Attendance of board meetings and company performance: Evidence from Taiwan

(1)

Attendance of board meetings and company performance:

Evidence from Taiwan

Hsin-I Chou

a

_{, Huimin Chung}

b

_{, Xiangkang Yin}

a,⇑

a

La Trobe University, Australia

b

National Chiao Tung University, Taiwan

a r t i c l e

i n f o

Article history:

Received 19 February 2013 Accepted 15 July 2013 Available online 25 July 2013

JEL classiﬁcation: G34

Keywords:

Board meeting attendance Family controlled ﬁrm Ultimate shareholder Firm performance

a b s t r a c t

This paper empirically investigates board meeting attendance and its effects on the performance of Tai-wanese listed corporations. Directors with higher qualifications attend board meetings more often by themselves. The ownership of the largest shareholder of a company also has a positive effect on director’s own meeting attendance. High meeting attendance by directors themselves can enhance a firm’s perfor-mance but high attendance by their representatives has an adverse effect. Independence of directors or a board is also positively associated with firm performance. These results largely hold even when the sam-ple is decomposed to count for different ownership structures and director types.

1. Introduction

The board of directors of a company has three main functions: monitoring, advising and contracting. It has the legal authority to ratify and monitor managerial initiatives, evaluate the perfor-mance of top managers, and reward or penalize that perforperfor-mance (Fama and Jensen, 1983a,b). The inside directors (executives of the company) provide valuable information about a ﬁrm’s activi-ties, while outside directors may provide both strategic input and objectivity in evaluating the top executives’ decisions. Hence, it is important to understand the behavior and work effort of directors and the behavioral difference between different directors. One of major duties of directors, especially for outside directors, is to at-tend board meetings because board meeting is the main vehicle for directors to collect information, make decisions and monitor the management (Adams and Ferreira, 2008). Moreover, it is quite difﬁcult to measure director work effort completely and directly in empirical studies. A straightforward way to partially identify direc-tor behavior and work effort is to investigate their board meeting attendance (Chou et al., 2010), which is the focus of this paper.

Existing empirical studies of board member activities are con-centrated on board meeting attendance by outside directors and

most studies are restricted to US companies. However, the data of board meeting attendance by the directors of US ﬁrms are not precise because the available data source only records whether a director attends more than 75% board meetings or not (e.g.,Adams and Ferreira, 2008, 2012; Lawler and Finegold, 2006). This paper intends to overcome this shortcoming of the existing empirical lit-erature by using a more comprehensive data set of board meeting attendance of Taiwanese companies. In contrast to the US compa-nies, companies listed in the Taiwan Stock Exchange must provide detailed information of board meeting attendance of all directors in their annual reports. It includes board meetings attended by directors themselves and attended by the representatives autho-rized by a director. With this more accurate information, we can have a closer look on board member activities. Particularly, we can empirically test the determinants of board meeting attendance with considerable accuracy. It is found that manager directors at-tend much more board meetings by themselves than outside direc-tors (inducting both independent and gray direcdirec-tors), while the attendance of family directors is between them.1_{Gray and family}

directors are also more likely to authorize a representative to attend board meetings on their behalf than other directors.

⇑ Corresponding author. Address: Department of Finance, La Trobe Business School, La Trobe University, Bundoora, Victoria 3086, Australia. Tel.: +61 3 9479 2312; fax: +61 3 9479 1654.

E-mail address:x.yin@latrobe.edu.au(X. Yin).

1

Directors coming from the controlling family of a company are called family directors in this paper. Furthermore, directors affiliated with the ultimate shareholder in terms of family relation or business relation are called ultimate directors. The details of firm and director classifications are given in the next section.

Contents lists available atScienceDirect

Journal of Banking & Finance

(2)

Another new feature of our database is that it contains compre-hensive information of the qualification of each director. We in-clude three proxies for the quality of a director in our analysis and find that all three measures are positively and significantly re-lated to directors’ own meeting attendance but negatively rere-lated to meeting attendance by their representatives. A more capable director seems more involving and is keener to play the director role by him/herself rather than to delegate the job to his/her repre-sentative. We also find that directors attend more board meetings by themselves and delegate fewer to their representatives if the largest shareholder of the firm has a greater proportion of cash flow rights. For other determinants of meeting attendance by directors themselves, our findings are consistent with the existing literature. Moreover, a determinant usually generates opposite ef-fects on a director’s own meeting attendance and authorized meet-ing attendance.

As an innovation, this study further explores work effort, or more specifically board meeting attendance, of various types of directors under different ownership structures. Concentrated own-erships in the forms of companies with a controlling family and/or a controlling ultimate shareholder are quite common in East Asia (e.g., Claessens et al., 2000; Claessens and Fan, 2002). To our knowledge, there is no literature to examine how directors play their roles differently due to ownership variation. According to the reality of the Taiwanese economy, this study considers five types of ownership structures. Our attention is on the contrast be-tween widely dispersed firms and family/ultimate shareholder controlled firms. Both independent and gray directors tend to at-tend more meetings by themselves if they seat on the board of a widely dispersed firm but fewer meetings if they are on the board of a family firm, although not all of these results are statistically significant. On the other hand, manager directors attend fewer board meetings if they are employed by a widely dispersed firm or by a firm with less divergence between the ultimate share-holder’s voting rights and cash flow rights. But this divergence makes family directors more likely reduce their own meeting attendance.

Whether the directors of a company properly play their moni-toring, advising and contract roles is ultimately testified by whether their work improving the company’s performance. With the relatively accurate information of board meeting attendance in our database, we can directly test the impact of directors’ meet-ing attendance on firm performance. To our knowledge, this is the first in the literature to quantitatively examine this relation.2_Our

findings indicate that the frequency of board meetings attended by directors themselves has a positive and significant effect on a firm’s profitability. However, the authorized meeting attendance is nega-tively correlated with performance. This negative effect is statisti-cally significant and economistatisti-cally comparable to the positive effect of directors’ own attendance.

While the typical agency problem of a widely dispersed firm is the conflict of interest between managers and shareholders, the main agency problem of a firm with concentrated ownership is the conflict of interest between controlling shareholders and minority shareholders. We study the director’s role in resolving these agency problems and improving performance by consider-ing further the attendance of board meetconsider-ings by different types of directors in these firms separately. Independent directors seem to play a more profound role in family or ultimate shareholder controlled companies than in widely dispersed firms, as

evi-denced by the findings that the effect of their own attendance to board meetings is significant on the profitability of family/ulti-mate controlled companies but insignificant on widely dispersed firms. The presence of family directors and ultimate directors in board meetings also has a significant impact on these firms’ performance.

The remainder of the paper is organized as follows. Section2

specifies the motivations and research questions of this paper. It also presents the regression models for testing. Section3describes the statistics of our sample and reports the main empirical results using firm-level and director-level data. The final section concludes the paper.

2. Research questions and methods

This research focuses on two questions. The ﬁrst is what factors determine a director to attend more (or less) board meetings. The second is whether and how a director’s work effort in terms of board meeting attendance affects his/her company’s performance. This section presents our motivations and research methods addressing these questions.

Because we want to address these issues by considering differ-ent firm ownership structures and differdiffer-ent types of directors, we need to identify the ultimate shareholder of a firm. FollowingLa Porta et al. (1999), we employ the cut-off of 20% control rights to trace who is the ultimate shareholder of a company. Direct voting rights are measured as the fraction of stocks held by a shareholder, and indirect voting rights are measured based on the latest link in the chain of stocks held by entities or nominal companies that are controlled by the shareholder. The ultimate shareholder is defined as the one who has the largest control rights by combining direct and indirect voting rights. Thus, the ultimate shareholder of a firm can be either an individual/family, a state agent, an institution or a widely held corporation. We call a company which has a control-ling ultimate shareholder ultimate controlled firm. Our focus is on family controlled firms, ultimate controlled firms and widely dispersed firms. Note that a family controlled firm is definitely an ultimate controlled firm but an ultimate firm is not necessarily a family controlled firm.

In the literature, directors of a firm are usually classified into three types: inside directors who are current employees of the firm, gray outsiders who are outsiders but have business ties with the company and independent outsiders who do not have any rela-tionship with the company (e.g., Baysinger and Bulter, 1985; Bhagat and Black, 2002). This classification is typical and well applicable to the US and the UK where the majority of the compa-nies are widely dispersed firms. Because a large proportion of the firms in our study are controlled by families or ultimate sharehold-ers, we classify directors into eight groups: family directors who are relatives of the controlling family, manager directors who are current employees of the firm, state directors who are agents of the government, institution directors who are agents of financial and investing institutions, gray directors who are outsiders but have business ties with the company, widely-held-corporation directors who are nominated by another widely held corporation, block outsiders who are large shareholders holding more than 1% of the firm, and independent directors who do not have any rela-tionship to the company.3 _{Moreover, we call a director ultimate}

director if he/she is afﬁliated with the ultimate shareholder of the company. As a group, ultimate directors include all family directors and some of state directors, institution directors and widely-held-corporation directors.

2

The only exception is the work byAdams and Ferreira (2009), which relates board meeting attendance to Return On Assets (ROA). However, their focus is on how female directors affect the governance and performance of US companies. They regress board meeting attendance on proportion of female directors in a ﬁrm and add ROA as one of control variables.

3

Institutional, gray and block outsiders are usually classiﬁed as gray outsiders in the US studies.

(3)

2.1. The determinants of board meeting attendance

We are interested in the characteristics of a director that can determine his/her board meeting attendance. We are also inter-ested in the question that whether directors behave differently in companies with different ownership structures and/or different features of large shareholders. To this end, we propose the follow-ing for empirical tests:

Attendi;t¼ f ðXi;t;Compensationi;t;Womeni;t;Tenurei;t;Interlocki;t;

Downi;t;Pledgedi;t;Meetingi;t;Qual1i;t;Qual2i;t;Qual3i;t;

Cashflow_i;t1;Excessi;t1;ROAi;t1;Foreigni;t;Domestici;t;

Bsize=Sizei;t;Le

v

eragei;t;Cashi;t;In

v

esti;t;Vari;t;Familyi;t;

Widelyi;t

e

i;tÞ ð1Þ

where f() represents a linear function. In the US, Securities and Ex-change Commission (SEC) only requires firms to disclose the names of directors who absent more than 25% of the board meetings dur-ing a fiscal year, and more detailed data on a director’s meetdur-ing attendance are not available. In this study, we take the advantage that Taiwanese firms must provide details of board meeting atten-dance for each director. Therefore, the dependent variable (Attend)

in (1) is measured in two ways: the percentage of board meetings attended by a director him/herself, and the percentage of meetings attended by the authorized representatives of the director. The descriptions of all variables in (1) can be found inTable 1.

We regress model (1) at individual director level so that sub-script i in (1) indexes individual director and subsub-script t indicates year. We consider two speciﬁcations of X. When the regression is over all directors in the sample, X is a vector with its elements being dummy variables indicating whether the director under con-sideration belongs to a particular type. Four types of directors, namely, independent, gray, manager, and family directors, are in-cluded as regressors. When regression is over directors of a partic-ular type, Xiin (1) is director i’s independence ratio (Indep), which

is used as a control variable here and its details are given in the next Subsection.

Compensation to directorship provides incentives to enhance the willingness of attending board meeting for a director (Chou et al., 2010). In Taiwan, there is no strict requirement of reporting compensation details for each director. Companies are only re-quired to report the total compensation to all directors in a com-pany. Also, there is no requirement of reporting board meeting fee. Therefore, we measure Compensation in (1) as the logarithm of average compensation to a director in a company in a year.

Table 1

Summary of variable deﬁnitions.

Variable Description

Board meeting attendance Own Meeting Attendance (Attend)

Percentage of board meetings attended by a director him/herself

Authorized Meeting Attendance (Attend)

Percentage of board meetings attended by representatives authorized by a director

Director variable Independence Ratio (Indep) A director’s independence status as speciﬁed byAppendix

Woman Director (Woman) Dummy variable equals one if the director is female and zero otherwise Compensation Logarithm of compensation to a director in a year

Tenure The number of years that a director has served on the board

Qualiﬁcation 1 (Qual1) Dummy variable equals one if a director is a lecturer or above in Business, Law, Finance, Accounting or Corporate Business related ﬁelds and zero otherwise

Qualification 2 (Qual2) Dummy variable equals one if a director has qualification of justice, procurator, attorney, CPA, specialist or technician of National Examination in Corporate Business related fields and zero otherwise Qualification 3 (Qual3) Dummy variable equals one if a director has five years experience in business, law, finance, accounting or

corporate business related ﬁelds and zero otherwise

CEO Duality (Duality) Dummy variable equals one if CEO and chairman is the same person and zero otherwise Director Ownership (Down) The percentage of shares held by all directors

Director Pledged Ratio (Pledged)

The percentage of all directors’ shareholdings that are pledged for loans and credits

Director Interlock (Interlock) The number of listed ﬁrm directorships held by a director Board variable Board Meetings (Meeting) The number of board meetings during a year

Board size (Bsize) Logarithm of the number of directors on a board

Ownership variable Ownership (Cashﬂow) The proportion of cash-ﬂow rights held by the largest shareholder group

Excess The divergence between control (direct and indirect voting rights) and ownership (cash-ﬂow rights) of the ultimate shareholder of a company

Foreign Institutional Shareholdings (Foreign)

The percentage of shares held by foreign institutions and funds

Domestic Institutional Shareholdings (Domestic)

The percentage of shares held by domestic institutions and funds

Family Dummy variable equals one if the ultimate shareholder of the company is a family group and zero otherwise

Widely Dummy variable equals one if the company is a widely dispersed ﬁrm and zero otherwise Firm variable Leverage The ratio of book value of debt to book value of assets

Growth Opportunity (R&D) Research and development expenses over sales Firm Size (Size) Logarithm of book value of assets

Volatility (Var) Variance of monthly returns stock over two prior years

Excess Cash (Cash) Cash and marketable security divided by the book value of total assets Investment Opportunity

(Invest)

Capital expenditure divided by the book value of total assets

Performance variable Return On Assets (ROA) Net income divided by the book value of total assets Earnings Per Share (EPS) Net income divided by the number of outstanding shares Sales To Assets Ratio (Sales) Sales divided by the book value of total assets

(4)

Directors and shareholders may use their own shareholdings as collaterals to increase control of the firm. They can collateralize their shareholdings to borrow money from banks and in turn buy more stocks of the firm. This, in turn, increases the deviation of their control rights to cash flow rights. They may also pledge their shareholdings for bank loans to invest in other opportunities. Pledging for loans effectively decreases the personal fund required for shareholdings (La Porta et al., 1999; Claessens et al., 2002). Thus, the regressors of (1) include the director pledged ratio, Pledged.

It is intuitive to expect that a director’s qualification can affect their board meeting attendance since the qualification may be re-lated to the effectiveness of a director’s functioning and impact. Our database enables us to measure the qualification of a director in three dimensions as specified inTable 1. Thus, the characteris-tics of a director are more comprehensively portrayed.

The conflict of interest exists between managers and sharehold-ers under the separation of ownsharehold-ership and control. However, own-ership around the globe tends to be more concentrated and large shareholders are likely to be directly involved in management. This structure shifts the focus away from management expropriating shareholders to majority shareholders who have the opportunity to expropriate wealth from minority shareholders (e.g.,Grossman and Hart, 1988; Shleifer and Vishny, 1997; Bebchuk et al., 2000). With large ownership and control, ultimate shareholders may re-solve the agency problem between shareholders and managers. But the power of ultimate shareholders can prevent effective mon-itoring by other directors and the possible expropriation by ulti-mate shareholders (mainly controlling families) is also likely to be harmful to firm performance (e.g.,La Porta et al., 1999, 2002; Claessens et al., 2000; Faccio and Lang, 2002). This potential of expropriation can be measured by divergence between control and cash flow rights of the ultimate shareholder (Excess), where control is based on direct and indirect voting rights of the ultimate shareholder (Claessens et al., 2000, 2006; Yeh and Woidtke, 2005). Another variable that proxies the controlling power of large share-holders in (1) is Cashflow, which measures the largest shareholder’s cash flow rights. Thus, Cashflow and Excess in (1) are not simply control variables; they also proxy for the features of large share-holders.4_{Introducing these variables into the regression enables us}

to resolve the question that how these most important features re-lated to the controlling power of large shareholders affects a direc-tor’s behavior in terms of board meeting attendance.

Lagged performance measure, Return On Assets (ROA), is in-cluded in (1) because it is possible that past performance affects current board meeting attendance. We ﬁrst regress (1) over all individual directors in the sample. In this regression, ﬁrm dummies (Family and Widely) are excluded because some types of directors do not exist under a certain ownership structure.5_{To investigate}

board meeting attendance by a particular type of directors, we sep-arately regress (1) again over the subsamples of independent, gray, manager and family directors. In these regressions, ﬁrm dummies are included to examine the impact of the type of the ﬁrm the direc-tor works for.6_{In (1), we add year dummies and two-digit industry}

dummies to control for year and industry effects, respectively. Each regression is examined based on robust standard errors clustered at the director level.

2.2. The effects of board meeting attendance on performance

An essential way that a board exerts its influence on its firm is coming through decisions and plans made in board meetings. In other words, the directors of a firm have to attend their board meetings to monitor, stipulate and supervise the firm or to make strategic decisions for it. Failure to regularly attend board meetings can be seen as a director is unwilling or unable to fulfill his/her duties.7_{Hence, attending board meetings is to accomplish a}

direc-tor’s responsibility and should be associated with subsequent higher ﬁrm performance. Thus, we posit the following hypothesis:

H1. Attendance of board meetings by directors themselves is positively correlated with ﬁrm performance.

Article 205 of Taiwanese Company Law stipulates that the quo-rum of any board meeting is a half of the number of board members and a resolution made by a board is eligible only if a half of or more attendees agree on it. Since authorized representatives are also counted, some directors may occasionally or quite often ask and authorize a shareholder or another director as their representatives to attend board meetings.8_{A plausible reason for directors sending}

representatives to board meetings can be that they are too busy to at-tend by themselves. Some directors, especially outside directors, have full-time jobs such as a CEO in other companies. Quite often, family directors of a family controlled firm are the managers or direc-tors of an affiliated firm controlled by the family. These busy direcdirec-tors may not have enough time to fulfill their duties and thus may not at-tend board meetings regularly (Fich and Shivdasani, 2006; Adams and Ferreira, 2008). Instead of not attending board meetings by them-selves, they may authorize a representative to attend meetings on their behalf to ensure meetings are eligible. Although busy directors usually have a good reputation and qualification (Kaplan and Reishus, 1990; Booth and Deli, 1996; Ferris et al., 2003), their representatives may not have the similar reputation and quality.9_These

representa-tives are likely to function not as well as directors themselves. Fur-thermore, the authorized representatives face an agency problem in the sense that they are more likely to shirk. Hence, the more meetings are delegated to representatives, the less effective is a director. These intuitions lead us to conjecture that

H2. Board meeting attendance by authorized representatives is negatively associated with subsequent ﬁrm performance.

To test these hypotheses, we speciﬁed a model as follows:

ROAi;tþ1¼ f ðAttendi;t;Indepi;t;Familyi;t;Widelyi;t;Bsizei;t;Dualityi;t;

Womani;t;Qual1i;t;Qual2i;t;Qual3i;t;Cashflowi;t;Excessi;t;

Foreigni;t;Domestici;t;Le

v

eragei;t;R&Di;t;Sizei;t;Vari;t;

In

v

esti;t;

e

i;tÞ ð2Þ

4

The largest shareholder of an ultimate ﬁrm is often the ultimate shareholder. But they are not always identical.

5

For instance, widely dispersed ﬁrms have no family directors.

6 _{There is an exception. Because that all family directors are working for family}

controlled ﬁrms so that there is no necessity of introducing Family dummy and Widely dummy for family directors.

7_{The Taiwan Stock Exchange and GreTai (over-the-counter) Securities Market}

specifies the following major duties for the board of directors of a listed company (Corporate Governance Best Practice Principles for TSE/GTSM Listed Companies, article 27): 1. Stipulation of an effective and appropriate internal control system; 2. Selection and supervision of managers; 3. Review of the management policy and business plan of the company; 4. Review of the financial goals of the company; 5. Supervision of the result of operations of the company; 6. Supervision and handling of the risks encountered by the company; 7. Ensuring the compliance with relevant laws and regulations by the company; 8. Planning the future development of the company; 9. Creation and maintenance of the company image and fulfillment of social obligations; 10. Appointment of Certified Public Accountants (CPA) or attorneys.

8_{According to Article 205 of the Taiwanese Company Law, only shareholders or}

directors can be the representative of a director to attend board meetings.

9

Many ultimate directors, especially family directors, do not have a good reputation and qualification. Holding several positions in affiliated firms as an agent of the ultimate shareholder can increase the control over the affiliated firms within the business group (Yeh and Woidtke, 2005).

(5)

We run regression (2) over all firms so that subscripts i and t index firm and year, respectively. The main performance measure of this study is ROA as presented in (2). In robustness check, we also con-sider other performance proxies. For meeting attendance (Attend) and director variables, the averages of individual director’s mea-sures over the firm is used in the regression.

In the existing literature the independence of a director is usu-ally proxied by whether the director is an outsider or not. Then, the independence of a board is measured by the proportion of inde-pendent directors seated on the board. This study has a much more detailed measure of board independence. We use ten director inde-pendence criteria speciﬁed by the Taiwanese company law to eval-uate the independence of each individual director and the independence of a board is proxied by the average independence of its members. The details of independence criteria and indepen-dent score calculation are presented inAppendixand each director can get a maximum of 10 points. We obtain data of independence score from the annual reports of companies in our sample. Because some companies report only 7 or 8 out of 10 independence criteria in their annual reports, the variable Indep for each director is the ratio of a director’s independence score to the number of total independence criteria reported by the company. A director’s inde-pendence ratio is thus a fraction varying from zero to one and it may take different values even for the same type of directors. Explanatory variable Indep in (1) is each director’s ratio while Indep in (2) is the average ratio across all directors within a company.

In addition to the aforementioned variables, regression (2) in-cludes a series of control variables and their deﬁnitions can also be found inTable 1.10_{In addition to commonly used control}

vari-ables, director qualification variables are unique to this study and we expect the quality of a firm’s directors makes a difference to its performance. We further add a dummy for family controlled firm (Family) and a dummy for widely dispersed firm (Widely) to examine whether these types of firms perform differently from other firms.11 Year dummies and two-digit industry dummies are used to control for year and industry effects, respectively. Each regression is exam-ined based on robust standard errors clustered at the firm level.

3. Data and empirical results

3.1. Sample selection and descriptive statistics of data

The sample for this study consists of all non-ﬁnancial ﬁrms listed on the Taiwan Stock Exchange (TWSE) in 2006 and 2007.12

The board composition, compensation, director ownership (i.e., cash-flow rights) and accounting data are collected from the Taiwan Economics Journal database. Control rights of an ultimate share-holder is calculated by thoroughly tracing the ultimate shareshare-holder’s direct and indirect voting rights, as specified byLa Porta et al. (1999). Based on this calculation, we can identify the classification of direc-tors (i.e., family director, state director, institution direcdirec-tors, etc.). The identification is a very complex and tedious process, which re-stricts the sample to cover only the first two years of the implemen-tation of the TWSE disclosure rule. Director information, including tenure, gender, interlock, qualifications and meeting attendance, are manually collected from annual reports of sample companies. The sample consists of 647 and 661 firms in 2006 and 2007,

respectively. Total numbers of directors are 4564 in 2006 and 4743 in 2007. In the regressions, we eliminate ﬁrms or directors where the data of ownership, accounting, meeting attendance or board are missing.

Before starting to examine the descriptive statistics of each var-iable, let us have a look at the aggregate figures of firms and direc-tors in the sample. This information is reported in Table 2. As mentioned in Section2, we use the cut-off of 20% control rights to trace the ultimate shareholder of a firm. Based on this criterion, Panel A ofTable 2shows that there are 221 (219) widely dispersed firms in 2006 (2007); 345 (360) family controlled firms; 16 (18) state controlled firms; 50 (51) firms controlled by another widely held corporation and 2 (1) institution controlled firms.13 _Overall,

there are 413 (430) firms have an ultimate shareholder (i.e., ultimate controlled firms) in 2006 (2007). According to the classification scheme mentioned in Section 2, directors are classified into eight groups and their distributions are given in Panel B ofTable 2. Overall, there are 1639 (1626) directors can be classified as ultimate direc-tors in 2006 (2007).

Panel A ofTable 3displays the descriptive statistics of the sam-ple. For the characteristics of directors, we ﬁnd that on average directors attend 77.6% of board meetings by themselves while 8.2% of the board meetings are attended by the representatives who are authorized by a director. There is a substantial difference between average compensation to a director (NT$ 1.59 million) and its median (NT$ 0.57 million).14_{The average independence}

ra-tio is 68.7% and the aggregate director ownership of a company is averaged at 20.2%. For the qualifications of directors, there are 5.7% of directors who are a lecturer or above in business, law, accounting, finance or corporate business related fields. About 4.2% of directors have a background of justice, procurator, attorney, CPA, specialist or technician of National Examination in Corporate Business related fields. Majority of directors have five-year experience in business, law, finance, accounting or corporate business related fields.

For ﬁrm performance, the average ROA is 4.8% per annum.15_The

average debt to asset ratio is around 37.1%. The average cash ﬂow

Table 2

Classiﬁcation of sample ﬁrms and directors.

2006 2007

Panel A. Firm classiﬁcation

Widely dispersed firm 221 219 Family controlled firm 345 360 State controlled firm 16 18 Widely-held-corporation controlled firm 50 51 Institution controlled firm 2 1 Total 634 649 Ultimate controlled firm 413 430

Panel B. Director classiﬁcation

Independent director 764 814 Family director 1290 1295 State director 99 103 Institution director 44 41 Manager director 814 724 Gray director 436 537

Widely-held-corporation controlled director 207 188

Block holder director 830 923

Total 4484 4625

Ultimate director 1639 1626

The definitions of firm classifications followLa Porta et al. (1999)andClaessens et al. (2000).

10

The selection of control variables is conventional except for the qualiﬁcations of directors, see for example,Yermack (1996), Baliga et al. (1996), Filatotchev et al. (2005), Yeh and Woidtke (2005), Dahya and McConnell (2007), Choi et al. (2007), Dahya et al. (2008), an Adams and Ferreira (2009).

11

We have also used a dummy for ultimate controlled ﬁrm in (2) but the results are quite similar. Therefore, we do not report these results inTable 7.

12

Since 2006, the TWSE disclosure rules require listed companies to report detailed information on director meeting attendance.

13

Note, the sample size here is smaller than what is reported in the previous paragraph because some ﬁrms are eliminated due to missing ownership data.

14

One US dollar was worth 29.06 New Taiwan dollars (NT$) as of January 1, 2013.

15 _{The main performance measure in this paper is ROA. In robustness check in}

Section 3.4, we also consider Earning Per Share (EPS), Sales to Assets Ratio (Sales) and Sales Growth Rate (Growth). Their descriptive statistics are also included inTable 3.

(6)

right is 23.3% and average excess right is 6.25%, which are larger than their medians. Such difference between average and median prevails for other ﬁrm variables too. The average (median) foreign tional shareholding is 0.19% (0%), which implies that foreign institu-tions only hold a very small portion of stocks of listed ﬁrms in Taiwan.

Panel B ofTable 3documents the average of director character-istics for four types of directors of our main interests. Not surpris-ingly, outside directors (including both independent and gray directors) attend fewer board meetings than other two types of directors on average. In terms of the first two qualifications, inde-pendent directors are much high than other directors but the dif-ference between their third qualifications is very small across four types of directors. We also note that family directors have the highest director interlock, which reflects the fact that a family member often seats on the boards of multiple firms controlled by the family.

Before conducting formal empirical analysis, we examine the correlations among main variables, which are reported inTable 4. The correlation coefﬁcients among the majority of explanatory variables are quite low. Only the family ﬁrm dummy, the widely

dispersed ﬁrm dummy and cash ﬂow rights are highly correlated to each other. For this reason, they enter into regression models separately. The correlation between own meeting attendance and authorized attendance is 0.504. Therefore we also test them sep-arately to avoid possible multicollinearity.

3.2. The determinants of attendance at board meetings

About the determinants of board meeting attendance, we test for two dependent variables: individual director’s own meeting attendance and authorized meeting attendance.Table 5 reports the results of the pooled regression of (1). As shown by speciﬁca-tions 1 and 2, independent and gray directors attend fewer board meetings by themselves relative to other directors, which is consis-tent with the statistics reported in Panel B ofTable 3. Manager directors on the other hand have the highest meeting attendance, which is hardly surprising given board meetings usually hold in their business time.

Director compensation is positively related to their meeting attendance and it is signiﬁcant at least at the 10% level. Although it is intuitive that directors appear in the board meetings more

Table 3

Descriptive statistics.

Variable Obs. Mean Std. Dev. Min Median Max

Panel A. Firm, director and board characteristics Director and board characteristic

Director’s Own Meeting Attendance 9276 0.776 0.293 0 0.900 1

Authorized Meeting Attendance 9276 0.082 0.183 0 0 1

Director compensation (in thousands NT$) 1289 1592.92 3408.81 0 571.43 42857.14

Number of Meetings 1308 9.60 5.719 2 8 69

CEO duality 1309 0.273 0.445 0 0 1

Board Size 1309 7.110 2.445 3 7 21

Director Independence Ratio 8978 0.687 0.227 0 0.700 1

Aggregate Director Ownership 1289 0.202 0.135 0.023 0.166 0.95

Aggregate Director Pledged Ratio 1289 0.101 0.195 0 0 0.97

Woman Director 9307 0.130 0.336 0 0 1 Tenure 9195 8.322 9.021 0 5 53 Qualification 1 9227 0.057 0.233 0 0 1 Qualification 2 9227 0.042 0.199 0 0 1 Qualification 3 9227 0.967 0.177 0 1 1 Director Interlock 9240 0.491 1.086 0 0 9 Firm characteristic ROA 1279 0.048 0.117 1.781 0.051 0.445 EPS 1308 1.909 3.695 9.380 1.270 50.48

Sales to assets ratio 1279 0.886 0.710 0.0004 0.717 7.863

Sales growth ratio 1306 0.112 0.359 5.087 0.101 2.494

Leverage 1279 0.371 0.174 0.0158 0.361 0.970

Market capitalization (in million NT$) 1230 22,100 8350 180.24 45486.26 1,310,000

Cash ﬂow right 1274 0.232 0.173 0 0.200 0.904

Excess right 1274 0.063 0.105 0 0.015 0.751

R&D/Sale 1279 0.010 0.051 0 0 1.635

Excess cash 1279 0.087 0.100 0.0004 0.049 0.733

Investment 1278 0.371 0.336 0 0.269 1.971

Volatility 1221 0.019 0.024 0.000011 0.013 0.43

Domestic Institutional Shareholdings 1286 0.019 0.035 0 0.002 0.41

Foreign Institutional Shareholdings 1286 0.002 0.009 0 0 0.09

Variable Independent director Family director Gray director Manager director

Panel B. Mean characteristics of four types of directors

Director’s Own Meeting Attendance 0.708 0.799 0.727 0.886

Authorized Meeting Attendance 0.076 0.086 0.106 0.043

Director Independence Ratio 0.986 0.525 0.705 0.666

Woman Director 0.118 0.148 0.146 0.086 Tenure 3.667 10.42 8.589 8.667 Qualification 1 0.189 0.026 0.042 0.016 Qualification 2 0.116 0.025 0.031 0.013 Qualification 3 0.938 0.973 0.966 0.994 Director Interlock 0.474 0.661 0.395 0.218

The board composition, compensation and ownership data such as control rights and cash-ﬂow rights and accounting data are collected from the Taiwan Economics Journal database. Director information, including tenure, gender, qualiﬁcations and director and supervisor meeting attendance are collected from companies’ annual reports. Panel A reports statistics over the whole sample while Panel B presents the averages of director characteristics over a particular type of directors.

(7)

often if they are paid higher, caution should be exerted in inter-preting this result. Recall that we have only firm-level compensa-tion data, the compensacompensa-tion variable used in the regression does not vary across directors within a firm. Thus, variation in incentive to directorship and variation in the effect of the incentive are par-tially masked.Table 5also demonstrates that directors who served on the board for a longer period usually attend more board meet-ings by themselves. This relation is significant at the 1% level. On the other hand, a director who holds more directorships is less likely to attend board meetings by themselves as the regression coefficient of director interlock is negative. Moreover, directors at-tend proportionately fewer meetings if a company has more board meetings. These results for director compensation, tenure and interlock, and the number of board meetings are consistent with previous studies (e.g.,Adams and Ferreira, 2008). For three qualifi-cation variables, all of them are significantly and positively corre-lated with director’s own meeting attendance. Thus, the higher is a director’s qualification, the more board meetings he/she attends. We like to bring the attention to the significantly positive effect of lagged Cashflow on director’s own meeting attendance. The re-sult implies directors generally attend board meetings more often when the largest shareholder has more interests in the firm. Nev-ertheless, the impact of lagged Excess on meeting attendance by directors themselves is negative and significant at the 10% level. Directors do not necessarily work harder if the ultimate share-holder has more power to entrench the resources away from the

minority shareholders. Note that these results are of aggregate ef-fects of large shareholders, which mask cross-sectional difference over different types of directors, which is the focus of next regres-sions. Finally, lagged ROA is positively but insigniﬁcant correlated to directors’ own meeting attendance. This implies that there is no strong evidence showing past performance affects directors attending board meetings by themselves.

Domestic institutional shareholdings are negatively and signif-icantly related to a director’s own meeting attendance only at the 10% levels. Foreign institutional shareholdings have the same sign but are not significant. In contrast, the effects of firm level control variables such as relative board size and investment oppor-tunity are statistically significant. Other control variables are insig-nificant in both specifications 1 and 2.

It is also interesting to examine the reasons for directors autho-rizing a representative to attend board meetings rather than attending the meetings by themselves. First, we can see from spec-ifications 3 and 4 inTable 5that family directors and gray directors are more likely to authorize a representative to attend board meet-ings on their behalf than other directors. While independent direc-tors are also more likely to do so, the result is not statistically significant. Director compensation is positively but insignificantly related to authorized meeting attendance. Director interlock is sig-nificantly and positively related to authorized meeting attendance, while director tenure is significantly and negatively correlated to it. Higher director interlock increases the director’s willingness of

Table 4

Correlation between main variables.

2 3 4 5 6 7 8 9 10 11 12 13 1. ROA 0.057b 0.058b 0.020 0.091c 0.012 0.028 0.008 0.067b 0.157c 0.138c 0.048a 0.011 2. Own attendance 0.504c 0.744c 0.008 0.183c 0.097c 0.031 0.022 0.034 0.065b 0.021 0.049a 3. Authorized attendance 0.202c _0.001 _0.240c 0.005 0.167c 0.045 0.144c 0.061b _0.013 _0.016 4. Total attend 0.008 0.022 0.114c 0.093c 0.060b 0.073c 0.027 0.033 0.068b 5. Indep 0.029 0.066b 0.085c 0.069b 0.008 0.173c 0.089c 0.012 6. Bsize 0.147c 0.185c 0.128c 0.260c 0.034 0.049a 0.030 7. Duality 0.143c 0.113c 0.077c 0.022 0.012 0.033 8. Interlock 0.146c 0.213c 0.029 0.035 0.043 9. Woman 0.100c 0.031 0.020 0.065b 10. Compensation 0.063b _0.048a _0.043 11. Qual1 0.165c 0.125c 12. Qual2 0.022 14 15 16 17 18 19 20 21 22 23 24 Widely 1. ROA 0.001 0.058b _0.119c _0.078c 0.142c 0.031 0.170c 0.172c 0.180c 0.027 0.004 2. Own attendance 0.074c 0.051a 0.001 0.014 0.049a 0.003 0.017 0.068b 0.082c 0.040 0.008 3. Authorized attendance 0.129c 0.083c 0.031 0.078c 0.069b 0.002 0.155c 0.017 0.073c 0.096c 0.034 4. Total attend 0.016 0.007 0.022 0.076c 0.109c 0.003 0.101c 0.091c 0.037 0.019 0.017 5. Indep 0.271c 0.084c 0.080c 0.048a 0.096c 0.064b 0.015 0.026 0.035 0.306c 0.261c 6. Bsize 0.170c 0.203c 0.135c 0.016 0.075c 0.009 0.301c 0.170c 0.183c 0.076c 0.042 7. Duality 0.025 0.171c 0.076c 0.022 0.008 0.004 0.064b 0.061b 0.027 0.043 0.098c 8. Interlock 0.214c _0.416c _0.069b _0.084c 0.078c 0.002 0.325c 0.110c 0.003 0.015 0.049a 9. Woman 0.084c 0.094c 0.066b 0.026 0.140c 0.062b 0.142c 0.215c 0.031 0.026 0.028 10. Compensation 0.182c 0.088c 0.091c 0.218c 0.095c 0.014 0.510c 0.127c 0.004 0.118c 0.131c 11. Qual1 0.032 0.023 0.127c 0.050a 0.075c 0.010 0.120c 0.043 0.088c 0.073c 0.046 12. Qual2 0.014 0.023 0.041 0.077c 0.021 0.013 0.058b 0.065b 0.094c 0.036 0.030 13. Qual3 0.013 0.031 0.017 0.025 0.002 0.001 0.032 0.025 0.072c 0.035 0.019 14. Cash ﬂow 0.266c 0.084c 0.074c _0.024 0.086c 0.228c _0.031 _0.016 _0.567c 0.575c 15. Excess 0.059b 0.017 0.043 0.021 0.221c 0.030 0.010 0.051b 0.250c 16. Domestic 0.103c 0.028 0.007 0.201c 0.077c 0.099c 0.057b 0.008 17. Foreign 0.044 0.003 0.227c 0.058b 0.072b 0.022 0.062b 18. Leverage 0.097c 0.120c 0.236c 0.038 0.029 0.026 19. R&D 0.004 0.029 0.028 0.029 0.034 20 Size 0.197c 0.047b 0.134c 0.092c 21. Var 0.041 0.032 0.038 22. Investment 0.034 0.067b 23. Family 0.798c In the table, a

Signiﬁcance at the 0.1 level.

b

Signiﬁcance at the 0.05 level.

c

(8)

sending a representative to attend board meetings, but a director with longer tenure is less likely to send a representative. These re-sults are complementary to the rere-sults of their counterparts of own meeting attendance we saw in speciﬁcations 1 and 2. On the other hand, the number of a company’s board meetings is negatively and signiﬁcantly related to authorized meeting attendance, which is similar to the previous result of director’s own attendance. Other director variables, including woman director, director ownership,

director pledged ratio and qualifications of directors, are insignifi-cant across specifications 3 and 4.

Opposite to the case of director’s own attendance, lagged Cash-flow is negatively and significantly related to authorized meeting attendance. Directors are not more likely to send representatives if the ultimate shareholder has more entrench power (i.e., greater Excess) or when past performance is poor (i.e., lower ROA). For other variables, only foreign institutional shareholding, board size over firm size, leverage, and investment opportunity are significant.

To investigate the effects of firm ownership structure and the controlling power of large shareholders in more detail, we run regression (1) again where regressions are over each director group rather than all directors. With respect to independent, grey and manager directors, we separately consider Family firm dummy and Widely dispersed firm dummy. Thus, corresponding to each specification inTable 5, there are two specifications in Panels A– C ofTable 6. For family directors we do not consider firm dummies since these directors serve only one type of firms—family con-trolled firms.Table 6only reports the results of the estimation of the coefficients of laggedCashflow and Excess andROA, and firm dummies while the coefficient estimates of other variables are not reported because they are qualitatively similar to their coun-terparts inTable 5.

As shown by Panel A ofTable 6, independent directors of family controlled firms (widely dispersed firms) attend fewer (more) board meetings than their counterparts in other firms. This pattern also appears in the attendance of gray directors. The only differ-ence is that the former is statistically insignificant while the latter is significant. These findings are supportive to the hypothesis that outside directors tend to miss more board meetings if the firm’s ownership is more concentrated. We conjecture that the reason for outsider directors are less interested in attending board meet-ings of family controlled firms is that they are less influential in these firms because family firms are often tightly or completely controlled by the dominant family. On the other hand, they can play a more effective role in wildly dispersed firms so that they are more attracted to board meetings of these firms. In contrast toTable 5, the effects of lagged Cashflow and Excess on the meeting attendance by both types of directors are not significant, while lagged ROA is still insignificant.

For manager directors, the meeting attendance pattern is opposite to that of outside directors, as specifications 1 and 2 show manager directors attend more meetings if they work for a family controlled firm but they attend fewer meetings if they are employed by a widely dispersed firm. Moreover, the effect of lagged Excess is positive and marginally significant while the effect of lagged Cashflow is also positive but insignificant. Think-ing of attendThink-ing more board meetThink-ings is one of forms of workThink-ing harder, these results in Panel C are consistent with the claim that agency problem between managers and shareholders is less se-vere in a firm with more concentrated ownership and/or a more powerful controlling shareholder. We note that past company performance in terms of ROA has a significantly positive effect on manager directors’ meeting attendance. It seems that out four types of directors considered inTable 6, only manager directors’ response to firm performance is statistically significant when they determine their work efforts. For family directors, Panel D reports that they attend fewer board meetings if the ultimate shareholder has greater lagged Excess. The reason for this result is that family directors are family members of the ultimate shareholder and they may delegate their responsibilities of monitoring the man-agement and decision-making for the firm to the ultimate share-holder. Their presence in the board meeting is likely to be symbolic or to merely meet some regulation requirement. As a re-sult, they do not want to be deeply involved in the business if the

Table 5

Determinants of board meeting attendance. Own meeting attendance Authorized meeting attendance 1 2 3 4 Independent director 0.108 0.107 0.016 0.014 (0.021)*** _(0.020)*** _(0.013) _(0.013) Gray director 0.095 0.091 0.035 0.030 (0.019)*** _(0.019)*** _(0.012)*** _(0.013)** Manager director 0.051 0.055 0.017 0.022 (0.019)*** _(0.019)*** _(0.013) _(0.013)* Family director 0.032 0.031 0.023 0.025 (0.019)* (0.019)* (0.013)* (0.012)** Compensation 0.0012 0.0014 0.0003 0.0002 (0.0007)* _(0.0007)** _(0.0005) _(0.0005) Women 0.001 0.001 0.0006 0.0002 (0.011) (0.011) (0.007) (0.007) Tenure 0.004 0.004 0.0005 0.0005 (0.0004)*** (0.0004)*** (0.0002)** (0.0002)** Interlock 0.008 0.005 0.011 0.009 (0.003)** _(0.004) _(0.002)*** _(0.003)*** Down 0.0002 0.0002 (0.0003) (0.0002) Pledged 0.0001 0.0001 0.0002 0.0001 (0.0002) (0.0002) (0.0002) (0.0001) Meeting 0.004 0.004 0.0014 0.0014 (0.001)*** _(0.001)*** _(0.0004)*** _(0.0004)*** Qual1 0.027 0.027 0.010 0.010 (0.015)* _(0.015)* _(0.009) _(0.009) Qual2 0.051 0.051 0.013 0.013 (0.016)*** _(0.016)*** _(0.010) _(0.010) Qual3 0.068 0.069 0.016 0.017 (0.022)*** (0.022)*** (0.015) (0.015) Lagged Cashﬂow 0.0006 0.0008 (0.0003)** _(0.0002)*** Lagged Excess 0.0006 0.00001 (0.0004)* _(0.0002) Lagged ROA 0.073 0.073 0.004 0.010 (0.046) (0.046) (0.024) (0.024) Foreign 0.646 0.648 1.127 1.103 (0.429) (0.430) (0.293)*** _(0.392)*** Domestic 0.241 0.236 0.011 0.018 (0.126)* _(0.127)* _(0.072) _(0.072) Bsize/Size 1.029 0.947 0.769 0.698 (0.205)*** _(0.205)*** _(0.138)*** _(0.138)*** Leverage 0.012 0.017 0.023 0.028 (0.025) (0.025) (0.016) (0.016)* Cash 0.022 0.025 0.014 0.010 (0.040) (0.040) (0.025) (0.025) Invest 0.023 0.023 0.017 0.016 (0.012)* _(0.012)* _(0.008)* _(0.008)* Var 0.244 0.246 0.073 0.085 (0.202) (0.201) (0.086) (0.084) Constant 0.871 0.852 0.025 0.001 (0.082)*** _(0.083)*** _(0.042) _(0.042) R-square 0.078 0.079 0.048 0.051 F-test 14.81*** _14.88*** _6.98*** _7.36*** Observations 8351 8351 8351 8351

The dependent variable is a director’s own meeting attendance (speciﬁcations 1–2) or the authorized meeting attendance (speciﬁcations 3–4). Each regression includes year and two-digit SIC code dummies. Numbers in parentheses are standard errors based on robust standard errors clustered at the director level.

*_{Significance at the 10% level.} ** _{Significance at the 5% level.} ***_{Significance at the 1% level.}

(9)

ultimate shareholder is sufﬁciently powerful as measured by Excess.

For authorized meeting attendance, the most notable result is of manager directors in Panel C. The pattern of their authorized meet-ing attendance is opposite to the pattern of their own attendance; i.e., a factor inducing them to attend more board meetings by themselves leads them to reduce authorized meeting attendance. The only exception is Family dummy in specifications 3 and 7, where the coefficients are both negative and insignificant. Lagged Cashflow and Excess are significant factors in affecting a manager director’s decision of delegating board meetings to a representa-tive. For other directors, the pattern of authorized meeting

atten-dance is not always contradictory to the pattern of own meeting attendance. There are quite a few cases where a factor making these directors attend more meetings by themselves also makes them delegate more meetings to their representatives although they are mostly insigniﬁcant.

3.3. The impact of board meeting attendance on firm performance After examining the determinants of directors’ board meeting attendance, we now investigate whether attending board meetings is associated with a firm’s performance.Table 7documents the esti-mation results of the relationship between firm performance and

Table 6

Determinants of directors’ meeting attendance by different types of directors.

Own meeting attendance Authorized meeting attendance

1 2 3 4 5 6 7 8

Panel A: Independent director

Lagged Cashﬂow 0.0003 0.0008 0.0003 0.0006 (0.0008) (0.0009) (0.0004) (0.0004) Lagged Excess 0.0004 0.0001 0.0005 0.0004 (0.0008) (0.0009) (0.0005) (0.0005) Lagged ROA 0.132 0.134 0.131 0.119 0.007 0.008 0.006 0.002 (0.093) (0.093) (0.094) (0.095) (0.053) (0.053) (0.053) (0.054) Family 0.002 0.007 0.022 0.017 (0.018) (0.023) (0.010)** (0.013) Widely 0.016 0.031 0.012 0.005 (0.018) (0.026) (0.010) (0.014)

Other controls Yes Yes Yes Yes Yes Yes Yes Yes

Observations 1381 1381 1381 1381 1381 1381 1381 1381

Panel B: Gray director

Lagged Cashﬂow 0.0003 0.0002 0.0003 0.0001 (0.0007) (0.0008) (0.0005) (0.0005) Lagged Excess 0.0009 0.0001 0.0016 0.001 (0.001) (0.001) (0.0008)* _(0.0008) Lagged ROA 0.056 0.052 0.052 0.052 0.056 0.053 0.051 0.053 (0.095) (0.096) (0.095) (0.095) (0.054) (0.054) (0.054) (0.054) Family 0.036 0.028 0.014 0.024 (0.016)** _(0.019) _(0.010) _(0.013)* Widely 0.044 0.046 0.004 0.001 (0.016)*** _(0.021)** _(0.010) _(0.013)

Observations 2507 2507 2507 2507 2507 2507 2507 2507

Panel C: Manager director

Lagged Cashﬂow 0.0008 0.0003 0.0009 0.0008 (0.0006) (0.0006) (0.0003)*** _(0.0003)** Lagged Excess 0.001 0.001 0.001 0.001 (0.0005)** (0.0006) (0.0003)*** (0.0003)*** Lagged ROA 0.147 0.147 0.119 0.127 0.062 0.062 0.039 0.041 (0.073)** _(0.071)** _(0.072)* _(0.072)* _(0.041) _(0.040) _(0.041) _(0.041) Family 0.018 0.003 0.023 0.003 (0.014) (0.016) (0.008)*** _(0.009) Widely 0.029 0.016 0.026 0.006 (0.013)** (0.017) (0.008)*** (0.010)

Observations 1383 1383 1383 1383 1383 1383 1383 1383

Panel D: Family director

Lagged Cashﬂow 0.0007 0.001 (0.0005) (0.0004)*** Lagged Excess 0.0016 0.0003 (0.0006)** _(0.0004) Lagged ROA 0.063 0.054 (0.082) (0.043)

Other controls Yes Yes

Observations 2689 2689

The dependent variable is meeting attendance of individual director or his/her representative, where regressiones cross over the same type of directors. Explanatory variable Indep replaces director type dummies inTable 5. Other explanatory variable include those in Table 5 and a dummy for firm ownership structure (except for Panel D) but only the estimates of the coefficients of Cashflow, Excess, ROA, and firm ownership structure are reported. Each regression includes year and two-digit SIC code dummies. Numbers in parentheses are standard errors based on robust standard errors clustered at the director level.

*_{Significance at the 10% level.} **_{Significance at the 5% level.} ***_{Significance at the 1% level.}

(10)

board meeting attendance as specified by regression model (2). We find that meeting attendance by directors themselves is positively related to the firm’s subsequent ROA and this relation is significant at the 5% level. This means that directors who attend more board meetings can lead a firm to be more profitable, which is consistent with hypothesis H1 that as the directors of a firm work harder (i.e. by attending more meetings) it performs better. AsTable 3 docu-mented, the standard deviation of own meeting attendance is 0.293 while the standard deviation of ROA is 0.117. Thus, the results inTable 7imply that a one standard deviation in director meeting attendance is, on average, associated with about 0.13 times of a standard deviation higher subsequent ROA. On the other hand, authorized meeting attendance is negatively associated with firm performance at least at the 10% significance level.16 _Authorizing

representatives instead of attending board meetings by directors themselves can harm a ﬁrm’s proﬁt performance. This implies that the representative of a director usually make no or little contribution during board meetings. Their appearance in the meeting cannot re-place the function of directors themselves as conjectured in hypoth-esis H2.

Independence of directors should enhance the monitoring power of the board (Dahya and McConnell, 2007; Choi et al., 2007), and therefore should also enhance the performance of the firm. Our regression results show that the proxy for board indepen-dence is positively correlated to firm performance, significant at the 5% or 10% level in some specifications.

For ownership variables, we note that Cashflow is positively cor-related with firm performance, and this relation is statistically sig-nificant at the 10% level. On the other hand, the effect of the ultimate shareholder’s Excess on subsequent ROA is positive but insignificant. These two results show that a firm performs better if the largest (ultimate) shareholder has greater ownership and is

Table 7

ROA and board meeting attendance: Pooled regression.

1 2 3 4 5 6 Own attendance 0.051 0.052 0.053 (0.024)** (0.025)** (0.024)** Authorized attendance 0.072 0.074 0.076 (0.037)* _(0.038)** _(0.038)** Indep 0.064 0.053 0.053 0.065 0.054 0.054 (0.032)** _(0.033) _(0.032)* _(0.032)** _(0.033) _(0.032)* Family 0.007 0.006 (0.008) (0.008) Widely 0.009 0.008 (0.008) (0.008) Bsize 0.004 0.003 0.003 0.004 0.004 0.003 (0.012) (0.012) (0.012) (0.012) (0.012) (0.012) Duality 0.0001 0.001 0.0005 0.001 0.002 0.002 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) Women 0.013 0.013 0.011 0.013 0.013 0.011 (0.028) (0.028) (0.029) (0.029) (0.028) (0.029) Qual1 0.081 0.086 0.085 0.081 0.087 0.086 (0.030)*** _(0.030)*** _(0.031)*** _(0.030)*** _(0.030)*** _(0.030)*** Qual2 0.002 0.0004 0.0008 0.0005 0.0007 0.0002 (0.039) (0.039) (0.039) (0.039) (0.039) (0.039) Qual3 0.010 0.011 0.010 0.013 0.013 0.013 (0.021) (0.023) (0.023) (0.020) (0.022) (0.022) Cashﬂow 0.048 0.047 (0.027)* _(0.027)* Excess 0.036 0.038 (0.038) (0.038) Foreign 0.411 0.371 0.392 0.459 0.421 0.442 (0.283) (0.279) (0.283) (0.286) (0.282) (0.286) Domestic 0.192 0.192 0.187 0.187 0.187 0.182 (0.106)* _(0.105)* _(0.108)* _(0.104)* _(0.103)* _(0.105)* Leverage 0.068 0.071 0.071 0.071 0.073 0.073 (0.041) (0.042)* _(0.042)* _(0.041)* _(0.032)* _(0.042)* R&D 0.119 0.125 0.124 0.121 0.127 0.126 (0.041)*** _(0.042)*** _(0.042)*** _(0.042)*** _(0.043)*** _(0.043)*** Size 0.009 0.008 0.009 0.010 0.009 0.009 (0.003)*** _(0.003)** _(0.004)** _(0.003)*** _(0.003)*** _(0.003)*** Var 0.512 0.510 0.513 0.520 0.519 0.522 (0.429) (0.446) (0.445) (0.424) (0.441) (0.439) Invest 0.053 0.054 0.055 0.053 0.055 0.056 (0.016)*** _(0.016)*** _(0.017)*** _(0.016)*** _(0.016)*** _(0.016)*** Constant 0.150 0.128 0.120 0.109 0.085 0.078 (0.061)** (0.063)** (0.063)* (0.059)* (0.061) (0.060) R-square 0.170 0.167 0.167 0.169 0.166 0.167 F-test 8.83*** _8.76*** _8.72*** _8.46*** _8.38*** _8.34*** Observations 1188 1177 1177 1188 1177 1177

The dependent variable is the Return On Assets (ROA) of subsequent year. Specifications 1 to 3 examine the relation between a firm’s ROA and average board meeting attendance by its directors, while specifications 4 to 6 test the relation between ROA and average authorized meeting attendance. Each regression includes year and two-digit SIC code (industry) dummies. Numbers in parentheses are standard errors based on robust standard errors clustered at the firm level.

*

Signiﬁcance at the 10% level.

** _{Signiﬁcance at the 5% level.} ***_{Signiﬁcance at the 1% level.}

16 _{Because own meeting attendance plus authorized meeting attendance plus}

meeting absence is identically equal to one, we do not report regressions against board meeting absence inTables 7–9but they are available upon request.

(11)

more powerful. They are consistent with the hypothesis that there is no substantial conflict of interest between controlling ers and minority shareholders. Domestic institutional sharehold-ings show a significantly positive impact which is consistent with previous findings (see, for instance,Filatotchev et al., 2005). We also note that in Taiwan family firms perform better than the other firms on average but widely-dispersed firms perform worse. However, these differences are statistically insignificant, as indi-cated by the coefficients of Family and Widely dummies in columns 2–3 and 5–6 ofTable 7.

The majority of control variables are statistically significant to firm performance. Leverage, R&D over sale and investment oppor-tunity are negatively and significantly related to firm performance across all specifications, which are similar to the previous findings (Cui and Mak, 2002; Yeh and Woidtke, 2005). Firm size is positively correlated to firm performance and significant at the 1% level. Note that one of qualification variables, director who is a lecturer or above in business related field, is positively related to firm perfor-mance and significant at the 1% level. Recalling that the standard deviation of Qual1 is 0.233 (seeTable 3), we find fromTable 7that a one standard deviation improvement in director Qualification 1 is associated with a better subsequent ROA by about 0.169 times of a standard deviation. This is in the similar order of the effect of direc-tors’ own meeting attendance. For the rest of variables, i.e., the var-iance of stock return, woman director, board size and duality of CEO, they are all insignificantly associated with firm performance.

To examine the effects of meeting attendance of different direc-tors, we further investigate the effects of board meeting attendance on company performance under various ownership structures. To this end, we consider three groups of companies: family controlled firms, widely dispersed firms and ultimate controlled firms. For each group of firms, we run multiple OLS regressions of (2) and each regression is against the meeting attendance of a particular type of directors. Indep in (2) now is the average independence ra-tio of this type of directors within a company. Panel A ofTable 8

reports the results of family controlled firms, where we separately test the attendance of four types of directors: family, manager, independent and gray directors. As inTable 7, each director’s meet-ing attendance is further decomposed into their own attendance and authorized meeting attendance. The results show that meeting attendance by family directors is positively related to firm perfor-mance and significant at the 10% level. This result can be inter-preted as the evidence that more involvement of controlling shareholders may mitigate the agency problem between share-holders and managers. The effect of meeting attendance by inde-pendent directors is also positively significant. In contrast, the authorized meeting attendances for all four types of directors have a negative but insignificant impact on the performance of family firms. For Indep, only the average independence ratio of indepen-dent directors has a significant effect on performance, i.e., the high-er is the independence of independent directors the betthigh-er the company performs.

Table 8

ROA and board meeting attendance of different directors under various ownership structures.

Family director/Ultimate director Manager director Independent director Gray director Panel A: Family controlled ﬁrm

Own attendance 0.037 0.015 0.033 0.012 (0.019)* _(0.049) _(0.019)* _(0.016) Authorized attendance 0.028 0.053 0.016 0.041 (0.036) (0.089) (0.041) (0.034) Indep 0.044 0.048 0.002 0.002 0.121 0.134 0.008 0.003 (0.060) (0.061) (0.034) (0.035) (0.039)*** _(0.041)*** _(0.029) _(0.029)

R-square 0.155 0.153 0.308 0.309 0.318 0.311 0.353 0.358 F-test 6.29*** 6.20*** 3.34*** 3.36*** 4.34*** 4.20*** 4.97*** 5.06*** Observations 625 625 299 299 331 331 364 364

Panel B: Widely dispersed ﬁrm

Own attendance 0.030 0.032 0.052 (0.068) (0.040) (0.037) Authorized attendance 0.127 0.044 0.078 (0.122) (0.077) (0.060) Indep 0.055 0.050 0.060 0.065 0.111 0.106 (0.072) (0.072) (0.172) (0.169) (0.055)** _(0.056)*

Other controls Yes Yes Yes Yes Yes Yes

R-square 0.350 0.354 0.256 0.256 0.279 0.279 F-test 4.17*** 4.24*** 2.46*** 2.45*** 3.89*** 3.89*** Observations 298 298 261 261 376 376

Panel C: Ultimate controlled ﬁrm

Own attendance 0.051 0.011 0.044 0.0004 (0.018)*** _(0.043) _(0.016)*** _(0.016) Authorized attendance 0.029 0.065 0.043 0.028 (0.032) (0.076) (0.032) (0.026) Indep 0.029 0.036 0.015 0.011 0.128 0.143 0.004 0.003 (0.047) (0.048) (0.031) (0.031) (0.029)*** _(0.032)*** _(0.028) _(0.027)

R-square 0.145 0.140 0.284 0.286 0.299 0.289 0.275 0.278 F-test 6.97*** 6.71*** 3.63*** 3.67*** 4.81*** 4.56*** 5.63*** 5.73*** Observations 742 742 357 357 405 405 452 452

The dependent variable is the Return On Assets (ROA) of subsequent year. Each regression includes year and two-digit SIC code (industry) dummies. Numbers in parentheses are standard errors based on robust standard errors clustered at the ﬁrm level.

*_{Significance at the 10% level.} **_{Significance at the 5% level.} ***_{Significance at the 1% level.}

(12)

Panel B ofTable 8reports the results of widely dispersed firms, which have manager directors, independent directors and gray outsiders. None of these directors’ own meeting attendance or authorized attendance has any significant impact. However, the independence of gray directors has a positive impact on firm performance, significant at the 5% or 10% level. The lack of

significance of board meeting attendance is a bit surprising. Never-theless, it may be premature to conjecture that directors do not play their roles in widely dispersed companies as in family con-trolled firms. If we compare independent directors’ own meeting attendance in widely dispersed firms (see Panel B ofTable 8) with its counterpart in family controlled firms (Panel A ofTable 8), we

Table 9

Performance and board meeting attendance: Pooled regression.

1 2 3 4 5 6

Panel A: Earnings per share

Own attendance 1.498 1.559 1.569 (0.615)** _(0.629)** _(0.630)** Authorized attendance 2.844 2.951 2.994 (1.207)** _(1.221)** _(1.219)** Indep 0.988 0.634 0.549 1.034 0.675 0.611 (1.172) (1.157) (1.156) (1.170) (1.155) (1.152) Family 0.193 0.159 (0.247) (0.247) Widely 0.141 0.124 (0.245) (0.244) Cashﬂow 1.538 1.494 (0.795)* _(0.796)* Excess 1.318 1.300 (1.515) (1.504)

R-square 0.249 0.245 0.245 0.251 0.247 0.247

F-test 10.05*** _10.14*** _10.11*** _9.90*** _9.99*** _9.99***

Observations 1188 1177 1177 1188 1177 1177

Panel B: Sales to assets ratio

Own attendance 0.456 0.465 0.464 (0.172)*** _(0.175)*** _(0.175)*** Authorized attendance 0.350 0.393 0.374 (0.270) (0.275) (0.274) Indep 0.410 0.343 0.387 0.423 0.355 0.402 (0.222)* (0.231) (0.226)* (0.224)* (0.232) (0.228)* Family 0.052 0.056 (0.056) (0.056) Widely 0.009 0.012 (0.058) (0.058) Cashﬂow 0.077 0.081 (0.169) (0.171) Excess 0.356 0.351 (0.306) (0.207)

R-square 0.254 0.255 0.254 0.250 0.250 0.249

F-test 6.58*** _6.23*** _6.33*** _6.26*** _5.80*** _5.96***

Observations 1188 1177 1177 1188 1177 1177

Panel C: Sales growth

Own attendance 0.159 0.163 0.161 (0.074)** _(0.075)** _(0.075)** Authorized attendance 0.190 0.196 0.195 (0.133) (0.133) (0.132) Indep 0.042 0.061 0.046 0.046 0.065 0.051 (0.104) (0.104) (0.102) (0.104) (0.104) (0.102) Family 0.010 0.012 (0.019) (0.019) Widely 0.033 0.034 (0.021) (0.021) Cashﬂow 0.015 0.016 (0.069) (0.069) Excess 0.059 0.062 (0.101) (0.101)

R-square 0.061 0.064 0.066 0.060 0.063 0.065

F-test 3.54*** _3.60*** _3.86*** _3.34*** _3.43*** _3.71***

Observations 1188 1177 1177 1188 1177 1177

The dependent variables are the Earnings Per Share (EPS) of subsequent year, Sales to Assets Ratio (Sales) of subsequent year and Sales Growth Ratio (Growth), respectively. Specifications 1 to 3 examine the relation between a firm’s performance and average board meeting attendance by its directors, while specifications 4 to 6 test the relation between performance and average authorized meeting attendance. Each regression includes year and two-digit SIC code (industry) dummies. Numbers in parentheses are standard errors based on robust standard errors clustered at the firm level.

*_{Signiﬁcance at the 10% level.} ** _{Signiﬁcance at the 5% level.} ***