An Empirical Examination of Mutual Fund Boards

Transcription

1 An Empirical Examination of Mutual Fund Boards J. FELIX MESCHKE * First Version: December 15, 2004 This Version: May 15, 2006 * Department of Finance, Carlson School of Management, University of Minnesota, th Ave. South, Minneapolis, MN meschke@umn.edu. I thank my dissertation co-chairs, Jeff Coles and Spencer Martin, and my committee members Stuart Gillan and Mike Lemmon for their unwavering support. I also thank George Cashman, Rich Evans, Ying Huang, Patrick Kelly, Laura Lindsey, and seminar participants at Arizona State University, Georgia Institute of Technology, HEC Montreal, McGill University, University of Cincinnati, University of Miami, University of Minnesota, University of Notre Dame, University of Oklahoma, University of Toronto, and especially at the Securities and Exchange Commission (SEC) for their helpful comments and suggestions.

2 An Empirical Examination of Mutual Fund Boards ABSTRACT Agency conflicts between mutual fund investors and fund sponsors have recently received close attention from regulators and legislators. In response to improprieties regarding pricing calculations and trading deadlines, the Securities and Exchange Commission has adopted a controversial new rule requiring fund boards to have at least 75% independent directors and an independent chair person. This study examines to what extend board independence and director incentives in the mutual fund industry are systematically related to fund expenses, performance, and compliance. It is based on a novel panel dataset that covers about 60% of assets listed in the CRSP mutual fund database over the period from 1995 through The study finds that funds overseen by an independent chair charge lower fees, while the relation between fund fees and the fraction of independent directors varies through time. Both measures of board independence do not affect fund performance in an economically significant way. Funds with higher director ownership and lower unexplained compensation charge lower fees and deliver higher returns. These empirical findings are consistent with both the watchdog hypothesis, stating that greater independence and better incentive alignment causes boards to bargain harder with fund advisors, which results in lower fees, and the clientèle hypothesis, suggesting that some sponsors strive to attract assets from relatively sophisticated investors who are concerned about conflicts of interest and sensitive to expenses. When attempting to empirically distinguish between these competing hypotheses, I find that boards with an independent chair negotiate lower total expenses, but approve higher management fees and that they do not have a lower probability of being subject to litigation by regulators and shareholders. In light of these results, this study finds only limited empirical support for recent initiatives to regulate mutual fund board structure. JEL Classifications: G32, G34, L20

3 I. Introduction Agency conflicts between mutual fund investors and fund sponsors 1 have recently received close attention from regulators and legislators. Fund sponsors and their advisors want to maximize discounted expected profits derived from fees. Fund investors usually do not like to pay these fees, since they reduce the risk-adjusted after-tax holding-period return. This conflict of interest is mitigated through two different mechanisms: First, under the most common contractual arrangement, fund sponsors and their advisors charge investors a percentage of assets under management and therefore have an incentive to increase those assets (and thereby fees) through investment performance. Second, fund investors are represented by the fund s board of directors, which has a fiduciary duty to negotiate with the advisor company in the investors best interest. Financial economists have emphasized the role of market forces, since investors can discipline under-performing fund advisors by redeeming their investments at the net asset value (NAV), thereby reducing the advisor s income derived from fees (Fama and Jensen (1983a); Tkac (2004)). In contrast, the Securities and Exchange Commission (SEC) and Congress emphasize the importance of board oversight, and especially the role of independent fund directors in mitigating conflicts of interest between fund sponsors and investors. In response to improprieties regarding pricing calculations and trading deadlines ( Market Timing and Late Trading ), the SEC and Congress are currently proposing an aggressive reform agenda that would, among many other requirements, mandate fund boards to have at least 75% independent directors as well as an independent chair. 2 1 Fund sponsors, like Fidelity or Vanguard, create mutual funds and operate them through their advisor companies. This article abstracts from the fact that sponsors and their advisor companies are different legal entities and throughout this paper fund sponsor, fund advisor, and fund family are used synonymously. 2 On April 7, 2006 a Federal Appeals Court for the District of Columbia vacated the independent chair and the 75% independent director provision, but withheld the issuance for 90 days to allow the SEC to provide cost estimates of the rule s implementation. Despite the legal limbo surrounding the governance rule, the SEC is currently proposing 1

4 An evaluation of these regulatory and legislative actions is hampered by the lack of structural models yielding predictions consistent with current rule proposals and the lack of empirical studies of fund governance. The few studies that examine the role of mutual fund boards document that more independent boards make better decisions from the fund investors perspective in carrying out discrete tasks such as fee negotiations (Tufano and Sevick (1997); Del Guercio, Dann, and Partch (2003)), protection against market timing (Zitzewitz (2003), and merger approvals (Khorana, Tufano, and Wedge (2004)). A key issue with these empirical results is that they are usually based on the examination of a single cross-section 3. Without the time dimension in the data it is impossible to tell whether more independent boards make decisions better aligned with fund share holder interests because of their greater independence, or whether this relation is largely the result of optimal contracting within the industry. The impact of an independent chair on mutual funds has not been examined at all. This study examines to what extend board independence and director incentives are systematically related to fund expenses, performance, and compliance. To that end, I first collect information on mutual fund boards and directors from mutual funds Statements of Additional Information (SAI) for the time period from 1995 through After obtaining fund-related data from the CRSP mutual fund database and calendar-yearend Morningstar disks, I construct a sample that contains 21,944 fund years, 1,832 board years and 9,227 director years, covering approximately one third of all mutual funds and about 60% of assets listed in the CRSP mutual fund database over the ten-year sample period. Second, I document that funds overseen by an independent chair charge 10 new rules covering many different aspects of fund operations. The U.S. House of Representatives has passed the Mutual Funds Integrity and Fee Transparency Act of 2003, and the Mutual Fund Reform Act of 2004 is now before the U.S. Senate. 3 Chen, Goldstein, and Jiang (2006) investigate directors ownership in mutual funds using data for 2002 and

5 lower fees, while the relation between fund fees and the fraction of independent directors varies through time. Both measures of board independence do not affect fund performance in an economically significant way. Third, I provide evidence that funds with higher director ownership and lower unexplained compensation charge lower fees and deliver higher returns. Fourth, these empirical findings are consistent with two different hypotheses: The watchdog hypothesis states that greater independence and better incentive alignment causes boards to bargain harder with fund advisors, which results in lower fees. Under the clientèle hypothesis, the empirical findings are driven by some sponsors aiming to attract assets from relatively sophisticated investors who are concerned about conflicts of interest and sensitive to expenses. I attempt to empirically distinguish between these competing hypotheses by examining two auxiliary implications of the watchdog hypothesis. Surprisingly, boards with an independent chair negotiate lower total expenses, but approve higher management fees. There is also no evidence that boards which were more independent prior to the mutual fund scandal have a lower probability of being subject to litigation by regulators and shareholders. Finally, I document that the measure of unexplained director compensation proposed by Sevick and Tufano (1997) is negatively related to fund performance and positively related to fees and litigation risk. Overall, the empirical results of this study provide little impetus for the focus of the current debate on restricting mutual fund board structure. This study contributes to the governance literature in several ways. First, in light of the large number of U.S. households that hold mutual funds and the secular trend towards increased mediated ownership of securities, a better understanding of mutual fund governance is important due to the economic impact of the $7.5 trillion dollar industry. The novel dataset constructed for this study allows to jointly examine director independence, director incentives, and external governance that 3

6 often have been studied separately for operating companies. Second, the relatively high degree of disclosure, the large number of funds, and the relative homogeneity of operations (compared to industrial firms) within the mutual fund industry allow to carefully control for the underlying nature of the firm and detect relations between governance variables and measures of performance. Lastly, examining the association between governance characteristics and fund fees, performance, and compliance allows for a first assessment of the empirical foundations for the contested SEC governance rule and new legislation currently before Congress. The remainder of this study is organized as follows: Section II provides institutional background about the mutual fund industry, relates this paper to previous research and develops the hypotheses examined in this study. Section III outlines the sample selection, describes the data, and explains how the variables are constructed. Section IV discusses univariate and multivariate results and how they relate to the hypotheses. Section V concludes. II. Institutional Background, Related Research and Hypotheses Development A. Institutional Background In 1990, one in four U.S. households owned mutual funds, and the entire mutual fund industry had just over $1 trillion in assets. As of March 2006, the industry had grown to $9.3 trillion - almost half of all U.S. household now own mutual funds, one in three households own funds through employer-sponsored retirement plans. 4 Mutual funds have become an important fixture in 4 See ICI Trends In Mutual Fund Investing March 2006 at 4

7 the financial landscape. Mutual fund sponsors organize a mutual fund or a series of funds either as an investment company or an investment trust to provide professional asset management to investors by gathering money from them and investing it collectively in financial assets. As a business enterprise, a mutual fund is no more than a legal shell. 5 The fund delegates all aspects of its investment and management to a separate corporation, an investment adviser, which operates under a contract approved by the fund s board of directors. 6 At least half of the directors need to be legally independent if the fund wants to rely on any of the SEC s Exemptive Rules 7 - the vast majority of mutual funds do and hence only a few boards have less than 50% independent directors. Overall, the complexity of mutual fund business reflects in its dual structure, on the one side consisting of the mutual funds and their shareholders, who are the suppliers of capital, and, on the other side, the investment management companies that manage all aspects of the pool of gathered assets. Mutual fund investors pay different fees. Fees related to distribution and redemption (sales loads) are paid at the time of a specific event, while fees related to management and service are paid by the fund on an annual basis. The management fee must be approved periodically by a majority of the fund s independent directors. 12b-1 fees are continuing distribution charges borne by investors as a percentage of their assets. More recently, many fund sponsors have created funds with several share classes. Although all classes hold the same portfolio securities and have the same portfolio manager, each class has its 5 Mutual funds are regulated by the Investment Company Act of 1940, which was amended in 1970 and 2001, and by Securities and Exchange Commission (SEC) Rules. 6 A mutual fund is overseen by a board of directors if it is organized as an investment company, and overseen by a board of trustees if it is organized as a trust. This legal distinction has no bearing on the analysis conducted in this study, and throughout the paper, the terms director and trustee as well as board of directors and board of trustees are used synonymously. 7 The SEC s Exemptive Rules are provided in the Appendix. 5

8 own distribution and service arrangements, resulting in different class expenses. For example, class A shares typically have a high front-end sales load and a 12b-1 fee that ranges from 25 to 35 basis points, while class B shares and class C shares tend to have no front-end load, but a 12b-1 fee of 1% and a contingent deferred sales charge (CDSC, or back-end load). Class B shares are usually converted into A shares after six to eight years, resulting in a reduction of the 12b-1 fee, while class C shares trigger a CDSC only if an investor redeems them during the first year of investment. 8 B. Related Literature and Hypotheses Development Fama and Jensen (1983a) argue that boards are less important for open-end mutual funds than for operating companies because investors can discipline underperforming fund advisors by redeeming their investment at net asset value (NAV), thereby directly reducing the advisor s fee income. In a competitive market with reasonably informed fund investors, market discipline imposed by their purchase and redemption decisions will sufficiently mitigate conflicts of interest between advisors and investors. In the presence of strong external market control, additional internal governance oversight will constitute a non-binding constraint with regards to fee-setting. The recent call of four former SEC commissioners to relieve mutual fund directors of their responsibility to negotiate fees is consistent with this reasoning 9, and some economists even suggest to contemplate eliminating the legal mandate to have mutual fund boards at all. 10 The case for greater board oversight of mutual funds is based on evidence suggesting that despite detailed disclosure requirements, a sizable fraction of fund investors seem to be ill-equipped 8 For a recent study on fund share classes, see Nanda, Wang, and Zheng (2004). 9 See Mutual fund fees best set by market rather than directors? By Sara Hansard, Investment News, March 13, See Tinker, Tailor, Mutual Fund Adviser... By Paula Tkac, Wall Street Journal, April 12, 2006; Page A14 6

9 to vigilantly monitor fees, performance, and internal policies and to rationally act on available information. Frazzini and Lamont (2006) find that retail investors direct their money to funds which invest in stocks that have low future returns and that investors would have done better sticking with their original investment. Many investors do not seem to understand the amount and impact of mutual fund fees, and especially expenses that are deducted directly from fund assets such as portfolio transaction costs, management fees, and distribution fees (see Barber, Odean, and Zheng (2003)). In addition, the disclosure of soft dollar arrangements, which permit fund managers to receive research and brokerage services by paying higher brokerage commissions, and revenue sharing, whereby advisers pay for broker-dealers to distribute fund shares, is quite limited. Investors who invest in mutual funds through their retirement accounts are often limited by their plan sponsors in moving their money to funds with more attractive fees or performance. In the presence of informational and institutional frictions, board oversight of mutual funds may potentially serve an important economic purpose. Requiring mutual funds to be monitored by boards may mitigate aspects of the agency conflict between fund advisors and shareholders, but this does not automatically imply cross-sectional or time-series variation in board structure. To the extent that the requirements of the Investment Company Act of 1940 constitute a binding constraint on the profit-maximization of fund sponsors, they choose a board structure that complies with legal requirements while minimally impeding sponsor profits. Why would one expect a systematic relation between mutual fund board structure and fund characteristics such as fees or performance? Two different mechanisms may cause such a relation, which I call the watchdog hypothesis and the clientèle hypothesis. According to the watchdog hypothesis, boards are initially designed by fund sponsors to min- 7

10 imally impede sponsor profits, but since the sponsor chooses directors based on incomplete information, some boards may turn out more independent than others. Some boards may increase their independence over time through additional appointments of truly independent directors and other organizational changes. These boards may develop into vigilant watchdogs that scrutinize the sponsor s behavior and negotiate at arms lengths with the advisor company. This improvement of the board s oversight effort and bargaining power results in lower fees, and conceivably even in better risk-adjusted performance, relative to less vigilant boards. In contrast, the clientèle hypothesis argues that a fraction of fund investors is concerned about conflicts of interest and prefers to invest in mutual funds whose governance structure appears to be better equipped to deal with these conflicts. For some sponsors, it will be profit-maximizing to adopt a more independent governance structure in order to attract assets from this clientèle. As an increasing number of investors becomes mindful of conflicts of interest, the clientèle growths and more sponsors find it in there best interest to adjust their governance structure. Both the watchdog and the clientèle hypothesis predict that more independent boards and boards with desirable director incentives are associated with lower fees. The watchdog hypothesis states that greater independence and better incentive alignment causes boards to bargain harder with fund advisors, which results in lower fees. Under the clientèle hypothesis, the negative relation between fees and governance characteristics appealing to investors is due to the fact that relatively sophisticated investors concerned about conflicts of interest are likely to be also more sensitive to fees charged by the fund. Trying to distinguish between watchdog and clientèle hypothesis is particularly interesting in light of the recent controversy surrounding the SEC s governance rule. Empirical support for the 8

11 watchdog hypothesis would give some credence to the regulator s attempt to impose governance characteristics on all fund boards that have caused desirable outcomes for fund investors. Empirical support more consistent with the clientèle hypothesis, however, would cast doubt on promises of lower fees and possibly better risk-adjusted performance as a result of mandating more independent mutual fund boards. The watchdog hypothesis implies that more vigilant boards will bargain not just about total fees, but also about individual elements such as Rule 12b-1 marketing fees and management fees. In addition, a more vigilant board will insist on compliance policies and procedures to mitigate a wide variety of conflicts of interest and incentives for abuse, resulting in a lower propensity of being involved in a shareholder lawsuit. Hence, the watchdog hypothesis has auxiliary implications regarding 12b-1 marketing fees, management fees, and compliance issues, which potentially allows to empirically distinguish between these two competing hypotheses. In summary, this study first examines whether mutual fund fees and performance are systematically related to board characteristics in a manner consistent with the watchdog and clientèle hypothesis. In a second step, the study attempts to empirically distinguish between these two hypotheses by analyzing auxiliary implications of the watchdog hypotheses. III. Data and Methodology A. Sample Selection Mutual fund sponsor can choose to have all their funds be overseen by a unitary board, or to employ a clustered structure, where different funds are overseen by different boards. Clustered boards often have some directors in common. The organizational structure can change over time 9

12 and is only observable from examining all mutual fund disclosure statements associated with a given sponsor. When attempting to construct a representative sample of mutual fund sponsors and their fund governance structures, one faces a tradeoff between economic relevance and appropriate representation of the population due to the fact that the vast majority of assets are managed by several dozen large fund sponsors like Fidelity, Vanguard, American Funds, and others. Tufano and Sevick (1997) examine a cross-sectional sample of the 50 largest mutual fund sponsors, thereby capturing 69% ($1.1 trillion) of all open-end mutual fund assets and 37% of all funds in While a random selection of mutual fund sponsors would yield a more representative sample, the economic relevance would be greatly diminished since many of the large sponsors would not make it into the sample. This study tackles the tradeoff by randomly selecting 400 investment companies from the March 2004 Morningstar Principia Disk. Since large sponsors are associated with multiple investment companies, they have a higher likelihood of being part of the sample, but the random selection of investment companies also includes smaller families. For the sample of 91 sponsors associated with the 400 randomly selected investment companies, I collect data on board characteristics from mutual fund Statements of Additional Information (Form 485APOS and Form 485BPOS), which are Part B of a mutual funds prospectuses and filed with the SEC, but not sent to shareholders unless explicitly requested. For each director on the fund board, I note whether the director is considered an interested person affiliated with the sponsor or advisor company and whether the director chairs the board. I also collect the name, age, principal occupation during the past five years, position held with the trust, length of time served on board, number of portfolios overseen, aggregate dollar range of equity securities in all registered investment companies overseen by the director, and total compensation from the fund 10

13 and fund complex paid to the director. Finally I obtain information on standing board committees and board committee meetings. Share-class level data about the funds associated with the funds sponsors are obtained from the CRSP Mutual Fund Database and supplemented by data from annual Morningstar Principia Discs for the years 1994 through Morningstar data are merged into CRSP, which has a wider coverage, doesn t suffer from survivorship bias and contains a unique fund identifier (icdi), allowing to reliably track fund observations through time. From 2003 on Morningstar provides the names of investment companies (trust names) associated with the funds, which is used to look up the corresponding disclosure statements on the SEC s webpage. Since several variables of interest are contained in both CRSP and Morningstar, the Morningstar data is used to identify data entry errors in CRSP and to complement missing information. To avoid severe multicollinearity concerns arising from regression analysis on the share-class level, this study aggregates all observations to the fund level by using the CRSP portfolio identifier, which is available from 2003 onwards. Share-class level names are stripped of their share class identifier to create a fund name variable, which is used to aggregate to the fund level share classes that where discontinued prior to The fund level observations are then merged to the governance data hand-collected from SEC disclosure statements by using trust and family names. Sponsors with a unified board of directors can be simply linked to funds by family name. For sponsors with clustered boards, Morningstar trust names are necessary to correctly link funds to their boards. For 541 funds that are not covered by Morningstar (such as money market funds) or got discontinued prior to 2003, fund-board associations are identified manually from SEC filings. The final sample consists of 18,945 fund years for the period between 1995 and

14 B. Descriptive Statistics Table I describes the variables used in this study and their data sources. Table II compares the sample assembled for this study to the CRSP universe. Although the sample contains funds from both small and large sponsors, sample funds tend to be larger, somewhat older and exhibit less negative risk-adjusted performance. Panel A displays pooled fund-year observations for the period from 1994 through 2005, while Panels B through D show the results for three different time periods. Average and median fund size, fund age, expenses, 12b-1 marketing fees, and portfolio turnover have increased for both the sample and the CRSP universe. Median risk-adjusted performance (measured by four-factor alpha, as described in the next section) for both data sets has improved from about -2.5% during the 1994 to 1997 period to about -1% for the 2002 to 2005 period. Table III displays summary statistics for sponsors, boards, and directors for the three time periods , , and Panel A shows that over the sample period the dollar amount of assets managed and the number of funds and share classes offered increased. Panel B indicates that the percentage of boards with an independent chair increased from an average 6% for the period to an average 12.6 % for the period.board size, the percentage of non-interested directors, and the percentage of female directors increased as well. Panel C shows that compensation, tenure and age of independent directors increased during the sample period. Table IV provides univariate statistics for sample funds along three different dimensions: Columns 2 and 3 show that funds overseen by an independent chair charge significantly lower fees, are smaller and younger, and have lower portfolio turnover. The median four-factor alpha is significantly higher, but average univariate performance does not differ along this partition. Columns 4 and 5 show that funds overseen by boards with at least 75% independent directors also charge lower fees, although 12

15 the magnitude is not as pronounced and median expenses including one-sevenths of total loads do not significantly differ. Columns 6 and 7 reveal that funds overseen by boards with above-median unexplained compensation charge lower fees and perform better than those overseen by boards with below-median unexplained compensation. Pearson and Spearman rank correlations between expenses, performance, and various board characteristics are reported in Table V. Expenses are negatively related to both measures of board independence and board size, and are positively related to unexplained compensation. Both raw return and risk-adjusted return are negatively related to unexplained compensation. Table VI examines to what extent board independence is related to sponsor characteristics. Whether a board has at least 75% independent directors doesn t seem to be driven by the sponsor characteristics examined in this table. But Table VI suggests that sponsors with a greater fraction of assets distributed by brokers may be less likely to have a board overseen by an independent chair, while sponsors with a greater fraction of assets invested in money market funds may be more likely to employ a board with an independent chair. Overall, this table does not unearth much evidence that board independence is strongly affected by sponsor characteristics. C. Variable Construction and Methodology Following Khorana and Servaes (2004), I calculate the objective-adjusted fund return, OAR i for fund i over the previous 36 months as follows: [ 36 ] 36 M OAR i = (1 + R i,t ) w j,t R j,t 1 (1) t=1 t=1 j=1 where M denotes the number of funds within a given investment objective, R j,t and w j,t are 13

16 fund j s return and weight within the investment object M for month t. R i,t denotes a given fund s return in month t, from which its corresponding investment-objective benchmark return is subtracted to yield the objective-adjusted return, OAR i. 11 Alphas for equity and bond funds are based on a four-factor model. For equity funds, I use the Fama and French (1992) factors (i.e., the excess return on the CRSP value-weighted index, the difference in returns across small and large stock portfolios (SMB), and the difference in returns across high and low book-to-market stock portfolios (HML)), and the Carhart (1997) momentum factor. For bonds funds, I use the excess returns on four Lehman Brothers indices: the government/corporate bond index, the mortgage-backed securities index, the long-term government bond index, and the intermediate-term government bond index. These factor-model specifications are consistent with those in Blake, Elton, and Gruber (1993) and Khorana, Tufano, and Wedge (2005). For money market funds, the style-adjusted excess return is used to estimate alpha. Monthly returns for the previous three years (i.e., 36 return observations) are used to estimate the regression parameters. To be included in the sample, a fund needs to have at least 12 non-missing monthly return observations. The monthly alphas are annualized. We calculated the average return for all MM funds (one of the detail style) each year and then substract that from the raw return to get the excess return for MM fund. To construct the redemption-to-performance sensitivity measure, I first calculate the percentage flow for each fund as 11 Following Khorana and Servaes (2004), I aggregate the 195 Strategic Insight investment objective codes into the following, broader investment objective categories: Specialty Natural Resources, Latin American Equity, Specialty Environment, Specialty Real Estate, Specialty Healthcare, Specialty Technology, Government Mortgage, Specialty Utilities, Aggressive Growth, Growth and Income, Principal Returns, Specialty Finance, Convertible Bond, Foreign Currency, Emerging Equity, European Equity, Government Bond, Corporate Bond, Municipal Bond, Specialty Gold, Canada Equity, Global Equity, Option Income, Money Market, Asia Equity, Global Bond, Small Stock, Balanced, Flexible Growth, and Sector. 14

17 %F low i,t = Assets i,t Assets i,t 1 (1 + r i,t ) Assets i,t 1 (2) for all months t and funds i. To capture the flow relation to negative performance β neg, I estimate the equation %F low i,t = α + β neg I(r i,t 1 < 0)r i,t 1 + β pos I(r i,t 1 0)r i,t 1 + ε i (3) and define the redemption-to-performance sensitivity measure as Redemption Sensitivity = ln(β neg ). (4) Unexplained compensation is calculated following Sevick and Tufano (1997). For each board, unexplained compensation is the average residual obtained from annually regressing director compensation on the number of boards a director sits on and the total assets overseen by that director. The remaining variables used in this study and their data sources a described in Table I. 15

18 IV. Results A. Fund Expenses and Governance Characteristics This section displays regression results of fund expenses and governance characteristics using pooled panel regressions, fixed-effect regressions, and Fama/MacBeth regressions. Petersen (2005) provides simulation evidence suggesting that in the presence of both time-series and cross-sectional correlation the combination of year dummies and clustered (Roger) standard errors avoids rejecting the null hypothesis too often. All OLS regressions in this paper follow this methodology, but to conserve space the coefficients on the annual year dummies are suppressed in the tables. 12 The regression specification controls for other board structure variables, fund and sponsor characteristics, fund investment style and distribution type. The watchdog hypothesis and the clientèle hypothesis both predict a negative relation between mutual fund expenses and measures of board independence. The two measures of board independence examined in this study are the percentage of non-interested directors on the board and an indicator variable equal to one if the chair of the board is an independent director. The contested SEC governance rule would require all boards relying on the exemptive rules to have an independent chair and at least 75% independent directors. Panel A of Table VII shows that funds of boards overseen by an independent chair charge lower fees. In contrast, the fraction of independent directors is not significant for the entire sample. The columns 3 through 5 display results for three subperiods. While the negative relation between an independent chair and lower fees holds for every subperiod, the fraction of independent directors 12 Fixed effects regressions do not control for cross-sectional correlation, while the Fama/MacBeth approach does not control for time-series dependence. These results are displayed as robustness checks. 16

19 is associated with higher fees during the first two subperiods and with lower fees during the last subperiod. The fund-fixed effects regression suggests that the negative relation between an independent chair and lower fees is not just driven by cross-sectional variation - funds that adopt an independent chair also lower their fees. The Fama/MacBeth (1973) approach leads to results that mirror these of the pooled OLS regression in column two. Panels B through D of Table VII display the same regressions separately for equity funds, bond funds, and money market funds. The negative association between an independent chair and fund expenses is more pronounced for equity than for bond funds, and does not exist for money market funds. Sevick and Tufano (1997) examine the relation between board independence and fund expenses for the 1992 cross-section of mutual funds associated with the 50 largest fund sponsors and report a negative regression coefficient. It appears that the results regarding fund expenses and the fraction of independent directors vary depending on the time period examined. If the regression specification adequately isolates the marginal relation between board independence and expenses, than this result is difficult to reconcile with the watchdog hypothesis. At the very least, the relation between board independence and fund fees is inconclusive. In contrast, the result that funds of boards overseen by an independent chair charge lower fees is robust across different time periods and regression specifications and is consistent with both the watchdog and clientèle hypothesis. Table VII also provides evidence that funds charge lower fees if they are overseen by independent directors whose interests are better aligned with those of fund shareholders. The unexplained compensation measure of Sevick and Tufano (1997) is reliably related to expenses. Funds overseen by independent directors who receive compensation not explained by the dollar amount of assets and the number of funds overseen charge higher fees. This finding is consistent with the results 17

20 that Sevick and Tufano (1997) find for the 1992 cross-section. For the period, there is evidence that funds charge lower fees if they are overseen by boards with greater reputation at stake (as proxied by their average tenure) and a larger fraction of directors who hold at least $100,000 in funds assets. In parallel work, Chen, Goldstein, and Jiang (2006) examine fund share ownership for directors of the 50 largest mutual fund sponsors for 2002 and They report that directors hold more shares in actively-managed funds such as equity, small-cap equity, and and lowgrade bond funds and fewer shares in funds with large holdings by institutional investors or funds with strong flow-performance chasing. While Chen, Goldstein, and Jiang (2006) document that directors tend to own shares in funds they oversee when the benefits from monitoring is expected to be higher, this study provides evidence consistent with the watchdog hypothesis which suggests that director ownership creates incentive alignment between shareholders and directors that results in closer scrutiny, harder bargaining with the investment advisor, and, ultimately, lower fees for fund shareholders. But one cannot rule out the clientèle hypothesis, which suggests that certain fund sponsors cater to an investment clientèle concerned about both fees and agency conflicts by offering funds at lower fees and by encouraging directors to invest in the funds they overseen. Some fund families have deferred compensation plans that requires independent directors to invest a portion of their annual compensation with the fund sponsor. As discussed earlier, external market discipline and internal board oversight are expected to both affect the level of fees mutual funds charge. It is therefore important to control for the degree of investor-imposed market discipline when estimating the effects of board characteristics on fees. This is done via the redemption sensitivity measure, which is the natural logarithm of a fund s flow-to-negative performance sensitivity (see Section 3.C for details.) The negative and significant 18

21 coefficient on this measure suggests that funds with an investor clientèle that is more likely to redeem their shares in response to negative performance charge lower fees. Other control variables are largely consistent with economic intuition: Older funds and funds from larger sponsors charge lower fees, while equity, speciality and international funds and funds with higher turnover charge higher fees. The regression model specified explains about two-thirds of the variation in fund fees. To summarize, funds overseen by an independent chair and funds overseen by boards that receive lower excess compensation, have more money invested in the funds they oversee, and more reputation at stake because of their longer average tenure, tend to charge lower fees. These results are broadly consistent with two different hypotheses: Under the watchdog hypothesis, independence and incentives are indicative of better functioning boards, while under the clientèle hypothesis, fees, board independence, and director incentives are jointly driven by the desire of some sponsors to appeal to a group of investors who are concerned about fees and conflicts of interest. B. Fund Performance and Governance Characteristics Other than through negotiating fees, there is no strong theoretical argument that links actions of mutual fund directors to performance of funds they oversee. Sevick and Tufano (1997) note that fund boards have at best an indirect impact on performance and limit their analysis to the relation between board characteristics and fund expenses. However, a concurrent study by Ding and Wermers (2005) examines 80 manager replacement decisions by mutual fund boards during the fiscal year and finds that poorly performing fund managers are more likely to be replaced by funds having larger numbers of outside directors on their boards. To the extend that more independent and better incentivised boards are less tolerant of underperformance, one 19

22 might expect a relation between fund performance and governance characteristics. In addition, one cannot rule out that certain boards of directors are able to identify talented portfolio managers. In that case, evaluating boards solely on how low they negotiate fund expenses may result in an incomplete picture of their activity since one would expect that in equilibrium more talented portfolio managers charge higher fees. Hence, this section examines the relation between mutual fund governance characteristics and fund performance, albeit with a high degree of scepticism. Measuring risk-adjusted fund performance requires an appropriate asset pricing model. One of the commonly used measures is Jensen s alpha, the regression intercept obtained from regressing mutual fund returns in excess of the risk-free rate on the excess market return. In response to evidence that the Capital Asset Pricing Model (CAPM) has limited ability to explain stock returns, researchers turned to multi-factor models such as the Fama/French three-factor model or the Cahart four-factor model. Whether these factors constitute risk factors is subject to debate. Fama and French (1993) put forth an argument for a risk-based interpretation of their size and book-to-market factors, while Daniel and Titman (1997) provide evidence more consistent with a characteristicsbased interpretation of cross-sectional return variation. Cahart (1997) proposes a momentum riskfactor in addition to the three Fama/French factors, while Grundy and Martin (2001) and Griffin, Ji, and Martin (2003) provide results inconsistent with a risk-based interpretation of momentum. Pastor and Stambaugh (2003) construct a liquidity factor and suggest a risk-based interpretation. When employing these factor models, the interpretation of fund performance varies, depending on whether or not one subscribes to a risk-interpretation. Under the risk-interpretation, the regression intercept from a multi-factor model is a measure of excess performance. Otherwise it is more appropriate to interpret the intercept term as a measure of performance relative to what can be 20

23 achieved through mechanic strategies. This study does take a position on the nature of multifactor models. Since fund characteristics may be related to the omitted variables in the excess performance measures employed, I attempt to alleviate the concern by employing several different performance measures. Panel A of Table VIII displays pooled OLS regressions of raw return, style-adjusted return, and four different factor models on board structure variables and fund-level controls for equity funds. There is no evidence that the funds of boards with an independent chair or a larger fraction of independent directors exhibit better performance. In fact, the coefficient on the board independence measure is negative and statistically significant for five out of six regression specification, although the economic magnitude is quite small: A one standard deviation decrease in board independence from 75% to 65% is associated with an increase in performance by a magnitude of 23 to 50 basis points ( %), depending on model specification. There is, however, a reliably negative relation between unexplained compensation and measures of fund performance for both equity funds and bond funds, but again the economic magnitude is small: Depending on model specification, a decrease in performance of one basis point is associated with an increase in unexplained compensation in the order of $50,000 - $100,000. For the period of , the regressions in Panel C and D provide some evidence that a greater fraction of independent directors invests at least $100,000 in the funds they oversee if these funds perform better. It is possible that boards with a greater ownership stake in the funds they oversee affect fund performance. Nevertheless, this finding is also consistent with evidence of performance-chasing behavior by mutual fund directors as reported by Chen, Goldstein, and Jiang (2006). 21

24 The results for the fund variables in Table VIII suggest a hump-shaped relation between fund assets and performance. There is also some evidence that increased competition within an investment style decreases performance, although the magnitude of the relation is very small. The adjusted R 2 of the models indicate that about half of the variation in raw returns can be explained by fund and board characteristics. In contrast, risk-adjusted performance is very hard to explain. Table VIII provides little evidence that more independent or better incentivized boards are associated with better fund performance by an economically significant magnitude. C. Distinguishing between Watchdog and Clientèle Hypothesis Without a structural model of mutual fund governance that can be calibrated to data it is difficult to reliably distinguish between the watchdog and the clientèle hypothesis. Yet this distinction is at the core of the current controversy surrounding the SEC s governance rule. As mentioned earlier, empirical support for the watchdog hypothesis would give some credence to the regulator s attempt to impose governance characteristics on all fund boards that have caused desirable outcomes for fund investors. This study therefore examines two auxiliary implication of this hypothesis. Table IX examines the relation between board characteristics and expenses, expenses plus 1 7 of maximum total loads (following Sirri and Tufano (1998)), 12b-1 fees, and management fees. The watchdog hypothesis predicts that more vigilant boards will bargain not just about total fees, but also about individual elements such as Rule 12b-1 marketing fees and management fees. The tables shows that higher unexplained compensation is positively related to all four measures of fees. In addition, an independent chair is associated with lower total fees that includes part of the front end and back end loads, and lower 12b-1 fees. Yet while boards with an independent chair approve expense rations that are about basis points lower, they also approve management fees that 22

25 are about 6-7 basis points higher. Section 15(c) of the Investment Company Act of 1940 requires fund boards to consider whether to approve the terms of the advisory contract, including the amount of the advisory fee. In a speech before the 2005 Mutual Fund Directors Forum, SEC Commissioner Annette Nazareth singled out the management fee as one of the biggest fee issues that impact investor returns and urged directors to carefully scrutinize the contract and the fees, using a broad range of information that advisers are required to provide. 13 That funds overseen by an independent chairman negotiate lower total expenses yet approve higher management fees, which are part of the total expenses, may casts some doubt on the broad applicability of the watchdog hypothesis. On the other hand, this finding is also consistent with a transparency argument: More independent boards may insist on greater transparency in the fee structure, which may result in higher direct compensation and lower indirect payments to the advisor company. Further work needs to be done to distinguish between these hypotheses. The management fee variable is only available through CRSP from 2003 onwards, which reduces the sample size by about 75% and therefore limits inferences. Table X examines whether more independent, better incentivized boards have a lower propensity of being involved in lawsuits stemming from the 2003 mutual fund scandal. Under the watchdog hypothesis, one would expect that more vigilant board will insist on compliance policies and procedures to mitigate a wide variety of conflicts of interest and incentives for abuse, resulting in a lower propensity of being involved in a shareholder lawsuit. I collect data on law suits from the Securities Class Action Clearinghouse at Stanford Law School and from the Securities and Exchange Com- 13 See Remarks Before the 2005 Mutual Fund Directors Forum, Investment Industry Trading Practices and Best Execution Workshop by Commissioner Annette L. Nazareth, 23

26 mission. The first two regression models of Table X report marginal effects of sponsor-level logistic regressions on an indicator variable equal to one if the fund family faces either SEC or class action litigation between 2003 and Since the results may be blurred by frivolous lawsuits, I also employ a more restrictive measure of legal troubles by only considering lawsuits if Morningstar s stewardship rating indicates a poor or very poor rating in the Regulatory Issues category. The results for this more restrictive variable are displayed in the last two columns of Table IX. The table does not provide any evidence that more independent boards or boards overseen by an independent chair prior to the fund scandal have a lower propensity of litigation. Unexplained compensation (aggregated on the sponsor level) and the fraction of sponsor assets distributed through broker distribution are both positively related to the likelihood of getting sued. Again, this finding does not provide support for the watchdog hypothesis. V. Conclusion Boards are the focus of many attempts to improve corporate governance. For industrial corporations, many institutional investors have joined the call for smaller, more independent boards, among them Institutional Shareholder Services, Inc., the Council of Institutional Investors, TIAA-CREF and CalPERS. For mutual funds, the Securities and Exchange Commission has adopted a new rule that requires 75% of fund directors to be independent. These attempts to improve board performance by mandating a certain board structure assume that boards are systematically organized in an inefficient manner. Recent accounting scandals, high-profile bankruptcies, and improprieties regarding pricing calculations and trading deadlines provide anecdotal support for this view. This study examines to what extend board independence and director incentives are in the 24