research typically treats the family behind the firm as a monolithic entity. However, family members roles in business differ and their interests are not always aligned (Tagiuri and Davis, 1996). When more family members participate in the business as owners and/or managers, the potential for conflicts increases. If not well contained, the family members conflicts will destroy firm value and hurt not only the family s but also public investors interests (Bertrand et al., 2008; Cheng et al., 2015; Miller et al., 2007). Ex ante, a family can engage governance mechanisms (Anderson and Reeb, 2004) to mitigate the potential conflicts and/or exercise mutual monitoring (Combs et al., 2010; Fama and Jensen, 1983) if the gain from such mechanisms is larger than the cost. As more family members become involved in the business, their divergence in objectives and associated conflict potential increase, and the potential benefits of strengthening governance that limits any family member s rent extraction also increase. Therefore, family participation and associated family governance protect firm value and benefit not just the family owners but also the minority shareholders, a beneficial spillover effect. However, an alternative and perhaps dominating view in the literature is that founding family participation per se is induced by rent extraction to some degree (Almeida and 2

4 Wolfenzon, 2006; Morck et al., 2005). When the family believes its self-interest is above the interests of other firm stakeholders, family members may not refrain from exploiting firm assets through their positions as owners and/or managers. To address this controversy, we investigate how monitoring among family members shapes firms related-party transaction decisions that are prone to conflict with minority shareholders. To be consistent with the positive governance spillover view, more family participation in firm ownership and management should be associated with higher quality firm governance traits, such as lower levels of abnormal related-party dealings, all else being equal. In contrast, if founding family involvement in the business is a symptom of weak family governance and rent extraction, more family participation should be associated with more pervasive suspicious related-party transactions. We manually collect data about the extent of founding family participation in ownership and management for 1,242 publicly traded private sector firms in China. As China s private sector opened up only about 30 years ago, firms are generally owned and managed by founders and their family members. For idiosyncratic reasons, the firms ownership and decision rights are distributed to different degrees among founding family members. Also, China s legal systems and capital markets are emerging, enabling us to investigate whether the importance of the family firm organization is related to weak market governance (Burkart et al., 2003). Consistent with the family monitoring spillover effects, we find that founding family participation is associated with fewer abnormal related-party transactions that are suspicious with respect to expropriation. Moreover, we find that the association between founding family participation and abnormal related-party transactions is more pronounced among firms subject to weaker market discipline, proxied by their lower stock trading volumes and fewer financial analysts following them. The evidence is consistent with the view that founding family participation lends credibility to firms constrained by weak market governance. 3

5 We examine the impact of family monitoring on abnormal related-party transactions using a quasi-natural experiment involving exogenous shocks to corporate governance. During our sample period, a change in government regulation occurred that aimed to curtail related-party loans of publicly traded companies. Our tests show that family participation curtails relatedparty loans pre-regulation, but the effects disappear after the regulation s implementation, suggesting that founding family participation in firms has been an important private enforcement mechanism that substitutes for the prior weak public governance. We further investigate whether and how the strength of family monitoring varies with family members position in the family hierarchy and relationship distance in the family. By cultural tradition and economic influences, the decisions of a founder are more likely to be questioned by senior family members, such as parents and siblings, than by junior family members such as children. Also, as family relationship distance increases, family goals and values diverge and trust dissipates. Hence, to safeguard their interests, more distant relatives likely have higher incentive to monitor the firms in which they have ownership or managerial positions than close family members such as the founders spouses and children. As expected, we find higher family monitoring effects as more siblings and parents of founders participate in the firms, while more children participation weakens family monitoring. While participation by the founder s spouse is not associated with abnormal related-party dealings, the participation by more distant relatives of the founder is associated with substantially fewer suspicious transactions. In addition to family position, the role of family members in firms matters to the strength of family monitoring. We find that suspicious related-party transactions are fewer when more family members are both owners and managers. However, family owners without managerial roles have little effect on abnormal related-party transactions, indicating that family owner managers have stronger ability in monitoring each other than family owners who do not act as 4

6 managers. Furthermore, the negative association between family participation and suspicious related-party transactions is more pronounced when family members are both owners and managers than when family members are managers and/or directors but not share owners, indicating that family owner managers have stronger incentive to monitor each other than family managers who do not own shares. Our key findings are robust to a host of diagnostic tests and several approaches to addressing endogeneity concerns, including difference-in-difference (DID) matching analysis and instrumental variable (IV) estimations. In the DID matching analysis, we identify observations with changes in family participation (treated group) and match them with observations without changes (matched group). Then we examine changes in abnormal relatedparty transactions from two years before the change in family participation to two years after the change. We find that more (less) family participation decreases (increases) abnormal related-party transactions. Moreover, we find similar family monitoring effects in dividend policy, another firm decision that is prone to conflict with and concern from public investors. This paper makes several contributions to the literature. First, it demonstrates family monitoring as a specialized input whose cost is potentially lower than formal enforcement mechanisms (Alchian and Demsetz, 1972; Burkart et al., 2003). Hence, it adds to the growing body of literature about the comparative advantages of the family firm organization, including moderating external financing constraints (Almeida and Wolfenzon, 2006; Masulis et al., 2011), human capital acquisition (Bhattacharya and Ravikumar, 2001), specialized values preservation (Bennedsen et al., 2015; Demsetz and Lehn, 1985), reputation building in economic and political markets (Khanna and Palepu, 2000; Morck and Yeung, 2004), management with a long-term perspective (Anderson and Reeb, 2003; D Aurizio et al., 2015), and trust among family members (Ferguson, 1998; Stacchini and Degasperi, 2015). Second, the weak legal system and public enforcement of property rights in China make 5

7 controlling owners rent extraction a serious problem (Jiang et al., 2010). This paper demonstrates that family governance can be a private mechanism against insiders rent extraction. Thereby, the paper extends the literature that largely focuses on formal enforcement (e.g., Burkart et al., 2003; Djankov et al., 2008; Doidge et al., 2009; Dyck and Zingales, 2004; Gopalan and Jayaraman, 2011; Leuz et al., 2003). Third, while most prior studies have treated a family as a homogenous group, this study allows for different roles and incentives of family members in the family and the firm. We investigate family members different monitoring incentives and abilities as indicated by their different positions in the family hierarchy and their different degrees of participation in firm ownership and management. This is an important first step toward understanding how and why family firms differ from firms that do not involve founding families. The remainder of this paper proceeds as follows. Section 2 develops the hypotheses. Section 3 presents the sample, variable patterns, and univariate test results. Section 4 reports multivariate regression results and Section 5 concludes the paper. 2. Hypothesis Development Conflicts between controlling and minority shareholders and their negative effects on corporate governance are well documented in global corporate governance research (Claessens et al., 2000; Faccio and Lang, 2002; Gopalan and Jayaraman, 2011; La Porta et al., 1999; Lins et al., 2013). As the majority shareholders of many publicly traded companies, founding families are often criticized and held responsible for the weak governance (Bertrand and Schoar, 2006; Morck et al., 2005). The rent extraction view of the controlling families is based on the assumption that they are monolithic groups of people with identical interests. However, in fact, family members have diverse interests and their roles in firm governance are typically unclear in the literature. 6

8 2.1. Governance Role of Family Family has long been recognized as the oldest and most fundamental economic organization. Economists view the family organization as a nexus of implicit contracts that delineate rights and responsibilities of family members to facilitate family cooperation and achieve productivity (Becker, 1981; Bunkanwanicha et al., 2013). Family contracts are typically enforced by customized internal governance mechanisms while depending little on formal laws (Mehrotra et al., 2013; Stacchini and Degasperi, 2015; Yan and Sorenson, 2006). Like any human organization, the family is subject to the problem of collective action (Olson, 1965). Sharing a common pool of resources, the family s productivity is constrained by its members divergence of interests, shirking, and free-rider problems (Demsetz, 1964, 1967). As family members grow in number, the problems loom larger. Efficiency enhancement dictates that some family members will be given decision rights and some incentive and monitoring mechanisms will be in place to ensure that their decisions are aligned with those of other family members. 1 Violation of the family contracts will result in sanctions that are sufficiently severe to deter the violation ex ante (Cheung, 1972; Williamson, 1983). The governance mechanisms in a family contract can include a family ownership structure to induce incentives (Cheng et al., 2015), a family committee to make family decisions (Villalonga and Amit, 2009), multiple family representation in management or the board for mutual monitoring (Combs et al., 2010; Fama and Jensen, 1983), as well as informal mechanisms such as family loyalty and trust to share family goals and values (Stacchini and Degasperi, 2015; Whyte, 1996). The family s internal governance can spill over to influence firm governance and public investors. For example, the quality of information required for 1 Due to various differences such as personality, upbringing, and position in the family, family members are not a coherent group of people by nature. On the other hand, division of labor and delegation of decisions are required to improve productivity. Of course, the decision rights allocation in a family is affected not only by economic reasons but also by non-economic reasons such as culture, altruism, and emotion. Whatever the reason, as long as a division of labor occurs in a family, the family has an incentive to engage governance mechanisms to mitigate family members conflicts of interest. 7

9 effectively monitoring managerial decisions is higher among family members than outsiders, not only because the family members live close by but also because they are committed to long-term relationships that depend on effective communication and cooperation (Schulze et al., 2003). Although as outsiders we are unable to observe the internal family governance in action, we can observe the roles and positions of family members in the firm and the family and make inferences about the effects of family governance on firm governance. To be consistent with the family governance view, we should observe firm decisions that vary with the degree of family participation in the firm. Putting this into the context of corporate governance, we examine firm related-party transactions as a class of managerial decisions that are prone to wealth transfer and therefore of concern to public investors (Bertrand et al., 2002; Cheung et al., 2006; Cheung et al., 2009; Jia et al., 2013; Jiang et al., 2010; Peng et al., 2011). If family participation enhances the monitoring of managerial decisions, we should find a low level of firm related-party activities that are likely to involve expropriation of minority shareholders, all else being equal. Formally stated, the family monitoring hypothesis predicts: H1: More founding family members participating in the firm is associated with fewer related-party transactions that are suspicious with respect to expropriation. In contrast, if rent extraction attracts family participation in the firm, we should find intensive related-party transactions associated with the family participation Market Governance Researchers have argued that family firms are organizations that substitute for weak formal institutions (Burkart et al., 2003; La Porta et al., 1999; Stacchini and Degasperi, 2015). In environments in which property rights are only weakly protected by laws, alternative institutions may arise to provide private enforcement. Where laws and formal enforcement are 8

10 weak in protecting firm contracts, enforcement and monitoring by founding families is valuable. In our context, the founding family enforcement manifests in participation in firm ownership and management. To evaluate the substitution role of the family firm organization, we examine the strength of capital market governance in the monitoring effects of founding families. We focus on a specific stock investor concern: the amount and quality of information about publicly traded firms. The information factor is influenced by numerous parties, including the firms, investors, intermediaries, and regulators. We call this market governance, the monitoring and enforcement provided by the various market participants as a whole. We expect the positive spillover effects of founding family monitoring to be stronger in firms subject to weak market governance, and vice versa. The related hypothesis is: H2: The negative association between founding family participation and related-party transactions that are suspicious with respect to expropriation is stronger in firms subject to weaker market governance. Alternatively, if weak market governance facilitates the controlling family s expropriation, we should find more problematic related-party transactions associated with family participation in such markets. 2.3 Family Positions and Relationship Distance Influenced by Confucian culture, Chinese family relationships are typically governed by hierarchical authority structures. Therefore, the founders power comes not only from their concentrated ownership and economic contribution to the family and the firm, but also from their seniority in the family hierarchy. Therefore, the founders decisions are more likely to be contested by their seniors and siblings than by their spouse and junior family members such as children. In addition to seniority, relationship distance may matter to the strength of family 9

11 monitoring. The degree of goal sharing and trust among distantly related family members is likely weaker than with close family members. The larger conflict potential of more distant relatives suggests their higher incentive to monitor the firm when they own shares or serve as managers of the firm. Stated formally: H3: More senior or distant family members in relation to the founder participating in the firm is associated with fewer abnormal related-party transactions Roles of Family Members in Firms The strength of family monitoring is likely affected by both the incentive and the ability of family members to monitor the business. We investigate these potential effects by examining family members ownership and managerial roles in the firm. 2 Family owners generally have stronger incentives to monitor their delegated managers than family members without shares. On the other hand, family managers are more effective monitors of their peers than family members not serving in any managerial role because the former have more expertise and better information about the firm than the latter. However, family managers may not have strong incentives to monitor if they are not significant owners. To facilitate our hypotheses development and empirical tests, we classify firm participating family members into two categories: (1) family owners and (2) family members who work as managers and/or directors (MD) in the firm but do not own shares. 3 Furthermore, we classify family owners into two types: family owners who also serve as senior managers and family owners who are not senior managers. 4 Then, we formally state the following 2 Family members may have other roles with the firm, such as non-managerial employee or business partner. We focus on the ownership and management roles as they are more relevant proxies for monitoring incentive and ability. 3 In addition to managers and directors, we include members of the supervisory board in subsequent empirical analyses. For convenience, we omit supervisory board members in the text. 4 Likewise, we classify family members who are MD but do not own shares into two sub-groups: family members as senior managers but not owning shares and family members as directors but not owning shares. As the size of these sub-samples is small, we omit this analysis for simplicity. 10

12 hypotheses: H4_1: Firms with more family owners engage in fewer suspicious related-party transactions than firms with fewer family owners. H4_2: The negative association between family owners and suspicious related-party transactions is stronger when the family owners are also senior managers, compared with when family owners do not hold managerial roles. The net effect of participation by non-shareholding family MD is not apparent. On the one hand, they have the ability to monitor peer family managers. On the other hand, they may lack the ownership incentive to monitor or even engage in problematic transactions themselves since they do not bear significant firm cash flow losses associated with the transactions. In contrast, it may be that family managers, even without share ownership, are willing to provide checks and balances because they are motivated by alternative family governance mechanisms, as discussed above. We leave this as an empirical issue. However, compared with family owners who are also senior managers, who have both incentive and ability to monitor their delegated managers, non-shareholding family MD should have weaker incentives to monitor. Therefore, we state the following hypothesis: H4_3: The negative association between family participation and suspicious related-party transactions is stronger when the family owners are also senior managers, compared with when family members work as MD but do not own shares. 3. Sample, Regression Models, and Basic Statistics This section describes the sampling process, regression models, key variables, and basic statistics Sample Selection and Patterns of Family Participation The history of private sector firms in China dates back to 1978, when the country s 11

13 economic reforms and opening-up policy began. Similar to firms in other East Asian markets (Claessens et al., 2000), private sector firms in China are nearly all owned and/or managed by founders and their family members. 5 The sample used in this study covers private sector firms that are publicly traded in China s A-share markets from 2002 through The sampling year begins in 2002 since few private sector firms are publicly listed before then. Table 1 summarizes the sample selection process. We start with 7,331 firm-year observations from 1,534 firms obtained from the Database of Private Sector Listed Firms, a sub-database of the China Center for Economic Research (CCER) Economic and Financial Database provided by SinoFin Information Services. Shareholders holding 5% or more shares of publicly traded firms are subject to mandatory disclosure in China and reported in the database. Based on the initial sample, we exclude observations whose financial data are subject to frequent controlling ownership turnovers and financial distress. Thus, we exclude 168 observations associated with firms experiencing two or more changes in controlling owners while being given special treatment status by the Chinese security regulator for more than half of the sample period. 6 To focus on a sample of firms influenced by founding families, we further exclude a small number (317) of observations associated with firms whose ultimate controlling owner is a company, a group of employees, or a group of co-founders who have no family relationship. The 1,509 observations associated with initial public offerings (IPOs) or privatization years are excluded because some key measures (e.g., the measures on related-party transactions and the measure on market governance) in the IPO year or privatization year may be distorted. 7 In addition, we exclude 73 observations whose financial, 5 A variety of definitions exists for family firms (Miller et al., 2007). The Chinese firms in our sample can be classified as family firms by most definitions, as they are generally owned and actively managed by founders and/or their families. 6 Special treatment (ST) is a unique concept of China s stock market applied since April 22, A firm is labeled ST if its net income is negative for two consecutive years or if its net assets per share in the most recent year are lower than its par value. The daily stock price of an ST firm is limited to the range of ±5%, and interim reports must be audited. Firms that are given special treatment status and experiencing controlling owner transfer are financially distressed firms. 7 The privatization can occur any day during a year and the privatized firm must have been in the family s control for less 12

14 ownership, or stock trading data are missing from the China Securities Market and Accounting Research (CSMAR) database. The final sample includes 5,264 firm-year observations from 1,242 firms. [Insert Table 1 here] Data on family member participation in firms are manually collected from IPO prospectuses and annual reports. From these sources, family relationships with firm founders are identified using the following criteria. First, for private sector firms that obtain public listing status through IPOs, we collect information about family relationships among owners and senior managers from the IPO prospectuses in which the information is reported as required by regulators. Second, for observations associated with firms whose controlling ownership changes after an IPO, we collect family relationship information from share transfer announcements. For each of the sample firms, we monitor entry and exit of family members by tracing turnover of owners, board members, and senior managers that are disclosed in corporate annual reports. As most of the firms do not declare family relationships when new family members enter the firms, we employ search engines (Google and Baidu) to identify any family relationships for each of the new entrants from various public information sources. We may miss the family relationship of a new entrant if it is not mentioned in any available source. This would cause noise in our proxies for family participation and bias against finding statistically significant results. Table 2 provides a description of the sample. As shown in Panel A, the number of firmyear observations increases over the sample period, consistent with the increasing prevalence of private sector firms in China. The average number of family members involved in business in the sample period is 2.22 and this figure increases monotonically from 1.61 in 2002 to 2.57 than a year. Therefore, the related-party transactions in the privatization year are not comparable to those in other years under the family s control. The investors trading behavior for the IPO firms also differs from that for other firms, making trading volume, a measure of market governance, not comparable to other observations. 13

15 in As described in the previous section, we decompose participating family members based on their relationship to the founders. Specifically, we classify firm participating family members into five categories based on their relationships with the founder: (1) siblings, which includes the founder s and the founder s spouse s siblings and their spouses, (2) parents, which includes parents of the founder, (3) children, which includes the founder s children and their spouses, (4) spouse, which is the spouse of the founder, and (5) extended, which includes extended family members such as uncles, cousins, nephews, and nieces. We use a dummy variable, Founder, to identify whether the founder is participating in the family business. The mean value of Founder is 99.3%, suggesting that almost all the observations in our sample are owned and/or managed by founders. On average, siblings, parents, children, and extended relatives of founders participate in the family firms. The mean value of Spouse is Since Spouse is a dummy variable that equals 1 if the founder s spouse is involved in the firm and 0 otherwise, the number suggests that founders spouses are involved in almost 20% of the firm-year observations. The degree of family participation increases over time across all the family relationship categories. In Panel B, we decompose participating family members based on their role in the family business. We first decompose them into two groups: Family owners and MD not owners; we further decompose the former family owner group into two sub-groups: Owners managers and Owners not managers. Of the family members participating in the firms, on average, 2.09 are family owners and 0.13 are managers or directors who do not own shares. The average number of family owners who are also managers is 0.38, while the average number of family owners who are not managers is With the exception of MD not owners, all the categories of family participation demonstrate increasing trends over time, suggesting increasing family participation in both ownership and management of publicly listed companies in China. 14

16 In Panel C, the majority (almost 60%) of the 1,242 firms in the sample has multiple family members participating in the family firms, with almost 37% having more than two members, 21% having more than three members, 12% having more than four members, and 8% having more than five members. On the other hand, 40% of the sample firms is associated with a single controlling owner but no other family members. [Insert Table 2 here] These above patterns demonstrate significant and increasing ownership and managerial participation in Chinese publicly traded firms by a variety of family members, in addition to firm founders, within only the short history of the capital market and private sector development Measuring Abnormal Related-Party Transactions and Regression Models Although related-party transactions are prone to the minority shareholder expropriation problem, many related-party trades can be induced for transaction cost reduction reasons (Fisman and Wang, 2010; Khanna and Yafeh, 2005) or for controlling shareholders who prop up the earnings of publicly listed firms when the listed firms have poor performance (Jian and Wong, 2010; Jia et al., 2013; Peng et al., 2011). It is not easy to disentangle abnormal and normal transactions between related parties, especially when publicly traded firms are affiliated with business groups (Fan et al., 2016; He et al., 2013). Cheung et al. (2006) classify related-party transactions into three categories, including transactions that are suspicious with respect to expropriation, transactions that are likely to benefit the firm, and transactions made for strategic purposes. Following their classification, transactions that are suspicious with respect to expropriation encompass several types of related-party transactions, including asset transactions, product trades, equity sales, and cash payments. A related-party asset transaction is the acquisition of assets by a listed company 15

17 from a connected party or the sale of assets from the listed company to a connected party. A related-party product trade is the trade of goods and services between the listed company and a connected party. A related-party equity sale is the sale of an equity stake in the listed company to a connected party. A related-party cash payment involves a direct cash payment by the listed company to a connected party. The transactions between a listed company and its subsidiaries and the transactions of related-party cash receipts are excluded because they are likely to benefit the listed firms, while transactions of equity purchases are excluded because they are more likely made for strategic purposes (Cheung et al., 2006). The result is several types of related-party transactions between the listed company and a connected person or a company controlled by the connected person that are prone to the expropriation problem. Therefore, we use this amount scaled by total assets (RPT_tunnel) to capture related-party transactions that are suspicious with respect to expropriation. Alternatively, we employ related-party loans as another measure of abnormal transactions. Jiang et al. (2010) document that inter-corporate loans have been used pervasively by controlling shareholders in China to transfer wealth out of publicly listed companies. Fisman and Wang (2010) and Jia et al. (2013) show that controlling shareholders of Chinese business groups transfer financial resources out of the publicly listed firms through inter-corporate loans when they experience a credit crunch. Following Jiang et al. (2010), related-party loans (Inter_loan) are estimated by other receivables deflated by total assets. We estimate the following equation to address whether a firm s abnormal related-party transaction level is sensitive to the extent to which founding family members participate in the firm: AbnormalRPT = α + α Familymembers + α ( V C + α Debt + α LagROA + α Sucession + α Cofounder + α RegionalGDP ) + Yeardummies + Industrydummies + ε α CV + α LogTA (1) where RPT represents abnormal related-party transactions, measured by RPT_tunnel (Cheung 16

18 et al., 2006) and alternatively Inter_loans (Jiang et al., 2010) as defined above. Family members is the number of controlling family members who are firm shareholders, senior managers, board directors, or members of the supervisory board in a given year. We expect α1 to be negative (H1). To separate the effects of family participation from those of ownership structures, we control for firm ultimate ownership structures in the regressions. Specifically, we include the cash flow rights of the ultimate controlling family (CR) to account for the incentive alignment effect of the controlling ownership and the difference between voting rights and cash flow rights of the controlling family (V-C) to capture the incentive and ability associated with the ownership structure to expropriate outside investors (Claessens et al., 2002). Firms with larger cash flow rights are expected to have fewer suspicious related-party transactions because of the alignment effect, while firms with larger divergence between the controlling voting rights and cash flow rights are expected to have more suspicious related-party transactions because of the controlling owners higher incentive and ability to expropriate outside investors. As Claessens et al. (2000) report, the separation of voting and cash flow rights in Asian firms is typically achieved through stock pyramids. As an alternative to the (V-C) variable, we include a dummy variable equal to 1 if a publicly listed firm is associated with and controlled by stock pyramids and 0 otherwise. The regression results based on the alternative pyramid dummy variable are similar. This is expected because the correlation between (V-C) and the pyramid dummy variable is high. Therefore, we do not tabulate the results based on the pyramid dummy variable. In addition, we control for firm size (LogTA), Debt, and ROA in the prior year as Jian and Wong (2010) and Jiang et al. (2010) suggest. We expect the level of related-party transactions to be higher in smaller, lower leveraged, and less profitable firms because these firms are more likely to be subject to conflicts of interest with public shareholders. Furthermore, the level of abnormal related-party transactions may change during and after the leadership succession of 17

19 the founder. To control for potential effects of succession, we include Succession, a dummy variable equal to 1 if a child of the founder takes the position of chairman, chief executive officer (CEO), or board member and 0 otherwise (Xu et al., 2015). The level of abnormal related-party transactions can be constrained by the existence of co-founders. We include in the regression model Cofounder, a dummy variable that equals 1 if there is at least a non-family cofounder and 0 otherwise. We include Regional GDP, the per capita gross domestic product (GDP) of the region (province) where a firm is registered, in the regressions to account for potential effects of regional economic and institutional development. We expect the coefficients of Regional GDP to be negative since abnormal related-party transactions are likely to be less prevalent in more developed regions. We include year and industry 8 dummy variables in the regressions to control for year and industry fixed effects. As we employ panel data, t statistics are based on standard errors adjusted for firm clustering in all regressions Basic Statistics and Univariate Tests Panel A of Table 3 presents the summary statistics of the variables employed in our empirical tests. Definitions of the variables are described in the previous sub-section and summarized in Appendix 1. Statistics of financial variables are based on data winsorized at the top and bottom 1% levels. During the sample period, the firms on average engage in 5.8 renminbi (RMB) of abnormal related-party transactions and 4 RMB of related-party loan transactions for every 100 RMB of total assets. The statistics of family participation are discussed earlier. The cash flow rights held by the largest ultimate shareholder/family is on average 28.7%. The average excess of voting rights over cash flow rights (V-C) is 7.90%. These ownership numbers are generally consistent with prior studies on publicly listed Chinese family firms. For example, Amit et al. 8 Following Guan et al. (2016), we use the two-digit CSRC industry code for the manufacturing sector and the one-digit code for other sectors. We follow the same rules for any industry-related classification. 18

20 (2015) report that the average controlling ownership is 26% with 9% voting control in excess of ownership. The sample observations have total assets of about 2,700 million RMB on average, have 21% of debt in total assets, and earn 4 RMB out of every 100 RMB of operating assets. Almost 12.8% of the sample observations experiences succession to the next generation, and about 10% of the sample observations has co-founders. The average per capita GDP of the firm s region (province) in a given year is 39,470 RMB. Table 3 also reports the basic statistics for variables capturing the strength of market governance, the expected agency costs, instrumental variables, and variables in DID analysis. [Insert Table 3 here] As an initial examination, we conduct univariate analysis by comparing the variables of related-party transactions among three groups of firm-year observations in which one, two, and three or more family members are involved in the business. In Panel B of Table 3, among the total firm-year observations, 2,679 observations are associated with a single family member, 1,021 with two family members, and 1,564 with three or more members. The group with a single family member (Family members=1) has the largest mean values of suspicious related-party transactions (RPT_tunnel and Inter_loans), while the group with three or more family members (Family members>=3) has the smallest mean values of intercorporate loans. The t-statistics between the group Family members=2 and the group Family members=1 and between the group Family members=3 and the group Family members=1 are all significant at the 1% level. Overall, the univariate statistics indicate that more founding family members participating in the firm is associated with fewer abnormal related-party transactions. We also compare firm characteristics among the three family groups differentiated by the extent of family participation. In rows 3 and 4, as more family members get involved in the business, controlling owners have less divergence between voting rights and cash flow rights 19

21 (V-C) but more cash flow rights (CR). Rows 5 through 7 compare financial attributes of firms in the different family groups. Firms with more family members involved in the business are larger with respect to the book value of total assets, as shown in row 5, use less debt, as shown in row 6, and have better accounting performance, measured by net income divided by assets, as shown in row 7. Firms with more family members participating in the business are more likely to have experienced succession to the next generation, as shown in row 8, and are more likely to be located in regions with better economic development, as shown in row 9. Overall, the comparison reveals substantial differences in firm and regional characteristics between the groups of firms differentiated by the level of family participation and suggests that these factors should be accounted for in subsequent regression analyses. The Pearson correlations among key variables are presented in Appendix 2. The two measures of suspicious related-party transactions are negatively correlated with the number of family members involved in the business at the 1% level, consistent with our expectation. Family members is also significantly positively correlated with cash flow rights (CR), but negatively correlated with the divergence between voting rights and cash flow rights (V-C), consistent with the mean comparison in Panel B, Table Regression Results In this section, we report on regression analyses performed to test our hypotheses regarding the role of founding family monitoring in firm governance Baseline Results Table 4, Panel A presents our baseline regression results with model (1), employing abnormal related-party transactions RPT_tunnel as the dependent variable. The estimated coefficient of Family members is negative and significant at the 1% level, regardless of whether the control variables are included (columns 1 and 2). Consistent with H1, having more family 20

22 members participating in the firm is associated with a lower level of abnormal related-party transactions. In terms of economic significance, the coefficient of Family members is in column (2), implying that having an additional family member as an owner or a manager is associated with a decrease in abnormal related-party transactions that are likely to be expropriation. Since the average RPT_tunnel in our sample is 0.057, the economic impact of one additional family member s participation in the firm is an 8.77% decrease in abnormal related-party transactions. [Insert Table 4 here] The estimated coefficients of the control variables are generally consistent with our expectations. The coefficients of V-C are significantly positive at the 10% level, suggesting that firm controlling owners with larger divergence between voting rights and cash flow rights tend to engage in more abnormal related-party transactions that are prone to expropriation. The coefficients of LogTA, Debt, and LagROA are all negative, suggesting that smaller, lower leveraged, and less profitable firms tend to have more abnormal related-party transactions. The significantly negative coefficients of Regional GDP indicate that firms in more developed regions engage in fewer suspicious related-party transactions. 4.2 Regulatory Changes and Family Monitoring In this sub-section, we examine the impact of family monitoring on corporate governance using a quasi-natural experiment involving a regulatory shock that changes the public protection of investor rights. On November 2, 2005, the State Council of China released The Notice of the State Council on Transmitting the Opinions of the China Securities Regulatory Commission on Improving the Quality of Listed Companies (No.34 [2005] of the State Council). The Notice prohibited capital occupation by controlling shareholders and required controlling shareholders to repay occupied capital to publicly listed firms by the end of

23 Otherwise, the top managers of the capital-occupying firms would be punished. Since then, inter-corporate loans of publicly listed firms have been less frequent. We expect that the regulation has enhanced public governance and thereby weakened the benefit of family monitoring. In the baseline regression (1), we add an indicator for the postregulation period and interaction terms between the indicator and Family members. To the extent that the regulatory change is unexpected and exogenous to the family participation in firms, the estimated coefficients of the interaction terms should be positive. Panel B, Table 4 reports the regression results. In column 1, the coefficient of Family members is insignificant. In the full sample period, no significant association exists between family participation and related-party loans. In column 2, PostReg2006 is the indicator variable that equals 1 for years since 2006, when inter-corporate loans are prohibited, and 0 otherwise. The coefficient of Family members is negative and significant at the 1% level, showing that family participation is negatively associated with related-party loans before the regulation change, consistent with the family monitoring hypothesis (H1). The estimated coefficients of PostReg2006 are negative and significant at the 1% level, suggesting that the regulation is effective in mitigating related-party loan activities. Furthermore, the interaction terms between PostReg2006 and Family members is positive and significant at the 1% level, consistent with our expectation that the benefit of family monitoring in curbing inter-corporate loans is smaller after the regulatory change and associated improvement of public governance Market Governance and Family Monitoring We next test H2 regarding the role of family participation in firm governance, conditioning on the strength of capital market governance. We employ firm stock trading volume and the extent to which analysts follow the stock as alternative proxies for market governance. Trading volume has been widely used as a reverse proxy for information asymmetry in prior studies 22

24 (e.g., Anderson et al., 2009; Leuz and Verrecchia, 2000). Following the literature, trading volume is defined as the average daily RMB stock transaction volume of a firm during a year. 9 The alternative proxy for market governance is the number of analysts providing forecasts on earnings per share of a firm nine months prior to the end of a year (Anderson et al., 2009). Analysts following captures the extent of scrutiny by a capital market. We decompose our sample firm-years into two groups based on the level of market governance. Every year, a firm is classified into the high group (having a high level of market governance) if its stock trading volume in the year is above the sample median value and in the low group otherwise. Low trading volume of a firm reflects a low level of market scrutiny and, therefore, private monitoring by the firm s founding family provides an alternative source of governance, and vice versa. Alternatively, we employ analysts following to divide the sample. A firm in a year is subject to weak market governance if the number of analysts following the firm is equal to or below the sample median number in the year; otherwise, the firm is grouped into the strong market governance sub-sample. Table 5 presents the results of the sub-sample regressions. The previously found negative association between Family members and RPT_tunnel is concentrated in the sub-sample of firms subject to weak market governance. The coefficients of Family members are both negative and significant at the 1% level in regression based on the below-median trading volume sub-sample (column 1) and in regression based on the below-median analyst following sub-sample (column 3), while the coefficients of Family members are insignificant in regression based on the above-median sub-samples (columns 2 and 4). The magnitude of estimated coefficients is larger in the sub-sample of firms subject to weak market governance than in the sub-sample of firms subject to stronger market governance, and the difference is significant at the 5% and 1% levels, respectively, based on the seemingly unrelated estimation. 9 The results remain the same if we use trading volume scaled by market capitalization to measure market governance. 23

25 Overall, the evidence from the sub-sample regressions in Table 5 demonstrates that family monitoring through participating in firm ownership and management is an important private governance mechanism that substitutes for weak market governance (Burkart et al., 2003). [Insert Table 5 here] 4.4 Family Positions and Family Monitoring In this section, we examine whether and how the strength of family monitoring is affected by family members position in the family hierarchy. Table 6 presents the results of the relationship between abnormal related-party transactions and different family positions. In column 1, we report the results of a regression including an independent variable, Sibling, the number of the founder s or the founder s spouse s siblings and/or their spouses. In column 2, we include Parents, the number of the founder s parents who participated in the family business. We include Spouse, a dummy variable that equals 1 if the founder s spouse participates in the family firm and 0 otherwise, in column 3 and Children, the number of the founder s children in the regression, in column 4. In column 5, Extended, the number of relatives of the founder other than siblings, parents, children, and spouse participating in the family firm is included in the regression. Then, in column 6, we include all the five variables on family position in the regression. Hypothesis 4 predicts that the monitoring effects of abnormal related-party transactions are from more senior or more distant family members in relation to the founder. Therefore, we expect the coefficients of Siblings, Parents, and Extended to be significantly negative. The coefficient of Siblings in column 1 is negative and significant at the 1% level, suggesting that more siblings participating in the family business curbs abnormal related-party transactions. The coefficient of Parents in column 2 is negative but the t value is slightly less than the t-critical value at the 10% level. The statistically insignificant effect of the parent 24