Ownership and fund performance evans

Portfolio Manager Ownershipand Mutual Fund Performance

Allison L. Evans*

This paper examines the association between a mutual fiind manager's personal fimd investmentand mutual fund performance. From a data set of newly released managerial ownership disclo-sures. I find that fund ownership levels are diverse and, in many instances, quite large. Mutualfund returns are increasing in the level of managerial investment, consistent with personal own-ership realigning decision-maker and shareholder interests. Also consistent with the reductionof agency costs, ¡find thai managerial ownership is inversely related to fund turnover However,there is no evidence of an association between managerial ownership and a mutual fiind s taxburden.

The purpose of this paper is to determine whether a fund manager's personal ownership ofmutual fund shares is associated with differences in mutual fund performance. Specifically,I explore whether mutual funds with minimally invested managers experience lower retumsand/or higher fund costs than do funds with managers who have a greater direct financialstake. Such a finding would be consistent with fund ownership causing a reduction in agencycosts between managers and fund shareholders in the mutual fund industry, beyond the currentcompensation structure. Over the years 2001 -2004,1 find that managerial ownership is positivelyrelated to fimd retums, inversely related to turnover levels, and unrelated to the fund's taxburden.

The relation between a manager's investment in a business and the performance ofthat businesshas generated interest across the fields of accounting, finance, and economics. Numerous papers,including Jensen and Meckling's ( 1976) seminal work, have analyzed the effects ofthe separationof ownership and control, as well as compensation structures that attempt to realign incentivesbetween the two. Many studies have emerged that examine the relation between personal own-ership and performance in the corporate setting. Studies have examined, for instance, the effectof firm managerial ownership on discretionary accounting adjustments (e.g.. Warfield, Wild, andWild, 1995). disclosure decisions (e.g.. Aboody and Kasznik, 2000; Nagar. Nanda. and Wysocki,2003). dividend policy (e.g.. Lambert, Lanen, and Larker, 1989; Brown, Liang, and Weisbenner,2005; Chetty and Saez, 2005). inventory choices (e.g.. Hunt, 1985; Niehaus, 1989), investmentdecisions (e.g.. Clinch. 1991; Ryan and Wiggins, 2002; Broussard, Buchenroth, and Pilotte,2004), fiiture earnings (Hanlon, Rajgopal, and Shevlin, 2005), and market valuation (e.g., Morck,Shleifer, and Vishny, 1988). This body of literature as a whole does document a relation between

This paper is hased on my dissertation at the University of North Carolina at Chapel Hill. I would like lo thank themembers of my dissertation committee for their valuable guidance and support: Douglas Shackelford (chair), JenniferConrad. Mark Lang. Ed Maydew. and Jana Raedy. I ahn appreciate helpful comments from Courtney Edwards. Jimtrving, Sudarshan Jayaraman. Christine Petniviti. and an anonymous referee. I am grateful to Upper. Inc. for providingcapital gain distribution data and to Kunal Kapoor. Director of Fund Analysis at Morningstar. Inc., for .several helpfuldiscussions.

'Allison Evans is an Assistant Professor of Accounting at Wake Forest University in Winston-Salem, NC.

Financial Management • Autumn 2008 • pages 513 - 534

514 Financial Management • Autumn 2008

ownership and performance in the corporate setting, although that relation is not necessarily astraightforward one. '

With the exception of concurrent research by Khorana. Servaes, and Wedge (2006. hereafterKSW). similar examinations ofthe relation between a manager's share ownership and perfor-mance have not been attempted in the mutual fund arena prior to 2007 because managerial fundownership data have only been available since 2005. Furthermore, since mutual fund managersare not required to own fund shares, many researchers have assumed that their investment issmall. Jin (2006), for instance, assumes that since mutual fund managers are not generally re-quired to own shares in their fund, they have little incentive to care about fiand tax consequences.This statement implies that the lack of an explicit requirement to own fiind shares is equiv-alent to managers not making material investments in their fiind. This assumption, althoughcommon, is contrary to the findings in this study. Ofthe 237 domestic equity mutual fund port-folios in my sample, 22% have managers who have personally invested over $1,000,000 in theirfund.

In addition to being a vehicle to extend the agency literature, mutual funds are important intheir own right. Mutual funds are playing an increasingly large role in domestic equity markets,as documented by the Investment Company Institute (2005, hereafter ICI). The ICI reports thatthe mutual fiind industry managed $8.1 trillion and held approximately 22% ofthe outstandingstock of US companies in 2004. At the same time, regulators and fund investors are increasingtheir focus on the trading behavior of fiind managers after the revelation of several recent tradingabuses. This increased scrutiny culminated in the Securities and Exchange Commission's (SEC)new regulation that requires mutual funds to disclose managerial ownership levels, compensationstructure, and conflicts of interest.^ Any empirical documentationof factors that might influencethe trading decisions of mutual fiind managers will benefit industry regulators, as well as currentand potential fund investors.

The paper is organized as follows. Section I develops the hypotheses. Section II describes thedata and sample selection process. Section III describes the empirical specifications. Sections IVand V present the resuhs fi"om univariate and multivariate tests, respectively. Section VIconcludes.

t. Hypothesis Development

There are multiple factors that could influence the governance of mutual fiinds. Two pri-mary govemance mechanisms are specified in Fama and Jensen (1983). One such factor is theexistence of the fund's board of directors (BOD). The BOD is responsible for matters suchas setting loads and fees, appointing committees, approving mergers, and approving generalinvestment objectives. Several papers study how the structure of the BOD (e.g.. size and in-dependence) affects those items under the board's control (e.g., Tufano and Sevick. 1997;Meschke, 2005; Qian, 2005; Khorana, Servaes, and Wedge, 2006). However, it is the fijnd

'Some studies (e.g., Morck, Shleifer, and Vishny. 1988; Stulz, 1988; McCotinell and Servaes. 1990) find that prof-itability is positively correlated with ownership in the cotT>orate setting up to a certain point, after which entrench-ment issues may actually erode profitability, creating an inverted U-shaped relation between owtiership and profitabil-ity. Coles, Meschke, and Lemmon (2008) find support for this relatioti, although their paper sets forth endogeneityconcerns.

'US Securities and Exchange Commission, 2004, Final Rule: Disclosure Regarding Portfolio Managers of RegisteredManagement Investment Companies, 17 Code of Federal Regulation Parts 239, 249, 270, 274. US Government PrintingOffice, Washington, DC.

Evans » Portfolio Manager Ownership and Mutual Fund Performance 515

manager per se who directly affects the day-to-day trading decisions of the fund, not the

Although Fama and Jensen (1983) specify this potential role the BOD could play in fundgovernance, their study acknowledges that this mechanism is relatively weak in the mutual fundsetting. A more influential form of govemance lies in the hands of fund investors themselves.Since fund shareholders are able to withdraw their investment at any time at the fund's full netasset value (NAV), they have substantial control over the fund. Any shareholder who feels thefund is underpertbrming may, on demand, withdraw the full market value of all shares owned.Such withdrawals could generate significant negative effects on the income of the fund and ofthe fund manager. Thus, portfolio managers have an incentive to act in a manner that will pleasefund investors and thereby minimize investor flight.

Whether managerial fund ownership is associated with performance in the mutual fund settingis an empirical question. Such a result should obtain to the extent that fund ownership alle-viates an incentive conflict between the manager and fund investors. Dow and Gorton (1997)model one such conflict, where a manager who actively searches for profitable trading oppor-tunities and finds none will still execute trades, although such trades might actually decreasefund value. This action results from incentives embedded in the manager's compensation con-tract, formulated based on the inability for outsiders to distinguish informed nontrading from amanager who shirks responsibilities or has no talent to pick good stocks. Thus, Dow and Gortonposit that a manager's compensation creates incentives to make trades, even those that reducevalue.

Although their compensation is based primarily on net assets under management, many fundmanagers also have at least a portion of their compensation (typically the annual bonus) tiedto fund retums. In theory, this compensation structure should help align the desires of fundshareholders (higher returns) with the incentives of fund managers (personal wealth). Only10% of the 237 fund portfolios in this sample disclosed that compensation is not tied tofund performance in some way. If this compensation structure is enough to alleviate any ex-isting agency issues, then fund ownership need not be associated with better fund returns.If, however, the compensation structure does not fully alleviate an agency problem betweenmanagers and shareholders, then I would expect fund retums to be positively associated withthe degree of managerial ownership. The SEC's stated motivation for enacting this disclo-sure requirement, to "help investors assess the extent to which portfolio manager's interestsare aligned with theirs," implies that current compensation arrangements have not alleviatedregulator concerns about potential agency problems in the mutual fund industry. My first hypoth-esis, stated in alternative form, is that managerial fund ownership is positively related to fundreturns.

In a related test, I also examine whether lower levels of managerial ownership are associatedwith higher levels of fund turnover, my second hypothesis. Finding an inverse relation betweenfund manager ownership and turnover would support the Dow and Gorton ( 1997) theory that oneincentive conflict between managers and shareholders lies in fund turnover. Aligning incentivesbetween the two parties should help offset the incentive of managers to make trades purely forthe sake of making trades, regardless of the potentially negative effect on fund value. Lowerturnover levels could also help reduce the fund's tax burden as well as administrative costs(brokerage commissions and trading expenses) associated with each sale. Managers who, as fundshareholders, would personally feel the effects of these costs (through lower fund returns) would

'Khorana, Servaes, and Wedge (2006) include three measures of board effectiveness in their tests and find no robustrelation between those measures and fund performance.


have greater incentive to minimize them.'' I will interpret a negative relation between ownershipand fund tumover as a reduction of agency costs in the mutual fund setting.

A manager's incentives can often diverge from the desires of investors with respect to the fund'stax burden. By trading shares held within the ftind's stock portfolio, the fund manager triggerscapital gains or losses. The SEC requires mutual funds to distribute net gains to fund shareholderseach year.̂ While all investors would favor a higher pretax retum, only taxable investors areinterested in the potential tax burden associated with fund distributions. Several papers (Huddartand Narayanan, 2002; Weiss, 2002; Jin, 2006) cite conversations with fund managers who assertthat taxes do not greatly affect trading behavior since shareholders vary in tax status and havedifferent (often unobservable) marginal tax rates. Yet empirical studies (Dickson and Shoven,1995; Bergstresser and Poterba, 2002) find evidence of significant heterogeneity in the taxliabilities associated with different mutual fund investments. This paper is the first to considerone possible explanation; the manager's personal stake in the fund.

It would require either an influential, tax-sensitive clientele or a personal incentive for a fundmanager to incorporate the potential tax burden of distributions into selling decisions. A fundmanager's compensation is rarely tied to a fund's tax burden (only one of the sample fundsdoes so), meaning that compensation does not provide a direct link between the fund's taxburden and the fiand manager's incentives.*" Personal ownership by the manager would align thatindividual's incentives with those of taxable individuals, regardless of whether they are the fund'sprimary clientele. However, even a manager with a strong personal desire to minimize taxes ondistributions should not alter selling behavior if such a strategy would be contrary to the desiresof the fund's key investors. It is quite plausible that a fund's clientele would be indifferent totaxes. The ICI (2005) estimates that 54% of 2004 mutual fund assets were held by tax-deferred ortax-exempt accounts and 37% were held in taxable household accounts.^ Hypothesis three, statedin altemative form, is that fund manager ownership is inversely related to a fund's potential taxburden. Such a relation wouid only be hypothesized to the extent that fund managers hold theirshares in taxable accounts.

Concurrent research by KSW also test the relation between mutual fund manager ownershipand ñind performance. Although the sample construction is different between this paper andKSW, both studies find a positive relation between managerial ownership and fund retums. Thecurrent work exclusively explores sole-managed domestic equity mutual funds. The KSW sampleis broader in that it examines sole- and multimanager funds and includes a wider selection of funds(e.g.. bond funds). However, KSW find that the performance-ownership relation is strongest insole-managed domestic equity funds (i.e., the setting that is used in this paper). Furthermore, mywork explicitly ties both returns and tumover tests to existing financial theory (Dow and Gorton,1997).

Other differences exist between KSW and this paper. For instance, KSW examine BOD'scharacteristics in their analysis, while the work presented here excludes or controls for items that

^Trading expenses are charged directly against the fund's assets, thus reducing NAV, the price at which fund shares aresold.

*To qualify as a flow-through entity, a fund must distribute at least 90% of its eamings to investors. After the Tax ReformAct of 1986, a fund must also distribute at least 98% of its realized capital gains net income, or be subject to an excisetax on the remaining undistributed portion.

"^Bergstresser and Poterba (2002} find that funds with a lower after-tax retum (a heavier tax burden) have lower subsequentinflows, which could lower fund net assets and, in turn, compensation.

^Household accounts can either be taxable (held directly by the individual(s)) or tax-deferred (held in a retirementaccount). Distributions received by a tax-deferred household account are not taxed until withdrawals arc made from theaccount, at the ordinary tax rates in place at the time of withdrawal.

Evans * Portfolio Manager Ownership and Mutual Fund Performance 517

were set by the board, since such items do not reflect what falls under the manager's control. Oneimportant area the KSW paper does not address is whether a fund's tax burden will vary based onmanagerial investment. I address this question by examining both after-tax returns and multiplemeasures ofthe fund's tax burden. The current research pertains to the wide range of investorswho hold any taxable (i.e., nonretirement) account.

II. Data and Sample

A. Ownership Data

Effective October 1, 2004, the SEC enacted a new federal rule requiring that mutual fundsdisclose certain items of information regarding their portfolio managers. In any prospectusreleased on or after February 28, 2005, funds must identify each portfolio manager {or team ofmanagers) responsible for the day-to-day management ofthe fund's portfolio. In its Statementof Additional Information (SAI), the funds must disclose each portfolio manager's beneficialownership of securities in the fund. The firms do not need to disclose the exact dollar amountof investment. Rather, they must disclose the portfolio manager's ownership stake as ofthe mostrecent fiscal year-end within seven dollar ranges: none, $1-$ 10,000, $10.001-$50,000. $50,001-$100,000. $100,001-$500.000. $500,001-$1,000.000, and over Sl.OOO.OOO.** Funds must alsodisclose other managerial information in the SAi, including the structure of, and method used todetermine each portfolio manager's compensation. In addition, the SAI must include informationregarding other accounts managed by its portfolio managers, including any material conflicts ofinterest. The Appendix presents a sample ofthe mandatory SAI disclosures.^

Figure 1 illustrates the disclosed ownership levels for the initial sample of portfolios. More thantwo-thirds ofthe observations fall into three ofthe seven SEC-mandated ownership categories:none (22%), SI00,001-$500,000 (24%), and over $1,000,000 (22%). For ease of exposition.Figure 1 displays the remaining four categories with the three primary ownership levels toillustrate, in essence, a low, medium, and high ownership level. Half of the sample portfolios(49%) have managers owning $100,000 or less. As stated, nearly 24% have managers owningbetween $100,000 and $500,000, while 28% have managers who have personally invested morethan $500,000.'" This figure reveals that ownership levels by fund managers are quite diverseand can be substantial. Subsequent tests will examine whether ownership in the "low" categoryis associated with a lower level of performance than being in the higher two categories.

B. Sample

For the purposes of this study, I obtain a list of sample funds and data for the fund characteristicsfrom Morningstar's Principia database. Following Bergstresser and Poterba (2002) and Plancich(2003). 1 extract data for domestic equity funds from each January release of Principia andmerge them into one data set. Detailed long-term and short-term capital gain distribution data are

"These personal ownership disclosures conform to the disclosures required for members ofthe BOD.

•"The disclosure is from Kobren Insight Funds, a fund in this paper's final sample. Fund families with multiple portfoliosunder management (e.g., Eaton Vance, Fidelity, and T. Rowe Price) often issue an SAI that reports data on multiple funds.As such, those disclosures are considerably longer and more intricate.

'"Within the "low" group, there is no statistical difference between [he "none" funds and the funds with ownership between$1 and $100,000 in any specification. I include the few funds that fall into the $500.001-$ 1.000,000 classification (6%)in the high-ownership group (HIGH), though inferences are robust to reclassifying those funds into the mid-ownershipgroup.


Figure!. Level of Mutual Fund Portfolio Manager Ownership

• $0-$100,000

D $100,001-$500,000

Dover$500,000

provided by Lipper (a Reuter's company that is a global provider of mutual fiind information andanalysis).

I hand-collect data from each fimd's SAI to construct a measure of managerial ownershipfor each fund in the sample. I only study funds with a single manager, which comprises nearlyone-half (47%) of all Morningstar funds at the end of 2004. My sample consists of funds thatissued ownership disclosures from the effective date ofthe SEC ruling (February 28, 2005) toSeptember 16, 2005. Over these six and one-half months, 1 collected data for 273 fund portfoliosof the types considered by this analysis (actively managed domestic grov/th, value, blend, orspecialty equity funds). ' ' Of those, I eliminate 36, since the manager data in the fund s SAI differfrom the manager information reported by Morningstar, leaving an initial sample of 237 fiindportfolios.'^''^

III. Empirical Specification

A. General

This paper examines the relation between fund manager ownership and various measures of fundperformance (retums and costs). Equation (1) captures the relation between each performancemeasure and managerial fund ownership:

"I exclude 13 tax-managed funds with ownership disclosures becau.se they are so few in number and because themanagers of those flinds may face a different incentive structure than managers of traditional funds. Another 13 portfoliosare excluded since their shares are restricted to itistitutional investors. Results are qualitatively unaltered when either ofthese groups is included in the analyses.

'-The data discrepancies could take two forms. Either the manager's name in the SAI is differenl from the name inMorningstar's database, or the SAI listed multiple managers, whereas Morningstar listed a sole manager

'^The mean (median) NAV for 2004 funds identified as having one manager is S17.62 ($13.99) versus $20.35 ($16.64)for sample funds. The mean (median) NAV for the universe of Momingstar funds is S17.78 ($ 13.75 ). The curTent samplehas a significantly larger NAV (at the 1% level) than both of these samples.

Evans • Portfolio Manager Ownership and Mutual Fund Performance 519

PERFu = a, + ßx LOWi + ß.Xu + £„. (1)

The variahle PERF takes on each individual performance measure, in turn, for ftind / in year /.For the purposes ofthe statistical tests, I include a LOW variable, which takes on the value I ifthe manager owns $100,000 or less in the fund and 0 otherwise. This variable construction willindicate whether managers with a minimal ownership level have statistically different performancelevels than managers who own more. The ideal ownership variable would be tbe manager'sownership as a percentage ofthat individual's net worth; unfortunately, such data are unavailable.Another potential ownership variable is ownership as a percentage ofthe fund's net assets. Theresulting ownership percentages are extremely small, using either the lower bound or endpointof each range to estimate fund ownership (with the exception of the "over $1,000,000" categoryin which the lower bound is always used). The median ftind has a managerial ownership level of0.000% using either estimation technique. The 75th percentile has a manager holding 0.002%(0.003%) of the fund's net assets, using the lower bound (midpoint) of the range. The 90thpercentile has a manager holding 0.001% (0.014%), respectively.

The variable X^ represents the set of covariates that could also affect a ftind's performance. Ineach test, I only use data representing years that the current manager is at the fund. Consequently,I delete any fund-year observations that fall during the sample period but before the currentmanager's tenure. I cluster the error term in each estimation by fund portfolio, since the sameftmd may be in tbe sample multiple years. '"̂

B. Ownership Measure

Funds disclose their manager's personal investment at tbe portfolio level. However, if the fundhas multiple share classes, the SEC does not require the fund to disclose in which share class(es)a manager has made an investment. The Eaton Vance Growth Fund, for instance, has class Ashares (ticker symbol EVGFX), class B shares (EMGFX), and class C shares (ECGFX). EatonVance disclosed that Arieh Coll. the Growth Fund portfolio manager, owned between $500,001and $1,000,000 in the portfolio as ofthe most recent fiscal year end. However, the disclosuredoes not detail whether the investment is in class A, B, or C shares (or some in all three).

Share classes do not represent fundamentally different investments. They simply representdifferent means of entry into the same underlying set of stocks (e.g., whether a fee is chargedupon an investor's entry or exit ft-om the ftind). Ofthe 237 sample portfolios, 59% have multipleshare classes. Momingstar treats each share class of a fund as a separate ftind, as do many mutualfund analyses. However, most fund characteristics are the same portfolio-wide (e.g., turnover,per-share distributions, and market capitalization) and thus are reported as the same for everyshare class.'^ Annual returns can also be the same between share classes, or only vary slightlyas a result of differences in expenses between classes. To avoid counting the same observationmultiple times, all analyses consider only the largest share class of a multiclass portfolio.

Ownership data have only been available since February 28, 2005 and are only disclosed forthe fund's most recent fiscal year-end, so I assume each manager's investment is the same acrossall years of this study (2001 -2004). '̂ It is possible that managers increased their ownership levels

'^SeePetersen (2005) fora discussion ofcrror-term clustering and entity fixed effects.

"Net assets vary betweeti share classes, as do expenses, sales loads, and Momingstar's calculation of capital gainsoverhang.

'"I re-estimate this paper's tesis over shorter time periods to reduce the number of years in which I assume the managers'ownership is the same as for the most recent (and only) disclosure. Results are robust to shortening the time frame to2003-2004. Using only 2004 as a sample period substantially reduces the .sample size. Although the trends found insubsequent tests are still present using just 2004 data, the results lose significance in thai specification.


immediately preceding the disclosure requirement. In that instance, my assumption would biasagainst finding significant differences in fund performance based on ownership. I should alsoemphasize that, for a fund to be in this paper's sample, it must have been in existence in 2005, theyear the disclosure requirement took effect. Any fund that did not survive until 2005 would nothave released ownership information and could not be included in this study.

C. Fund Returns

In this section, 1 study the relation between a fund's retums (adjusted for the mean return forthe fund's style category) and managerial ownership. The first dependent variable, ANNRTN,is the difference between the ftind's simple pretax annual retum, as computed by Momingstar,and the mean return for all Momingstar funds within the same fund-style category (i.e., large-capvalue, mid-cap growth) in a given year.'^ I also examine two after-tax return measures: annualreturns after taxes on distributions, and after taxes on distributions and the sale of fund shares.The SEC, since 2001, requires funds to present both after-tax retums numbers, setting forth acommon formula for funds to use to compute each measure.'" Only taxable investors wouldbe subject to the fund's tax costs. Each retum measure is presented and analyzed before salesloads.'^

Morningstar reports retums after adjusting for fund expenses, which are taken directly out ofeach fund's assets.^" A portion of those costs are captured by the fund's expense ratio and do notreflect any decision made by the fund manager. I consequently include the fund's expense ratioas an independent variable to control for the portion of net retums that are out ofthe manager'scontrol. Other costs, which are not included in the expense ratio but which also decrease fundreturns (brokerage commissions and transaction costs), are directly related to the amount of tradesthe fund manager executes.

1 include additional covariates tbat could be associated with a fund's returns. To account fora fund's systematic risk, I include BETA as a control variable. I add LAGINFLOWS to proxyfor a fund's need to sell additional shares to meet shareholder redemptions.'' Such liquidationslikely took place over this time period, as mean inflows for sample funds were negative in 2001and 2002.^^ I also include an indicator variable to represent funds with a first-year manager{NEWMGR), since new managers often make portfolio changes when they take over a fund

"The fund-style categories are defined by Momingstar, based on analysis of each fund's trading behavior, rather than thefund's self-reported investment style.

'"The calculations assume that distributions are received by individual shareholders in the top statutory tax bracket andthus represent an upper bound on the tax on distributions. The formula for the tax on sale assumes a ftind shareholderexits the fund at the end ofthe year and thus incorporates any unrealized gain or loss on fund shares. Funds calculate thattax burden based on a hypothetical shareholder with a $1,000 initial investment.

'''The return figures disclosed in ftjnd prospectuses are reported after incorporating the maximum sales load ofthe fund.However, these sales charges are not set by the fund manager and have consequenlty been removed from the currentanalysis. When sales loads are incorporated, the difterences between mean returns based on ownership are magnified.implying that sales loads are lower for funds with higher levels of ownership. Untabulated univariate statistics supportthis implication. The mean total sales load for the low-ownership funds is 2.4%, while it is only 0,64% for the highestownership group, a difference that is significant at the 1% level.™Those expenses include management, administrative, and t2-bl (marketing) fees.

^'The focus of this paper is on the manager's behavior when buying and selling stocks held within the fund. However, itis possible for fund shareholders to sell their fund shares, which funds are required to repurchase. If a fund docs nol haveenough cash to meet redemptions, it may have to sell some of its stock. Negative inflows mean redemptions exceed newstock purchases within a fund. Inflows are calculated following Bergstresser and Poterba (2002). Inflows = [Total NetAssets,/Total Net Assets,_, -(NAV, + DIV, + GA¡NS,)INAV,.,]I{I -I- ANNRTNJl).^^The ICI (2005) also estimates that net new cash flow for equity funds was negative in 2002.

Evans ' Portfolio Manager Ownership and Mutual Fund Performance 521

(Scherbina and Jin, 2006). BOND, the percentage of fund assets that is invested in bonds, is acovariate since stocks and bonds exhibit different patterns of retums. AGE., the age ofthe fundin years, is included to capture any systematic differences between older and younger funds.TENURE is included since the degree of the agency problem may be enhanced for seasonedmanagers wbo are nearing retirement (Chevalier and Ellison, 1999). The natural log ofthe netassets ofthe fund family (FFAMNETA) also serve as a control variable, since funds belonging tolarger fund families may have higher performance (Chen, Hong, Huang, and Kubik, 2004). I addyear and fund-style indicators (e.g., large-cap value, mid-cap growth) to capture time differencesand differences between mean fund returns based on investment objectives.

Controls for the after-tax specifications are largely the same as for the pretax retums specifi-cations. I also include a control variable representing the one-year lag ofthe fund's capital gainsoverhang {percentage of net assets that represent appreciation). Prior research (Barclay, Pearson.,and Weisbach, 1998; Bergstresser and Poterba, 2002) finds evidence consistent with the level ofappreciation witbin a fund influencing a manager's selling decisions.

D. Fund Turnover

In this section, 1 analyze whether higher levels of fund ownership are associated with differencesin the frequency of a fund's trades. An inverse relation would be consistent with the Dow andGorton (1997) theory that managerial compensation encourages noise trading, even when novalue-enhancing trades are found. If managerial ownership reduces this agency problem, thenhigher share ownership by mutual fund managers should result in lower levels of share turnover.Minimally invested managers, however, would have no such incentive to minimize turnover-related costs. Low turnover helps minimize selling expenses (brokerage and transaction costs)and potentially a fund's tax burden, since fewer sales may decrease the capital gains or losses afund would trigger In fact, after avoiding net taxable gains altogether. AllianceBernstein (2004)lists low turnover as the next-best way for managers to minimize the potential tax burden onmutual fund distributions. Both transaction and tax costs reduce a shareholder's return on amutual ftind investment.

The variable TURN, a fund's turnover ratio, can be interpreted as the percentage ofthe portfolio'sholdings that have changed over the past year. A low tumover (e.g., 20%-30%) would indicate abuy-and-hold strategy, whereas a high turnover (e.g., over 100%) indicates an investment strategyinvolving considerable buying and selling of securities. To address the possibility that turnoveris a proxy for ftind style, I again use the mean ratio for the fund's style category as a benchmark.The resulting TURN variable represents the mean-adjusted turnover volume relative to the meanfor the fund's peer group. As in all previous tests, the fund's investment style is classified byMorningstar, based on its analysis ofthe fund's trading behavior

I include INFLOWS as a covariate, since ftinds with low or negative net inflows may baveto sell additional stock to meet shareholder redemptions. I add a fund's current and laggedretums to control for the fact that managers might sell shares to lock-in gains or rebalancetheir portfolios following market upswings. I also add a one-year lag ofthe fund's capital gainsoverhang (LAGCGOH), since the level of unrealized gains could influence a manager"s sellingdecisions. I include NEWMGR because a new manager may make more sales to rebalance thefund's portfolio. AGE is included since newer funds may have increased trading needs until tbeyare more established, and TENURE captures differences in the agency problem between youngerand older managers. Year and fund-style indicators also serve as covariates, consistent with priorspecifications.


E. Fund Tax Costs

Finally, I examine the association between a manager's ownership in the fund and the tax burdenofthat fund. I use several different measures designed to capture different aspects of a fund'stax sensitivity. Consistent with prior tests, each measure represents the difference between thefund's tax measure and the average for the same fund-style category. The first construct of amanager's sensitivity to taxes on fund distributions is LTTOT, the percentage of total gains paidout as long-term gains. Plancich (2003) uses LTTOT to test whether mutual funds shifted moretoward long-term gains after the Tax Reform Act of 1997 reduced the capital gains tax rate froma maximum of 28% to 20%. She finds evidence consistent with such a shift. The quantity LTTOTrepresents a fund's sensitivity to individual tax incentives, because in all years of this study, short-term capital gains are taxed at a higher rate than long-term gains.^^ Since most dividends are nowalso taxed at a 15% tax rate, short-term capital gains are currently the most tax-disadvantageousform of mutual fund distribution.

Although favoring long-term gains would decrease a taxable individual's tax burden, a bettermeasure of tax-sensitivity may be whether funds distribute gains at all. AllianceBemstein (2004)lists avoiding short-term capital gains and offsetting realized capital gains with capital losses asthe primary tax-management techniques for muUial fiind managers. Dickson, Shoven, and Sialm(2000) find that active realization of capital losses has a significant effect on after-tax returnsfor actively managed funds. Specifically, instead of distributing capital gains realized within thefund, the manager could offset those gains with fund losses.^'' To see whether higher ownershipis associated with this tax-sensitivity measure, PCTGA/N, the per share dollars of gains paid outas a percentage of each share's net asset value, is used as a dependent variable.

The final tax-cost measure, FCTTAX, captures the effects of dividend, short-term gain, andlong-term gain distributions. PCTTAX represents the maximum per share tax burden a taxableindividual would face on all fund distributions as a percentage of tbe fund's net asset value,relative to the average for funds in the same category.^- If managers with larger personal fundinvestments want to minimize their current tax bill, PCTGAIN and PCTTAX should decreasewith fund ownership. However, it is possible that desires of a tax-insensitive clientele wouldoutweigh the manager's personal tax situation, or that the managers themselves would hold theirfund investments primarily in tax-deferred accounts.

I include INFLOW, since negative cash inflows could prompt the fund manager to increasestock sales to meet redemptions. The variable NETA., the natural log of a fund's total net assets, isincluded to control for scale effects. I add LAGCGOH, since the fund's level of unrealized gainsmay affect a manager's decisions, consistent with prior specifications, and I include NEWMGR,since a first-year manager might make more stock sales to rebalance the portfolio, increasingthe potential tax burden ofthe fund. The quantities AGE and TENURE are covariates to capture

^̂ Another disadvantage of short-tenn gains is that they become classified as ordinary income on an individual's lax retum,meaning that the investor cannot utilize that distribution to ofFset any capital losses. An individual's capital losses can beoffset by capital gains, then up to $3.000 of ordinary income. Any remaining capital losses must be carried forward foruse in fiiture years.

'•"A ftind cannot distribute the losses it generates. However, it may use those losses to offset gains within a portfolio forup to eight future years.

ïiPCTTAX = {{Itcg X /„) -f {stcg X f,,) 4- {div x tj))INAV. The tax rate on long-term capital gains (/„) is 20% in theyears before the enactment ofthe 2003 JGTRRA and 15% thereafter. The tax rate on short-term gains (f,,) is set equalto the maximum ordinary tax rate, and thus represents an upper bound on the potential tax burden from short-term gaindistributions. The dividend tax rate (fj) is equal to the maximum ordinary tax rate in all years before the Jobs and GrowthTax Relief Reconciliation Act, and 15% in 2003 and 2004. The variables Itcg, stcg, and div represent long-term capitalgains, short-term capital gains, and dividends, respectively.

Evans » Portfolio Manager Ownership and Mutual Fund Performance 523

differences in trading activity based on the establishment of the fund or the manager's agencyissue relative to career stage, i also include year and fund-style indicators to control for time andinvestment objective differences. Finally, PCTGAIN is a regressor in the LTTOT specification tocontrol for the amount of gains distributed, following Plancich (2003).

IV. Univariate Results

A. Dependent Variables

Panel A of Table I presents the mean, median, and standard deviation of each performancemeasure that I use as a dependent variable (returns, turnover, and tax costs, respectively), bythe binary ownership variable {LOW). Before deletions due to missing values, the initial sampleconsists of 812 fund-year observations ( 136 observations for the cumulative four-year return test).Table I also shows the value ofthe difference in the variable means between the two ownershipgroups, and presents /-values denoting the statistical significance ofthat difference. Fund retums,however defined, are positively related to ownership. Funds with minimally invested managershave significantly lower mean-adjusted returns than fiinds with more highly invested managers, atthe 1 % leve!. Mean-adjusted after-tax retums are also significantly lower for LOW funds (after-taxreturns on distributions is significant at the 1% level, while after-tax returns on distributions andthe sale of fund shares is significant at the 5% level). Overall, these results lend initial supportto the hypothesis that both pretax and after-tax retums are lower for ftinds with the low levels ofmanagerial investment.

Panel A also illustrates that funds with minimally invested managers have a significantlyhigher level of costly mean-adjusted fimd turnover (at the 1% level). This relation is consistentwith the reduction of an agency cost that often results in an inflated level of executed trades, aswell as the minimization of selling expenses and/or tax costs, which also have a negative effecton shareholder retums. The tax-cost statistics in Panel A provide initial evidence of whetherfunds with minimally invested managers have a higher tax burden, which would be consistentwith the finding that those funds appear to have higher levels of fund tumover. However, theunivariate results do not clearly support this implication. A tax-sensitivity hypothesis would bavepredicted that LTTOT means would increase with increasing managerial ownership, since tax-sensitive funds would favor long-term capital gains. The LTTOT mean is actually higher for theLOW group, although the difference is not significant. Similarly, the mean PCTGAIN is lowestfor the low-ownership group, which is also inconsistent with more highly invested managersattempting to minimize the current-year tax burden. However, there is a significant differencein tbe percentage of NAV that represents the total tax on distributions in a direction predictedby tax sensitivity. The quantity PCTT.AX is significantly higher, at the 5% level, for funds withminimally invested managers. In sum. univariate tax cost results do not clearly support a negativerelation between ownership and the fund's current tax burden.

B. Control Variables

Panel B of Table I presents univariate statistics for the control variables. It illustrates that thereare significant diñerences between funds with different managerial investment levels. Both netassets and capital gains overhang (the percentage of net assets representing capital appreciation)are increasing with ownership, with differences between groups significant at the 1% level.The fund's expense ratio, capturing costs outside the fund manager's control, is significantlydecreasing in managerial ownership, consistent with managers choosing to invest in funds with


Table I. Descriptive Statistics by Level of Managerial Ownership

The number of observations is 812 before missing-data deletions (136 for the 4YR retum variable). Allvariables are in excess of the fund-style mean. Variable definitions: LOW = 1 if manager owns $0-$100,000 in the fund, 0 otherwise; Difference = the excess value ofthe variable mean for LOW = 0 overlOlV = 1; ANNRTN - fund's percentage annual return: ANNUALIZED 4YR RTN = ftind's cumulativereturn annualized over a four-year period computed by Morningstar methodology; /íF7"7>íA'AriV-Distrib(Distrib & Sale) = fund's percentage annual return including tax burden from distributions (and assumingshareholder exits fund at year-end), calculated per SEC regulations); TURN = fund's turnover ratio; LTTOT =ratio of a fund's long-term to tota! capital gain; PCTGAIN — percentage of a fund's net asset value paid outas capital gains; PCTT4X = per-share maximum tax burden on distribution as a percentage of a fund's netasset value. A i-test denotes the significance of the difference in means between LOW = 1 and LOW = 0using a two-tailed test.

Panel A. Mean-Adjusted Fund Performance

L0tV=1 LOW = Difference (f-Value)

ANNRTNMeanMedianSD

ANNUALIZED 4YR RTNMeanMedianSD

AFTTAX RTN-DistnbMeanMedianSD

-0.41-0.54

8.8

0.440.036.4

0.05-0.24

9.4

AFTTAX RTN-Distnh & SaleMeanMedianSD

TURNMeanMedianSD

LTTOTMeanMedianSD

PCTGAINMeanMedianSD

PCTTAXMeanMedianSD

0.410.146.2

10.9-33.0192.8

0.0870.170.30

0.0014-0.0039

0.03

0.0013-0.00092

0.01

2.21.0

10.3

2.71.87.5

2.71.3

10.8

1.50.637.3

-52.8-59.6

71.1

0.0600.140.31

0.0006-0.0039

0.03

-0.00016-0.0010

O.OI

2.(3.

2.(3.

2.(3,

1,

.6*"

.79)

.3"*

.94)

. 7 ' "

.70)

, 1 "(2.36)

-63.:i.27)

-0.027(-0.64)

0.008(0.35)

-0.0014**(-2.34)


Table I. Descriptive Statistics by Level of Managerial Ownership {Continued)

The number of observations is 812 before missing-data deletions. A /-test denotes the significance ofthedifference in means between LOU' = 1 and LOW = 0. Variable definitions: LOW = I if manager owns$0-$ 100,000 in the fund, 0 otherwise; Differenee = the excess value ofthe variable mean for LOW = 0over LOW = \; ANNRTN — fund's percentage annual return; CGOH = percentage ofthe fund's assets thatrepresent capital appreciation; EXPRATIO = fund's expense ratio; NETA = the fund's total net assets, inmillions; FFAMNETA = the ftind family's total net assets, in millions; BETA — the fund's systematic risk asreported by Morningstar; BOND = percentage of fund assets invested In bonds; .4GE = fund age in years;INFLOW = net inflows weighted by beginning net asset value; TENURE = length of time managing thefund, in years; L/M/S GROW/VALUE/BLEND = 1 if the fund is a largc/mediunVsmall growtli/value/blendfund, and 0 otherwise; SPECIALTY = 1 if the fund is a specialty ftmd, and 0 otherwise; NEWMGR = 1 ifthe manager is in his first year, and 0 otherwise.

Panel B. Control Variables

LOW^O Difference (r-Vaiue)

Continuous VariablesCGOH

MeanMedianSD

EXPRATIOMeanMedianSD

NETAMeanMedianSD

FFAMNETAMeanMedianSD

BETAMeanMedianSD

BONDMeanMedianSD

AGEMeanMedianSD

INFLOWMeanMedianSD

-20.8-1.0101

1.61.50.71

306100612

23.2136.871

56. 782

1.40.896.4

0.8303.6

11.27.4

12.8

0.560.161.7

-0.1613.059.1

1.31.30.45

1.278293

2.524

31.9216,804

77. 525

1.90.908.7

0.5502.5

15.611.213.9

0.440.1977.4

20.6'"(3.44)

-0.30'*'(-7.28)

972*"(7.II)

8. 708'(1.78)

0.50(0.69)

-0.28(-1.29)

4.4'"(4.74)

-0.12(-0.02)

526

Tabie i. Descriptive Statistics

Panel B.

L

Financial Management • Autumn 2008

by Level of Manageriai Ownership {Continued)

Control Variables (Continued)

.OW = 1 LOW = 0 Difference (r-Vaiue)

TENUREMeanMedianSD

Discrete VariablesLGROW

MGROW

SGROW

LVALUE

MVALUE

SVALUE

LBLEND

MBLEND

SBLEND

SPECIALTY

NEWMGR

No. observationsPercentage of all observations

4.63.74.4

14.8%

13.7%

14.6%

5.8%

2.5%

5.2%

15.1%

6.6%

1.1%

20.6%

100%

12%

44855%

7.25.566.0

10.9%

13.8%

8.7%

11.4%

4.9%

5.4%

18.1%

9.6%

5.1%

12.1%

100%

6%

36445%

2.6" '(6.98)

-3.9%-(-1.66)

0.1%(0.04)

- 5.9%**'(-2.63)

5.6%'"(2.81)2.4%'(1.80)0.2%

(0,09)3%

(1.13)3%

(1.55)4%*"(3.21)

-8 .5%—(-3.33)

0%

- 6 % " *(-3.06)

" • SignifÈcant al the 0.01 level.

" Significant at the 0.05 level.

' Significant at the 0.10 level.

lower expenses. Fund age and manager tenure are also significantly lower for managers with alow level of personal investment. These findings are consistent with managers amassing moreinvestment over time. Similarly, managers in their first year (NEWMGR) are more likely to havelittle to no fund investment. There are also some differences in the types of fund represented byLOW funds. Most notably, significantly more LOW funds are specialty funds. Overall. Table Iillustrates that there are significant differences in the univariate statistics for both the primaryand control variables.^''

*̂'Of the !0% of funds that disclose that fund manager compensation is not tied to fund performance. 59% are LOW funds.However, this estimate is based on a very small number of funds and thus might not generalize to all such funds.


V. Multivariate Results

A. Fund Returns

Table II presents the results for the regression of mean-adjusted returns on the dichotomousownership variable. Annual fund returns in excess of fund-style means are statistically lower,at the 1% level, for funds where the fund manager has a low level of personal investment.•^•'Economically speaking., this result indicates that funds where the manager owns $100,000 orless generate excess annual returns that are approximately 2.6% lower than funds with managersthat own more than that amount. This finding is consistent with managerial investment aligningdecision maker and shareholder interests.̂ ** The results are similar for the regression that usesmean-adjusted returns after taxes on distributions. The after-tax retums are significantly lower,at the 5% level, for funds with minimally invested managers. Results for the after-tax retums ondistribution and sate variable are in the predicted direction, but fall out ofthe range of statisticalsignificance. This variable assumes investors sell shares at year-end and thus is not as relevant forlong-term investors. In addition to fund investment style and year, which explain a large portion ofmutual fund's annual retums, prior-period inflows are negatively related to mean-adjusted retumsat the 5% level. This result is consistent with the difficulty of managing an increasing asset base.

B. Two-Stage Least Squares Estimation

Of concern in the foregoing analysis is the potential endogeneity of the ownership choice;specifically, that low (high) fund returns could be causing low (high) managerial ownership.To address this concem, I estimate a two-stage least squares (2SLS) regression, the results ofwhich are presented in Table III. The first stage is a logistic estimation in which I regressthe binary LOW variable on fund style, lagged mean-adjusted fund returns, and a measure ofmanager compensation, since managers with higher income will have more to invest. Becausefiind managers are largely compensated based on net assets under management, I use total netassets as the compensation proxy. The Pearson correlation coefficient between the dichotomousLOW variable and total net assets is 0.70, which is significant at the 1% level. Total net assetsare uncorrelated with mean-adjusted returns (a Pearson correlation coefficient of 0.004) and thuscan be considered an appropriate instrument for the first stage.

Estimating 2SLS with this structure does not result in a change in inferences from the originalordinary least-squares (OLS) estimation.'̂ '̂ The coefficient on the predicted LOW value fromthe first stage is significant at the 5% level. Thus, even after removing the effect of prior-periodretums on managerial ownership, personal investment is positively related to fund retums. Thisresult is consistent with higher managerial ownership leading to higher returns, rather than thereverse. Alternatively stated, this result supports the idea that ownership reduces agency costs

*'Using ownership as a percentage of fund net assets as the independent variable does not yield significant results.However, when Ihe returns analysis is estimated using a rank regression, the results are significant using the percentageownership variable (consistent with the LOW analysis) at the 5% level, using either the lower bound or midpoint of thedollar range in the estimation of managerial ownership.

'̂'In a separate analysis, I examined whether cumulative retums are lower for low-ownership funds over a cumulativefour-year period. I used the subset of funds that were in existence and had complete data over the entire sample period.While the resulting .sample is quile small (57 observations), the results are consistent with the tabulated returns analyses,LOW ftinds had significantly lower cumulative retums than the other sample funds at the 1% significance level,

^'Although the primary endogeneity concem is between ownership and pretax returtis, 1 also estimate 2SLS using eachofthe two after-tax retum figures. Inferences remain unchanged.


Table II. Mean-Adjusted Fund Returns Regression Results

All dependent variables are in excess of the fiind-style mean. Variable definitions: ANNUAL RTN =ftind's annual return percentage; ^F7T/lX/?7?V-Distrib (Distrib & Sale) = fund's percentage annual returnincluding tax burden from distributions (and assuming shareholder exits fund at year-end), calculated perSEC regulations); LOW = I if the manager owns $0-$ 100,000 in the fund. 0 otherwise; LAGINFLOWS =one-year lag of net inflows weighted by beginning net asset value; EXPRATIO = ftind's expense ratio;BETA = fund's systematic risk, reported by Morningstar; BOND — percentage of fund assets invested inbonds; AGE = fund age in years; TENURE = length of time managing the fund in years; NEWMGR — 1 iftenure is less than one year, 0 otherwise; FFAMNETA = the natural log ofthe fund family's total net assets;LAGCGOH — one-year lag ofthe percentage ofthe ftind's assets that represents capital appreciation.

InterceptLOWLAGINFLOWSEXPRATIOBETABONDAGETENURENEWMGRFFAMNETALAGCGOH

No. observationsYear fixed effects?Fund-style indicators?Adjusted R^

*" Significant at the 0.01 level" Significant at the 0.05 level

ANNUALRTN

Coefficient I

2.67-3.14"*-0.018*'-0.72-0.014

0.14-0.0014-0.071-0.56

0:11-

515YesYes

0.045

r-Value

0.94-2.63-2.46-0.87-0.44

1.38-0.06-0.97-0.38

0.56-

AFTTAXRTN(Distribution)

Coefficient3.98

-2.73**-0.018*'*-0.72-0.014

0.12-0.0031-0.017-0.49

0.13-0.0029

515YesYes

0.048

f-Value1.33

-2.10-2.65-0.78-0.42

0.90-0.13-0.17-0.34

0.66-0.31

AFTTAXRTN(Distribution

Coefficient1.58

-1.26-0.012*'-0.098-0.0046

0.0210.013

-0.00870.10

-0.029-0.0012

515YesYes

0.037

& Sale)

i-Value0.77

-1.45-2.51-0.15-0.19

0.260.75

-0.12O.IO

-0.23-0.18

in the mutual fund industry. In additioti to the SEC's motivation for enacting the new disclosurerequirements (to help investors to assess incentive alignment with managers), anecdotal evideticealso supports the agency theory. Kunal Kapoor, Director of Fund Analysis at Morningstar, Inc.,places mutual fund managers into two distinct categories: money managers and fiind shareholders.This categorization reflects the opinion that personally invested fund managers likely possessmore conviction, taking more care in executing appropriate trades, than are money managerswho simply perform investing services in exchange for compensation. The results of this 2SLSestimation are compatible with both the SEC's motivation for requiring the managerial ownershipdisclosures and the opinion of leading analysts in the mutual fund industry.

C. Fund Turnover

Results from the turnover regression are reported in Table IV. Consistent with the univariateresults, mean-adjusted turnover levels are inversely related to fund ownership. Specifically, LOWfiands have significantly higher excess tumover levels (approximately 61% higher) than the other


Table III. Two-Stage Least Squares Estimation

AU dependent variables are in excess ofthe fund-style mean. Variable definitions: LOW = 1 if the man-ager owns $Q-$100.000 in the fund, 0 otherwise; NETA = the natural log ofthe fund's total net assets;(LAG)ANNRTN = fund's (one-year lag of its) annual retum percentage; LOWPREDICT = predicted valueof LOW from the first stage ofthe 2SLS estimation; LAGINFLOWS = one-year lag of net inflows weightedby beginning net asset value; EXPRATIO = fund's expense ratio; BETA — fund's systematic risk, reported byMomingstar; BOND = percentage of fund assets invested in bonds; AGE = fund age in yeat^; TENURE =length of time managing ihe fund, in years; NEWMGR = 1 if tenure is less than one year, 0 otherwise;FFAMNETA = the natural log ofthe fund family's total net assets.

Dependent Variable =LOW

Dependent Variable =ANNRTN

Coefficient Wald Coefficient i-ValueInterceptNETALAGANNRTNLOWPREDICTLAGINFLOWSEXPRATIOBETABONDAGETENUREFFAMNETANEWMGR


0.52"-0,025"*-0 .030 ' "

4.1630.1615.25

6.78" 3.08

726YesYes

12.37*"-0.019" '-0.38-0.018

0.13-0.0095-0.038-0.044-0.27

726YesYes

0.043

-2.51-2.66-0.96-0.88

1.26-0.83-1.01-0.13-0.33

"* Significant at the O.OI level.

" Significant at the 0.05 leve!.

satnple futids.^" These results are consistent with personal fund ownership leading to a reductionin a manager's incentive to inflate fund tumover in an effort to appear competent as a managerand thereby increase compensation. Following Dow and Gorton (1997). decreasing the amountof noise trading should also result in an increase of fiind retums, which conforms to the resultspresented in Table II. Executing fewer trades reduces transaction costs and possibly decreasesthe fund's tax burden. In addition, results from the control variables reveal that tumover isnegatively related to a fund's prior-year returns., as well as the one-year lag of the fund's capitalgains overhang, consistent with selling to lock in gains. Thus, the relation between managerialownership and both returns {a positive relation) and tumover (a negative relation) in this paper isconsistent with ownership reducing agency costs in the mutual fund setting.

D. Fund Tax Costs

Since the significance between ownership coefficients is somewhat less in the after-tax returnsspecifications than in the pretax specification, one may conclude that the fund's tax burden might

•""Using ownership as a percentage of net assets yields significant results at the 5% level in the predicted relation (lowertumover in funds with higher ownership percentages). This result persists using either the lower bound or midpoint oftheownership ranges (lower bound only for the highest category) as a dollar estimate for managerial investment.

-6.3461.4"*-0.0029-2.74- 1 . 1 8 ' "

0.32^0.29"*

0.17- 2 . 8 8 "

571YesYes0.18

0.233.10

-0.17-0.76-2.61

0.02-2.76

0.44-2.06

530 Financial Management * Autumn 2008

Table IV. Mean-Adjusted Fund Turnover Regression Results

The dependent variable is the fund's turnover ratio in excess ofthe mean for the fund-style category. Variabledefinitions: LOW ^ 1 if the manager owns $O-$IOO.OOO in the fund, 0 otherwise; INFLOWS ~ net inflowsweighted by beginning net asset value; NETA = the natural logofthe fund's total nel assets, LAGANNRTN =fund's one-year lagged annual return percentage; NEWMGR = I if the manager is in his first year, and0 otherwise; LAGCGOH = one-year lag of the percentage of the ftind's assets that represents capitalappreciation; AGE = fund age in years; TENURE = length of time managing the fund, in years.

Coefficient t-ValueInterceptLOWINFLOWSNETALAGANNRTNNEWMGRLAGCGOHAGETENURE


•" Significant at the 0.01 level." Significam at the 0.O5 level.

not be reduced as a result ofthe executioti of trades by a highly invested manager. However, thosefunds with more substantially invested managers {LOW = 0) also have lower fund turnover, acommon strategy for tax minimization in mutual funds. As a direet test of whether the tax burdenis reduced in funds with highly invested managers, I analyze three tax-sensitivity measures:LTTOT, PCTGAIN. and PCTTAX.

The results from these regressions are presented in Table V. The coefficient on LOW is notsignificantly different from zero in two ofthe three specifications and is only weakly significantin the third. Such results suggest that a fund manager's ownership and the fund's tax btirden haveno relation to one another. These findings are consistent with fund managers trading on behalfof a tax-insensitive clientele, or a clientele of unknown tax status. They are also consistent withthe anecdotal claims that managers trade irrespective of the tax implications of their behavior.The fact that funds managed by owners with more than a negligible investment do not appear totax-manage could also indicate that managers hold mutual fund shares in tax-deferred accounts,although ownership disclosures are not required to denote in what kind of account(s) the managerholds a personal investment.^' Thus, while certain research has shown that managers, as a whole,engage in some degree of tax planning (Bhabra, Dhillon, and Ramirez, 1999; Gibson, Safieddine,and Titman, 2000; Huddart and Narayanan, 2002; Plancich, 2003), the results of tbis analysisdo not support the hypothesis that managers with higher ownership engage in more tax planningthan do managers without a substantial personal fund investment.

^'Additionally, a portion ofthe managers' compensation could be direclly píaced into their retirement/tax-deferredaccounts, although none ofthe sample funds have disclosed whether this process has occurred. This treatment would alsoreduce a manager's incentive to lower the tax burden ofthe fund.


Table V. Mean-Adjusted Fund Tax Costs Regression Results

All dependent variables are in excess ofthe fiind-style mean. Variable definitions: LTTOT = ratio of afund's long-tenn to total capital gain distributions; PCTGAIN = fund's per-share capital gain payout as apercentage of NAV; PCTTAX = fund's per-sharc tax due by taxable investor as a percentage of NAV, LOW =1 if the manager owns $0-$ 100.000 in Ihe fund, and 0 otherwise; INFLOWS = net inflows weighted bybeginning net asset value; NETA = the natural log of Ihe fund's total net assets; LAGCGOH - one-year lagofthe percentage ofthe fund's assets that represent capital appreciation; NEWMGR = I if the manager is inhis first year, and 0 othei^vise; {LAG)ANNRTN = fund's (one-year lagged) annual retutn percentage; AGE =fund age in years; TENURE = length of time managing the fund, in years; PCTGAIN = percentage of a

net asset value paid out as capital gains.

InterceptLOWINFLOWSNETALAGCGOHNEWMGRANNRTNLAGANNRTNAGETENUREPCTGAIN


"' Significant at the 0.01 level

" Significant ill ihe 0.05 level

• Significani al the 0.10 level

LTTOT

Coefficient-0.19*

0.28-0.00014

0.00750.0041"0.19"'

-0.00067-0.0024-0.00083

0.0072'0.81"

199YesYes0.12

f-Value

-1.810.57

-1.310.612.453.75

-0.21-1.29-0.46

1.792.17

PCTGAIN

Coefficient-0.00093

0.0010-0.00OOI8*-0.00076

0.000042"0.0057

-0.0000290.000430.0000890.00020

-

572YesYes

0.063

r-Value-0.11

0.36-1.72-0.74

2.23I.4I

-0.161.340.960.68

-

PCTTAX

Coefficient-0.0022

o.ooir-0.0000014-0.0000037

0.00000430.000260.00000750.000089'0.000015

-0.0000085-

572YesYes

0.094

r-Vaiue-1.36

1.69-0.65-0.02

!,280.330.231.780.78

-0.12-

Results frotn control variables indicate that a fund's tax burden is positively related to thecapital gains overhang at the beginning of the year. This relation is consistent v̂ 'ith overhangaffecting trading decisions, consistent with prior research (Barclay, Pearson, and Weisbach, 1998;Bergstresser and Poterba, 2002). The LTTOT variable is positively related to the percentage ofnet asset value paid out as gains. It is also significantly higher for funds with new managers whomake initial portfolio changes when they take over the fund.

VI. Conclusion

As the share ofthe market owned by mutual funds continues to rise., determining the factorsthat affect fund manager trading behavior has become increasingly important. This paper usesnewly disclosed managerial ownership data and details of other fund characteristics from theyears 2001-2004 to study the relation between a manager's personal wealth and the performanceofthe associated fund. I find that managerial fund ownership varies widely, with approximatelyone out of every two managers owning over $100,000 in the managed fund, and one out of everyfive managers owning over $1,000,000.

532 Financial Management » Autumn 2008

I find that fiinds run by minimally invested managers have significantly lower style-adjustedretums (approximately 2.6%) than do funds where managers own more than $100,000 of theirftind. They also have significantly higher turnover levels. Both findings are consistent with thereduction ofthe agency costs set forth in the Dow and Gorton (1997) model, where managersmake value-reducing trades in lieu of making no trades when they cannot identify any suitableinvestments. This paper's tax analyses do not find evidence of decreased tax sensitivity in fundswith minimally invested managers. This result is consistent with more substantially investedmanagers trading for a tax-insensitive clientele, a clientele of unknown tax status, or with amanager holding shares in a tax-deferred retirement account.

Investors should consider many variables when they choose to invest in mutual ftind shares.The optimal ftind choice for each individual depends on personal goals, investment horizon, andrisk profile. However, the results of this study suggest that investors of any tax status might needto add managerial ownership to the list of variables to consider when choosing a mutual fundinvestment.• ' '

References

Aboody, D, and R. Kasznik, 2000, "CEO Stock Option Awards and the Timing of Corporate VoluntaryDisclosures," Journal ofAccounling and Economics 29, 73-100.

AllianceBernstein, 2004, Biting Back at Taxes: Exploring Tax-Managed Mutual Funds, New York, NY,AllianceBernstein Investment Research and Management.

Barclay, M., N. Pearson, and M. Weisbach, 1998, "Open End Mutuai Funds and Capital Gains Taxes,"Journal of Financial Economics 49, 3-43.

Bergstresser, D. and J. Poterba, 2002, "Do After Tax Retums Affect Mutual Fund Inflows?" Journal ofFinancial Economics 63, 381-414.

Bhabra, H., U Dhillon, and G. Ramirez, 1999, "A November Effect? Revisiting the Tax Loss SellingHypothesis," Financio/A/onogemeni 28, 5-15.

Broussard. J., S. Buchenroth, and E. Pilotte, 2004, "CEO Incentives, Cash Flow, and Investment," FinancialManagement 33, 51-70.

Brown, J., N. Liang, and S, Weisbenner, 2005, "Executive Financial Incentives and Payout Policy; FirmResponses to the 2003 Dividend Tax Cut," University of Illinois and Federal Reserve Board WorkingPaper.

Chen. J.. H. Hong. M. Huang, and J.D. Kubik, 2004, "Does Fund Size Erode Mutual Fund Performance?The Role of Liquidity and Organization," American Economic Review 94, 1276-1302.

Chetty, R. and E. Saez, 2005, "Dividend Taxes and Corporate Behavior: Evidence from the 2003 DividendCut," Quarterly Journal of Economics 120, 791-833.

Chevalier, J. and G. Ellison, 1999, "Career Concerns of Mutual Fund Managers," Quarterly Journal ofEconomics \\4,i34-S56.

Clinch, G., 1991, "Employee Compensation and Firms' Research and Development Activity," Journal ofAccounting Research 29, 59-78.

Coles, J., F. Meschke. and M. Lemmon. 2008, "Structural Models and Hndogeneity in Corporate Finance: TheLink between Managerial Ownership and Corporate Performance," Arizona State University, Universityof Utah, and University of Minnesota Working Paper.


Dickson. J, and J. Shoven, 1995, "Taxation and Mutual Funds: An Investor Perspective," in J. Poterba, Ed.Tax Policy and ¡he Economy, Vol. 9, Cambridge. MA, MIT Press, 151-181.

Dickson. I , J, Shoven, and C. Sialm, 2000, "Tax Externalities of Equity Mutual Fimds," National TaxJournal 53, 601-62^.

Dow, J. and G. Gorton, 1997, "Noise Trading, Delegated Portfolio Management, and Economic Welfare,"Journal of Political Economy 105, 1024-1050.

Fama, E. and M.Jensen, 1983, "Separation ofOwnership and Contro^'Vourna/o/LawanrfEconomiCi 26,301-326.

Gibson, S., A. Safieddine, and S. Titman, 2000, "Tax Motivated Trading and Price Pressure: An Analysisof Mutual Fund Holdings," Journal of Financial and Quantitative Analysis 35. 369-386.

Hanlon, M., S. Rajgopal, and T. Shevlin, 2005, "Are Executive Stock Options Associated with FutureEamings?" Journal of Accounting and Economics 36, 3-43.

Huddart. S. and VG. Narayanan. 2002, "An Empirical Examination of Tax Factors and Mutual Funds StockSales Decisions," Review of Accounting Studies 7, 319-341,

Hunt, H.. 1985, "Potential Determinants of Corporate Inventory Accounting Decisions," Journal of Ac-counting Research 23,448-467.

Investment Company Institute, 2005, The Investment Company Fact Book, Washington, DC, InvestmentCompany Institute.

Jensen, M. and W. Meckling, 1976, "Theory ofthe Firm: Managerial Behavior, Agency Costs and OwnershipStructure," Journal of Financial Economics 3, 305-360.

Jin, L., 2006, "Capital Gain Tax Overhang and Price Pressure," Jburna/o/Finance 61, 1399-1431.

Khorana, A.. H. Servaes, and L. Wedge, 2006, "Portfolio Manager Ownership and Fund Performance,"Journal of Financial Economics 85, 179-204.

Lambert, R., W. Lanen, and D. Larker, 1989, "Executive Stock Option Plans and Corporate Dividend Policy,"Journal of Financial and Quantitative Analysis 24,409-425.

McConnell, J, and H. Servaes, 1990, "Additional Evidence on Equity Ownership and Corporate Value,"Journal of Financial Economics 20, 293-315.

Meschke, F., 2005, "An Empirical Examination of Mutual Fund Boards." University of Minnesota andArizona State University Working Paper.

Morck, R., A. Sblcifer, and R. Vishny, 1988, "Management Ownership and Market Valuation," Journal ofFinancial Economics 20, 293-315.

Nagar. V. D. Nanda, and P Wysocki, 2003, "Discretionary Disclosure and Stock-Based !ncentives,'Vourrta/of Accounting and Economics 34, 283-309.

Niehaus, G., 1989, "Ownership Strucnire and Inventory Method Choice," The Accounting Review 64,269-284.

Petersen, M., 2005, "Estimating Standard Errors in Finance Data Sets: Comparing Approaches," North-western university Working Paper.

Plancich, S., 2003, "Mutual Fund Capital Gain Distributions and the Tax Reform Act of 1997," NationalTax Journal 56, 27 \-296.

Qian, M., 2005, "Who Can You Trust? A Smdy on Mutual Fund Govemance," Boston College WorkingPaper.


Ryan, H. Jr. and R. Wiggins III, 2002, "The Interactions between R&D Investment Decisions and Compen-sation Policy," Financial Management 31, 5-30.

Scherbina, A. and L. Jin, 2006. "Change is Good or the Disposition Effect among Mutual Fund Managers,"Harvard Business School Working Paper.

Stulz, R., 1988, "Managerial Control of Voting Rights: Financing Policies and the Market for CorporateControl," Journal of Financial Economics 21, 25-54.

Tufano, P. and M. Sevick, 1997, "Board Structure and Fee-Setting in the US Mutual Fund Industry," Journa/of Financial Economics 46, 321 -335.

Warfieid, T, J. Wild, and K. Wild, 1995, "Managerial Ownership, Accounting Choices and Informativenessof Earnings," Journal of Accounting and Economics 20, 61 -91.

Weiss, I., 2002, "Discussion of'An Empirical Examination of Tax Factors and Mutual Funds' Stock SalesDecisions," Aev/evf o//íccotínííng S/Wi//ej 7, 343-347.

Ownership and fund performance evans

Economy & Finance

mutual fund industry

mutual fund performanceallison

mutual fund arena

mutual fund managersare

mutual fund returns

fund ownership levels

fund turnover

fund investors