01- board composition and firm value

Upload: anonymous-nosl188b

Post on 09-Apr-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/8/2019 01- Board Composition and Firm Value

    1/24

  • 8/8/2019 01- Board Composition and Firm Value

    2/24

    1. Introduction

    Corporate boards of directors have the fiduciary duty to ensure that the management acts in the interest of all shareholders. However, there is mounting evidence questioningthe ability of boards of directors to mitigate agency problems in corporations. Therefore,many view boards of publicly traded firms to be relatively passive entities, oftendominated by the managers whom they are charged with monitoring.

    A properly constituted board should effectively monitor management and help enhancefirm value. The effect of board composition on firm value has been addressed by a number of researchers. 1 In studies examining U.S. firms, there is little evidence that firms withmore independent boards outperform other firms. For example, Morck et al. (1988) ;Hermalin and Weisbach (1991) ; Mehran (1995) , and Klein (1998) report an insignificant

    relation between corporate board independence and various measures of firm performance.Further, Agrawal and Knoeber (1996) and Bhagat and Black (2001) document a negativerelationship between board independence and firm value using various measures of firm performance including Tobins Q . In particular, by conducting a large sample, longhorizon study, Bhagat and Black (2001) report evidence that low profitability firmsincrease the independence of their boards of directors and firms with independent boardsappear to underperform other firms. Bhagat and Black (1999) suggest that supermajority-independent boards may be suboptimal and that an optimal board contains a mix of inside,independent, and even affiliated directors, who bring different skills and knowledge to the board.

    The effect of board composition on firm value is an issue in many countries. Becausehigh ownership concentration is the norm rather than the exception around the world(LaPorta et al., 1999 ), an important issue that needs to be addressed is the effect that outside directors may have on firm value in the presence of a high degree of ownershipconcentration. Canadian capital markets provide an opportunity to study the relationship between firm value and board composition in the presence of significant ownershipconcentration. Similar to the United States, Canada is a country where public equitymarkets are well developed. At the same time, however, a large number of publicly tradedfirms in Canada are controlled by a large block holder or an affiliated group of investors.By using Canadian data, it may be possible to determine the effectiveness of outsidedirectors in the presence of dominant shareholders.

    When ownership is concentrated in a firm, there are potential conflicts of interest between the dominant shareholders and dispersed minority shareholders. That is,controlling shareholders may expropriate wealth from minority shareholders. Canadianlawmakers provide minority shareholders with various forms of legal protection fromdominant shareholders (see Cheffins, 1999 , for a review). 2 Dyck and Zingales (in press)show that the private benefits of control, measured by the control block premium, are

    1 See Hermalin and Weisbach (2002) for a survey of the literature on the board of directors.2

    An example of the protection of minority shareholders is the Ontario Security Commission regulations that stipulate that a major transaction between a corporation and a related party such as a controlling shareholder must be disclosed in detail to all investors, scrutinized by a committee of outside directors, and approved by a majorityof all disinterested shareholders (see Ontario Securities Commission Policy No. 9.1, Part V, 1994 ).

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410388

  • 8/8/2019 01- Board Composition and Firm Value

    3/24

    present but are relatively small for Canadian firms suggesting that Canada does have arelatively transparent governance system when considering both the legal and extra-legalenvironment. 3 Thus, Canadian boards operate in a unique jurisdiction where ownership ishighly concentrated but there is considerable protection of minority shareholders. In viewof this, our paper addresses the following question: Does an improvement in boardcomposition enhance firm value under concentrated ownership?

    The contribution of this paper is to show that in an environment characterized byownership concentration and significant outside shareholder protection, attention to boardcomposition can improve firm value. Our findings suggest that firm value declines whenadditional outside directors are appointed without regard to their expertise. We also findthat ownership concentration reduces firm value. However, when outside directors who areofficers of financial institutions are on the board or the proportion of outside directors on

    the audit committee increases, the negative impact of concentrated ownership on firmvalue is mitigated. This suggests that board composition may affect how well the boardmonitors management.

    The remainder of this paper is organized as follows. Section 2 discusses the relevant literature and develops the research questions. Section 3 reviews the Canadian corporategovernance environment. Section 4 provides an overview of results. Section 5 provides adescription of the model, while Section 6 includes a description of the data. Employing both univariate analysis and multivariate regression analysis, Section 7 investigates theeffect of board composition on firm value. Section 8 concludes the paper.

    2. Related literature

    Boards of directors are typically composed of inside directors and outside directors.Outside directors are often viewed as representing outside shareholders while insidedirectors are assumed to represent the management or controlling shareholders. However,it is unclear whether outside directors represent outside shareholders more effectively thanmanagement or controlling shareholders. The role of outside directors in the protection of shareholders has long been a subject of much debate. Fama (1980) and Fama and Jensen(1983) observe that outside directors compete in the outside directors labor market. Theyhave incentives to develop reputations as experts in monitoring management because thevalue of their human capital depends primarily on their performance as monitors of the topmanagement of other organizations. However, the empirical evidence on the monitoringeffectiveness that outside directors provide is somewhat mixed. While several authors findthat independent outside directors protect shareholders in specific instances in which thereis an agency problem ( Weisbach, 1988; Byrd and Hickman, 1992; Xie et al., 2003 ), othersfind either no relation or a negative relation between outside directors and shareholder welfare.

    3 Dyck and Zingales (2004) note, however, that the private benefits of controlling shareholders such as prestigeand the reputation of being a controlling shareholder of a public firm do not always reduce firm value.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410 389

  • 8/8/2019 01- Board Composition and Firm Value

    4/24

  • 8/8/2019 01- Board Composition and Firm Value

    5/24

    Outside directors in Canada operate in a different environment than those in the UnitedStates. Many Canadian CEOs are the founders and controlling shareholders of the firmsthey manage, unlike in the Un ited States and U.K. where most public companies are widelyheld by individual investors ( Daniels and Halpern, 1996 ). Since the agency problem faced by minority shareholders is particularly acute under strong inside ownership concentration,improvement in board composition may have a significant effect on firm value. Seen in thislight, our investigation deals more specifically with whether outside directors play asignificant role in improving firm value even in the presence of large block holders.

    While outside directors are expected to curb agency costs, those with financial expertisemay be able to do so more effectively. Using evidence on earnings management, Xie et al.(2003) document that the monitoring that outside directors provide improves when theyare financially sophisticated. Since officers of financial intermediaries are financially

    sophisticated, the expectation is that they can be particularly helpful to the board inreducing agency costs ( Booth and Deli, 1999 ).In addition, outside directors dispatched from Canadian financial institutions are more

    likely to have a greater interest in and greater influence in monitoring management because of the strong concentration of financial power in a relatively small number of institutions. These institutions are large both in absolute terms and relative to the size of the Canadian economy. The industry concentration became more pronounced after 1987,when the structure of Canadian financial institutions moved from a separate system to auniversal banking system. For the study period of 1993 to 1997, the seven largest deposit-taking institutionsthe Big Six banks (Bank of Montreal, Bank of Nova Scotia, Canadian

    Imperial Bank of Commerce, National Bank of Canada, Royal Bank of Canada, andToronto-Dominion Bank) and Movement des Caisses Desjardins of Quebecspan allaspects of financial intermediation. In 1999 the big six banks of Canada managed assets of C$277, C$263, C$227, C$226, C$215, and C$72 billion, respectively. In the same year Morgan Stanley MSCI Country Statistics reports an equity capitalization of C$783 billion(U$502 billion) for Canada.

    The market concentration in the life insurance industry is more extreme. Among lifeinsurers five independent companiesManufacturers Life Insurance, Sun Life Assurance,Clarica Life Insurance, Great West Life Assurance, and Canada Life Insurance accounted for more than 90% of the life insurance assets by the end of 1997. Pensionassets are also highly concentrated. The largest three pension fundsCaisse de Depot et Placement du Quebec (CDPQ), Ontario Teachers Pension Plan Board (Teachers), andOntario Municipal Employee Retirement System (OMERS)dominate the pension fundmarket. Finally, the highly visible and concentrated investment banking industry wasessentially absorbed by the big banks.

    With these kinds of resources, large financial institutions are able to provide credibleexternal monitoring of the managers of firms in which they invest. Because of a limitednumber of investment candidates in the Canadian capital market, large financialinstitutions can effectively monitor the managements of Canadian firms at a competitivemonitoring cost. Furthermore, large Canadian financial institutions have an incentive tomonitor corporate management because it is problematic for these financial institutions tosimply sell underperforming holdings given the relatively small size of the economy wherethey primarily invest.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410 391

  • 8/8/2019 01- Board Composition and Firm Value

    6/24

  • 8/8/2019 01- Board Composition and Firm Value

    7/24

    dependent variables are interrelated. As such we consider each dependent variable as afunction of the other.

    In order to deal with the endogeneity problem that may exist between boardcomposition and firm value discussed earlier (see Bhagat and Black, 2001 ), thespecification of the firm value model in Eq (1) examines firm value at time t as afunction of the lagged independent variables at time t 1.

    Firm value t f ownership concentration t 1 ; dual class of common stock t 1 ; board independence t 1 ;directors from financial institutions on the board t 1 ;audit committee independence t 1 ; board size t 1 ; firm size t 1 ;

    financial leverage t 1: 1

    Ownership concentration is a control variable of considerable importance given that theCanadian economy is structurally different from that of the United States and where highownership concentration is the rule rather than the exception. We anticipate a negativerelation between ownership concentration and the firm value if dominant shareholder expropriation of private benefits exceeds the benefits of dominant shareholder monitoringof executives.

    We also expect a negative relation between the existence of dual class common stock andfirm value since dual class common stock will tend to increase the control of dominant

    shareholders and, therefore, increase their ability to extract private benefits. Furthermore,since dual class common stock indicates a divergence of cash flow rights and control rights,controlling shareholders have greater incentive to extract the private benefits of control inthe presence of dual class common stock as they will bear a smaller fraction of the cost.

    If an independent board can reduce agency problems, there should be a positive relation between board independence and firm value. Likewise, we expect a positive relation between the independence of the audit committee and firm value since an independent audit committee can reduce agency problems in both widely held firms and closely heldfirms. Alternately, the argument that firms add independent directors to placateshareholders in the face of poor performance suggests that a more independent board or committee could be associated with lower firm value.

    Directors who are officers of financial institutions can provide monitoring benefits tothe board because of their financial expertise. Furthermore, they could be moreindependent than other outside directors because, unlike the United States, Canada has ahighly concentrated financial services sector, so there may be less pressure to avoidmonitoring in exchange for business. Therefore, we expect a positive relation betweenthe presence of directors who are officers of financial institutions and firm value.

    We control for board size because Yermack (1996) and others have shown that boardsize has a negative influence on firm performance. Additionally, firm size and financialleverage are widely used in the literature as control variables for firm specific effects. Weincorporate them here to control for firm-specific effects on firm value.

    In order to examine the potential effect of firm value on board independence wealso propose a board independence equation. Similar to Eq. (1), this equation examines

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410 393

  • 8/8/2019 01- Board Composition and Firm Value

    8/24

    board independence as a function of lagged firm value and governance variables to dealwith the endogeneity problem that may exist between firm value and board independence.

    Board Independence t f ownership concentration t 1 ;dual class common stock t 1 ; firm value t 1 ;directors from financial institutions on the board t 1 ;audit committee independence t 1 ; board size t 1 ;firm size t 1 ; financial leverage t 1: 2

    Ownership concentration is likely to be an important determinant of boardindependence as dominant shareholders have significant input on who sits on the board.We anticipate a negative relation between ownership concentration and board independ-

    ence if dominant shareholders attempt to use their voting power to reduce independent monitoring. Since the existence of dual class common stock will also tend to increase thecontrol of the dominant shareholders, we expect the sign on the relation between theexistence of dual class common stock and board independence to be the same as that of ownership concentration.

    We expect a positive relation between the independence of the audit committee and boardindependence. Certainly having more outside directors on an audit committee is likely tomean more outside directors overall. However, it is possible that a firm that feels it isimportant to have strong outside director representation on the board may also feel the sameabout audit committee representation. Likewise, we expect a positive relation between the presence of directors from financial institutions and the degree of board independence sincehaving more outside directors allows more room for a director from such institutions.

    Finally, we examine how changes in firm value are related to contemporaneous changesin board structure variables. Having examined the relation between the variables across time,it would be useful to look at observations taken during the same period to see if there is arelation between the variables when measured concurrently as well. By looking at changes inthe board variables it should be possible to see if the parameter estimates in the earlier equations are significant only in a lead/lag relation or if they are also significant con-temporaneously. First differences are also used as many of the board parameters tend to befairly stable over time. The relation between these variables is provided in Eq. (3) below.

    D Firm value t f D ownership concentration t ; D board independence t ;D directors from financial institution on the board t ;D audit committee independence t ; D board size t ; D firm size t ;D financial leverage t : 3

    6. Data

    Board composition and ownership data are collected from proxy documents returned byCanadian firms found in the Global Vantage database for the period between 1991 and1997. Financial information for firms is collected from the Global Vantage database and

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410394

  • 8/8/2019 01- Board Composition and Firm Value

    9/24

    market information for firms is from the TSE-Western database. Canadian financial firmsare excluded from the sample since they are more closely regulated and their operation isquite different than industrial firms.

    Measures of firm value, ownership concentration, and board structure were calculatedfor the 1993 to 1997 period. The resulting unbalan ced panel consists of 679 firm years(236 unique firms) as shown in Panel A of Table 1 . We then construct a balanced paneldata set using only those firms for which a complete set of variables are available over thefull five years selected for the regression analysis. The paper reports only the results usingthe balanced panel because the results using the unbalanced panel are qualitatively thesame. The balanced panel, which is also shown in Panel A of Table 1 , has a total of 330

    Table 1Sample distribution

    Panel A. Panel data structure

    Year Unbalanced panel

    Balanced panel

    1993 78 661994 102 661995 126 661996 156 661997 217 66Total 679 330

    Panel B. Industry Distribution

    Industry Number of firms Number of firm years Percentage

    Balanced panel

    Unbalanced panel

    Balanced panel

    Unbalanced panel

    Balanced panel

    Unbalanced panel

    Metals and minerals 8 31 40 96 12.12 14.14Oil and gas 13 37 65 117 19.70 17.23Paper and forest products 7 17 35 52 10.61 7.66Consumer products 6 15 30 52 9.09 7.66Industrial products 15 61 75 181 22.73 26.66Real estate and construction 0 0 0 0 0 0Transportation and environment 3 11 15 33 4.55 4.86Utilities 2 7 10 17 3.03 2.50Communications and media 8 24 40 71 12.12 10.46Merchandising 1 11 5 23 1.52 3.39Others 3 22 15 37 4.55 5.45Total 66 236 330 679 100 100

    Panel A shows the distribution of both balanced and unbalanced panel data by year over the full period 1993 1997. Note that results from the unbalanced panel are not reported in the paper since the unbalanced panel givesqualitatively identical results as the balanced panel.Panel B shows the industry distribution of the 66 Canadian firms with complete ownership, board, and accountingdata over the full period 19931997. The data for the firms are collected from the Global Vantage database for

    financial information and the TSE-Western database for market information. Financial firms are excluded due tostructural differences associated with the Canadian financial institutions. The utilities in Canada appear to be far less regulated than is typical in the United States; the standard deviation of the annual returns for the TSE utilitiesis comparable to that of the TSE industrials.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410 395

  • 8/8/2019 01- Board Composition and Firm Value

    10/24

    firm years (66 unique firms observed over five consecutive years from 1993 to 1997).Using the twelve Toronto Stock Exchange in dustry classifications, we show the industrydistribution of the sample firms in Panel B of Table 1 . We use this industry classification inorder to find industry adjusted Q ratios in the following section.

    6.1. Variable definitions

    The data include ownership and governance variables as well as control variables. Thecontrol variables include not only two different measures of firm value, but also firm sizeand leverage. The ownership and governance variables include ownership concentration, board size and independence, the existence of financial directors, outside directors on theaudit committee, and the existence of a dual class common stock issue.

    Ownership concentration is measured by the fraction of voting rights held by thedominant shareholders (M1) where dominant shareholders are defined as those who havethe largest block of shares. We also include a dummy variable indicating the existence of dual class common stock (DC). Dual class common stock can allow a dominant shareholder to attain control of a firm with a disproportionately smaller investment andincrease the ability to extract private benefits.

    The board size (BSIZE) variable is included to control for the potential impact of differential board size. Four additional board variables are included: board independence,the existence of financial directors, outside directors serving on the audit committee, andaverage board tenure.

    Board independence is measured by the fraction of outside directors on the board (BI).While outside directors have been defined in a number of ways in the literature (see, e.g.,Rosenstein and Wyatt, 1990 ) we define company officers, family members of thecontrolling shareholders, and related company officers as inside directors and others asoutside directors.

    Directors who are officers of financial institutions are expected to be better monitors because of their expertise. Furthermore, since they are likely to represent the interest of financial institutions, they may be more independent than other outside directors. Thiseffect is likely to be far stronger in Canada than in the United States given the relativelyhigh concentration of the Canadian financial industry. The representation of financialinstitutions on the board is measured by a dummy variable, FID, which takes on the valueof one if there are one or more directors who are officers of financial institutions. 5

    We also measure the degree of outside influence on the audit committee. Anindependent audit committee could reduce agency problems in both widely held firms andclosely held firms through more effective monitoring. The independence of the audit committee is measured by the proportion of outside directors on the committee (ODAC).

    We use industry adjusted Q ratios and raw Q ratios as measures of firm value. The Qratio is measured as the sum of the market value of equity, the book value of preferredstock and the book value of total debt divided by the book value of total assets. To

    5 Using the number of directors as the proxy for the monitoring of financial directors we obtain essentiallyidentical results as reported in the paper. We further break FID down into variables representing different types of financial institutions. Such variables do not provide significantly different results than FID alone.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410396

  • 8/8/2019 01- Board Composition and Firm Value

    11/24

  • 8/8/2019 01- Board Composition and Firm Value

    12/24

    Table 3Pearson correlation coefficients

    ADJQ M1 DC BSIZE BI FID ODAC FSIZE LEV

    ADJQ 1.000M1 0.132*** 1.000DC 0.311*** 0.259*** 1.000BSIZE 0.204*** 0.163*** 0.083 1.000BI 0.221*** 0.349*** 0.016 0.194*** 1.000FID 0.007 0.107* 0.110* 0.323*** 0.195*** 1.000ODAC 0.117* 0.023 0.079 0.312*** 0.470*** 0.078 1.000FSIZE 0.229*** 0.143*** 0.113* 0.658*** 0.181*** 0.234*** 0.400*** 1.000LEV 0.222*** 0.050 0.014 0.245*** 0.153*** 0.260*** 0.188*** 0.292*** 1.000

    The table shows the correlation between the industry adjusted Q ratio and the lagged explanatory variables usinga panel data consisting of 66 firms observed for five consecutive years from 1993 to 1997. Q ratios are calculatedas the ratio of the sum of the market value of equity, the book value of preferred stock and the book value of debt to total assets. The industry adjusted Q ratio (ADJQ) is calculated by subtracting the industry average Q ratiofrom the firms Q ratio. M1 is the fraction of voting rights of the dominant shareholder. DC is the dummy variablefor dual class common stock. BSIZE is the natural log of the total number of directors on the board. BI is thefraction of outside directors on the board. FID is the dummy variable for the presence of directors who are officersof financial institutions. ODAC is the fraction of outside directors on the audit committee. FSIZE is the natural log

    of net sales. LEV is the ratio of the sum of short-term debt and long-term debt to total assets. The statisticalsignificance is based on the probability that the correlation is equal to zero.

    * Significant at 10 percent.*** Significant at 1 percent.

    Table 2Descriptive statistics

    Variables Mean S.D. Minimum Median Maximum

    ADJQ 0.055 0.695 1.220 0.191 3.893Q ratio 1.094 0.725 0.140 0.919 5.323M1 0.252 0.258 0.000 0.171 0.906DC 0.239 0.428 0.000 0.000 1.000 NUMDIR 10.685 2.904 4.000 11.000 18.000BSIZE 2.327 0.300 1.386 2.398 2.890BI 0.686 0.156 0.286 0.700 0.938FID 0.385 0.487 0.000 0.000 1.000ODAC 0.866 0.162 0.500 1.000 1.000LEV 0.253 0.154 0.000 0.243 0.736SALES 1833 3763 8 684 33,191

    FSIZE 6.471 1.489 2.079 6.527 10.410The table shows descriptive statistics of variables used for the analysis. It uses panel data consisting of 66 firmsobserved for five consecutive years from 1993 to 1997. Q ratios are calculated as the ratio of the sum of themarket value of equity, the book value of preferred stock and the book value of debt to total assets. The industryadjusted Q ratio (ADJQ) is calculated by subtracting the industry average Q ratio from the firms Q ratio. M1 isthe fraction of voting rights of the dominant shareholder. DC is the dummy variable for dual class common stock. NUMDIR is the total number of directors on the board. BSIZE is the natural log of NUMDIR. BI is the fraction of outside directors on the board. FID is the dummy variable for the presence of directors who are officers of financial institutions. ODAC is the fraction of outside directors on the audit committee. LEV is the ratio of thesum of short-term debt and long-term debt to total assets. SALES is net sales in millions of dollars. FSIZE is thenatural log of SALES.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410398

  • 8/8/2019 01- Board Composition and Firm Value

    13/24

    The ownership concentration is also negatively correlated with board independence.The negative correlation appears to suggest that the dominant shareholders avoid outsidemonitoring. However if outside directors are added to the board to placate non-dominant shareholders when the firm does poorly, the negative correlation may merely reflect alesser need for such appeasement when there is heavy ownership concentration. Not surprisingly, there is a fairly strong positive correlation between board independence andaudit committee independence.

    7. Empirical results

    7.1. Firm value analysis

    An examination of firm value should indicate the impact of board composition onminority shareholders. Table 4 shows a univariate analysis between firm value and boardindependence. The observations are ranked according to the level of board independencefor each year. The firms in the quartiles with the greatest and the least independence arechosen for the analysis for each year. Firm value, estimated using industry-adjusted Qratios, is calculated within each quartile for the following year. Tests of differences inmeans between the firm values in the two quartiles are run for the years 1994 to 1997.Despite consistent differentials between the two groups, the annual differences aregenerally not significant due to a small number of observations. However, when

    aggregated over the full period, the results become highly significant. The results in Table4 suggest a potential negative relation between the two variables.

    Moving to multivariate analysis, Table 5 reports the result of estimating the Eq. (1)using the Parks (1967) generalized least squares method for panel data as described in

    Table 4Firm value quartile analysis

    Difference in mean firm value

    Year Top quartile firms in BI Bottom quartile firms in BI Difference

    1994 0.219 [16] 0.204 [16] 0.423 (1.41)1995 0.201 [17] 0.123 [16] 0.324 (1.37)1996 0.185 [16] 0.095 [16] 0.281 (0.94)1997 0.206 [19] 0.400 [17] 0.607 (1.85)*Aggregate 0.214 [68] 0.201 [65] 0.415 (2.84)***

    This table reports univariate analysis of firm value between the top quartile subsample and the bottom quartilesub-sample ranked in terms of board independence (BI). Across the sample period firm value is measured at timet and board composition is measured at t 1. Firm value is measured by the industry adjusted Q ratio and boardindependence is measured by the proportion of outside directors on the board. The industry adjusted Q ratio iscalculated by subtracting the industry average Q ratio from the firms Q ratio. Q ratios are calculated as the ratioof the sum of the market value of equity, the book value of preferred stock and the book value of debt to totalassets. The number of firms in the quartile appears in square brackets to the right of the mean firm value for each

    year. The t -statistics, calculated assuming unequal variances, appear in parentheses underneath the differences.The statistical significance reported is two-tailed.

    * Significant at 10 percent.*** Significant at 1 percent.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410 399

  • 8/8/2019 01- Board Composition and Firm Value

    14/24

    Kmenta (1986, pp. 616625) , which adjusts for cross-sectional heteroskedasticity andautocorrelation. In the first regression, we use the industry adjusted Tobins Q as themeasure of firm value. In the second regression the unadjusted Tobins Q ratio is used asthe dependent variable. All the independent variables are measured at time t 1 whenestimating the Q ratio or adjusted Q ratio at time t .

    The coefficient of dominant shareholder voting rights in the regression of the industryadjusted Q is negative and significant, although the coefficient is not significant in theunadjusted Q ratio regression. The findings suggest that the existence of a dominant shareholder has a significant negative effect on firm value compared to other firms withinthe industry. This result provides some support for the argument that there is expropriationof wealth from minority shareholders to dominant shareholders in the firm. Also in support of this interpretation the existence of dual class common stock (DC) has a highlysignificant negative relation with firm value in both regressions.

    A significant negative coefficient for the board independence variable (BI) indicatesthat firm value has a negative relationship with the proportion of outside directors inCanada. It also implies that outside directors are generally not effective monitors in a

    Table 5Firm value regression analysis

    Ind. variables Industry adjusted Q ratio Q ratio

    Constant 1.618 (5.41)*** 3.061 (8.32)***M1 0.190 ( 2.36)** 0.068 ( 0.91)DC 0.421 ( 13.14)*** 0.365 ( 15.37)***BSIZE 0.373 ( 3.27)*** 0.627 ( 5.87)***BI 1.208 ( 5.72)*** 0.837 ( 5.23)***FID 0.240 (3.56)*** 0.231 (3.52)***ODAC 0.550 (3.90)*** 0.115 (0.70)FSIZE 0.032 ( 2.84)*** 0.015 (0.97)LEV 0.837 ( 6.24)*** 0.843 ( 3.41)*** R-square 0.24 0.44Sample size 264 264

    The table uses regressions on pa nel data to estimate firm value as a function of the lagged board structure. Toestimate the model we use t he Parks (1967) generalized least squares method for panel data as described inKmenta (1986, pp.616625) . M1 is the fraction of voting rights of the dominant shareholder. DC is the dummyvariable for dual class common stock. BSIZE is the number of directors on the board. BI is the fraction of outsidedirectors on the board. FID is the dummy variable for the presence of directors from financial institutions. ODACis the fraction of outside directors on the audit committee. FSIZE is the natural log of total sales. LEV is the ratioof the sum of short-term debt and long-term debt to total assets. The dependent variable for the first regression isthe industry adjusted Q ratio. The industry adjusted Q ratio is calculated by subtracting the industry average Qratio from the firms Q ratio. Q ratios are calculated as the ratio of the sum of the market value of equity, the book value of preferred stock and the book value of debt to total assets. In the last regression unadjusted Q ratios areused as the dependent variable and industry dummies are used to control for industry fixed effects. The regressioncoefficients of industry dummies are not reported. Q ratios are calculated as the ratio of the sum of the market value of equity, the book value of preferred stock and the book value of debt to total assets. The t -statistics in parentheses are adjusted for cross-sectional heteroskedasticity and autocorrelation. The statistical significancereported is two-tailed.

    ** Significant at 5 percent.*** Significant at 1 percent.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410400

  • 8/8/2019 01- Board Composition and Firm Value

    15/24

    Canadian-type governa nce environment. This is con sisten t with the results of Hermalinand Weisbach (1991) ; Agrawal and Knoeber (1996) , and Bhagat and Black (1999) who point out that poorly performing firms are likely to add additional outside boardmembers. This can be explained if poorly performing firms add outside directorssimply to appease unhappy investors rather than for the purpose of increasing firmvalue.

    This Canadian result is consistent with that of Bhagat and Black (2001) who alsodocumen t the negative effect of b oard independence on firm value in the United States.Further, Bhagat and Black (1999) suggest that an optimal board contains a mix of inside,independent, and even affiliated directors, who bring different skills and knowledge to the board. Similar to Bhagat and Black (1999) , we document that a shift in board compositionhas also occurred in Canada from boards dominated by the inside directors to boards

    dominated by outside directors.Our results also suggest that the type of outside director on the board can make adifference. The existence of financial institution directors on the board has a significant positive effect on firm value. This is true for both regressions. This result suggests that inthe Canadian corporate governance environment it is not the proportion of outsidedirectors on the board that affects firm value but the type of director. More specifically, thisresult is consistent with the hypothesis that directors from financial institutions providemonitoring benefits to the firm.

    We note that this result is also consistent with the lender monitoring hypothesisdiscussed in Fama (1985) . The positive and significant correlation between FID and LEV

    in Table 3 suggests that there is a lending relation between financial institutions withwhich FIDs are affiliated and firms where FIDs serve as directors. Therefore, FID mayalso proxy for lender monitoring.

    Moreover, there is also evidence that the greater the proportion of outside directorson the audit committee, the more they contribute to firm value. This result is consistent with the hypothesis that expertise counts for directors since the members of the audit committee are generally chosen from those members of the board that have superior accounting and financial expertise. 7 Thus, even with a high degree of dominant shareholder ownership, adding knowledgeable board members who have greater expertiseand greater incentives to participate in corporate governance will tend to increase firmvalue.

    In unreported regression estimations, we divide the sample into larger firms andsmaller firms by size to investigate any possible differences between larger firms andsmaller firms. The parameter estimates of the sub-samples are consistent with those of the overall regressions. Similarly, unreported regressions examine the non-linear effect of board independence on firm value using the square of the board independence

    7 To determine whether the outside directors on the audit committee in our sample are indeed selected for their financial expertise, we measure the proportion of those outside directors with considerable business and financial

    expertise on the audit committee. We consider unrelated company officers, financial institutions officers, former bankers, consultants, and c orporate directors as well as corporate lawyers as directors with financial expertise.Similar to Xie et al. (2003) , who use essentially the same classification scheme, we find that most of the outsidedirectors on the audit committee (88.8%) in our sample appear to have financial expertise.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410 401

  • 8/8/2019 01- Board Composition and Firm Value

    16/24

    variable (BIP2). Since BIP2 turns out to be highly collinear with BI, we regress BIP2on BI and extract the residuals of BIP2. We then use these residuals in the place of BIP2 in the regression equation. The coefficient estimate of the residuals of BIP2 isnot significant suggesting that the effect of board independence on firm value does not have a non-linear component.

    7.2. Board independence analysis

    A further question is that of the determination of board independence. The prior section showed evidence that board independence reduces firm value in the following period. An obvious corollary would be to determine if firm value impacts boardindependence in the following period. Table 6 examines a univariate analysis in which

    the firms are ranked by firm value and divided into quartiles. The top and bottomquartiles are chosen and board independence is determined for the following period. Wethen test for differences in mean board independence between the two quartiles. Again, by aggregating over all years to deal with the small annual sample, the results areconsistent with the notion that poorly performing firms have more outside directors insubsequent years.

    Table 7 examines board independence in a multivariate regression framework. Wefind that the coefficient of firm value is negative and significant suggesting that weaker firm performance tends to lead to greater board independence in the next year.We also find that board independence is highly negatively related to dominant

    shareholder ownership. The results suggest that dominant shareholders attempt to usetheir voting power to reduce the degree of board independence for entrenchment purposes.

    Table 6Board independence quartile analysis

    Difference in mean board independence

    Year Bottom quartile firms in firm value Top quartile firms in firm value Difference

    1994 0.686 [16] 0.605 [16] 0.081 (1.77)*1995 0.733 [16] 0.658 [16] 0.075 (1.52)1996 0.724 [16] 0.636 [16] 0.088 (1.70)*1997 0.718 [16] 0.683 [16] 0.035 (0.64)Aggregate 0.715 [64] 0.645 [64] 0.070 (2.78)***

    This table reports univariate analysis of board independence between the top quartile subsample and the bottomquartile sub-sample ranked in terms of firm value. Across the sample period, board composition is measured at time t and firm value is measured at t 1. Firm value is measured by the industry adjusted Q ratio and boardindependence is measured by the proportion of outside directors on the board. The industry adjusted Q ratio iscalculated by subtracting the industry average Q ratio from the firms Q ratio. Q ratios are calculated as the ratioof the sum of the market value of equity, the book value of preferred stock and the book value of debt to totalassets. The number of firms in the quartile appears in square brackets to the right of the mean board independence

    for each year. The t -statistics, calculated assuming unequal variances, appear in parentheses underneath thedifferences. The statistical significance reported is two-tailed.

    * Significant at 10 percent.*** Significant at 1 percent.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410402

  • 8/8/2019 01- Board Composition and Firm Value

    17/24

    Board independence is not significantly related to the existence of dual class stock inany of the specifications. Board independence is, however, significantly related to having adirector from a financial institution on the board and having outside directors on the audit committee. We note that both of these regressors could be casually related. For example, if financial directors were added to the board, it would seem reasonable that they would bemore likely to be added to the audit committee given their financial expertise, therebyincreasing audit committee independence.

    If board composition is required to have a strict majority of outside board members,the lower limit of the BI variable will be 0.5 and this variable will be censored at this point. However, the corporate governance regulation during the study period was not coercive but was based on moral suasion. Consistent with this regulatory framework about 12% (41 out of 330) of the observations fall below the b limiting Q observationwith the minimum percentage of outside directors on the board in our sample being28.6% ( Table 2) showing that the BI is not strictly censored at 50%. However, sincethere is a strong tendency for the firms to maintain a majority of outside directorsconsistent with the censoring argument, we re-estimate the board independence

    Table 7Board independence regression analysis

    Ind. variables Model 1 Model 2 Model 3

    Constant 0.437 (9.93)*** 0.443 (8.64)*** 0.389 (4.79)***ADJQ 0.054 ( 7.30)*** 0.077 (4.90)***Q ratio 0.051 ( 5.71)***M1 0.268 ( 12.17)*** 0.235 ( 10.63)*** 0.295 (7.52)***DC 0.007 (0.68) 0.001 (0.14) 0.012 (0.54)BSIZE 0.013 (0.77) 0.062 (3.55)*** 0.009 (0.23)FID 0.025 (3.32)*** 0.036 (3.86)*** 0.034 (1.76)*ODAC 0.269 (6.40)*** 0.228 (5.78)*** 0.298 (5.09)***FSIZE 0.008 (1.49) 0.009 (1.54) 0.011 (1.40)LEV 0.016 ( 0.64) 0.062 (1.72) 0.005 ( 0.09) R-square 0.40 0.46 0.40

    Log likelihood 90.39Sample size 264 264 264

    This table examines board i ndependence at time t as a function of firm value at t 1 using a panel d ata. For Models 1 and 2, we use the Parks (1967) generalized least squares method for panel data as described in Kmenta(1986, pp. 616625) . For Model 3, we use tobit regression where the limit observation is 0.5. The dependent variable, BI, is the fraction of outside directors on the board. ADJQ is the industry adjusted Q ratio. Q ratio is theunadjusted firms Q ratio. The Q ratios are calculated as the ratio of the sum of the market value of equity, the book value of preferred stock and the book value of debt to total assets. The industry adjusted Q ratio is calculated by subtracting the industry average Q ratio from the firms Q ratio. M1 is the fraction of voting rights of thedominant shareholder. DC is the dummy variable for dual class common stock. BSIZE is the natural log of thenumber of directors on the board. FID is the dummy variable for the presence of directors who are officers of financial institutions. ODAC is the fraction of outside directors on the audit committee. FSIZE is the natural log of total sales. LEV is the ratio of the sum of short-term debt and long-term debt to total assets. The tests control for industry fixed effects in the second regression model. However, the coefficients on the industry dummies are not reported. The t -statistics adjusted for cross-sectional heteroskedasticity and autocorrelation appear in parentheses.The statistical significance reported is two-tailed.

    * Significant at 10 percent.*** Significant at 1 percent.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410 403

  • 8/8/2019 01- Board Composition and Firm Value

    18/24

    equation using tobit regression with the limiting value of 50%. The tobit estimationresult, which is reported in the last column of Table 7 , gives qualitatively the sameresult as the OLS panel regression.

    7.3. Contemporaneous relations

    We also examine contemporaneous changes in the board control variables as arobustness check. In the previous tables firm value is significantly related to the prior periods board independence variable while board independence is similarly significantlyrelated to the prior periods firm value. An obvious question, however, is whether therelations survive when examined contemporaneously. To the extent that the prior resultsare robust, the significant relation between the two variables should remain when

    contemporaneous changes are examined. Table 8 examines changes in firm value and board independence. The dual class variable was not included due to the lack of significant year-to-year changes in the variable.

    In Table 8 panel data from 1993 to 1997 is used to estimate the relation. Clearly, thenegative relation between firm value and board independence survives even when lookingat year-to-year changes. Similarly, the positive estimates for directors from financial

    Table 8Effect of changes in board independence on changes in firm value

    Ind. variables D Industry adjusted Q ratio t

    Constant 0.012 (1.44)*D M1 t 0.138 (1.84)*D BSIZE t 0.003 (0.38)D BI t 0.463 (3.15)***D FID t 0.109 (10.11)***D ODAC t 0.444 (4.22)***D FSIZE t 0.045 (1.53)D LEV t 0.155 (1.50)Log likelihood 46.51Sample size 264

    This table examines the effect of chang es in board c omposition variables on changes in firm value using paneldat a. To estimate the model, we use the Parks (1967) generalized least squares method for panel data as describedin Kmenta (1986, pp. 616625) . Change in firm value is measured by the change in industry adjusted Q ratio. Theindustry adjusted Q ratio is calculated by subtracting the industry average Q ratio from the firms Q ratio. Q ratiosare calculated as the ratio of the sum of the market value of equity, the book value of preferred stock and the book value of debt to total assets. D M1 is the change in the voting interest of the largest block. D BSIZE is the change inthe number of directors. D BI is the change in the fraction of outside directors on the board. D FID is the change inFID, which is the change in the number of directors who are officers of financial institutions. D ODAC is thechange in the fraction of outside directors on the audit committee. D FSIZE is the percentage change in the firmsize where the firm size is measured by net sales. D LEV is the change in the financial leverage where the financialleverage is measured by the ratio of the sum of short-term debt and long-term debt to total assets. All changes are

    contemporaneous. The t -statistics adjusted for cross-sectional heteroskedasticity and autocorrelation appear in parentheses. The statistical significance reported is two-tailed.

    * Significant at 10 percent.*** Significant at 1 percent.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410404

  • 8/8/2019 01- Board Composition and Firm Value

    19/24

    institutions and for outside directors on the audit committee also are consistent with our previous estimates.

    7.4. Endogeneity

    In an effort to circumvent the endogeneity problem that affects the OLS estimationmodels, a heuristic approach is used to investigate qualitatively the likely direction of causality. The heuristic consists of comparing the variabilities of the Q ratios and adjusted

    Q ratios with those of board and ownership variables. If the former are clearly larger thanthe latter, it is more plausible that the variation in Q is b caused Q by the variation in boardand ownership structure rather than the converse, i.e., direct causality is more likely thanreverse causality. 8

    Table 9 shows the time series variabilities of the Q ratios as well as the adjusted Qratios as opposed to board and ownership variables of the sample firms. It uses panel dataconsisting of 66 firms observed for five consecutive years from 1993 to 1997. Theestimated time series standard deviations of the Q ratios and the adjusted Q ratios for eachfirm are averaged across these firms. We do the same for the estimated standard deviationsof the board and ownership variables. Finally, we calculate the coefficients of variation by

    8 We thank a referee for the suggestions leading to Table 9 and discussion concerning it.

    Table 9Comparison of time series variations of Q ratios vs. board and ownership variables

    Variables Mean S.D. Coefficient of variation

    Percent of firmswith zero S.D.

    ADJQ 0.055 0.232 4.182 0%Q ratio 1.094 0.254 0.232 0%M1 0.252 0.057 0.226 18%DC 0.239 0.007 0.029 98% NUMDIR 10.685 0.428 0.040 21%BSIZE 2.327 0.069 0.030 3%BI 0.686 0.056 0.082 2%FID 0.385 0.154 0.400 68%ODAC 0.866 0.060 0.069 42%

    The table shows the time series variability of Q ratios as opposed to board and ownership variables of the samplefirms. It uses a panel data consisting of 66 firms observed for five consecutive years from 1993 to 1997. We first estimate the time series standard deviation of Q ratios and adjusted Q ratios for each firm and average these acrossfirms. Then, we estimate the time series standard deviation of board and ownership variables for each firm andaverage these across firms. Finally, we calculate the coefficients of variation by dividing the standard deviation bythe sample average. We also report the percentage of firms with zero time series standard deviations.Q ratios are calculated as the ratio of the sum of the market value of equity, the book value of preferred stock andthe book value of debt to total assets. The industry adjusted Q ratio (ADJQ) is calculated by subtracting theindustry average Q ratio from the firms Q ratio. M1 is the fraction of voting rights of the dominant shareholder.DC is the dummy variable for dual class common stock. NUMDIR is the total number of directors on the board.BSIZE is the natural log of NUMDIR. BI is the fraction of outside directors on the board. FID is the dummyvariable for the presence of directors who are officers of financial institutions. ODAC is the fraction of outsidedirectors on the audit committee.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410 405

  • 8/8/2019 01- Board Composition and Firm Value

    20/24

    dividing the standard deviation by the sample average. We also report the percentage of firms with zero time series standard deviations.

    The CV of the adjusted Q is 4.182 and that of the raw Q ratio is 0.232, which aresubstantially larger than those of the ownership and board variables (with the exception of FID). The percentage of firms with a standard deviation of zero for both the adjusted andraw Q ratios is zero while those for the ownership and governance variables range from2% to 98%. This comparison of variabilities as measured by the coefficient of variation aswell as the percentage of firms with zero standard deviation indicates that Q ratios andadjusted Q ratios are clearly more variable than board and ownership variables. This result is consistent with the interpretation that direct causality is more likely than reversecausality.

    An alternative and more direct way to deal with endogeneity in this context is to carry

    out an event study.9

    For example, an event study which examines the effect of announcements of outside directors joining the board may allow us to better determinethe existence or non-existence of direct causality. Using the announcements of director appointments for the study period of 1993 to 1997 found in the CBCA (Canadian Businessand Current Affairs) as well as Canadian Newsstand databases we identify 11 usableappointment announcements of outside directors. 10

    We use the standard market adjusted return model to find the cumulative averageabnormal return (CAAR) of the sample where the market index used is the TSE 300 TotalReturn Index. 11 The abnormal return is simply the total daily return of the issue minus thecorresponding market return. A cumulative abnormal return over [ 1, 1] event window

    was calculated for each firm. The 3-day CAAR is 0.848%, which is statisticallyinsignificant ( p-value is 0.16). This result, albeit based on a very small sample size, is not inconsistent with the result from the regression analysis, which indicates that outside boardappointments in the period did not improve firm value. 12 Clearly, however, this event study does not address reverse causality at all. Furthermore, the very small number of usable announcements found in the CBCA and Canadian Newsstands databases does not allow us to investigate the announcement effect of directors from financial institutions joining the board or events which reduce inside voting interests.

    9

    Again, we thank the referee for this suggestion regarding how we might better address the endogeneity problem.10 Using search tools, we identified 210 new director appointment announcements in the databases during the

    study period. After removing announcements made by firms which are not found in the TSE/Western databaseand inside director announcements related to appointments of CEOs or other officers of the firm to the board aswell as announcements of outside director appointment which occur simultaneously with other major corporateannouncements, the sample of outside director announcements is reduced to 11.11 The TSE 300 daily price index is a float weighted price index of the top 300 (ranked by dollar value of float

    outstanding) stocks listed on the Toronto Stock Exchange. The TSE Total Return Index is identical to the TSE 300Price Index in terms of securities included. However, the value of the index incorporates dividends as wellchanges in price. When we use the CFMRC (Canadian Financial Market Research Center) value weighted indexto calculate the market return adjusted abnormal return, we also get qualitatively the same result. Finally, when we

    use the market model, we get qualitatively the same result.12 We believe the reasons why we capture only 210 announcements are (1) most Canadian firms did not d advertise T these announcements in the newspapers during the period and (2) the CBCA and Newsstand data bases did not make an effort to capture director announcements during the period.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410406

  • 8/8/2019 01- Board Composition and Firm Value

    21/24

    A further robustness check is to use two-stage least squares regressions to examine the potential endogeneity of board independence. While simultaneous regressions can dealwith the bias induced in OLS results by endogeneity, they are often more sensitive thanOLS to model misspecification. Furthermore, since it is difficult to identify appropriateinstrument variables to carry out the two stage regression analysis correctly for the twoequations on hand, we rule out this approach.

    7.5. Outside directors and dual class common stock

    Given the strong significance of dual class stock in the firm value regressions, whichlends support to the dominant shareholder expropriation hypothesis, we pursue a more in-

    Table 10Firm value, outside directors and dual class common stock

    Ind. variables Model 1 Model 2 Model 3 Model 4

    Constant 1.830 (5.81)*** 1.663 (5.73)*** 1.752 (5.77)*** 1.480 (4.57)***M1 0.227 ( 2.62)** 0.148 ( 1.97)** 0.192 ( 2.27)** 0.179 (2.32)**DC 1.618 ( 7.81)*** 0.591 ( 13.73)*** 1.646 ( 4.50)*** 0.096 (0.40)BSIZE 0.365 ( 3.59)*** 0.363 ( 3.16)*** 0.296 ( 2.49)*** 0.367 (3.10)***BI 1.554 ( 6.45)*** 1.151 ( 5.47)*** 1.169 ( 5.56)*** 1.238 (5.47)***DC*BI 1.806 (6.31)***FID 0.221 (4.50)*** 0.146 (1.84)* 0.207 (3.05)*** 0.231 (3.41)***DC*FID 0.405 (4.66)***

    ODAC 0.532 (4.09)*** 0.472 (3.10)*** 0.219 (1.25) 0.571 (3.74)***DC*ODAC 1.385 (3.59)***LTENURE 0.069 (1.55)DC*LTENURE 0.156 (1.27)FSIZE 0.028 ( 2.59)*** 0.031 ( 2.86)*** 0.036 ( 3.73)*** 0.034 (2.88)***LEV 0.783 ( 6.17)*** 0.903 ( 6.50)*** 0.909 ( 6.52)*** 0.805 (7.23)*** R-square 0.27 0.26 0.26 0.24Sample size 264 264 264 264

    The models of firm value as a function of outside dir ectors in dual class common stock firms use balanced paneldat a. To estimate the model, we use the Parks (1967) generalized least squares method for panel data as describedin Kmenta (1986, pp. 616625) . The panel consists of 66 Canadian firms observed over the 19931997 period.

    We measure firm value at time t and explanatory variables at t 1. We measure the firm value using the industry-adjusted Q ratio. The industry adjusted Q ratio is calculated by subtracting the industry average Q ratio from thefirms Q ratio. Q ratios are calculated as the ratio of the sum of the market value of equity, the book value of preferred stock and the book value of debt to total assets. M1 is the fraction of voting rights of the dominant shareholder. DC is the dummy variable for dual class common stock. BSIZE is the natural log of the number of directors on the board. BI is the fraction of outside directors on the board. FID is the dummy variable for the presence of directors who are officers of financial institutions. ODAC is the fraction of outside directors on theaudit committee. LTENURE is the natural log of the average tenure of all directors. FSIZE is the natural log of net sales. LEV is the ratio of the sum of short-term debt and long-term debt to total assets. In all regressions industryadjusted Q ratios are used as the dependent variable. The interaction variables tested are as follows: DC *BI is theinteraction of DC and BI; DC *FID is the interaction of DC and FID; DC *ODAC is the interaction of DC andODAC; DC *LTENURE is the interaction of DC and LTENURE. The t -statistics adjusted for cross-sectional

    heteroskedasticity and autocorrelation appear in parentheses. The statistical significance reported is two-tailed.* Significant at 10 percent.

    ** Significant at 5 percent.*** Significant at 1 percent.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410 407

  • 8/8/2019 01- Board Composition and Firm Value

    22/24

    depth analysis of the effect o f the boar d composition on the relation between firm valueand the dual class variable. Table 10 reports the regression results of four alternativespecifications using the adjusted Q ratio as the dependent variable.

    In Model 1 the interaction of dual class common stock and board independence isexamined by including an interaction term between the two. 13 The coefficient estimate of DC*BI is positive and significant. Indeed the introduction of DC *BI actually increases thesignificance of BI. The positive interaction term suggests that greater board independencecan alleviate some of the agency problems associated with dual class common shares.Although independent directors do not appear to benefit the firm overall, they appear to beable to reduce some of the loss in value from the existence of dual class common equity.

    In Model 2, we test the hypothesis that directors who are officers of financialintermediaries enhance firm value by mitigating the agency problem of concentrated

    ownership by examining the interaction of DC with FID. Much like the boardindependence interaction term, the results indicate that a firm that has a director from afinancial institution will partially alleviate the negative impact of the existence of dualclass common stock on firm value.

    The fraction of outside directors on the audit committee may have a positive effect onfirm value since they are likely to be more independent and more sophisticated inaccounting and finance than other outside directors. The interaction of DC and ODAC inModel 3 is positive and significant, suggesting again that more sophisticated directors tendto temper the negative impact of the existence of dual class common stock.

    Directors with long tenure should become more knowledgeable about the firm and

    could have a positive influence on firm value. Model 4 examines the effect on firm valueof the interaction of DC and LTENURE, where LTENURE is the natural log of the averagetenure of all directors on the board. The coefficient of the interaction term lackssignificance. This suggests that longer tenured directors do not reduce agency problems indual class common stock firms.

    For three of the four models in Table 9 , the results are highly suggestive that a properlydesigned board can alleviate some of the agency problems associated with dual classcommon shares. The evidence in Table 9 suggests the complexity of the interactions of the board composition variables.

    8. Summary and conclusions

    Even in an economy with strong legal protections for shareholders, a well-structured board can enhance shareholder value through monitoring managerial behavior. Further-more, a well-structured board is not necessarily an outsider-dominated board as iscommonly assumed. The evidence presented in this paper indicates there is a negativerelation between board independence and firm value. This result is inconsistent with the

    13Since DC *BI is highly collinear with both DC and BI, we also use its residuals as the regressor. First, we

    regress the cross product on both DC and BI, collect the residuals, and then use them as the regressor. Theresiduals of DC *BI are expected to contain all the information which is not collinear with DC and BI. The result is essentially identical.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410408

  • 8/8/2019 01- Board Composition and Firm Value

    23/24

  • 8/8/2019 01- Board Composition and Firm Value

    24/24

    References

    Agrawal, A., Knoeber, C.R., 1996. Firm performance and mechanisms to control agency problems betweenmanagers and shareholders. Journal of Financial and Quantitative Analysis 31, 377397.

    Bhagat, S., Black, B., 1999. The uncertain relationship between board independence and firm performance.Business Lawyer 54, 921963.

    Bhagat, S., Black, B., 2001. The non-correlation between board independence and long-term firm performance.Journal of Corporation Law 27, 231274.

    Booth, J.R., Deli, D.N., 1999. On executives of financial institutions as outside directors. Journal of CorporateFinance 5, 227250.

    Byrd, J., Hickman, K., 1992. Do outside directors monitor managers? Evidence from tender offer bids. Journal of Financial Economics 32, 195221.

    Cheffins, B., 1999. Current trends in corporate governance: Going from London to Milan via Toronto. DukeJournal of Comparative and International Law 10, 542.

    Daniels, R.J., Halpern, P., 1996. Too close for comfort: The role of the closely held public corporation in theCanadian economy and the implications for public policy. Canadian Business Law Journal 26, 11 62.

    Dyck, A., Zingales, L., 2004. Private benefits of control: International comparison. Journal of Finance (in press).Fama, E.F., 1980. Agency problems and the theory of the firm. Journal of Political Economy 88, 288307.Fama, E.F., 1985. What is different about banks? Journal of Monetary Economics 15, 2939.Fama, E.F., Jensen, M.C., 1983. Separation of ownership and control. Journal of Law and Economics 26,

    301325.Hermalin, B.E., Weisbach, M.S., 1991. The effects of board composition and direct incentives on firm

    performance. Financial Management 20 (4), 101 112.Hermalin, B.E., Weisbach, M.S., 2002. Boards of directors as an endogenously determined institution: A survey

    of the economic literature. Economic Policy Review. Federal Reserve Bank of New York.Klein, A., 1998. Firm performance and board committee structure. Journal of Law and Economics 41, 275299.

    Kmenta, J., 1986. Elements of Econometrics, 2nd ed. MacMillan, New York, NY.LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., 1999. Corporate ownership around the world. Journal of Finance

    54, 471517.LaPorta, R., Lopez-de-Silanes, F., Shleifer, A., 2000. Investor protection and corporate governance. Journal of

    Financial Economics 59, 327.Mehran, H., 1995. Executive compensation structure, ownership, and firm performance. Journal of Financial

    Economics 38, 163184.Morck, R., Shleifer, A., Vishny, R., 1988. Management ownership and market valuation: An empirical analysis.

    Journal of Financial Economics 20, 293315.Ontario Securities Commission Policy No. 9.1, Part V, 1994 Canadian Security Law Reporter (CCH), 471901.Parks, R.W., 1967. Efficient estimation of a system of regression equation when disturbances are serially and

    contemporaneously correlated. Journal of American Statistical Association 62, 500509.

    Rosenstein, S., Wyatt, J.G., 1990. Outside directors, board independence, and shareholder wealth. Journal of Financial Economics 26, 175192.

    Toronto Stock Exchange Committee on Corporate Governance in Canada, 1994. Where were the directors?Toronto Stock Exchange, Toronto.

    Toronto Stock Exchange, Institute of Corporate Directors, 1999. Five years to the Dey: Report on corporategovernance. Toronto Stock Exchange, Toronto.

    Xie, B., Davidson III, W.N., DaDalt, P.J., 2003. Earnings management and corporate governance: The role of the board and the audit committee. Journal of Corporate Finance 9, 295316.

    Weisbach, M.S., 1988. Outside directors and CEO turnover. Journal of Financial Economics 20, 431460.Yermack, D., 1996. Higher valuation of companies with a small board of directors. Journal of Financial

    Economics 40, 185212.

    J. Erickson et al. / Pacific-Basin Finance Journal 13 (2005) 387410410