Is Manager Gender Important in the Performance of Mutual Funds?

Transcription

1 College of Sain Benedic and Sain John s Universiy DigialCommons@CSB/SJU Accouning and Finance Faculy Publicaions Accouning and Finance Is Manager Gender Imporan in he Performance of Muual Funds? Seven Welch College of Sain Benedic/Sain John's Universiy, swelch@csbsju.edu Yunqing Wang Sain Cloud Sae Universiy Follow his and addiional works a: hp://digialcommons.csbsju.edu/acc_pubs Par of he Finance and Financial Managemen Commons Recommended Ciaion Welch, Seven and Wang, Yunqing, "Is Manager Gender Imporan in he Performance of Muual Funds?" (2013). Accouning and Finance Faculy Publicaions. Paper 6. hp://digialcommons.csbsju.edu/acc_pubs/6 This Aricle is brough o you for free and open access by DigialCommons@CSB/SJU. I has been acceped for inclusion in Accouning and Finance Faculy Publicaions by an auhorized adminisraor of DigialCommons@CSB/SJU. For more informaion, please conac digialcommons@csbsju.edu.

2 IS MANAGER GENDER IMPORTANT IN THE PERFORMANCE OF MUTUAL FUNDS? Seven J. Welch 1, College of S. Benedic/S. John s Universiy Yunqing Wang, S. Cloud Sae Universiy Absrac We invesigae wheher here are differences in characerisics and performance of muual funds caused by he manager s gender. Through examining a large sample of U.S. domesic equiy muual fund, we find some evidence ha suggess female managers have a lower risk olerance han males. This leads o he observaion ha females end o hold a higher oal number of asses (socks) and fewer asses in heir op 10 holdings han do male managers. We hen analyze performance wihin funds over ime in order o evaluae he impac of changes in managemen s gender composiion on funds performance. We find some evidence ha he percenage of female managers managing a fund is negaively relaed o he fund s performance over ime. JEL Classificaions: C91, G11 Keywords: Muual funds, manager characerisics, gender differences, fund performance This version: March 1, Conac auhor: Deparmen of Accouning & Finance, S. John s Universiy, P.O. Box 2000, Collegeville, MN 56321; Phone: (320) , Fax: (320) , swelch@csbsju.edu.

3 I. INTRODUCTION Wih recen advances in he undersanding he psychology of invesing, i is now commonly undersood ha differen groups of invesors have varying characerisics. Some sysemaic differences beween male and female invesors have been documened. These include ha women end no o have as much confidence as male invesors, and women are generally more risk averse han men 1. As a popular invesmen vehicle, muual funds have grown remendously over he pas few decades. If hese behavioral phenomena found in individual invesors also prevail among muual fund managers, here are many poenial ramificaions for invesors. For example, if women managers are more risk averse, we should see women managing more conservaive funds, such as value funds. In combinaion wih being conservaive, if women managers are less confiden wih heir sock picks, hey should have a lower urnover in heir porfolios because hey would be more cauious abou moving money ino a new securiy. As a resul, female managers ransacion fees would be lower. The combinaion of more conservaive sock picks and lower urnover may imply eiher higher or lower reurns depending upon which rai dominaes. According o modern porfolio heory, aking greaer invesmen risk should resul in receiving a greaer reward, so invesing in less risky asses would lead o lower reurns. This implies ha men may have beer performance in heir invesmens since hey are less risk averse han women. However, numerous empirical sudies (Lakonishok e al. (1994), La Pora e al. (1997), Lee and Swaminahan (2000) and Barberis and Shleifer (2003)) find ha value socks ouperform growh socks in he long run 2. If females end o inves in value-oriened socks, hen heir long-erm performances should be 1 A few references regarding gender differences include Barber and Odean (2001), Eses and Hosseini (1988), Bajelsmi and VanDerhei (1997) 2 However, Chan e al. (2002) find evidence ha growh managers ouperform value managers afer adjusing for syle. Page 1

4 greaer han males. No o be forgoen, lower urnover implies lower ransacions coss, and herefore higher reurns on he porfolio (Barber and Odean (2001, 2000)). So, if female managers express boh he lack of confidence rai and a endency oward conservaive invesing, women should make higher nominal reurns han men. There are a few reasons o believe ha muual fund managers, in general, behave differenly han average individual invesors. Firs, fund managers invesmen choices are ofen consrained by limis se forh in he fund prospecus. Wih some excepions, such as secor funds, prospecuses ypically have diversificaion requiremens ha do no allow fund managers o inves large porions of he funds in individual socks, secors, or oher high-risk invesmens. Second, educaion makes a difference. Almos all fund managers have college degrees, whereas only 28% of aduls in he Unied Saes had aained bachelor s degrees. 3 Addiionally, more han 50% of muual fund managers hold MBAs and more han 35% hold CFAs (Akinson e al. (2003) and Chevalier and Ellison (1999)). Third, female muual fund managers have chosen a career pah in porfolio managemen, a male-dominaing field. No like average individual female invesors who may no willingly paricipae in rading aciviies, female fund managers have he knowledge and experise in invesmen and managing risk and may be more confiden and compeiive in order o esablish hemselves in he field. Because of hese reasons, invesmen behaviors of male and female fund managers may no be as differen as migh be observed in he general invesing public. I is imporan o invesigae he gender issue in muual fund managemen, because if gender does maer hen i should be a facor for invesors o consider when choosing funds. So far, a few gender sudies have been conduced on muual funds. However, he resuls have no been consisen: some researchers find ha female fund managers are more risk averse, rade less 3 U.S. Deparmen of Educaion (2005) Page 2

5 and shy away from compeiion (Beckmann and Menkhoff (2008)) while ohers do no idenify any significan gender differences (Akinson, e al (2003). This implies ha furher examinaion on he issue is necessary. Our sudy is an exension of he exising ones. Firs, our sample includes over wo housand fund managers among which over 10% are female and covers he ime period From March 1984 o February I is among he mos recen and includes he larges sample in he similar sudies. Akinson, e al (2003) include only 72 female fund managers (5.6% of he overall sample), and Beckmann and Menkhoff (2008) only have oal 148 U.S. fund managers paricipaing heir survey. We also mach a male manager o each female manager using hree differen crieria including one wih a conrol for self-selecion bias. Using hese various filers allows us o be more focused and beer idenify he gender differences, if any exis. Second, we examine changes in gender composiions of fund managemen and invesigae wheher he change in composiion has a significan impac on he muual fund performance. To he bes of our knowledge, his sudy is he firs such invesigaion of his issue. Specifically, we focus on U.S. domesic sock muual fund managers and assess he gender differences in performance and invesmen behaviors. Our sample includes 2,217 fund managers of which 223 are female managers. Our empirical resuls sugges ha female muual fund managers have a lower risk olerance (sandard deviaion) han male managers and hey spread heir risk ou over more socks and hold a lower percenage of asses in heir op 10 holdings. Compared o male managers, females end o manage no-load funds in he value caegory. We also find ha he gender difference is no a source of abnormal reurns. When we examine performance wihin funds over ime, we idenify he relaionship beween changes in he gender composiion of a fund s managemen and is performance. There is srong evidence ha a fund experiences beer reurns when here is a major change in he Page 3

6 gender composiion of is managemen. There is weak evidence ha alhough boh increasing and decreasing he percenage of female managers will resul in beer fuure reurns in he fund, a larger increase in performance is usually associaed wih a major decrease in he percenage of is female managemen. The res of his paper is organized as follows. The nex secion covers previous lieraure relevan o he ideas presened in his paper. The hird secion gives a descripion of he daa and daa sources. We presen our hypoheses and models in he fourh secion. Secion V. enumeraes he empirical resuls, and he las secion summarizes and concludes. II. LITERATURE REVIEW Our review begins wih he exising research which focuses on he differences beween males and females in he general public making financial decisions. Using a psychology survey experimen conduced among shareholders, securiy analyss, insiuional invesors and general business persons, Eses and Hosseini (1988) find ha women have considerably less confidence han men when i comes o invesmen decisions. Through a behavioral finance experimen run among undergraduae and graduae business sudens, Powell and Ansic (1997) also find ha females were less confiden in heir predicions han males wih a similar level of prior experience and educaion. Sundén and Suree (1998) use he Survey of Consumer Finance sponsored by he Federal Reserve Board o sudy defined conribuion reiremen plans. They observe ha gender and marial saus significanly affec allocaion decisions, i.e., women, especially married women, end o allocae heir invesmens more conservaively han men. Page 4

7 Single men end o choose mosly socks in heir porfolios while married women choose mosly bonds. 4 In conras, Schuber e al. (1999) believe ha survey daa may no be a reliable source o idenify he gender difference in risk-aking behavior and previous sudies ha amplify gender sereoyping may be erroneous. Raher, hey conduc an experimen ha brings ou he risk endencies of each gender in financial siuaions. 5 They find ha men and women s risk-aking behavior depends significanly on he financial decision seing. In a conrolled seing, hey find ha even hough women exhibi more risk-aversion o losses in absrac gambles, here are no differences in risk-aking behavior beween gender when facing invesmen and insurance-relaed decisions. While all he lieraure provides useful insighs upon he gender issue, Powell and Ansic and Schuber e al. used sudens as heir primary subjec for heir surveys and experimens. Due o he capive naure and he age and educaion level of heir subjecs plus he relaively small sample sizes, heir resuls may be more difficul o exrapolae o he general public. Using accoun daa from a large brokerage firm, Barber and Odean (2001) find ha men rade 45 percen more han women, which leads o a greaer annual reducion in heir ne reurns. They aribue he performance and rading aciviy of men o overconfidence in heir invesmen abiliies. The second half of he review focuses on lieraure ha deals more direcly wih managerspecific characerisics ha may influence muual fund reurns. Golec (1996) analyzes wheher non-gender manager characerisics explain fund performance, risk and fees. He examines Bajelsmi and VanDerhei (1997), Hinz, McCarhy, and Turner (1997) and Jianakoplos and Bernasek (1998) have also used surveys and oher economic daa and similarly discover ha women inves more conservaively han men. 5 The insrucions and procedures for he experimen conduced in Schuber e al. (1999) may be found (in English) a hp:// Page 5

8 muual funds over he ime period of 1988 o The manager characerisics being invesigaed include age, enure wih he fund, years of educaion, wheher or no he manager has an MBA degree, managemen eam size, fund age, fund asses, load charge, and fund objecive. He reveals ha manager age, enure and educaion are imporan facors in deermining a fund s performance, risk and fees. 6 Chevalier and Ellison (1999) conduc a similar sudy over a ime frame of 1988 o 1995 during which socks were in he mids of a significan bull marke in he Unied Saes. They examine he characerisics of fund managers such as age, he average composie SAT score a he manager s undergraduae insiuion, and wheher he manager has an MBA. Alhough hey find ha only he undergraduae college aended by he manager urns ou o be relevan, his finding may be due o he sample ha hey used. Some of heir daa was back-filled. Oher daa hey used was manually enered based on assumpions abou he manager s age when graduaing college or abou he schools aended by hese managers. These assumpions may inroduce biases ha lead he auhors o incorrec conclusions. Akinson, Baird, and Frye (2003) fully analyze he gender issue in muual fund managementhey look a fixed-income muual fund managers and compare male and female managers o see if here is a difference in performance, risk, or oher characerisics. According o heir resuls, here are no significan gender differences in muual fund managemen. This suggess o hem ha differences found in he lieraure may be aribued o invesor knowledge and wealh consrains. However, hey do find ha he manager s gender influences he invesors decision on wheher o inves in a fund. Female fund managers receive lower fund inflows in he firs year managing a paricular fund. 6 Goesman and Morey (2006) emphasize he significance of manager educaion as an imporan deerminan in fund performance. They find ha muual fund managers ha received MBAs from a op school achieved significanly greaer performance han oher managers. Page 6

9 Beckmann and Menkhoff (2008) conduc a survey among 649 fund managers in US, Germany, Thailand, and Ialy. There are 148 response from US in which 11.4% is female. While examining he risk behavior in U.S, hey find ha female professionals are more risk averse ha he males. They find no evidence of gender difference in overconfidence. III. DATA All daa were obained from he Morningsar Principia Advanced Muual Funds Module daed February 29, From his CD, he following variables were obained: monhly reurns, ne asses, expense raio, sandard deviaion, oal number of holdings, percenage of asses in op 10 holdings, P/E raio, urnover, he saring dae for he curren managemen, and wheher he fund was a no-load fund or no. The CD includes oal 9,327 U.S. domesic sock muual funds. The following crieria are used o selec he funds for his paper: (1) Funds ha are closed o new invesmen, Bear Marke and Converible Bond funds are eliminaed because heir performance characerisics are differen from oher ypes of muual funds. (2) Exchange-raded funds (ETFs), unmanaged, and eam-managed funds are eliminaed because here is no discernable manager. (3) To assure enough of a performance rack record for he monhly analysis, a minimum hree-year performance measure is required. 7 (4) When muliple classes exis for he same fund, he class wih he longes hisory is seleced (Carhar (1997)). 8 If wo or more classes have idenical incepion daes, he one wih larger ne asses is seleced. The filered sample has 2,217 oal funds of which 233 are primarily managed by women (10.51%). Where here is missing daa, he 7 The hree-year ime frame is he same ha Morningsar uses o provide heir firs Morningsar Raing. 8 The classes represen a common pool of asses bu differ in erm of how disribuion-relaed fees are paid. Page 7

10 mos recen fund repor (prospecus, annual, and/or semi-annual repors) was used o correc his deficiency. Manager informaion is also available on he disk including he names of all curren and prior fund managers and a shor biography of mos curren managers. Ou of he 2,217 fund managers seleced, only 909 of hem (41.0%) have educaion background informaion available on he Morningsar CD. We find ha he availabiliy of he educaion-relaed maerial was largely dependen upon he fund family o which he muual fund belonged. I appears ha many of he larger fund families do repor managers educaion daa (such as Fideliy and Vanguard) and many smaller families do no. We would expec he educaion saisics o be biased oward higher educaion han he indusry average because presumably he larger groups have he abiliy o selec op candidaes a premium salaries. Therefore, we choose no o inroduce educaion variables ino our paper. Among he filered sample of 2,217 funds, wo caegories of sub-samples are creaed. In he firs caegories, each of 233 female managers are mached o a male counerpar based upon size and enure, sandard deviaion (risk) and enure, and he percen of asses in he op 10 holdings and enure wih heir closes male counerpars. The size-enure mached sample aemps o alleviae bias based on diseconomies of scale and manager experience. This maching is similar o Akinson, e al. (2003). The sandard-deviaion-enure mached sample aemps o alleviae a poenial self-selecion bias as well as conrol for manager experience. The selfselecion bias occurs because muual fund managers self-selec o highly compeiive and risky field. They may be less risk-averse in naure compared o he res of general populaion regardless of gender. We aemp o assure a comparable level of risk aversion among paired sample by using he sandard deviaion. The asse-holding-enure mached sample is creaed in Page 8

11 order o conrol for a poenial bias in asse concenraion beween he mached sample and he full sample of male domesic equiy fund managers. In shor, each mached sub-sample consiss of he 233 female managers and heir 233 male counerpars. In he second caegory of sub-samples, we selec funds ha had (since March 1984) a major change in managemen gender composiion, eiher from female o male or male o female. A major change is defined as a gender change of a leas 50% of he managemen composiion of a fund a a specific ime. For example, a fund managed by one male manager (100% male and 0% female) changes o one female manager (0% male and 100% female). Or a fund managed by hree female managers (100% female and 0% male) swiches o one female manger and wo males (33% female and 67% male). Addiionally, he same managemen composiion mus remain in conrol of he fund for a leas 12 monhs boh before and afer he changing dae. As a resul, he changing composiion sub-sample includes 102 funds wih a oal of 150 gender composiion changes (90 are male o female changes and he remainder are female o male changes). IV. HYPOTHESES AND MODELS Hypohesis one: wihin syles of muual funds, fund characerisics do no vary due o fund managers gender. The majoriy of academic research agrees ha, in general, men and women have sysemaic behavioral differences in he way ha hey inves: women end o be more risk averse han men. Beckmann and Menkhoff (2008) observe he same risk aversion differen beween genders in fund managers using a survey daa. Unforunaely, heir sample size is raher small. Addiionally, Schuber e al. (1999) poin ou ha conclusions drawn from a survey may be Page 9

12 erroneous since survey daa lack adequae conrols for individual risk-aking behavior. Using non-survey daa, Akinson, e al. (2003) are unable o find a significan difference in performance beween male and female domesic fixed-income muual fund managers I is reasonable o believe ha given heir superior raining and knowledge of invesmens relaive o he general public, i is likely ha performance differences beween male and female fund managers are insignifican. Therefore, we expec ha here are no differences in fund characerisics beween genders. To es his hypohesis, paired -ess ha compare he means of female and male managers are performed on he following well-known fund characerisics using he hree mached subsamples: expense raio, sandard deviaion (proxy for risk), number of holdings (proxy for diversificaion), percenage of asses in op 10 holdings (proxy for concenraion and diversificaion), P/E raio (proxy for value), urnover defined by purchases or sales (whichever is less) divided by he average monhly ne asses, wheher he fund was a no-load fund or no, ne asses (fund size) and manager s enure (o conrol for experience). If he resuls of paired -ess are insignifican, hypohesis one holds, i.e., here are no gender differences in fund characerisics. If he -es resuls are significan, he hypohesis is rejeced. This means ha muual fund managers inves differenly due o gender. Compared o heir male counerpars, female managers more likely have lower urnover (due o less overconfidence) and inves more in conservaive socks (due o more risk aversion), which leads o higher long-erm reurns. Hypohesis Two: Reurns vary in a sysemaic way due o fund managers gender. We expec ha he similariies in he educaion and experience of fund managers will ouweigh he differences in gender, which means ha here should no be significan differences Page 10

13 in fund performance due o gender. However, if daa ell us oherwise, i will inform invesors how o increase reurns based upon he gender characerisics of muual fund managemen. Furher, i will inform he muual funds, hemselves, how o adjus heir managemen gender composiions o increase performance. A one-index model and wo muli-facor models are applied o examine he performance of he male- and female-managed muual funds. Specifically, r = α + β I + γ Gender + ε p, p p p p p, r 1 2 = α + β I + β SMB + β HML + γ Gender + ε p, p p p p p p p, r = α + β I + β SMB + β HML + β UMD + γ Gender + ε p, p p p p p p p p p, (1) (2) (3) where r is he excess reurn on he p h fund during monh, α is he abnormal excess reurn p, on he porfolio, β is he sandard marke-model risk measure, I is he excess reurn on a p marke index,,, and are he sensiiviies of he excess reurns on facors SMB, HML, β p β p β p p and UMD, respecively, and Gender p is a dummy variable equal o one if he manager is female and zero if male. We use he Russell 1000 index and he S&P 500 index as a proxy for he marke. The Russell 1000 is chosen because i is he bes-fi index for all domesic equiy funds according o Morningsar. The S&P 500 is chosen because some of our funds go back as far as March 1984, while he Russell 1000 is no available unil January As a much more widely used proxy for he marke in he lieraure, we also use he S&P 500 for robusness comparisons. SMB (Small Minus Big) is he average reurn on hree small porfolios minus he average reurn on hree big porfolios. I encapsulaes he idea ha small company socks ouperform large company socks over ime. HML (High Minus Low) is he average reurn on wo value Page 11

14 porfolios minus he average reurn on wo growh porfolios. HML encompasses he asserion ha value socks (high book-o-marke) ouperform growh socks (low book-o-marke) over ime. UMD (Up Minus Down) is he momenum facor, defined as he average reurn on he wo high prior reurn porfolios minus he average reurn on he wo low prior reurn porfolios from six value-weighed porfolios formed on size and prior reurns. 9 Pooled-OLS regression resuls are obained for all hree models using he full sample of 2,217 funds. The focus is on he dummy variable, Gender p. If he coefficien of he gender dummy is insignifican, i indicaes ha here is no evidence of he difference in fund performance due o gender. However, if he dummy is posiive and significan, he hypohesis is rejeced. This resul will imply ha female managers produce beer reurns han male managers. We ake our analysis one sep furher. Wihin a given fund, we examine wheher here is any performance change due o he gender composiion change in he managemen using he changing composiion sub-sample. The one-index model and wo muli-facor models along wih a new dummy variable are used. Specifically, r = α + β I + γ Change + ε p, p p p p p, r 1 2 = α + β I + β SMB + β HML + γ Change + ε p, p p p p p p p, r = α + β I + β SMB + β HML + β UMD + γ Change + ε p, p p p p p p p p p, (4) (5) (6) where Change p is equal o one when here is a higher percenage of female managers beween he wo periods, and zero denoes a relaively higher percenage of male managers beween he wo periods. We are direcly comparing wo ime periods wihin he managemen of a single 9 The UMD facor and he Fama-French facors are available o he public from Kenneh French s web sie a hp://mba.uck.darmouh.edu/pages/faculy/ken.french/daa_library.hml Page 12

15 muual fund. As shown in he following diagram, in order o allow new managemen o be fully in conrol of he fund s asses (and herefore, reurn), we skip a six-monh window righ afer every change in he gender composiion in he managemen. We use boh 12- and 24-monh periods for robusness and o be sure ha we are using enough daa. Old Managemen Period New Managemen Period 12- or 24-monh esed period 6-monh window (no esed) 12- or 24-monh esed period A panel regression is run on he hree models o deermine if he change in gender is significan. The changing composiion sub-sample is used here. We furher separae he sample ino male-o-female changes and he female-o-male changes in managemen composiion. Since in all cases, he change dummy is equal o one for an increase in female managemen, a posiive and significan resul for his dummy will indicae ha female managemen ouperforms male managemen over ime wihin a fund. Finally, we analyze changes in manager gender uses he acual percenage of female managers for each monh of he fund s exisence, or as far back as March 1984, whichever is less. The models used are as follows: Page 13

16 r = α + β I + γ Female + ε p, p p p p p, r 1 2 = α + β I + β SMB + β HML + γ Female + ε p, p p p p p p p, r = α + β I + β SMB + β HML + β UMD + γ Female + ε p, p p p p p p p p p, (7) (8) (9) where Female p is equal o he percenage of female managers managing he fund in any given monh. If he coefficien for Female p is posiive and significan, hen he percenage of female managemen is posiively relaed o he performance of he fund over ime. V. EMPIRICAL RESULTS A descripion of female domesic equiy muual fund managers relaive o oal funds in our sample is illusraed in Table 1. All funds are separaed ino four caegories: value, blend, growh and oher. Panel A shows ha female managers comprise abou 10.5% of he seleced domesic equiy muual fund managers. Addiionally, among he four fund caegories females are slighly over-represened in he blend caegory and under-represened in he value caegory: almos 12% of blend managers are female, while only a lile over 9% of value managers are female. Panel B of Table 1 compares he percenage allocaion of seleced female fund managers in he four caegories o he percenage of he overall sample. The number and percenage of female managers in value funds is lowes. Also, 29% of female mangers run blend funds. This is somewha higher han he proporion of blend fund managers o he overall sample (26%). The proporions of female managers in growh and oher caegories are almos idenical o he proporion of female managers o oal funds. These wo panels aken ogeher sugges ha here is a endency for females o manage blend funds and no o manage value funds. Page 14

17 Comparisons of 10 fund characerisics beween male and female fund managers relaed o hypohesis one are presened in Tables 2, 3 and 4. The funds were separaed ino he value, blend, growh, and oher syle caegories. The comparison resuls for he size-enure mached sample are shown in Table 2.The resuls indicae ha female value fund managers have a higher urnover raio and a higher propensiy o manage a no-load fund. The higher urnover is conrary o he finding ha females in he general populaion end o rade less. Female blend fund managers end o have a slighly higher P/E raio han mached males. Female growh fund managers have a lower propensiy o manage a no-load fund. And, female managers managing he oher funds inves significanly less asses in heir op en holdings. Comparisons beween male and female managers across all four caegories show ha male fund managers have a higher sandard deviaion in reurns and end o concenrae heir asses more in heir op en holdings suggesing hey may be less diversified han female managers. Addiionally, male managers have a significanly higher urnover raio. Table 3 reveals he comparison resuls using he sandard-deviaion-enure sub-sample. For he value funds, he gender difference on urnover disappears bu female managers are sill more likely o manage a no-load fund hen male ones. Moreover, here is a significan difference in he oal number of holdings, which implies ha female value fund managers under he curren mach hold more socks han male managers. The resuls also show ha female managers in he blend, growh and oher caegories hold less asses in heir op 10 holdings. Furhermore, female blend fund managers have a higher urnover raio, female growh fund managers have a higher P/E raio, and, females in he oher caegory hold more socks han heir male counerpars. When we look a comparisons across all four caegories, evidence shows ha females hold more Page 15

18 socks, have a lower concenraion in heir op 10 holdings and a higher P/E raio han male managers. The resuls of gender comparisons among he holding-enure mach sub-sample are presened in Table 4. The gender differences of he fund characerisics in value funds are consisen wih he sandard-deviaion-enure mached sample: females hold more socks and are more likely o manage a no-load fund han males. There are no significan differences beween male and female blend fund managers. In he growh caegory, males are more likely han females o manage a no-load fund, which is same as shown in size-enure sub-sample. In he oher caegory, female managers have somewha lower sandard deviaion in he fund reurns. When look a he gender differences for all caegories combined, female managers have slighly lower sandard deviaions in reurns and rade much less han male managers. Hypohesis one is rejeced because we do observe cerain gender difference in fund characerisics wihin syles of muual funds. However, when reviewing and analyzing Tables 2, 3, and 4 ogeher, i is difficul o draw an unambiguous conclusion based upon he paired -es resuls. The gender differences in fund characerisics ofen end no o reain among all hree sub-samples wih differen maching crieria. The mos robus resul in all hree sub-samples is ha females are more likely o manage a no-load fund in he value caegory han are males. Also, he following gender differences are shown in any wo ou of hree mached samples: sandard deviaion is significan in boh ables ha are no mached by sandard deviaion. Female managers ypically achieve less sandard deviaion in reurns, have a significanly lower percenage of asses in heir op 10 holdings, and have more holdings han males. Each of hese findings seems o be consisen wih he findings in he general public (see Barber and Odean (2001) among ohers) ha females have a lower risk olerance, which means ha female Page 16

19 managers ry o spread heir risk ou over many socks and hold few asses in heir op 10 holdings. Finally, i is worh poining ou ha ne asses and expense raio are never significanly differen beween he genders in all sub-samples. The remaining discussion of he empirical resuls is focused on he fund performance differences due o managers gender. We analyze panel daa of managers currenly in conrol of all 2,217 funds using equaions (1) o (3). The resuls presened in Table 5 show ha he gender dummy is no saisically significan in all hree models. This implies ha under curren managemen, fund manager s gender does no affec fund performance. Addiionally, one-index and hree-facor models show ha muual funds ouperform he marke regardless under male or female managers. However, afer considering he momenum facor, he significan posiive abnormal reurns disappear. Now, we urn our analysis o gender differences in performance when here is a major change in he gender of fund managemen wihin a single muual fund. We apply he changing composiion sub-sample o equaions (4) o (6). We analyze he concluding 12-monh period for previous managemen, skip a six-monh window so ha new managemen can adjus he fund s porfolio according o is preferred mehod of invesing, and analyze he following 12-monh period for new managemen. 10 The resuls of pooled regressions are shown in Table Panel A displays he resuls from all 129 changes in managemen gender. The coefficiens of dummy variable for changing composiion in managemen are negaive bu insignifican in all equaions. The abnormal reurn (alpha) is no significan in he single index equaions, bu i is in he mulifacor equaions. We furher explore he assumpion ha he direcion of changes in he 10 To run a robusness check, he same analysis was conduced using 24-monh esing period. There are no fundamenal changes in he resuls. Therefore, hey are no presened in he paper. 11 We perform SBIC es for pooled and fixed effecs. All of he SBIC resuls favor using he pooled regression. Furher, none of he fixed-effecs resuls were maerially differen from he pooled regressions. Page 17

20 managemen gender composiion may affec he fund performance. Panel B shows he resuls of a change from a lower percenage of female managers o a higher percenage of female managers (i.e. male-o-female). The coefficiens of dummy for changing composiion are now posiive bu again insignifican. The abnormal reurns are negaive and significan in all equaions. In Panel C, changes in managemen gender from a higher percenage of female mangers o a lower percenage of female managers (i.e. female-o-male) are shown. The coefficiens of dummy for change are now significan and negaive in all equaions while he abnormal reurns are no longer significan. In shor, hypohesis wo is rejeced under cerain condiions. I appears ha abnormal reurns are significan only when managemen changes from a lower percenage o a higher percenage of female managers, suggesing ha, in his case, all managers have significanly negaive reurns and here is no significan difference in he gender of he manager. However, when managemen changes from a higher percenage o a lower percenage of female managers, he gender of he manager is significan. The resuls show ha fund reurns become beer afer new managemen wih lower percenage of females is in place. Finally, we examine he impac of gender composiion change in managemen hrough he percenage of female manager in a given fund using he same changing composiion subsample. Raher han looking only a he manager enures immediaely before and afer a major change in manager gender composiion, we assign a percenage of female managers o each monh of a fund s exisence or as far back as 20 years (March 1984), whichever is less. As an example, his number is equal o 1 if he only managers of a fund are female or 0.25 if here are four managers of a paricular fund and one of hem is female. The variable, Femalein equaions (7) o (9), picks up changes in gender composiion of fund managemen regardless wheher i is a major or minor change. Page 18

21 The resuls of pooled regressions are shown in Table 7. The coefficien of Female is negaive bu insignifican in he single-index models. However, i becomes negaive and significan in he muli-facor models. The muli-facor models may be beer since hey consisenly have higher adjused R-squareds, indicaing ha hese models can explain a higher percenage of he excess reurns in he funds. So, he regression resuls sugges ha he percenage of female managers of a paricular fund is inversely relaed o is performance. VI. CONCLUSION Lieraure in recen years suggess ha females are more conservaive invesors han males. Bu fund managers ypically have more invesmen educaion han he average person. So, he quesion of wheher or no female fund managers exhibi more conservaive endencies is an ineresing one. Along wih his quesion, we explore wheher or no he invesmen decisions of female fund managers affec he performance of he funds ha hey manage in a differen way han do he invesmen decisions of male managers. We find a few differences in he mached comparison ha sugges ha female managers have a lower risk olerance. They end o spread heir risk ou over many socks and hold few asses in heir op 10 holdings. Also, females managing value funds manage no-load funds significanly more han males. When examining curren managers for all 2,217 domesic equiy funds, we find ha here are no significan differences in performance of muual funds due o managemen gender. This is consisen wih Akinson e al. (2003) where hey examine fixed-income funds. We hen analyze performance wihin funds over ime in order o examine he impac of changes in managemen s gender composiion on he fund performance. When considering all Page 19

22 funds wih major gender composiion changes, here is no evidence shown ha he composiion changes lead o a corresponding performance change. However, afer we break down he issue, i is raher clear ha he performance improves afer a reducion in he percenage of female managemen in a given fund. We also find ha using he muli-facor models he percenage of female managers in a fund is negaively relaed o he fund s performance over ime. Finally, even hough we show evidence here ha an inverse relaionship exiss beween he percenage of female managers in a fund and is performance, here are hree reasons why one should be cauioned in inerpreing and applying our empirical resuls. Firs, alhough significan, he absolue value of he differences beween funds wih male managers and funds wih female managers is small. So, an invesor may no benefi simply by seeking a male muual fund manager. Second, he resuls should no be blindly applied o any individual fund manager. Many female managers ouperform male managers. Third, all of our resuls are based upon hisorical daa. If muual fund managers abiliies have evolved over he las weny years, resuls may be differen for oday s managers. Page 20

23 REFERENCES Akinson, Sanley M., Samanha Boyce Baird, and Melissa B. Frye, Do Female Muual Fund Managers Manage Differenly?, The Journal of Financial Research, 2003, pp Bajelsmi, Vickie L. and Jack. L. VanDerhei, Risk Aversion and Pension Invesmen Choices, in Michael S. Gordon, Olivia S. Michell, and Marc M. Twinney, eds., Posiioning Pensions for he Tweny-Firs Cenury, Universiy of Pennsylvania Press, Philadelphia, 1997, pp Barber, Brad M. and Terrance Odean, Boys Will Be Boys: Gender, Overconfidence, and Common Sock Invesmen, Quarerly Journal of Economics, 2001, pp Barber, Brad M. and Terrance Odean, Trading is Hazardous o Your Wealh: The Common Sock Invesmen Performance of Individual Invesors, The Journal of Finance, 2000, vol. 55, no. 2, pp Barberis, Nicholas and Andrei Shleifer, Syle Invesing, Journal of Financial Economics, 2003, vol. 68, pp Beckmann, Daniela and Lukas Menkhoff, Will women be women? Analyzing he gender difference among financial expers, KYKLOS, 2008, vol. 61, no. 3, pp Carhar, Mark M., On Persisence in Muual Fund Performance, The Journal of Finance, 1997, vol. 52, no. 1, pp Chevalier, Judih and Glenn Ellison, Are Some Muual Fund Managers Beer Than Ohers? Cross-Secional Paerns in Behavior and Performance, The Journal of Finance, June 1999, pp Eses, R. and J. Hosseini, The Gender Gap on Wall Sree: An Empirical Analysis of Confidence in Invesmen Decision Making, Journal of Psychology, 1988, pp Golec, Joseph H., The Effecs of Muual Fund Managers Characerisics on Their Porfolio Performance, Risk and Fees, Financial Services Review, 1996, vol. 5, no. 2, pp Goesman, Aron A. and Mahew R. Morey, Manager Educaion and Muual Fund Performance, Journal of Empirical Finance, 2006, vol. 13, no. 2, pp Lakonishok, Josef, Andrei Shleifer and Rober W. Vishny, Conrarian Invesmen, Exrapolaion, and Risk, The Journal of Finance, 1994, vol. 49. no. 5, pp Page 21

24 Powell, Melanie and David Ansic, Gender Differences in Risk Behaviour in Financial Decision- Making: An Experimenal Analysis, Journal of Economic Psychology, 1997, pp Schuber, Renae, Marin Brown, Mahias Gysler, and Hans Wolfgang Brachinger, Financial Decision-Making: Are Women Really More Risk-Averse?, American Economic Review, 1999, pp Sundén, Annika E. and Brian J. Suree, Gender Differences in The Allocaion of Asses in Reiremen Savings Plans, American Economic Review, 1998, pp Page 22

25 TABLE 1. Female Domesic Equiy Muual Fund Managers Panel A. Female Managers as a Percenage of Funds by Caegory Caegory # of Female Managers Toal Funds Percenage Value % Blend % Growh % Oher % Toal 233 2, % Panel B. Female Managers by Caegory as a Percenage of All Female Managers and Toal Funds by Caegory as a Percenage of All Funds Caegory # of Female Managers Percenage Toal Funds Percenage Value % % Blend % % Growh % % Oher % % Toal % 2, % Noe: This able illusraes he relaive proporion of domesic equiy fund managers ha are female. Only funds acively managed primarily by a single male or female manager were considered when compiling his able. All daa was obained from he Morningsar Principia Advanced Muual Funds Module daed February 29, Page 23

26 TABLE 2. Male and Female Muual Fund Manager Comparison of Means (Mached by Ne Asses and Tenure) Caegory (Observaions) Expense Raio Sandard Deviaion Toal Number of Holdings % Asses in Top 10 Holdings P/E Raio Turnover Raio No-load Ne Asses Tenure (Days) Value (35 pairs) Male Female p-value ** 0.003*** Blend (68 pairs) Male Female p-value * Growh (75 pairs) Male Female p-value ** Oher Funds (55 pairs) Male Female p-value *** All Funds (233 pairs) Male Female p-value * ** * Noe: This able shows a comparison of means and p-values (o repor he significance of he difference beween he wo paired sample means) for several imporan variables ha are relaed o a fund s invesmen objecive or o is managemen syle. In each caegory, a porfolio of fund reurns is averaged and he means are compared wih hose of he opposie gender. The number in parenheses nex o each caegory name is he number of disinc female fund managers ha exis in ha caegory. These were mached by oal asses and enure o he lis of funds managed by male managers, and he closes maches were seleced for he male porfolio. All daa was obained from he Morningsar Principia Advanced Muual Funds Module daed February 29, * Significan a he 10% level, ** Significan a he 5% level, *** Significan a he 1% level. Page 24

27 TABLE 3. Male and Female Muual Fund Manager Comparison of Means (Mached by Sandard Deviaion and Tenure) Caegory (Observaions) Expense Raio Sandard Deviaion Toal Number of Holdings % Asses in Top 10 Holdings P/E Raio Turnover Raio No-load Ne Asses Tenure (Days) Value (35 pairs) Male Female p-value * * Blend (68 pairs) Male Female p-value *** ** Growh (75 pairs) Male Female p-value *** 0.045** Oher Funds (55 pairs) Male Female p-value * 0.019** All Funds (233 pairs) Male Female p-value *** 0.000*** 0.082* Noe: This able shows a comparison of means and p-values (o repor he significance of he difference beween he wo paired sample means) for several imporan variables ha are relaed o a fund s invesmen objecive or o is managemen syle. In each caegory, a porfolio of fund reurns is averaged and he means are compared wih hose of he opposie gender. The number in parenheses nex o each caegory name is he number of disinc female fund managers ha exis in ha caegory. These were mached by oal asses and enure o he lis of funds managed by male managers, and he closes maches were seleced for he male porfolio. All daa was obained from he Morningsar Principia Advanced Muual Funds Module daed February 29, * Significan a he 10% level, ** Significan a he 5% level, *** Significan a he 1% level. Page 25

28 TABLE 4. Male and Female Muual Fund Manager Comparison of Means (Mached by % Asses in Top 10 Holdings and Tenure) Caegory (Observaions) Expense Raio Sandard Deviaion Toal Number of Holdings % Asses in Top 10 Holdings P/E Raio Turnover Raio No-load Ne Asses Tenure (Days) Value (35 pairs) Male Female p-value * *** Blend (68 pairs) Male Female p-value Growh (75 pairs) Male Female p-value * Oher Funds (55 pairs) Male Female p-value * All Funds (233 pairs) Male Female p-value ** * Noe: This able shows a comparison of means and p-values (o repor he significance of he difference beween he wo paired sample means) for several imporan variables ha are relaed o a fund s invesmen objecive or o is managemen syle. In each caegory, a porfolio of fund reurns is averaged and he means are compared wih hose of he opposie gender. The number in parenheses nex o each caegory name is he number of disinc female fund managers ha exis in ha caegory. These were mached by oal asses and enure o he lis of funds managed by male managers, and he closes maches were seleced for he male porfolio. All daa was obained from he Morningsar Principia Advanced Muual Funds Module daed February 29, * Significan a he 10% level, ** Significan a he 5% level, *** Significan a he 1% level. Page 26

29 TABLE 5. Gender Significance in Curren Muual Fund Managers One-index Regression 3-Facor Regression 4-Facor Regression Russell 1000 S&P 500 Facor Es. Coeff Facor Es. Coeff Facor Es. Coeff Facor Es. Coeff ALPHA *** ALPHA *** ALPHA ** ALPHA *** GENDER GENDER GENDER GENDER Russell *** S&P *** RMRF *** RMRF *** SMB *** SMB *** HML *** HML *** UMD *** Adjused R-squared Pooled regressions were performed on reurns from all 2,217 curren domesic equiy managers using he equaions: = and r p, = α p + β p, iii, + γ pgenderp + ε p,, where GENDER is a dummy equal o 1 if he r p, α p + β pi + γ pgenderp + ε p, J i= 1 curren manager is female for all reurns in monhs of he curren manager's enure and 0 if he curren manager is male. Sandard errors are below coefficiens. * Significan a he 10% level; ** Significan a he 5% level; *** Significan a he 1% level. Page 27

30 TABLE 6. Dynamic Change in Manager Gender (12-Monhs Before and Afer) PANEL A: All Major Changes in Managemen Gender RUSSELL 1000 S&P 500 Fama-French 3 Facors Carhar 4 Facors Facor Es. Coeff. Facor Es. Coeff. Facor Es. Coeff. Facor Es. Coeff. Alpha Alpha Alpha *** Alpha *** CHANGE CHANGE CHANGE CHANGE RUS1000EX *** SP500EX *** RMRF *** RMRF *** SMB *** SMB *** HML ** HML UMD *** Adjused R-squared Fixed-Effecs F-Tes Fixed-Effecs P-Value SBIC - Pooled OLS SBIC - Fixed Effecs No. Observaions Page 28

31 TABLE 6. Dynamic Change in Manager Gender (12-Monhs Before and Afer) (coninued) PANEL B: Major Changes in Managemen Gender from Male-o-Female RUSSELL 1000 S&P 500 Fama-French 3 Facors Carhar 4 Facors Facor Es. Coeff. Facor Es. Coeff. Facor Es. Coeff. Facor Es. Coeff. Alpha ** Alpha *** Alpha *** Alpha *** CHANGE CHANGE CHANGE CHANGE RUS1000EX *** SP500EX *** RMRF *** RMRF *** SMB *** SMB *** HML HML *** UMD *** Adjused R-squared Fixed-Effecs F-Tes Fixed-Effecs P-Value SBIC - Pooled OLS SBIC - Fixed Effecs No. Observaions Page 29